impacts of improved irrigation and drainage systems of the yinchuan … · yinchuan plain it was...

39

From: Willett, I.R. and Zhanyi Gao, ed., 2006. Agricultural water management in China. Proceedings of a workshop held in Beijing, China, 14 September 2005. Canberra, ACIAR Proceedings No. 123.

Impacts of improved irrigation and drainage systems of the Yinchuan Plain, northern China

Riasat Ali1 and Jiamin Wu2

Abstract

The Yinchuan Plain, located on the banks of the Yellow River, is historically one of the largest irrigation areasin northwestern China, having a large network of irrigation and drainage channels. Irrigation from the YellowRiver over a long period has caused waterlogging, secondary salinity, shallow groundwater levels andenvironmental pollution. In addition, over-exploitation of deeper groundwater in some parts of the plain iscausing groundwater pollution due to leakage from shallow unconfined aquifers. This paper summarisesresearch to meet these problems.

The research project identified areas that are at high risk of salinity and shallow groundwater leveldevelopment. The shallow groundwater is saline, and extensive areas of the Yinchuan Plain have medium soilsalinity risk. There is widespread pollution of both surface and groundwater from nutrients and salts. Shallowgroundwater in more than 50% of the Yinchuan Plain has been polluted. There is excessive seepage fromirrigation channels to the surrounding land, resulting in the development of shallow watertables and soilsalinity.

It was determined through field trials and geochemical modelling that up to 50% shallow groundwater canbe mixed with surface water for irrigation without any significant losses in crop productivity. Reduction in cropyield is expected if groundwater alone is used for irrigation. Field experiments and modelling suggested that,by replacing flood irrigation with furrow irrigation, about 35% of the irrigation water can be saved withoutsacrificing productivity. Deep, open drains are effective for lowering the shallow watertables and reducing soilsalinity. In some areas around Yinchuan city, the groundwater abstraction from the first confined aquifershould be reduced to avoid leakage and pollution from shallow groundwater. The surface water levels in SandLake should be lowered by 0.5 m to help arrest the spread of salinity in the surrounding areas.

Based on these research findings, investment plans for salinity management, surface water pollution controland irrigation and drainage management have been prepared by the local government and agencies. The localgovernment has already installed 3000 shallow groundwater production wells in the region to help controlshallow groundwater levels and supplement irrigation. The Yellow River water quota for the region has beenreduced by about 30%.

1 CSIRO Land and Water, Private Bag 5, Post Office,Wembley, Western Australia 6913, Australia. Email:<[email protected]>.

2 Ningxia Remote Sensing Center, Yinchuan, Ningxia750021, People’s Republic of China.

Chinawater.book Page 39 Monday, July 24, 2006 11:33 AM

40



41


Introduction

Ningxia Autonomous Region is located in the north-west of China and the Yinchuan Plain occupies itsnorthern parts (Figure 1). It covers an area of approx-imately 7790 km2. The Yinchuan Plain is surroundedby the Helan Mountains in the west and the E’erduosihighland in the east (Figure 2). The Yellow Riverenters the plain from the southwestern boundary andfollows along its southern boundary to the east andcontinues northward. The plain slopes gently towardsthe east and north. The plain is approximately 165 kmlong and varies in width between 40 and 60 km. Thepredominant soil types are alluvial and Podzolic (Wuand Yao 2004). There are numerous lakes andswamps in the interior parts of the plain.

The climate of the Yinchuan Plain is arid, with amean annual rainfall of approximately 200 mm,about 60–75% of which falls between July and Sep-tember. Precipitation provides only about 10–20% oftotal crop water requirements. Average annual poten-tial evaporation is around 1400 mm and there are140–160 frost-free days. The average annual temper-ature is approximately 9˚C.

Around 2.7 million people live on the plain, whichhas a history of more than 2000 years of agricultureand irrigation development. The first canal was con-structed at the southern end of the Yinchuan Plainduring 214 BC. Early in the Qing Dynasty (214 BC)up until the 1960s, more than 10 main canals and 13main drains were constructed to meet the irrigationand drainage requirements of the plain (Figure 3).Total irrigation water withdrawal from the YellowRiver is around 5600 GL/annum for 450,000 ha ofirrigated area on the plain (Feng 2003). The agricul-tural sector consumes 93% and the remainder is used

by industry and for domestic water supplies. Themain crops are rice, wheat and maize, but fruits andvegetables are also grown. Raising fish in the lakesand ponds is also common.

Irrigation occurs mainly through flood and furrowirrigation. The efficiency of the irrigation system is, asa whole, very low (36%). Application rates of 15 ML/ha (1000 m3/mu) to 24 ML/ha (1600 m3/mu) arereported for crops where a crop evapotranspirationdemand of 7.5–9.0 ML/ha (500–600 m3/mu) isexpected. Rice irrigation is reported to be up to 49.5ML/ha (3300 m3/mu); that is, 2–3 times the actualcrop need. The cause of this poor performance iscomplex but includes inadequate and poor manage-ment of water distribution systems to farms and onfarms, inefficient irrigation practices and low water

Figure 1. Location of the Yinchuan Plain in China

Figure 2. Map of the Yinchuan Plain, NingxiaAutonomous Region, northwestern China


42


prices, the last discouraging investment in highervalue crops, more efficient production methods orgroundwater pumping (Wang and Gan 2003). Exces-sive use of irrigation water over time, poor subsurfacedrainage and excessive seepage from poorly main-tained irrigation and drainage systems has resulted inincreased accessions to groundwater, causing thewatertable to rise at an alarming rate on some parts ofthe plain (0.2 m/year). As a result, shallow waterta-bles (1–2 m) have developed in most of the northernparts of the plain. This has led to the development ofsecondary salinity in about 40% of the plain (Anyin2002). In some areas, the problem of salinisation hasbecome very severe. Excess water from irrigationwater overuse has been discharged into the drainagesystem. This drainage water exports nutrients,causing pollution of surface water resources of theplain. The drainage system of the plain also receivesrural sewage and effluent from local industries.

Drainage water carrying excessive levels of nutri-ents and industrial and sewage contaminants is dis-charged into the Yellow River, polluting it and

creating potential problems for the downstream users.The inland-irrigated areas of northern China are cur-rently suffering from acute problems of (a) watershortage for irrigation and domestic water supplies,(b) rapid expansion of waterlogged areas, (c) degrada-tion of soils due to alkalinity, sodicity and salinity, (d)increase in the rate of salt discharge to rivers andlakes, (e) depletion and contamination of groundwaterand (f) increasing rates of discharge of nutrient andother pollutants to surface-water systems.

Sustainable agricultural production in the Yin-chuan Plain requires urgent improvement in waterand salinity management; otherwise the agriculturesector is likely to face serious water availability,water quality, land degradation and environmentalpollution problems. The overall aim of the researchreported on here was to improve water and salinitymanagement on the Yinchuan Plain to increase grainproduction and the availability of good quality waterfor agricultural and other uses, reduce soil salinisa-tion and minimise environmental pollution.

Figure 3. Map showing the Yellow River and theirrigation and drainage network on theYinchuan Plain

Figure 4. Locations of the network of observationwells on the Yinchuan Plain

Chinawater.book 2 Monday, July 24, 2006 11:33 AM

43


Material and methods

To improve water and salinity management on theYinchuan Plain it was first necessary to evaluate thewater resources of the region, identify mechanisms ofwater wastage, quantify the main processes causingsoil salinisation and development of shallow water-tables, trace sources of surface water pollution, andthen suggest strategies for their improvement.

To determine hydrological response units (HRU) ofthe Yinchuan Plain, groundwater level data for thepast 20 years were collected, collated and analysed.The regional geology, landform characteristics,lithology and aquifer structure were also collected andcollated. The HRU were constructed with the help oftime series of groundwater level data and other hydro-logical and physical properties of the plain and aqui-fers, following the procedure described by Salama etal. (2002). The plain has an extensive network of newand existing observation wells (Figure 4). Thegroundwater data from these wells were analysed forthe past 20 years to assess the rates of water level rise(RR) and depth to water table (DTW) in various partsof the Yinchuan Plain. A weighting factor (HZ) wasintroduced for any differences in the regional geo-graphic position, landform unit, lithology of stratumand aquifer structure. Accordingly, the Yinchuan

Plain can be divided into eight hydrological units(Figure 5) and a weighting factor for each unit is givenin Table 1. Zhang et al. (2004b) give further details ofthe methodology. The HRU was calculated as:

HRU = DTW × RR × HZ (1)

Over 300 groundwater and 200 soil samples werecollected during 2002 and 2003 to assess and mapgroundwater and soil salinity on the Yinchuan Plain.These data were also required for geochemical mod-elling and analyses. Remote sensing was used toassess the soil salinity, with soil sampling data usedfor ground-truthing.

Water pollution is a major issue on the YinchuanPlain where contaminants from farms, rural sewageand industrial waste is polluting the surface waterresources of the plain and the Yellow River. To quan-tify the rates of flow and pollutant discharge intodrains, canals and the Yellow River, and assess thesources and types of pollutant discharge, Huinongcanal and associated drains were selected for study.Huinong canal is one of five main major canals onYinchuan Plain. Drain No. 5 associated with thiscanal discharges into the canal 10 km upstream of itsoutlet into the Yellow River. Electrical conductivity(EC) loggers were installed at various locations in thecanal, drain and Yellow River to monitor water

Table 1. Hydrogeological conditions and weighting factors for hydrological zones in the Yinchuan Plain

ID Name of zone Lithology of aquifer

Depth(m)

Aquiferyield

(m3/day)

Total dissolved solids(g/L)

Depth towatertable

(m)

Weighting factor

a Qintongxia pluvial fan Sand and gravel

10–300 2000–5000 <1 0.5–4 3

b Pluvial gradient plain of Helan mountain foot

Gravel >100 1000–5000 <1 5–30 3

c Wu-Lin plain east Yellow River

Sand and fine sand

10–50 500–2000 <3 1–5 6

d Alluvial plain of south Yinchuan Plain

Mid-sand and fine sand

10–40 500–2000 <1 1–10 5

e Alluvial plain of Yinchuan 10–40 500–2000 <1 1–5 5

f Fluvial and lacustrine plain of Yinchuan

10–50 500–2000 1–3 0.5–5 8

g Fluvial and lacustrine plain of north Yinchuan Plain

20–50 500–2000 1–5 0–5 9

h Taole plain zone 100 500–2000 <3 0.5–5 7


44


salinity, and ultrasonic doppler instruments (UDIs) tomonitor flow. Water quality was monitored throughregular sampling.

Groundwater modelling was conducted to assessover-exploitation, if any, of the groundwaterresource within the Yinchuan city area, determine therates of groundwater abstraction that will be sustain-able and identify optimal location of the well field toavoid pollution of deeper groundwater. All previousgroundwater abstraction data and groundwater leveldata from various aquifers were collected and col-lated. These data were used to calibrate the ground-water model MODFLOW (McDonald and Harbaugh1988). After its calibration, the model was used tosimulate various management scenarios as detailedin Zhang et al. (2004a).

To assess the feasibility of conjunctive water useand use of shallow groundwater, a site was selectedin the Huinong District of Shizuishan, Ningxia,where field experiments were conducted by using (1)Yellow River water; (2) Yellow River water andpumped groundwater alternately; (3) shallowgroundwater; and (4) a 50:50 mixture of the two. Arandomised block design was adopted and there werethree replications for every treatment. Each plot hadan area of 129.6 m2 (3.6 × 36 m). Crops grown were

spring wheat and maize planted as mixed cropping.The spring wheat was seeded on 7 March 2004 andharvested on 15 July 2004. The maize was seeded on10 April 2004 and harvested on 1 October 2004. Fiveirrigations were applied during the growing season.Approximately 90 mm depth of water was applied ineach irrigation. In April and October, soil sampleswere collected from various depths of the soil profileabove the watertable for soil moisture, soil chemistryand salinity analysis. Groundwater samples were alsocollected to assess any changes in the quality ofgroundwater as a result of irrigation with varyingquality water. Crop yields were measured from eachplot and crop.

To assess water-use efficiency of furrow and floodirrigation methods, a site was selected in the HuinongDistrict of Shizishan, Ningxia, and field experimentswere conducted to assess any water savings fromusing furrow irrigation rather than the traditionalflood irrigation for growing Chinese wolfberry andevaluate impacts, if any, on crop yields and salinitybuild-up in the soil profile. An extensive area withinthe Yinchuan Plain is planted with Chinese wolf-berry. The field experiments were conducted during2003. There were two treatments (flood and furrow)with one replication for each treatment. Both treat-ments were planted with Chinese wolfberry. Each ofthe two plots in the flood irrigation treatment had anarea of about 700 m2 (36.6 × 19.1 m), whereas eachof the two plots in the furrow irrigation had an area of620 m2 (35.5 × 17.5 m). The soil moisture fromvarious segments of the soil profile up to 1.2 m depthwas measured regularly during the growing season.Both treatments were irrigated on the same day butirrigation depths varied between two treatments.

The Junmachang and Jinshan sites, affected byboth salinity and waterlogging, were selected toexamine the changes in soil salinity adjacent to adrainage system, patterns of groundwater seepageinto the drain and monitor changes in the soil andwater chemistry. The sites were instrumented withtransects of shallow piezometers for the continuousmonitoring of groundwater levels adjacent to thedrainage system and EC meters to monitor drainwater quality. Soil was sampled twice a year at twolocations along each transect of piezometers and ana-lysed for major ions and pH.

Sand Lake and surrounding brackish areas wereselected to study groundwater discharge patterns andtheir effect on the development of surroundingbrackish areas and to identify management options

Figure 5. Hydrological zones of the Yinchuan Plain


45


that will help minimise adverse impacts to theseareas. Sand Lake is one of the most important touristsites in the Ningxia Autonomous Region. Due to thecontinuously rising water levels and the increasingsalinity of the unconfined aquifer, groundwater dis-charge to the lake is increasing. A brackish area northof the lake is another groundwater discharge zone.Although there are about 50 monitoring wells cov-ering a 50 km radius, there is not enough informationto reliably identify the aquifers that are discharginginto the brackish area. Transects of both deep andshallow piezometers were installed to assess the ratesof recharge and discharge between various aquifers,and groundwater discharge to the soil surface. Elec-trical conductivity meters were installed to monitorthe lake water salinity. Soil sampling of the sur-rounding brackish area was conducted regularly tomonitor changes in the soil chemical composition. Aweather station was installed to monitor weatherparameters, and a database was established for thestudy area. All previously collected data were col-lated and stored in the database. Groundwater mod-elling was conducted to assess the impacts of variouswater levels in the lake on the rates of groundwaterdischarge to the surrounding areas.

Results and discussion

Hydrological response units (HRU) of the Yinchuan Plain

The regional HRU were determined based on theweighing factors in various zones, rates of water levelrise and DTW. The HRU for 1992 and 2000 areshown in Figure 6. The higher the HRU value, thebigger is the risk of developing a shallow watertable.Accordingly, the zones of shallow groundwater riskwere located in the north of the Yinchuan Plainduring 1992 and 2000. These areas need urgent atten-tion to arrest the development of shallow water levelsand risk of soil salinisation. Zhang et al. (2004b)provide further details about the HRU.

Water resource pollution

Total water resources available in the YinchuanPlain are 42.33 × 108 m3/annum, with a surface waterresource component of 25.40 × 108 m3/annum and agroundwater component of 16.93 × 108 m3/annum(Zhao 1999; Zhang and Wang 2003). However, thequality of the Yellow River water deteriorates on the

Figure 6. Hydrological response units (HRUs) of the Yinchuan Plain in 1992 and2000


46


Yinchuan Plain due to the discharge of drainagewater containing excessive salts, nutrients and otherpollutants (Figure 7). Salt concentrations alsoincreased: at the inlet of the plain (Qingtongxia) in2002, the average concentration of total dissolvedsalts (TDS) of the Yellow River was 440 mg/Lwhereas at the outlet it was 555 mg/L.

The main drains discharged the most polluted waterinto the Yellow River. The pollutants included NH4,phosphorus, chloride, sulfate, biological oxygendemand (BOD) and volatile hydroxybenzene. TheTDS at the outlets of main drains was in the range 750–1740 mg/L. Water quality in the Sand Lake and WestLake also deteriorated over time. The shallow ground-water in more than 50% of the Yinchuan Plain hasbeen polluted to some extent. Over 40% of the watersamples from first confined aquifer also containedNH4 at concentrations between 0.01 and 1.95 mg/L.

Over-exploitation of groundwater

Groundwater modelling showed that there wasover-exploitation of groundwater resources in urban

Yinchuan city. Over-exploitation from confinedaquifers has caused excessive drawdown in the con-fined aquifers, resulting in downward pressure gradi-ents and leakage of polluted water from shallowaquifers. The local government plans to increasegroundwater abstraction to 2355 × 108 m3/annumduring the next five years (Zhu and Xia 2002).Pumping these volumes from confined aquifers willcause excessive drawdown in the second and thirdconfined aquifers (Table 2). To control the rate ofwater level decline, and therefore pollution in thesecond and third aquifers, the location of some of thewater fields should be changed and pumping yieldsfrom the third aquifer should be reduced. The model-ling scenarios and results are explained in Zhang etal. (2004a).

Soil and groundwater salinityAccording to analyses of the soil sampling data,

the main soil types in the Yinchuan Plain includealkaline and calcareous silt and clay loam soils. Theaverage soil pH is around 8.2. The main primary min-erals are hydromica, chlorite, kaolinite and smectite.The secondary minerals are calcite, gypsum, dolo-mite and halite. The salt content of the soil tends todecrease with depth from soil surface to the ground-water level. There was a medium salinity hazard inextensive areas of the Yinchuan Plain. The areas ofhigh salinity hazard were relatively small.

The groundwater had an average pH of about 8.The TDS of most of the groundwater ranged between1 and 3 g/L (Figure 8). The areas of either less than 1g/L or greater than 3 g/L TDS groundwater were rel-atively small. The chemical characteristics ofshallow groundwater were similar to that of soil, indi-cating geochemical interactions between soil andshallow groundwater.

Figure 7. Yellow River water quality at the inlet andoutlet of the Yinchuan Plain

Table 2. Impact of pumping on groundwater levels in the Yinchuan city area

Aquifer ID 2000 2005 2010

Pumping (104 m3/day) IIIII

14.16.2

14.16.2

2421

Area of cone of depression (km2) IIIII

350248

1020700

1023992

Depth to water in the cone centre (m) IIIII

25.417.2

2514

3340

Annual mean decline rate (m) IIIII

0.310.39

1.242.32


47


Conjunctive water use

Groundwater quality data were also used in the geo-chemical modelling to assess the feasibility of mixing itwith surface water and using it for irrigation. The plainwas divided into four zones (A, B, C, D) dependingupon the groundwater quality (Figure 9). ThePHRREQC (Parkhurst 1995) model was used for thispurpose. This model has been used for a similar studyin the Ord River Irrigation Area (ORIA), northern Aus-tralia (Ali et al. 2002). The chemical composition ofYellow River water was used for canal water (CW, 498mg/L). Representative groundwater qualities in thezones A, B, C and D were from wells W13 (2365 mg/L), W89 (1135 mg/L), W46 (955 mg/L) and W43(1270 mg/L) respectively. The Yellow River water wasmixed with the groundwater from these zones invarious proportions (80:20 to 50:50) in the PHREEQCmodel and suitability for irrigation of the resultant

water quality (salinity and sodicity risks) assessed. Themodelling results showed that there will be high risk ofsalinity development if these mixtures are used for irri-gation over long periods without providing adequateand regular leaching. Both salinity and sodicity riskswill be higher in zone A of the Yinchaun Plain (Figure9). In this zone it will be feasible to mix up to only 20%groundwater with the surface water to irrigate rela-tively salt-tolerant crops only. In other zones it will besafer to mix 20% groundwater with surface water forirrigation of crops. However, salt-sensitive crops willbe affected by this water. Only in zones B, C and D willit be feasible to irrigate salt-tolerant crops if 50%groundwater is mixed with surface water. Heavy irriga-tion applications at regular intervals will be required toavoid salinity build-up in the soil profile. The use of thisresource for irrigation will reduce pressures on theYellow River water and help lower watertables on theYinchuan Plain. Further details on geochemical mod-elling and results are given in Shi (2004).

Taking crop yields obtained by using the YellowRiver water as a benchmark, there was significantreduction in the crop yields when pumped ground-water alone was used for irrigation (Table 3). Therewere no significant differences in the crop yieldswhen either alternate irrigation with Yellow Riverwater and pumped groundwater, or a mixture ofYellow River water and groundwater was applied.The results show that it is feasible to mix shallowgroundwater with surface water up to 50% withoutany significant reductions in the crop yields. Saltaccumulation will be expected over time which canbe avoided using regular leaching applications.Using the pumped groundwater alone would not befeasible and would result in crop yield reductionsalong with salt accumulation in the soil profile.

Figure 8. Zones (A, B, C and D) of distinct ground-water quality in the Yinchuan Plain

Table 3. The impact of use of irrigation water ofvarying quality on average crop yields

Treatments Mean yield of wheat and maize

(t/ha)

Irrigation with Yellow River water 19.0

Mixture (50:50) of pumped water and Yellow River water

18.9

Irrigation with, alternately, pumped water and Yellow River water

17.9

Irrigation with pumped water 15.4


48


Irrigation techniques

The averages of pH and EC of the top 1.2 m ofloamy soil in the experimental plots were 8.3 and 80mS/m, respectively. The soil moisture variationduring the growing season in the two treatments isshown in Figure 10. The average water content wassimilar in both furrow and flood treatments duringinitial stages of crop growth. It was lower for furrowirrigation during later stages of the crop growth. How-ever, it always remained above 50% of field capacitythroughout the monitoring period for both treatments.

The irrigation amounts were the same for the firsttwo irrigations but significantly less for furrow irriga-tion treatment afterwards (Table 4). Total water appliedin the furrow irrigation treatment was 35% less thanthat for flood irrigation but the average yields of wolf-berries were similar for both treatments (13.3 and 13.1t/ha, respectively). Therefore, a significant amount ofwater can be saved if the furrow irrigation technique isused instead of flood irrigation. The irrigation water

productivities for the flood and furrow irrigation treat-ments were 2.2 and 3.4 kg/m3, respectively. For thefurrow irrigation method, salt would be expected toaccumulate in the soil profile over time. This could beavoided by applying regular leaching applications.

Figure 9. Potential salinity and sodicity risks of the use of various mixtures of surface and groundwater forirrigation (GW: groundwater; CW: canal water; 8: 80% surface water and 20% groundwater)

Figure 10. The dynamics of soil water contentfollowing use of one or other of twoirrigation techniques


49


Seepage from channels and drainage effectiveness

Several sites were selected and instrumented tomonitor the rates of seepage from various irrigationchannels to surrounding agricultural land. Analysesof the collected data suggested that there was exces-sive seepage from the channel network. This seepageresults in wastage of water and also causes risinggroundwater levels. Higher groundwater levels nearirrigation channels, and groundwater gradientssloping away from channels, suggest that there iscontinuous seepage from channels. The main causesof seepage include poor maintenance, higherhydraulic heads (water levels in the canals are higherthan ground surface levels to serve the commandarea) and insufficient protection from farm animalsand machinery.

The effectiveness of deep, open drains was evalu-ated at two sites by installing transects of both shallowand deep piezometers. The groundwater levels inthese piezometers were monitored from 2001 to 2004.Analyses of the monitoring data suggested that thedrains were effective in lowering the groundwaterlevels surrounding deep open drains. Their areal influ-ence was between 100 and 150 m at two sites. Soilroot-zone salinity at various depths up to watertablelevel was also monitored. It took two years for the soilsalinity to improve to levels suitable for crop growth.The crops were grown three years after drain con-struction at both sites. The biomass productivity atboth sites was comparable to that at adjacent sitesunaffected by soil salinity and waterlogging.

Surface and groundwater interactions in the Sand Lake and surrounding brackish areas

The monitoring of groundwater levels surroundingthe Sand Lake site, and surface water levels in SandLake, suggested that the water levels in the lake areusually very high during the rainy season. Thisresults in groundwater recharge, causing the ground-water levels to rise in the surrounding areas anddevelopment of additional brackish areas. Ground-

water levels gradually fall away from the Sand Lake,suggesting groundwater recharge. During the dryseason, the water levels in Sand Lake are generallylower, and surface water from canal and drainagesystem is usually pumped into the lake to maintainthe water levels required. The groundwater model-ling suggested lowering the currently maintainedwater levels in the Sand Lake by 0.5 m to reducegroundwater recharge and therefore the developmentof additional brackish areas. Groundwater modellingalso showed that the higher, shallow groundwaterheads surrounding the lake have resulted in the devel-opment of downward pressure gradients. This iscausing leakage of low-quality shallow groundwaterinto the first confined aquifer leading to its pollution.The groundwater samples from the first confinedaquifer showed higher levels of NH4 and other pol-lutants, supporting the modelling results.

Conclusions

The hydrological response units for the YinchuanPlain identified high-risk areas that need urgentattention to arrest the development of shallow water-table levels and soil salinisation. The quality of theshallow groundwater is low, and there is a mediumsoil salinisation risk in extensive areas of the Yin-chuan Plain. The geochemical modelling suggestedstrong interactions between surface water, soil andshallow groundwater. On most of the Yinchuan Plainit is feasible to mix 20% shallow groundwater withsurface water for irrigation, provided adequateleaching applications are applied regularly. In someparts of the Yinchaun Plain it is feasible to mix up to50% shallow groundwater with surface water for irri-gation of crops that are relatively salt tolerant.

There is widespread pollution of both surface andgroundwater resources of the Yinchuan Plain. Thepollution of surface water resources occurs fromnutrients and industrial waste, along with excessivedischarge of salts into the Yellow River. The surfacewater quality in the major lakes has also deterioratedover time. The shallow groundwater in more than

Table 4. Irrigation amounts applied in flood or furrow irrigation

Treatment Water applied (mm)

15/11/02 19/04/03 11/05/03 11/06/03 28/06/03 11/09/03 Total

Flood irrigationFurrow irrigation

135135

105105

10533

10242

6739

8429

598383


50


50% of the area of the Yinchuan Plain has been pol-luted, and over 40% of the water samples from thefirst confined aquifer contain NH4 above acceptablelimits. In some areas of the Yinchaun Plain over-exploitation from the first and second confined aqui-fers is causing leakage of polluted shallow ground-water into the first confined aquifer.

A reduction in crop productivity is expected ifshallow groundwater alone is used for irrigation. Nosignificant differences in the crop productivity areexpected if either alternate irrigation of the YellowRiver water and pumped groundwater, or a mixtureof Yellow River water and groundwater, is applied asirrigation. It was shown through field experimentsthat a significant quantity (35%) of water can besaved if furrow irrigation is used instead of flood irri-gation, with no significant impacts on yield.

Field experiments on the rates of seepage from irri-gation channels suggested excessive seepage fromthese channels, resulting in the development ofshallow watertables and soil salinity. From the resultsof trials on the effectiveness of deep, open drains, itwas concluded that these drains are effective in low-ering the water levels and reducing soil root-zonesalinity. Their areal effectiveness ranges from 100 to150 m. Groundwater modelling showed that, toreduce groundwater recharge and avoid the develop-ment of additional brackish areas surrounding SandLake, it is necessary to lower the water levels in theSand Lake by 0.5 m.

The impacts of this research are significant.Farmers and state agencies now treat irrigation anddrainage management as one of the most seriousissues. Based on these research findings a new plan isbeing prepared to control salinity and improve cropyield. There is already an increased investment by thegovernment in the irrigation and drainage sector. Fol-lowing reduction in irrigation water withdrawal fromthe Yellow River, about 3000 shallow groundwaterwells have been constructed to supplement irrigation,lower shallow watertables and reduce the risk ofsalinisation. A process for establishing associationsof farmers who use irrigation water has been initiatedfor the purposes of encouraging them to apply water-saving techniques to reduce water wastage, introducegradual irrigation water price increase, educatefarmers about the optimal use of fertilisers and pesti-cides, and increase their awareness of the sources ofwater resource pollution and potential risks.

Acknowledgment

The research reported was conducted in projectLWR/1998/130, ‘Water resources and salinity man-agement in agricultural areas of inland northernChina and northern Australia’ of the AustralianCentre for International Agricultural Research.

References

Ali, R., Salama, R., Pollock, D., and Bate, L. 2002.Geochemical interactions between groundwater and soil,groundwater recycling and evaporation in the ORIA.CSIRO Australia, Land and Water Technical Report 21/02.

Anyin, H.A. 2002. The techniques of research on the watersaving and rebuilding of Qingtongxia irrigation region.The Yellow River of People, 41–42.

Feng, Y. 2003. A study on the regional water carryingcapacity. Progress of Water Sciences, 109–113.

McDonald, M.G. and Harbaugh, M.G. 1988. A modularthree dimensional finite difference groundwater flowmodel. US Geological Survey, Techniques of water-resources investigations, Book 6, Chapter A1.

Parkhurst, D.L. 1995. User’s guide to PHREEQC; acomputer program for speciation, reaction path,advective transport, and inverse geochemicalcalculations. US Geological Survey, Water ResourcesInvestigations Report, 33, 8.

Salama, R., Bekele, E., Bates, L., Pollock, D., Byrne, J.,Wayne, H., Watson, G. and Bartle, G. 2002. Hydrologicalresponse units of the Ord Stage I Irrigation Area and thedynamic filling of the aquifer of the Ivanhoe andPacksaddle Plains. CSIRO Australia, Land and WaterTechnical Report 07/02, 78p.

Shi, X. 2004. The formation mechanism of salt and alkalihazard in Northern area of Yinchuan Plain. Proceedingsof an international conference on water security for futuregenerations, Changchun, China, 26–31 July.

Wang, L. and Gan, H. 2003. The circulation of water and saltin the irrigation region of Ningxia: evolvement andcontrol. China Conservancy and Hydroelectricity Press,82–83.

Wu, J. and Yao, J. 2004. Salinity development of YinchuanPlain 1960s–2000. Proceedings of an internationalconference on water security for future generations,Changchun, China, 26–31 July.

Zhang, L. and Wang, L. 2003. Groundwater resources ofNingxia. Ningxia People Press, 10p.

Zhang, Y., Wu, J. and Yao, J. 2004a. Numerical simulationof groundwater flow in the Yinchuan Area. Proceedingsof an international conference on water security for futuregenerations, Changchun, China, 26–31 July.


51


Zhang, Y., Wu, J. and Wang, Z. 2004b. Dynamic responserelationships between the shallow groundwater table andsurface water and hydrology response units in YinchuanPlain. Proceedings of an international conference onwater security for future generations, Changchun, China,26–31 July.

Zhao, W. 1999. The evaluation of Ningxia water resources.Geography Progress, 121–132

Zhu, Y. and Xia, J. 2002. A study on the theory and methodsof water carrying capacity. Geography Progress, 180–188.


52


Effects of groundwater depth and water-saving irrigation on rice yield and water balance in the Liuyuankou Irrigation System, Henan, China

Liping Feng1,2, B.A.M. Bouman1, T.P. Tuong1, Yalong Li3, Guoan Lu4, R.J. Cabangon1 and Yuehua Feng5

Abstract

Declining water availability has led to the development of water-saving technologies for rice, such as alternatewetting and drying (AWD) and aerobic rice (high-yielding rice grown in nonflooded soil). Little is known aboutthe performance of these systems under different hydrological conditions, and their impacts on the water balanceof rice fields. This study quantified the effects of groundwater depth (GWD) and irrigation management on yield,water productivity, and water balance for lowland and aerobic rice in the Liuyuankou Irrigation System (LIS) inHenan, China, using a modelling approach. We parameterised and evaluated the crop model ORYZA2000 using4 years of field experiments. ORYZA2000 was sufficiently accurate in simulating the observed crop and soilwater variables. We ran ORYZA2000 with 24 years of historical weather data for different groundwater depthsand water management options: continuous flooding (CF); AWD with re-irrigation at 5 and 10 days after thedisappearance of ponded water; aerobic with re-irrigation when soil water potentials in the root zone droppedbelow –10, –15, –20, –30, and –50 kPa; and rainfed. In lowland rice, AWD gave yields similar to those of CF butsaved 30–60% of irrigation water by reducing percolation flows by50–80%, and had little effect on evaporationand transpiration. AWD irrigation management increased total water productivity (WP) by 30–60%. Shallowergroundwater depth had a significantly higher grain yield than deeper groundwater depth in lowland riceproduction. In aerobic rice, yield declined by 10%, going from a –10 to –50 kPa irrigation threshold level, butirrigation inputs decreased by 80% because of an equivalent reduction in percolation. WP increased 60% when thethreshold soil water potentials decreased from –10 to –50 kPa. Average yield at lower groundwater depth (20–120cm) was 39% higher than at deeper groundwater depth (1000 cm). The results indicate that AWD and aerobic ricemaintain high yields and save irrigation water with shallow GWD in the area.

1 International Rice Research Institute, Los Baños,Philippines.

2 China Agricultural University, Beijing 100094,People’s Republic of China.

3 Wuhan University, Wuhan, People’s Republic of China.4 Huazhong Agricultural University, Wuhan, People’s

Republic of China.5 Huibei Experiment Station, Kaifeng, People’s Republic

of China.


53



54


Introduction

The lower Yellow River Basin is one of the mostimportant food production areas in China. The Liuy-uankou Irrigation System (LIS), one of the typicalirrigation systems in this region, is located in easternKaifeng City, south of the Yellow River, and encom-passes an area of about 40,700 ha.

With the declining water availability in irrigationsystems, water-saving technologies at the field scaleare being developed to reduce the rice water require-ment. These technologies include alternate wettingand drying (AWD), flush irrigation (FI), and aerobicrice. In AWD, the rice field is allowed to dry for a fewdays between irrigation events, including a mid-season drainage in which the field is allowed to dryfor 7–15 days at the end of the tillering stage. TheAWD system has been reported to save water and tomaintain or even increase yield, and to be widelyadopted by farmers (Li 2001). Belder et al. (2004)compared the effects of AWD and continuousflooding (CF) irrigations on rice performance andwater use at different levels of nitrogen (N) input intypical lowland environments and concluded thatbiomass and yield did not differ significantlybetween AWD and CF, but AWD could reduce water

use up to 15% without affecting yield when theshallow groundwater stayed within about 0–30 cm. Anew development in water-saving technologies is theconcept of ‘aerobic’ rice (Bouman et al. 2005). In theaerobic rice system, special high-yielding aerobicrice varieties are grown in unsaturated aerobic soilsthroughout the season, just like an upland crop suchas wheat or maize, with supplementary irrigation andsufficient external inputs to reach high yields. Theaerobic condition is maintained by using flush irriga-tion or sprinklers, so that ponding occurs for onlyshort periods just after irrigation or rain. Aerobic riceis targeted at water-short irrigated lowlands wherethe availability of water is too low to grow rice underthe AWD regime and at favourable uplands withaccess to supplementary irrigation. Research inChina and the Philippines suggested that yields ofaerobic rice of around 70% of that realised under CFcan be obtained using about 50% of the water used inCF systems (Bouman et al. 2005; Yang et al. 2005).

The potential of water-saving technologies toreduce water inputs and their effect on yield andwater productivity depend on soil type, groundwatertable depth, and climate (Bouman and Tuong 2001).The adoption of water-saving technologies mayaffect environmental conditions, which may haverepercussions on the performance of these water-saving technologies themselves. For example, AWDand aerobic rice will change percolation, which mayaffect regional hydrological conditions such asgroundwater recharge and level. Little is known sofar about the performance of these water-saving irri-gation systems under different hydrological condi-tions, however, or of their impacts on the crop yieldand water balance of rice fields. Most research onAWD and aerobic rice has been limited to individualfield experiments, and it has been suggested that sim-ulation models should be applied to synthesise exper-imental findings and extrapolate them to differentenvironments and agro-ecological conditions(Bouman and Tuong 2001; Belder et al., 2005).

Since the mid-1990s, the International RiceResearch Institute and Wageningen University andResearch Centre have been developing the ‘ORYZA’series of models to simulate the dynamics of ricegrowth and development in potential (Kropff et al.1994), N-limited (Drenth et al. 1994), and water-limited (Wopereis et al. 1996) situations. Recently,these models were integrated and updated in thesingle model ORYZA2000 (Bouman et al. 2001a,b).The model worked well for lowland rice under irri-

Abbreviations used

AWDCCFCVDDAEDATEFIGWD

ILAILISPPRFRRFSDTWMO(s)WP

alternate wetting and drying capillary risecontinuous floodingcoefficient of variationdeep drainagedays after emergencedays after transplantingevaporationflush irrigationgroundwater depth; depth of the top of the watertableirrigationleaf area index Liuyuankou Irrigation Systempercolationpartially rainfed with survival irrigationrainfallrainfedstandard deviationtranspirationwater management option(s)water productivity


55


gated and rainfed conditions at various levels of Nfertilisation in Indonesia (Boling et al. 2006).

The objectives of this study reported here are (1) toevaluate the rice crop growth model ORYZA2000for traditional inbred and aerobic rice varieties underdifferent water conditions in the LIS, and (2) to useORYZA2000 to quantify the effects of differentgroundwater table depths (GWD) and irrigation man-agement on grain yield, water balance components,and water productivity for lowland and aerobic riceconditions in the LIS. The implications of the find-ings for water management of the LIS are discussed.

Materials and methods

Field experiments

The field experiments were conducted at Kaifeng,Henan Province, China, in the summer seasons of2001–2004. The site of the experiment in 2001 was alowland rice-growing environment, Gaozhai village(34˚82'N, 114˚51'E, altitude 69.17 m). In 2002–2004,the experiments were in Panlou village (34˚78'N,114˚52'E, altitude 68.05 m). The change in locationspermitted the evaluation of the model in differenthydrological conditions and the testing of the aerobicrice system.

The experiments were conducted in a split-plotdesign, with three replicates in 2001 and four in 2002and 2004. In 2001 and 2002, the inbred rice varietyXD90247 was used. In 2003 and 2004, aerobic ricevariety HD297 was used. In 2001, the main plotswere water treatments: (1) CF in puddled soil, (2)AWD in puddled soil, and (3) FI at –50 kPa soil waterpotential threshold level. In FI, there was no standingwater in the field for most of the time. In 2002, thewater treatments were three FI treatments with soilwater potential threshold levels at –10, –30, and –70kPa. A fourth treatment of ‘partially rainfed with sur-vival irrigation’ (PRF) was included, in which irriga-tion was applied only when the rice crop showed verysevere drought symptoms. In 2003 and 2004, themain plots were two irrigation regimes: (1) FI at athreshold level of –30 kPa and (2) PRF. The subplotswere two N rates: N1, 225 kg/ha in 5 splits, and N2,300 kg/ha in 5 splits. The subplots were two rowspacings: D1 – spacing between rows at 30 cm, andD2 – spacing between rows at 24 cm.

All experiments were kept as free as possible fromweeds, pests, and diseases. We measured phenology(date of transplanting, panicle initiation, flowering,

maturity, harvest); leaf area index (LAI) and greenleaf, stem, and panicle biomass, and total dry matterat 15 days after emergence (DAE) and at 30 daysafter transplanting (DAT), panicle initiation, flow-ering, and grain filling; and grain yield. We measureddaily standing water depth, soil water potential usingtensiometers, irrigation input, drainage water, dailypercolation rate, groundwater depth, and soil phys-ical and hydraulic properties in different soil layers.These soil characteristics were used to determine vanGenuchten parameters (van Genuchten 1980; vanGenuchten et al. 1991) for the soil-water balance sub-model in ORYZA2000.

Daily meteorological parameters (rainfall, panevaporation, hours of sunshine, maximum andminimum temperature, wind speed etc.) were col-lected from the meteorological station at the Huibeiexperiment station some 8 km away from the site in2001 and at 1 km from the 2002, 2003, and 2004 site.

Rice model ORYZA2000

Model description

ORYZA2000 simulates the growth, development,and water balance of lowland rice in situations ofpotential production, water limitations, and nitrogenlimitations. It is assumed that, in all these productionsituations, the crop is well-protected against diseases,pests, and weeds. Bouman and Van Laar (2005) havepresented a summary description and evaluation ofORYZA2000 for potential and N-limited situations.For water-limited conditions, the model includes soil-water balance modules PADDY (Bouman et al.2001a,b) and SAWAH (Ten Berge et al. 1995).PADDY is suitable for typical, poorly drained lowlandrice soils. SAWAH can be used for both lowland ricesoils and regular ‘upland’ soils. In PADDY, a puddledlowland rice soil is modelled as a layer of muddytopsoil on top of a 3–5 cm plough sole, which overliesa nonpuddled subsoil. With ponded water on the sur-face, vertical water flow is either a fixed percolationrate or it can be calculated dynamically from hydraulicconductivity characteristics from the plough sole andthe nonpuddled subsoil. The conductivity characteris-tics are expressed by either van Genuchten parametersor by parameters of a power function. With no pondedwater on the surface, incoming water is redistributedby calculating gain and loss terms for all layers. Allwater in excess of field capacity is drained from thelayer, with a maximum rate equal to the saturatedhydraulic conductivity of the layer. The water reten-


56


tion characteristics are model input data and can besupplied either as measured data or as van Genuchtenparameters. In SAWAH, the general flow equation issolved numerically under given boundary conditions,using explicit and implicit solution schemes forunsaturated and saturated layers of the soil profile,respectively. The pressure head is defined as zero at allsaturated–unsaturated interfaces, and this condition isused as an internal boundary condition to calculateflow through the soil layers. For each soil layer, theconductivity curve and the water retention curve haveto be specified by the van Genuchten parameters.

Model evaluation

The performance of ORYZA2000 for XD90247and HD297 under lowland and aerobic soil condi-tions was evaluated using all four years of experi-mental data. Since two years of experimental datawere needed for each variety to arrive at a goodparameterisation, we could not distinguish an inde-pendent ‘validation’ data set. Following Bouman andVan Laar (2005), we used a combination of graphicalanalysis and statistical measures. We graphicallycompared the simulated and measured above-groundbiomass, leaf area index, ponded water depth, soilwater potential, and grain yield. For the same varia-bles, we computed the slope (α), intercept (β), andcoefficient of determination (R2) of the linear regres-sion between simulated (Y) and measured (X) values.We also calculated the Student’s t-test of meansassuming unequal variance (P(t)), and the absolute(ARMSE) and normalised (NRMSE) root meansquare errors between simulated and measuredvalues:

ARMSE = (1/n Σ(Yi – Xi)2)0.5 (1)

NRMSE = 100*(1/n Σ(Yi – Xi)2)0.5/ΣXi/n (2)

where n is the number of observations.A model reproduces experimental data best when

α is close to 1, β close to 0, R2 close to 1, and P(t)larger than 0.01, ARMSE is similar to the standarddeviation of measured values, and NRMSE is similarto the coefficient of variation of measured values.

Model scenarios

We took the calibrated crop parameters of aerobicvariety HD297 and inbred variety XD90247, and thesoil properties and parameters representing theexperimental conditions. We ran the ORYZA2000model with 24 years of historical daily weather datafrom 1981 to 2004. These data were from the mete-

orological station at the Huibei experiment station inthe LIS.

For the lowland rice system, we used the PADDYmodel to explore the performance of inbred ricevariety XD90247 and the water balance under 49 sce-narios: the combinations of seven water managementoptions (WMO) and seven GWD. The water manage-ment scenarios were CF_25 10mm (CF, re-irrigate 25mm when ponded water is lower than 10 mm), CF_7510mm (CF, re-irrigate 75 mm when ponded water islower than 10 mm), AWD_50mm10d (AWD, re-irri-gate 50 mm 10 days after disappearance of pondedwater), AWD_50mm5d (AWD, re-irrigate 50 mm 5days after disappearance of ponded water),AWD_75mm10d (AWD, re-irrigate 75 mm 10 daysafter disappearance of ponded water), AWD_75mm5d(AWD, re-irrigate 75 mm 5 days after disappearanceof ponded water), and rainfed (RF): purely rainfed, noirrigation. Seven generic GWDs were 20, 60, 90, 120,150, 190, and 1000 cm below the soil surface, contin-uous from day 1 to day 365.

For the aerobic rice system, aerobic rice varietyHD297 and the SAWAH model were used for 42combinations of six WMOs and seven GWDs. Sixwater management options were used: FI –10 kPa(FI, irrigate 50 mm when soil water potential at 15 cmdepth is –10 kPa), FI –15 kPa (FI, irrigate 50 mmwhen soil water potential at 15 cm depth is –15 kPa),FI –20 kPa (FI, irrigate 50 mm when soil water poten-tial at 15 cm depth is –20 kPa), FI –30 kPa (FI, irri-gate 50 mm when soil water potential at 15 cm depthis –30 kPa), FI –50 kPa (FI, irrigate 50 mm when soilwater potential at 15 cm depth is –50 kPa), and RF.Seven generic GWDs were used: 30, 60, 90, 120,150, 190, and 1000 cm below the soil surface, contin-uous from day 1 to day 365.

Simulation outputs were seasonal sums of evapo-ration (E), transpiration (T), irrigation (I), rainfall(R), growth duration (DAE for direct-seeded crop,and DAT for transplanted crop), and yield of roughrice. The water balance components (expressed inmm of water) in the field were computed according to

I + R + C = E + T + D + dW

where R = rainfall, I = irrigation, C = capillary rise,T = transpiration, E = evaporation, D = deepdrainage, and dW is the difference in soil waterstorage in the field. Total water use at the field scaleis the sum of the terms on either the left or right sideof the equals sign. We defined percolation (P) as P =I + R – E – T; P includes dW, capillary rise, and


57


deep drainage. If P was negative, this indicated thatcapillary water and water from the soil layer wereused. If P was positive, this indicated that water wasadded to the groundwater table.

Calculations and statistical analysis

Water productivity

Water productivity (kg grain/m3 water) was calcu-lated as grain yield per unit of irrigation water(WP(I)) and per unit of total water (irrigation + rain-fall) (WP(I + R)) from transplanting to harvest forlowland rice or from emergence to harvest foraerobic rice.

Statistics

Analysis of variance (ANOVA) was performed onwater balance components and yield to determine theeffects of water management options and differentgroundwater depths, and their interactions, using theIRRISTAT software. Differences among treatmentmeans were evaluated. The level of confidence (Pc)was set at 95% or 99% depending on the parameters.

Results

Evaluation of ORYAZ2000 model

Model evaluation for crop variables

Figure 1 presents typical graphics of comparisonsbetween simulated and measured biomass of totalabove-ground dry matter, green leaves, stems, andpanicles, and of LAI of treatment FI30 of XD90247in 2002 and treatment W1N1D2 of HD297 in 2003.The dynamics of LAI were simulated quite well forXD90247. Generally, LAI was under-simulated inthe early growing season but quite well simulatedafter the mid growing season for HD297. Thedynamics for biomass of total above-ground drymatter, green leaves, stems, and panicles were simu-lated well, except that simulated values for stembiomass exceeded the measured values at maturityfor XD90247. The dynamics for biomass of totalabove-ground dry matter, green leaves, stems, andpanicles were simulated quite well for HD297.

Figure 1. Simulated (lines) compared with measured biomass of total above-ground dry matter (◆),leaves (❊), stems (+), and panicles (�), and of leaf area index (●) of typical treatments forXD90247 (above, FI30: flush irrigation (FI) at threshold level of –30 kPa) in 2002 andHD297 (below, W1N1D2: FI at threshold level of –30 kPa, 225 kg/ha N rate, 24 cm of rowspacing) in 2003


58


Figure 2 compares simulated with measured grainyields for all the experiments. For reference, the 1:1line plus and minus the SD of the measured variableis also shown. For XD90247, most of simulated yielddata points fell within or very close to the 1:1 ± SDlines, indicating good simulation of yield. ForHD297, the simulated yield data points fell within ornear the 1:1 ± SD lines of measured ones. Generally,we see that the ORYZA2000 model provides a satis-factory simulation of grain yield for the inbred riceand aerobic rice varieties.

Table 1 gives the statistical parameters of good-ness-of-fit for crop growth variables and grain yieldsof the whole experimental dataset. The means of sim-

ulated values were close to the measured ones. Thebiases (means of simulated values – means of meas-ured values) of yield and final biomass were –121and 1998 kg/ha for XD90247 and 199 and 405 kg/hafor HD297. The standard deviations (SDp) of simu-lated values were also close to the measured ones.Student’s t-test showed that the simulated and meas-ured values of biomass of crop organs, LAI, and grainyield were all similar at the 99% confidence level.The linear relationship between simulated and meas-ured values was highly significant, with R2 values of0.912–0.986 for XD90247 and 0.635–0.955 forHD297. The linear slope α was close to 1, except forLAI, with a value of 0.622, and yield, with a value of0.751 for HD297. The intercept β approached 0 andthe R2 was close to 1, except for LAI, with a value of0.752, and yield, with a value of 0.635 for HD297.

The values of ARMSE were 2201 kg/ha for totalbiomass and 660 kg/ha for yield of XD90247 and 740kg/ha for total biomass and 560 kg/ha for yield ofHD297. The values of NRMSE were 11–29%, withan average of 21% for XD90247, and 12–59%, withan average of 31% for HD297. The ARMSE values ofcrop growth variables and grain yield were 1.8 (1.5–2.2) times the SD in measurements for XD90247 and2.1 (1.2–3.0) times SD in measurements for HD297.The NRMSE values of these crop variables were 2.1(1.6–2.6) times the coefficients of variation (CV) inmeasurements for XD90247 and 2.2 (1.0–3.7) timesthe CV in measurements for HD297.

Model evaluation for soil water dynamics

Figure 3 presents typical graphics of comparisonsbetween simulated and measured depth of pondedwater in AWD and FI treatments for XD90247 in2001.

Figure 2. Simulated versus measured grain yield forXD90247 (filled circles) of 2001–02 andHD297 (open circles) of 2003–04experimental data set. The thicker line isthe 1:1 relationship; thinner lines are ±SDof the measured values around the 1:1 line

Figure 3. Simulated (lines) and measured (symbols) depth of ponded water of alternate wettingand drying (AWD) and flush irrigation (FI) for XD90247 in 2001


59


Tab

le 1

. E

valu

atio

n re

sults

for

OR

YZ

A20

00 s

imul

atio

ns o

f cr

op g

row

th v

aria

bles

ove

r th

e gr

owin

g se

ason

for

the

two

vari

etie

s of

the

expe

rim

ents

fro

m 2

001

to20

04, L

iuyu

anko

u Ir

riga

tion

Syst

em, C

hina

Cro

p va

riab

leN

Xm

ea (

SDp)

Xsi

m (

SDp)

P(t

)α

βR

2A

RM

SEN

RM

SE

(%)

SDC

V

(%)

XD

9024

7 da

tase

t (20

01–0

2)T

otal

bio

mas

s (k

g/ha

)B

iom

ass

of p

anic

les

(kg/

ha)

Lea

f ar

ea in

dex

Fina

l bio

mas

s (k

g/ha

)Y

ield

(kg

/ha)

39 21 39 8 8

4998

(49

71)

2597

(28

65)

2.33

(1.

94)

1071

7 (4

547)

6100

(27

53)

5382

(54

90)

2591

(30

46)

2.63

(2.

10)

1271

5 (5

355)

5979

(28

31)

0.37

**0.

50**

0.26

**0.

22**

0.47

**

1.07

51.

049

1.03

21.

170

0.99

7

8–1

340.

22 178

–102

0.95

0.97

0.91

0.99

0.94

1350 50

20.

6922

01 660

27 19 29 21 11

622

257

0.43

1121 43

9

12 8 17 9 7

HD

297

data

set (

2003

–04)

Tot

al b

iom

ass

(kg/

ha)

Bio

mas

s of

pan

icle

s (k

g/ha

)L

eaf

area

inde

xFi

nal b

iom

ass

(kg/

ha)

Yie

ld (

kg/h

a)

96 48 96 17 17

2453

(259

7)12

51(1

272)

1.91

(1.9

5)63

67(1

733)

2873

(869

)

2451

(289

9)15

44(1

374)

1.42

(1.4

0)67

72(1

813)

3072

(819

)

0.50

**0.

14**

0.02

**0.

26**

0.25

**

1.09

11.

029

0.62

20.

979

0.75

1

–225

.525

7.2

0.2

536.

291

4.0

0.95

50.

908

0.75

20.

876

0.63

5

653

507

1.12 74

056

0

27 41 59 12 19

365

167

0.40 59

335

3

14 13 16 12 14

N =

num

ber

of d

ata

pair

s; X

mea

= m

ean

of o

bser

ved

valu

es in

who

le p

opul

atio

n; X

sim

= m

ean

of s

imul

ated

val

ues

in w

hole

pop

ulat

ion;

SD

p =

stan

dard

dev

iatio

n of

pop

ulat

ion;

P(t

) =

sign

ific

ance

of

pair

ed t-

test

, P(t

) >

0.01

mea

ns th

at s

imul

ated

and

mea

sure

d va

lues

are

the

sam

e at

the

99%

con

fide

nce

leve

l; α

= s

lope

of

linea

r re

latio

n be

twee

n si

mul

ated

and

obs

erve

d va

lues

; β =

inte

rcep

t of

linea

r rel

atio

n be

twee

n si

mul

ated

and

obs

erve

d va

lues

; R2

= ad

just

ed li

near

cor

rela

tion

coef

fici

ent b

etw

een

sim

ulat

ed a

nd o

bser

ved

valu

es; N

RM

SE (%

) = n

orm

alis

ed ro

ot m

ean

squa

re e

rror

(%);

AR

MSE

= a

bsol

ute

root

mea

n sq

uare

err

or. S

D =

sta

ndar

d de

viat

ion

of m

easu

red

vari

able

s; C

V =

coe

ffic

ient

of

vari

atio

n of

mea

sure

d va

riab

les.

Tab

le 2

. E

valu

atio

n re

sults

for

OR

YZ

A20

00 s

imul

atio

ns o

f w

ater

dyn

amic

s of

the

who

le e

xper

imen

tal

data

set

for

XD

9024

7 an

d H

D29

7 fr

om 2

001

to 2

004,

Liu

yuan

kou

Irri

gatio

n Sy

stem

, Chi

na

Var

iabl

eN

Xm

ea (

SDp)

Xsi

m (

SDp)

P(t

)α

βR

2A

RM

SEN

RM

SE

(%)

SDC

V

(%)

XD

9024

7 da

ta s

et (

2001

–02)

Dep

th o

f po

nded

wat

er (

mm

)So

il w

ater

tens

ion

(kPa

)31

533

858

(37

)9.

236

(10.

353)

37 (

52)

10.3

53 (

9.19

6)0.

04**

0.48

**0.

833

0.67

94.

072.

930.

511

0.62

832

6.33

755 69

235.

0760 81

HD

297d

ata

set (

2003

–04)

Soil

wat

er te

nsio

n (k

Pa)

1320

12 (

12)

14 (

13)

0.00

0.48

8.15

90.

222

1310

73.

8152

N =

num

ber

of d

ata

pair

s; X

mea

= m

ean

of o

bser

ved

valu

es in

who

le p

opul

atio

n; X

sim

= m

ean

of s

imul

ated

val

ues

in w

hole

pop

ulat

ion;

SD

p =

stan

dard

dev

iatio

n of

pop

ulat

ion;

P(t

) =

sign

ific

ance

of

pair

ed t-

test

, P(t

) >

0.01

mea

ns th

at s

imul

ated

and

mea

sure

d va

lues

are

the

sam

e at

the

99%

con

fide

nce

leve

l; α

= s

lope

of

linea

r re

latio

n be

twee

n si

mul

ated

and

obs

erve

d va

lues

; β =

inte

rcep

t of

linea

r re

latio

n be

twee

n si

mul

ated

and

obs

erve

d va

lues

; R2

= ad

just

ed li

near

cor

rela

tion

coef

fici

ent b

etw

een

sim

ulat

ed a

nd o

bser

ved

valu

es; N

RM

SE (

%)

= no

rmal

ised

roo

t mea

n sq

uare

err

or (

%);

AR

MSE

= a

bsol

ute

root

mea

n sq

uare

err

or. S

D =

sta

ndar

d de

viat

ion

of m

easu

red

vari

able

s; C

V =

coe

ffic

ient

of

vari

atio

n of

mea

sure

d va

riab

les.


60


Although the simulated values of ponded waterdepth exceeded the measured values in the earlygrowing season and were below the measured valuesin the late growing season, the fluctuation of simu-lated field water depth agreed well with the dynamicsof measured values.

Figure 4 presents typical graphics of comparisonsbetween simulated and measured soil water poten-tials at 15 cm depth. The fluctuation of simulated soilwater tension agreed well with the dynamics of meas-ured values, although there was some spread in themeasured data points. The dynamics of soil watertension were simulated well for the varieties by theORYZA2000 model in the LIS.

Table 2 gives the statistical parameters of goodness-of-fit for ponded water depth and soil water tension ofthe whole dataset of XD90247 and HD297. As forcrop variables, the means of the measured values wereclose to the simulated values. The bias (means of sim-ulated values – means of measured values) of ponded

water depth was –21 mm for XD90247; the biases ofsoil water tension were –1.117 kPa for XD90247 and2.00 kPa for HD297. The standard deviations (SDp) ofmeasured values were also close to the simulatedvalues. Student’s t-test showed that the simulatedvalues of ponded water depth and soil water tensionand the measured values were similar at Pc = 99% forXD90247. The linear relationship between simulatedand measured values was highly significant, with R2

values of ponded water depth of 0.511 for XD90247and R2 values of soil water tension of 0.629 forXD90247D502 and 0.222 for HD297 at Pc = 99%.The slope α of the regression line was close to 1 andthe intercept β was near 0 for XD90247.

The ARMSE value of ponded water depth was 32mm. The NRMSE value was 55%. The ARMSE valuewas close to the SD in measurements and the NRMSEvalue of ponded water depth was similar to the coeffi-cients of variation in measurements. The ARMSEvalues of soil water tension were 6.337 kPa for

Figure 4. Simulated (lines) and measured (symbols) soil water potential at 15 cm depth of soillayer of typical treatments for XD90247 (above, FI30, FI70: FI at threshold level of –30kPa, –70 kPa) in 2002 and HD297 (below, W1N1D1: FI at threshold level of –30 kPa,225 kg/ha N rate, 30 cm of row spacing; W2N2D1: rainfed, 300 kg/ha N rate, 30 cm ofrow spacing) in 2003


61


XD90247 and 13 kPa for HD297. The NRMSE valueswere 69% for XD90247 and 107% for HD297. TheARMSE values were close to the SD in measurementsfor XD90247. The NRMSE values of soil water tensionwere close to the CV in measurements for XD90247and twice the CV values in measurements for HD297.

Effects of water-saving irrigation and groundwater depths: lowland rice system

The ANOVA results showed that the major waterbalance components (percolation, transpiration,evaporation, and irrigation), grain yield, and waterproductivity were significantly affected by WMO,GWD, and their interactions at Pc = 95%.

Effects of water-saving irrigation

Grain yield

Table 3 shows the means of grain yields under dif-ferent WMOs. The mean yield ranged from 7188 kg/hain RF to 7949 kg/ha in CF irrigation. The ANOVAresults (Table 3) show that there were no significantdifferences between yields of AWD and CF irrigation,but that AWD and CF gave significantly higher grainyields than RF. This indicated that AWD is a water-saving practice, which has a yield similar to that of CFirrigation. From this scenario, the irrigation of AWDwas 245–413 mm, while that of CF was 620–653 mm.

Water productivity

Table 3 shows total water productivity under dif-ferent WMOs. WP(I+R) ranged from 0.79 kg/m3 inCF with 75 mm re-irrigation to 2.38 kg/m3 in rainfed.The ANOVA results (Table 3) showed that RF had asignificantly higher WP(I+R) than both AWD andCF. AWD had a significantly higher WP(I+R) than

CF. AWD with re-irrigation of 50 mm 10 days afterthe disappearance of ponded water had a signifi-cantly higher WP(I+R) than other AWDs, whereasAWD with re-irrigation of 50 mm 5 days after thedisappearance of ponded water and with re-irrigationof 75 mm 10 days after the disappearance of pondedwater had a significantly higher WP(I+R) than AWDwith re-irrigation of 75 mm 5 days after the disap-pearance of ponded water. There were no significantdifferences in WP(I+R) between the different CF irri-gations. Water productivity fell with increasing irri-gation. Although the water productivity of RF ismuch higher, the RF yield is much lower.

Water balance components

Figure 5 shows water balance components (andstandard errors) under different WMOs. CF had ahigher irrigation input, 620–653 mm, whereas AWDhad a lower irrigation input, 245–413 mm. Onaverage, AWD saved about half (49%) of the water.The largest irrigation input was for continuousflooding with re-irrigation of 75 mm when pondedwater was lower than 10 mm. The smallest irrigationinput was for AWD with re-irrigation of 50 mm 10days after the disappearance of ponded water. Evap-oration ranged from 81 mm in rainfed to 114 mm incontinuous flooding irrigation. Transpiration rangedfrom 426 mm in rainfed to 440 mm in continuousflooding irrigation. Percolation ranged from 68 mmin AWD with re-irrigation of 50 mm 10 days after thedisappearance of ponded water to 451 mm in contin-uous flooding with re-irrigation of 75 mm whenponded water is lower than 10 mm. CF had a perco-lation rate 70% higher than that of AWD. Percolationwas negative, which indicated that the soil watersupply (capillary rise) was used to satisfy evapotran-spiration (ET). The trends of total water input underdifferent water management options were the sameas those of irrigation input.

Effects of groundwater depths

Grain yield

Table 4 shows the means of grain yields under dif-ferent depths of the groundwater table (GWD). Themean yield ranged from 7573 kg/ha at GWD of 1000cm to 7941 kg/ha at GWD of 20 cm. Yield decreasedwith increasing GWD. The ANOVA results (Table4) showed that the yields at GWD of 20–190 cmwere significantly different from the yield at GWDof 1000 cm at Pc = 95%, but there were no significantdifferences between yields at GWD 20–150 cm and

Table 3. Mean of yield and total water productivity(WP(I+R)) under different water manage-ment options (WMO)

WMO Yield (kg/ha)

WP(I+R) (kg/m3)

AWD_50mm10dAWD_50mm5dAWD_75mm10dAWD_75mm5dCF_25 10mmCF_75 10mmRF

7915 a7939 a7925 a7941 a7949 a7949 a7188 b

1.35 a1.15 b1.22 b1.04 c0.82 d0.79 d2.38 e

Note: statistical differences (P ≤ 0.05) among numbers in thesame column are indicated by different lower-case letters.


62


60–190 cm. Shallower groundwater depth gavehigher grain yields than deeper groundwater inlowland rice production.

Water productivity

Table 4 shows total water productivity under dif-ferent GWDs. WP(I+R) ranged from 1.13 kg/m3 atGWD 1000 cm to 1.35 kg/m3 at GWD 20 cm.WP(I+R) fell with increasing GWD. The ANOVAresults (Table 4) showed that the WP(I+R)s at GWD20–190 cm were significantly different from theWP(I+R) at GWD 1000 cm at Pc = 95%, but therewere no significant differences between WP(I+R)s at

GWD 20–150 cm and 60–190 cm. WP(I+R) at shal-lower groundwater depths was higher than that atdeeper groundwater depth in lowland rice production.


Figure 6 shows water balance components (andstandard errors) under different GWDs. The rainfallvalue was constant. Irrigation input ranged from 349mm at groundwater depth of 20 cm to 375 mm at1000 cm. Irrigation at GWD 20 cm was significantlydifferent from that at 60–1000 cm, but there were nosignificant differences between among 60 to 1000 cmGWDs at Pc = 95%. Irrigation at 1000 cm GWD wassignificantly different from that at 20–190 cm at Pc =95%. Shallower groundwater depth saved irrigationinput.

Evaporation was almost the same value, althoughevaporation at 20 cm GWD was significantly dif-ferent from that at 60–1000 cm at Pc = 95%. Transpi-ration at 1000 cm GWD was significantly lower thanthat at 20 to 190 cm. Percolation ranged from 163mm at 20 cm GWD to 197 mm at 1000 cm GWD.

Percolation at 20 cm GWD was significantly lowerthan that at 20–1000 cm, and percolation at 1000 cmGWD was significantly higher than that at 20–190cm at Pc = 95%. Percolation increased withincreasing GWD. The total water input at 20 cmGWD was small because of less irrigation input. Thetotal water input at 1000 cm GWD was also small,

Table 4. Mean of yield and total water productivity(WP(I+R)) under different groundwaterdepths

GWD (cm)

Yield (kg/ha)

WP(I+R) (kg/m3)

206090

120150190

1000

7941 a7867 ab7867 ab7864 ab7857 ab7839 b7573 c

1.35 a1.26 ab1.26 ab1.26 ab1.26 ab1.25 b1.13 c


Figure 5. Water balance components under different water management options.See text for details of options. Vertical and capped lines are standarderrors.


63


but because of less capillary rise. The total waterinput at 60–190 cm GWD was the same.

Effects of water-saving irrigation and groundwater depths: the aerobic rice system

The ANOVA results showed that the major waterbalance components (percolation, transpiration,evaporation, and irrigation), grain yield, and waterproductivity were significantly affected by WMO,GWD, and their interactions at Pc = 95%.

Effects of water-saving irrigation

Grain yield

Table 5 shows the means of grain yields under dif-ferent WMOs. The mean yield ranged from 3566 kg/hain RF to 4288 kg/ha at the FI –10 kPa threshold level.

Yield decreased from the FI –10 kPa threshold level to–50 kPa and then to rainfed. The ANOVA results(Table 5) show that yields among different irrigationthreshold levels and rainfed conditions had significantdifferences at Pc = 95%. Average irrigation yield was12.4% higher than rainfed yield, and yield at the –10kPa irrigation threshold level was 16.8% higher thanrainfed yield.

Water productivity

Table 5 shows total water productivity under dif-ferent WMOs. WP(I+R) ranged from 0.59 kg/m3 atthe –10 kPa irrigation threshold level to 1.38 kg/m3 inRF. The ANOVA results (Table 5) show no signifi-cant differences in water productivity betweenWP(I+R)s at the –30 and –50 kPa flush irrigationthreshold levels and RF. WP(I+R)s at the –10, –15,and –20 kPa flush irrigation threshold levels weresignificantly different at Pc = 95%, and were signifi-cantly different from that at the –30 and –50 kPa flushirrigation threshold levels and RF. Flush irrigation atthe threshold levels of –10 and –15 kPa had higherwater productivities and yields. Although RF gavehigh water productivity, yield was low.


Figure 7 shows water balance components (andstandard errors) under different WMOs. There weresignificant differences among the irrigation watermanagement options. Irrigation ranged from 296 mmat –50 kPa to 1682 mm at the –10 kPa threshold level.

Table 5. Mean of yield and total water productivity(WP(I+R)) under different watermanagement options (WMO)

WMO Yield (kg/ha)

WP(I+R) (kg/m3)

FI –10 kPaFI –15 kPaFI –20 kPaFI –30 kPaFI –50 kPa

RF

4288 a4159 b4068 c3975 d3862 e3566 f

0.59 a1.01 b1.24 c1.39 d1.39 d1.38 d


Figure 6. Water balance components under different groundwater depths.Vertical and capped lines are standard errors.


64


Evaporation ranged from 146 mm in rainfed to 181mm at the –10 kPa irrigation threshold level. Transpi-ration ranged from 200 mm in rainfed to 238 mm atthe –10 kPa irrigation threshold level. Evaporationand transpiration differences were not large. Percola-tion ranged from 226 mm at the –50 kPa thresholdlevel to 1580 mm at the –10 kPa threshold level. Per-colation in the RF treatment was negative, which indi-cated that the soil water supply (capillary rise) wasused to satisfy evapotranspiration (ET). The trends oftotal water input under different water managementoptions were the same as those of irrigation input.

Effects of groundwater depths

Grain yield

Table 6 shows the means of grain yields under dif-ferent GWDs. The mean yield ranged from 2654 kg/haat GWD 1000 cm to 4645 kg/ha at GWD 20 cm. Yieldfell with increasing GWD. The ANOVA results(Table 6) showed that the yields among differentGWDs had significant differences at Pc = 95%. Yieldat GWD 20 cm was 43% higher than that at 1000 cm.Average yield at shallower GWDs (20–120 cm) was39.2% higher than at deeper GWD (1000 cm).

Water productivity

Table 6 shows total water productivity under dif-ferent GWDs. WP(I+R) ranged from 0.08 kg/m3 atGWD 1000 cm to 1.74 kg/m3 at GWD 30 cm. TheANOVA results (Table 6) showed that WP(I+R)s at

GWD 30 and 60 cm were significantly different fromthose at GWD of 90–1000 cm. There were significantdifferences among the WP(I+R)s at GWD 90–1000cm. Water productivity at shallower GWDs wasmuch higher than that at deeper ones. Water produc-tivity fell with increasing GWD.


Figure 8 shows water balance components (andstandard errors) under different GWDs. Percolation,transpiration, evaporation, and irrigation were signif-icantly affected by different groundwater depths.Rainfall was not affected and kept the same value.Irrigation input ranged from 0 mm at GWD 60 cm to

Figure 7. Water balance components under different water management options.See text for details of options. Vertical and capped lines are standarderrors.

Table 6. Mean of yield and total water productivity(WP(I+R) under different groundwaterdepths

GWD(cm)

Yield (kg/ha)

WP(I+R) (kg/m3)

30 60 90

120 150 190

1000

4645 a4506 b4250 c4069 d3949 e3832 f2654 g

1.74 a1.74 a1.42 b1.29 c1.07 d0.82 e0.08 f



65


2371 mm at GWD 1000 cm. There were no signifi-cant differences in irrigation between 30 and 60 cmGWD at Pc = 95%. Irrigation at 30 and 60 cm GWDwas significantly different from that at 90–1000 cm.Irrigation increased with increasing GWD. Shal-lower groundwater saved irrigation input.

Evaporation was almost the same for GWD 30–190cm. Evaporation at GWD 1000 cm was higher thanthat for shallower depths. Transpiration ranged from154 mm at 1000 cm GWD to 270 mm at 30 cm. Tran-spiration fell with increasing groundwater depths.

Percolation ranged from 121 mm at 90 cm GWD to2324 mm at 1000 cm. Percolation at shallowergroundwater depth was significantly lower than whenthe top of the watertable was deeper (1000 cm). Perco-lation increased with increasing groundwater depths.The total water input at 30 and 60 cm GWD was smallbecause of less irrigation input. The total water inputwas large at 1000 cm of groundwater depth.

Discussion and conclusions

Agreement is relatively close between simulated andmeasured values of crop and soil variables. Themajority of the statistical parameters and graphicalcomparisons show that the ORYZA2000 model issufficiently accurate in simulating, over time, cropgrowth variables and yield, and soil water dynamicsfor inbred rice and aerobic rice under rainfed and irri-gated production situations in the region. The model

can be used to support field experiments in exploringthe effects of management interventions (such aswater management, fertiliser application, and plantdensity) and environmental conditions (weather,depth of the groundwater table) on the growth anddevelopment of inbred and aerobic rice varieties andwater balance, with quantifiable errors of simulation.

For the puddled rice system in the northern part ofthe study area, irrespective of groundwater depth, allAWD irrigation gave yields similar to those of con-tinuous flooding. AWD saved 30–60% of irrigationwater, mostly because of reduced percolation. It hadlittle effect on evaporation and transpiration. AWDirrigation increased total WP by 30–60%. Yields hadno significant differences from 20 to 190 cm GWD,but all the yields at 20–190 cm of groundwater depthwere significantly different from the yield at ground-water depth of 1000 cm. Shallower GWDs gave sig-nificantly higher grain yields than deeper GWDs inlowland rice production.

For the aerobic rice system in the southern part ofthe study area, irrespective of groundwater depth,yield decreased from the –10 to –50 kPa irrigationthreshold, and so did the irrigation input. The yield ofrainfed was the lowest. There was a 10% yield lossfrom the –10 to –50 kPa threshold, but an 80% irri-gation water saving from the –10 to –50 kPathreshold, mostly because of reduced percolation.WP increased 60% from the –10 to –50 kPa irrigationthreshold; WP was the lowest at the –10 kPa irriga-

Figure 8. Water balance components under different groundwater depths(vertical and capped lines are standard errors)


66


tion threshold. There were falling trends in yield andtranspiration from 30 to 1000 cm GWD for all watermanagement options, and the differences of yield andtranspiration increased from the –10 to –50 kPa irri-gation threshold, and to rainfed conditions.

The results indicate that AWD and aerobic ricemaintain high yields and save irrigation water withshallower groundwater depths in the area. This high-lights the importance of water-saving techniques andshallower groundwater depths for increasing riceyields and water productivity.

Acknowledgments

This paper reports on part of the work in ACIARproject LWR1/2000/030, ‘Growing more rice withless water: Increasing water productivity in rice-based cropping systems’. The authors are grateful forACIAR’s financial support to implement this study.The authors thank Ms A. Boling and Mr Dule Zhaofor their help with statistics.

References

Belder, P., Bouman, B., Cabangon, R., Guoan Lu, Quilang,E.J.P., Li, Y., Spiertz, J.H.J. and Tuong, T.P. 2004. Effectof water-saving irrigation on rice yield and water use intypical lowland conditions in Asia. Agricultural WaterManagement, 65, 193–210.

Belder, P., Spiertz, J.H.J., Bouman, B., Guoan Lu andTuong, T.P. 2005. Nitrogen economy and waterproductivity of lowland rice under water-savingirrigation. Field Crops Research, 93, 169–185.

Boling, A., Bouman, B.A.M., Tuong, T.P., Murty, M.V.R.and Jatmiko S.Y. 2006. Modelling the effect ofgroundwater depth on yield-increasing interventions inrainfed lowland rice in Central Java, Indonesia.Agricultural Systems (in press).

Bouman, B.A.M., Kropff, M.J., Tuong, T.P., Wopereis,M.C.S., ten Berge, H.F.M. and van Laar, H.H. 2001a.ORYZA2000: modeling lowland rice. Los Baños,Philippines, International Rice Research Institute.

Bouman, B.A.M., Peng, S., Castaneda, A.R. and Visperas,R.M. 2005. Yield and water use of irrigated tropicalaerobic rice systems. Agricultural Water Management,74, 87–105.

Bouman, B.A.M. and Tuong, T.P. 2001. Field watermanagement to save water and increase its productivity inirrigated rice. Agricultural Water Management, 49, 11–30.

Bouman, B.A.M., Tuong, T.P., Kropff, M.J. and van Laar,H.H. 2001b. The model ORYZA2000 to simulate growthand development of lowland rice. In: Ghassemi, F.,White, D., Cuddy, S. and Nakanishi, T., ed.,MODSIM2001, Proceedings of the internationalcongress on modelling and simulation. Canberra,Australian National University, 4, 1793–1798.

Bouman, B.A.M. and van Laar, H.H. 2005. Description andevaluation of the rice growth model ORYZA2000 undernitrogen-limited conditions. Agricultural Systems, 87,249–273.

Drenth, H., ten Berge, H.F.M. and Riethoven, J.J.M., ed.1994. ORYZA simulation models for potential andnitrogen limited rice production. SARP ResearchProceedings, Wageningen: DLO Research Institute forAgrobiology and Soil Fertility; Wageningen: WAUDepartment of Theoretical Production Ecology; LosBaños, Philippines, International Rice Research Institute,223p.

Kropff, M.J., van Laar, H.H. and Matthews, R.B., ed. 1994.ORYZA1, an ecophysiological model for irrigatedlowland rice production. SARP Research Proceedings,Wageningen: DLO Research Institute for Agrobiologyand Soil Fertility; Wageningen: WAU Department ofTheoretical Production Ecology; Los Baños, Philippines,International Rice Research Institute, 110 p.

Li, Y.H. 2001. Research and practice of water-savingirrigation for rice in China. In: Barker, R., Li, Y.H. andTuong, T.P., ed., Water-saving irrigation for rice.Proceedings of an international workshop held in Wuhan,China, 23–25 March 2001. Colombo, Sri Lanka,International Water Management Institute, 135–144.

ten Berge, H.F.M., Metselaar, K., Jansen, M.J.W., de SanAugustin, E.M. and Woodhead, T. 1995. The SAWAHriceland hydrology model. Water Resources Research,31, 2721–2732.

van Genuchten, M.Th. 1980. A closed-form equation forpredicting the hydraulic properties of unsaturated soils.Soil Science Society of America Journal, 44, 892–898.

van Genuchten, M.Th., Leij, F.J. and Yates, S.R. 1991. TheRETC code for quantifying the hydraulic functions ofunsaturated soils. Washington, DC, United StatesEnvironmental Protection Agency, 85p.

Wopereis, M.C.S., Bouman, B.A.M., Tuong, T.P., tenBerge, H.F.M. and Kropff M.J. 1996. ORYZA_W: ricegrowth model for irrigated and rainfed environments.SARP Research Proceedings, IRRI/AB-DLO,Wageningen, Netherlands,159p.

Yang Xiaoguang, Bouman, B.A.M., Wang Huaqi, WangZhimin, Zhao Junfang and Chen Bin 2005. Performanceof temperate aerobic rice under different water regimes inNorth China. Agricultural Water Management, 74, 107–122.


67


Tracking fallow irrigation water losses using remote-sensing techniques: a case study from the

Liuyuankou Irrigation System, China

M. Hafeez1 and S. Khan1,2

Abstract

Efficient water use is the key for sustainable management of water resources. Currently, water resources arenot managed efficiently, especially in developing countries, due to the non-availability of reliable informationabout the actual water used by different agricultural crops within a large irrigation system or at the basin scale.Therefore, an estimation of spatially distributed crop water consumption is important and challenging fordetermining water balance at different scales to promote efficient management of water resources. The use ofremote sensing data can resolve difficulties in determining water balance due to scientific developments in thecalculation of actual evapotranspiration. In this study, seven TERRA/MODIS satellite images were acquiredon different dates (6 April, 31 May, 12 June, 12 July, 23 August, 22 September and 24 October) during thesummer cropping season of 2002 over Liuyuankou Irrigation System (LIS) located along the Yellow Riverbasin of China.

The surface energy balance algorithm for land (SEBAL) was applied to MODIS sensor data for theestimation of crop water requirement. The actual evapotranspiration (ETa) was integrated for 24 hours on apixel-by-pixel basis from the instantaneous evapotranspiration (ET). Later, temporal integration (April–October 2002) was done to get the seasonal actual evapotranspiration (ETs) map of the LIS area.

Volumes of crop water consumption at different scales were compared through statistical analyses. Thecomparison provided better decision-making for identifying, at different spatial scales in an irrigation system,areas (e.g. fallow lands) that have high non-beneficial evapotranspiration. The result showed a uniquecombination of derived ETa from different MODIS images for water consumption in the LIS. The results werefurther compared with the crop potential evapotranspiration (ETc) calculations at a meteorological station inthe LIS. This showed a deviation of –5% between ETa and ETc. which is within an acceptable range. However,the accuracy of this comparison of modelled ETa against measured data of ETc needs to be considered withrespect to scale. Modelled area data were derived from discrete areas of one square-kilometre (the spatialresolution of a MODIS pixel for thermal bands) and would therefore contain reflectance attributes from manydifferent physical mediums (mixed spectral signatures) and a resulting combined evapotranspiration rate. Thediscussion provides the research orientation for ET assessment at different scales and its further implicationsin applied research for water management aided by satellite images.

1 CSIRO Land and Water, Locked Bag 588, WaggaWagga, New South Wales 2678, Australia. Email: <[email protected]>.

2 School of Science and Technology, Charles SturtUniversity, Locked Bag 588, Wagga Wagga, New SouthWales 2678, Australia. Email: <[email protected]>.


68



69


Introduction

The world’s thirst for water is likely to stay as one ofthe most pressing resource issues of the 21st century.Agriculture is the largest consumer of water in theAsian region as compared to other sectors i.e.domestic, municipal, industrial and environmental.However, the water-use efficiency of agriculture isvery low. The improvement of water-use efficiencyrequires a complete understanding of all the terms ofthe water balance at various scales such as field, farm,irrigation system and basin level. The dominant aspectof water balance is evapotranspiration (ET), which isone of the most difficult parameters to measure in thefield. A number of researches undertaken in the pasthave estimated reference ET from meteorological dataand converted this to actual ET. The major disadvan-tage of this approach is that most methods generateonly point values, resulting in estimates that are notrepresentative of larger areas. These methods are alsobased on crop factors under ideal conditions andcannot therefore represent actual crop ET.

The use of remote-sensing techniques for the esti-mation of the water evaporation component of waterbalance is achieved by solving the energy balance ofthermodynamic fluxes at the surface of the earth. Theuse of these remote-sensing techniques has becomeincreasing popular since 1990 due to the relativelylow cost of data collection — $0.03/ha for irrigatedlands (Sakthivadivel et al. 1999).Various methodsfor the estimation of actual evapotranspiration havebeen developed by combining satellite images andground meteorological data for large areas (Vidal andPerrier 1989; Choudhury 1994; Granger 1997).Another method of estimating actual ET is thesurface energy balance algorithm for land (SEBAL).SEBAL is a thermodynamically based model, usingthe partitioning of the fluxes of sensible heat andlatent heat of vaporisation.

SEBAL was originally developed in Spain andEgypt with Landsat 5TM in 1995 by Bastiaanssen(1995). Further applications to irrigation perform-ance were later found for the same sensor in Argen-tina (Roerink et al. 1997). Water consumption oflarge irrigation systems has been addressed also withNOAA AVHRR in Pakistan (Bastiaanssen et al.1999, 2001). Farah (2001) studied modelling ofevaporation under various weather conditions in theNavaisha Basin, Kenya. Farah’s results extendedSEBAL calculations of NOAA AVHRR underclouds with a Penman–Monteith approach supported

by a Jarvis–Stewart type model. Combinations ofLandsat and NOAA are reported in Timmermans andMeijerink (1999) where Landsat 5TM was used, andin Chemin and Alexandridis (2001) where Landsat7ETM+ was used. Later, Hafeez (2003) appliedSEBAL for the estimation of seasonal actual eva-potranspiration using TERRA/ASTER, TERRA/MODIS and Landsat 7ETM+ sensors in UPRIIS,Philippines.

The main constraint in using remote sensing-basedmodels is that ETa is calculated on only the satelliteoverpass days. The non-availability of cloud-freeimages, intensive computing procedure and cost alsopose major constraints in processing visible/thermal-infrared satellite images for daily ETa estimation.Temporal integration strategies have to be used inorder to interpolate the missing satellite data andobtain the integrated ETa information for a season, sothat ET results can be used in water balance studies.

The primary aim of the study reported here was tocalculate the daily actual evapotranspiration in theLiuyuankou Irrigation System (LIS) area, using theSEBAL model applied to a remote sensing TERRA/MODIS sensor. A second objective was to estimateseasonal actual evapotranspiration of the LIS area forthe summer season of 2002. These results were inte-grated with a MODFLOW-based water balance ofthe LIS which will be reported on separately.

Description of the Liuyuankou Irrigation System

LIS is located in northeastern China, in KaifengCounty of Henan Province (Figure 1). Irrigation forcrop production is met by water drawn from the chan-nels diverted from the Yellow River, mainly in thenorthern area of LIS (situated north of (above) therailway line and hereinafter known as ARL), as wellas by groundwater pumping in the southern area ofLIS (situated south of (below) the railway line andhereinafter known as BRL). The total area of LIS is55,512 ha with a net irrigated command area of40,724 ha. The Liuyuankou major canal is located inthe upper part of the irrigation district, and feedsthree main canals and fourteen branch canals. Meanannual temperature is 14.1°C and mean annual pre-cipitation 627 mm (of which 70–80% falls duringJune–September). Mean annual evaporation is1316 mm. Maximum evaporation occurs from Marchto August. The major crops cultivated in the LIS


70


include winter wheat, summer maize, cotton, rice andsoybean.

Materials and methods

Satellite data

MODIS is the key instrument aboard the TERRA(EOS AM-1) satellite, which began operation inMarch 2000. TERRA/MODIS views the entire earthsurface every 1–2 days, acquiring data in 36 spectralbands (different wavelengths of electromagneticradiation). The bands have 250–1000 m spatial reso-lution. An overview of the sensor is provided inTable 1. This study uses seven MODIS images of theLIS area acquired on different dates (see details inTable 1) to estimate seasonal actual evapotranspira-tion for the summer season of 2002.

Specificity of porting SEBAL to MODIS sensor

Acquisition of image data ‘level 1B’ (L1B) wasdone through the Red Hook Eros Data Center websiteusing the ftppull file transfer protocol. Extraction ofthe binary file was performed for two bands (1 and 2),five short-wave infrared bands (3, 4, 5, 6 and 7) andtwo thermal bands (31 and 32). A subset image forthe study area was created from the whole image ofChina for better visualisation and geo-referencing.The L1B data were already calibrated for radiometricvariations, while geo-referencing was done in theTM/WGS/84/Zone 50 with a root mean square error(RMSE) of less than 1 pixel.

The pre-processing parameters required forSEBAL include the normalised difference vegetationindex (NDVI), emissivity, broadband surface albedo,and surface temperature for both sensors. The NDVI

Figure 1. Layout of Liuyuankou Irrigation System (LIS) in China

Table 1. Overview of TERRA/MODIS sensor and details of satellite image used in the study

Sensor Dates in 2002 of satellite images used

Subsystems Number of bands

Spectral range(µm)

Spatial resolution

(m)

TERRA/MODIS 6 April, 31 May, 12 June, 12 July 23 August, 22 September, 24 October

VNIRSWIRSWIRTIR

25

1116

0.62–0.8760.459–2.1550.405–0.9653.66–14.385

250 × 250500 × 500

1000 × 10001000 × 1000


71


was calculated from bands 1 and 2 of MODIS data,and the broadband albedo was calculated usingweighing factors of all visible, near infrared and shortwave infrared bands of MODIS (Liang et al. 1999).Surface emissivity of the sensor was calculated fromthe NDVI of the sensors. Surface temperature fromMODIS sensors was calculated from thermal bands31 and 32 using the split-window techniquedescribed in Wan (1999).

Running SEBAL

Calculation of the net incoming radiation and thesoil heat flux were done after Bastiaanssen (1995),while the later development of Tasumi et al. (2000)was incorporated to determine the sensible heat flux.However, to calculate the first temperature differencebetween air and soil for the ‘hot’ pixel (i.e. where thelatent heat flux is assumed null), a first estimation ofthe air density was required. This was achieved bygeneralising meteorological data on relative humidityand maximum air temperature from a meteorologicalstation at the time of satellite overpass. Iterations ofsensible heat flux were conducted five times. Opera-tional observation showed that this method does notstabilise the air–soil temperature difference as fast asthe earlier method in Bastiaanssen (1995). In SEBAL,manual sampling of hot pixel values of the previousiteration’s output image files are required before thenext iteration can be done, which is a practical con-straint in using this technique. This constraint can beresolved by automation (after hot pixel identification)of the data collection. The sensible heat flux can beimproved by repeating the iteration five times, but thisprocess is time and space consuming.

The ET is calculated in SEBAL (Tasumi et al. 2000;Hafeez 2003) from the instantaneous evaporative frac-tion, Λ, and the daily averaged net radiation, R

n24. The

latter has to be transformed from W/m2 to mm/day bythe T0-dependent latent heat of vaporisation equationinserted in the main equation (Equation 1):

ET24 = Λ{Rn24[(2.501 – 0.002361T0)106]} (1)

where ET24 = daily ET actual (mm/day) R

n24 = average daily net radiation (W/m2)

T0 = surface temperature (°C).

The evaporative fraction, Λ, is computed from theinstantaneous surface energy balance at the momentof satellite overpass for each pixel (Equation 2):

Λ = λΕ/Rn – G0 = λE/λE + H0 (2)

where λE = latent heat flux (the energy allocated forwater evaporation) (λ can be interpreted in irrigatedareas as the ratio of actual evaporation to crop poten-tial evaporation. It is dependent on the atmosphericand soil moisture condition equilibrium.)

Rn = net radiation absorbed or emitted from theearth’s surface (radiative heat) (W/m2)

G0 = soil heat flux (conduction) (W/m2)

H0 = sensible heat flux (convection) (W/m2).

The ETa calculation through remote sensing onspecific dates provided a good indication of its spatialdistribution in the irrigation system. However, thisinformation could not be used directly, as ETa directlydepends upon weather conditions and water availa-bility in the field, which varies from day to day. It wastherefore necessary to simulate daily values to get anaccurate estimation of seasonal ETa. A larger sampleof timely ET observations is necessary to obtain anaccurate result and to adjust the daily fluctuation ofETa for integration of seasonal ETa. As proposed byTasumi et al. (2000), missing values of ETa could beobtained by daily calculation of reference evapotran-spiration (ETo) using the modified Penman–Monteithmethod. Temporal integration was undertaken in foursteps: (1) determination of the period represented byeach satellite image, e.g. selecting the 6 April 2002image to represent the month of April etc.; (2) com-putation of ETo using the modified Penman–Monteithmethod for the whole period represented by eachimage (average monthly ETo values collected fromHuibei meteorological station, summarized in Table2); (3) computation of Km values for each period asshown in Table 3; and (4) computation of cumulativeseasonal ETa using Equation 3:

where ETa = actual daily ET value computed bySEBAL for each pixel of image ‘i’(mm)

Km = multiplier for ET for the representa-tive period

n = number of satellite images processedET = seasonal ET (mm).

Results and discussion

Daily meteorological data were collected from theHuibei weather stations and used to calculate ETothrough the modified Penman–Monteith equation

ETs ETa( ) Km( )ii 1=

n

∑= (3)


72


(Allen et al. 1998). ETo was converted into potentialcrop transpiration, ETc, by multiplication with thecrop coefficient Kc (ETc = Kc × ETo). Based on theactual cropping calendar, the weighted crop coeffi-cient Kc for different satellite overpass dates wasused in this study. The major crops in the LIS wererice, maize and cotton.

In this study, the pixel values of ETa, calculatedthrough SEBAL, in the area surrounding Huibeimeteorological station were compared with themeasured evaporation through class A pan (Epan),estimated ETo and crop evapotranspiration (ETc)from the Huibei meteorological station for thesummer season of 2002 in LIS as shown in Figure 2.There is a significant difference in ET valuesobtained from remote sensing and classical tech-niques based on weather station data. The former pro-vides spatial distribution results, whereas the latterprovides only point values.

Epan indicates the evaporation from water bodies.As validated in Figure 2, Epan value from Huibeimeteorological station was always higher (on average26%) than ETa values for all seven image acquisitiondates during the season. For pixels assumed to beunder a cropping area, the estimated ETa was onaverage 11% and 5% lower than the average ETo andETc calculated from meteorological station data. Thecomparison provides an indication of the amount ofconfidence that can be placed on the values of ETaderived from the remote-sensing images. The ETa isestimated from all the physical media within one pixel(250 m as in the case of MODIS), which might havemixed spectral signatures of roads, settlements, waterbodies, and rice fields. Due to the large pixel size ofMODIS (250 m), it is difficult to absolutely comparesuch information with the classical point data frommeteorological data, even though Figure 2 shows a

good trend regarding the accuracy of ETa derivedfrom the SEBAL model.

Nevertheless, the accuracy of this comparison ofmodelled against measured data needs to be consid-ered with respect to scale. Modelled area data werederived from discrete areas of one square-kilometre(spatial resolution of a MODIS pixel for thermalbands) and would therefore contain reflectanceattributes from many different physical media(mixed spectral signatures from rice fields, barefields and roads) and a resulting combined eva-potranspiration rate. Comparison between the mod-elled data and the point-based measured data fromclass A pans or meteorological stations introducesthe possibility of scale-related errors. Even though acomparison of ETa with Epan, ETo and ETc does notbring sufficient absolute elements for validation, itdoes contribute to a consistency cross-checking ofthe method of ETa calculation from SEBAL. Thecomparisons show similar trends in ET in the timedomain of the summer season 2002.

The integration of daily ETa raster maps was doneusing the straightforward method described in the

Table 2. Average ETo and cumulative ETo values representing meteorological stations in the LiuyuankouIrrigation System, northwestern China

ET 6 April 31 May 12 June 12 July 23 August 22 September 24 October

Cumulative ETo (mm)Average ETo (mm)

97.953.53

96.774.92

138.96.16

117.35.72

106.334.23

96.452.23

72.661.93

Table 3. Values of Km for the summer season 2002 in the Liuyuankou Irrigation System, northwestern China

1–30 April 1–31 May 1–30 June 1–31 July 1–31 August 1–30 September 1–30 October

Km 27.75 19.67 22.55 20.51 25.14 43.25 37.65

Figure 2. Comparison of ETa of rice pixels withEpan, ETo, and ETc, at Huibei weatherstation for the summer season 2002 in LIS


73


temporal integration part of the previous section. Theseasonal actual evapotranspiration (ETs) map on apixel-by-pixel basis was produced through integra-tion of all daily ETa images for the summer season2002. The statistics of ETs and volume of water con-sumption for different areas in the LIS during thesummer season 2002 (1 April–31 October) are sum-marised in Table 3. Results show that ARL area has ahigher volume of water consumption through actualevapotranspiration than the BRL area. This is mainlydue to higher ET from shallow watertables caused byinefficient surface water irrigation and lateral seepageof river water. The comparison provided better deci-sion-making for crop water requirements in the ARLand BRL areas for this irrigation system.

From a water management perspective, the mostimportant model output of SEBAL was the spatially

distributed estimation of the seasonal actual eva-potranspiration (ETs) which was later used with aMODFLOW model of the area. These volumes werecross-validated by the water balance of the LIS pro-vided by Khan et al. (2004). For example, Figure 3depicts a range from 300 mm to 855 mm of ETs in theLIS region for the summer season of 2002.

Low seasonal actual evapotranspiration is mod-elled for the bare fields and settlements, while the irri-gated areas range from medium to high ETs. Theagricultural areas in the LIS just above the railwayline (ARL) have higher ETs values due to a shallowwatertable, lateral seepage from the Yellow River anda network of leaky irrigation canals. Higher ETsvalues are indicated by darker blue colour in Figure 3.

The areas below the railway line (BRL) have lowerETs values because the watertable is quite deep and

Table 3. Seasonal evapotranspiration (ETs) in LIS

Area Area

(ha)

Minimum ETs

(mm)

Maximum ETs

(mm)

Mean ETs

(mm)

Standard deviation

(mm)

Volume

(million m3)

Above rail line (ARL) areaBelow rail line (BRL) areaLIS

27,18728,32555,512

300.2327.3300.2

852.8751.3852.8

577.1537.7557.4

160.3124.6160.7

170.9147.6317.3

Figure 3. Seasonal actual evapotranspiration (ETs) map of the Liuyuankou Irrigation System,northwestern China using MODIS sensor (250 m) for the summer season of 2002


74


there is no surface water irrigation network. LowerETs values are coloured yellow in Figure 3. The ETsmap further shows a spatial gradient of decreasingevapotranspiration from the northern (ARL) partstowards the southern parts (BRL) of the irrigationsystem. In Figure 3, the irrigated rice fields (blue) canbe differentiated from the non-irrigated fields (red) ata spatial resolution of 250 m.

The histograms of ETs for LIS, ARL and BRLfrom an image representing the whole summerseason of 2002 over the irrigation system of LIS areshown in Figure 4. The histogram of ETs shows thewater consumption pattern has many peaks with themain peak features in a covered area is 518 mm @1531 ha for BRL area, 516 mm @ 1087 ha for LISarea and 745 mm @ 437 ha for ARL area.

Conclusion

This study focused on the evaluation of multi-tem-poral MODIS data to calculate actual and seasonalevapotranspiration, applying the SEBAL model.Optical satellite imagery and the SEBAL algorithmprovide estimates of spatially distributed ETa on thedays of satellite overpass. The spatial patterns couldgenerally be explained by the cropping patternsobserved in the field. The problem of spatially dis-tributed seasonal ETa estimation can be overcome byintegrating daily ETa from satellite images acquiredon different dates in a cropping season with the ref-erence evapotranspiration. The seasonal actual eva-potranspiration provides a good indicator of cropwater consumption throughout the study area. Acomparison of ETa estimated from SEBAL withpotential crop evapotranspiration (ETc) measure-ments showed a deviation of –5%, which is within an

acceptable range. The possible reason for deviationof the ETa estimates using the MODIS sensor is pixelsize for small agriculture fields, because the thermalbands of MODIS provide surface temperature overone square kilometre. Estimation of actual ET fromremote sensing indicated relatively good accuracyand potential for use in the water balance and waterproductivity of the LIS for the summer season of2002. In the next step of this study, volume of waterconsumed for different land use types is being esti-mated. This will provide information about thevolume of water lost through fallow land in the LIS.The quantification of non-beneficial ET will helpirrigation managers to develop new strategies forwater saving from the fallow land, which will helptowards sustainable management of water resourcesin the LIS area.

Acknowledgment

The authors are grateful to Australian Centre forInternational Agricultural Research (ACIAR) forproviding a financial support to complete the study.The authors are also thankful to Miss Yaqiong Hu fordownloading the MODIS data.

References

Allen, R., Pereira, L.S., Raes, D. and Smith, M. 1998. Cropevapotranspiration. guidelines for computing crop waterrequirements. Rome, Food and Agriculture Organizationof the United Nations, FAO Irrigation and Drainage PaperNo. 56.

Bastiaanssen, W.G.M. 1995. Regionalization of surface fluxdensities and moisture indicators in composite terrain, aremote sensing approach under clear skies inMediterranean climates. Wageningen, The Netherlands,Agricultural Research Department, Report No. 109.

Bastiaanssen, W.G.M., Ahmad, M.D. and Chemin, Y. 2002.Satellite surveillance of water use across the Indus Basin.Journal of Water Resources Research, 12, 1273.

Bastiaanssen, W.G.M., Chemin, Y., Ahmad, M.D., Ali, S.,Asif, S. and Prathapar, S.A. 1999. Patterns of cropevaporation in the Indus Basin recognized from NOAA–AVHRR satellite. In: Proceedings of the 17th

International Conference on Irrigation and Drainage(ICID), Lausanne, Switzerland, September 1999, 9p.

Chemin, Y. and Alexandridis, T. 2001. Using remotesensing for deriving water productivity indicators in lowdata environments. A case study for different irrigationunit sizes in Zhanghe Irrigation District, China. AsianJournal of Geoinformatics, 5, 3–11.

Figure 4. Sensor histogram of seasonal actualevapotranspiration (mm)


75


Choudhury, B.J. 1994. Synergism of multispectral satelliteobservations for estimating regional land surfaceevaporation. Remote Sensing of Environment, 49, 264–274.

Farah, H.O. 2001. Estimation of regional evaporation underdifferent weather conditions from satellite andmeteorological data: a case study of the Navaisha Basin.Doctoral thesis, Wageningen University and ITC, CIP-Data Koninklijke Bibliotheek, Den Haag, 170p.

Granger, R.J. 1997. Comparison of surface and satellite-derived estimates of evapotranspiration using a feedbackalgorithm. Applications of remote sensing in hydrology.In: Proceedings of the Third International WorkshopNHRI Symposium 17, 16–18 October 1996. Greenbelt,Maryland, USA, NASA Goddard Space Flight Center.

Hafeez, M.M. 2003. Water accounting and productivity atdifferent spatial scales in a rice irrigation system: a remotesensing approach. In: Vlek, P., Denich, M., Martius, C.and Van De Giessen, N., ed., Ecology and development,Volume 8. Cunillier Verlag, Göttingen.

Khan, S., Rana, T., Yuanlai, C. and Blackwell, J. 2004. Canirrigation be sustainable? Proceedings of 4thInternational Crop Science Congress, Brisbane,Australia, 26 September–1 October 2004. At <http://www.regional.org.au/au/cs/2004/symposia/1/7/1399_shahbazkhan.htm>.

Liang, S., Strahler, A.H. and Walthall, C.W. 1999. Retrievalof land surface albedo from satellite observations: asimulation study. Journal of Applied Meteorology , 38,712–725.

Roerink, G.J., Bastiaanssen, W.G.M., Chambouleyron, J.and Menenti, M. 1997. Relating crop water consumptionto irrigation water supply by remote sensing. WaterResources Management. 11, 445–465.

Sakthivadivel, R., Thiruvengadachari, S., Amerasinghe, U.,Bastiaanssen, W.G.M. and Molden, D.J. 1999.Performance of the Bhakra Irrigation System, India,using remote sensing and GIS techniques. Colombo, SriLanka, International Water Management Institute,Research Report No. 28.

Tasumi, M., Bastiaanssen, W.G.M. and Allen, R.G. 2000.Application of the SEBAL methodology for estimatingconsumptive use of water and stream flow depletion in theBear River Basin of Idaho through remote sensing.Appendix C: A step-by-step guide to running SEBAL.Raytheon Systems Company and the University of Idaho,USA, EOSDIS project final report.

Timmermans, W.J. and Meijerink, A.M.J. 1999. Remotelysensed actual evapotranspiration: implications forgroundwater management in Botswana. Journal ofApplied Geohydrology, 1, 222–233.

Vidal, A. and Perrier, A. 1989. Analysis of a simplifiedrelation used to estimate daily evapotranspiration fromsatellite thermal IR data. International Journal of RemoteSensing, 10, 1327–1337.

Wan, Z. 1999. MODIS land-surface temperature algorithmtheoretical basis document (LST ATBD). Santa Barbara,University of California.


76


Implications of environment and institutions for water productivity and water savings: lessons from

two research sites in China

D. Molden1, Dong Bin2, R. Loeve3, R. Barker1 and T.P. Tuong4

Abstract

This paper is based on research conducted at two irrigation systems in China situated in strikingly differentenvironments. The Zhanghe Irrigation System (ZIS) is located just north of the Yangtze River approximately 200km west of Wuhan. The Liuyuankou Irrigation System (LIS) lies south of the Yellow River, just to the east ofKaifeng City. ZIS is situated in hilly terrain with clay loam soil and relatively abundant water resources butincreasing competition for water for other uses. LIS is situated in flat terrain with loam soil and good groundwaterresources in the physically water-scarce Yellow River Basin. What can be learned by contrasting these cases? Thelessons about water productivity and savings form part of a changing trend in thinking about irrigation thatconsiders the analysis of scales, multiple uses, and practices of irrigation in the context of water scarcity.

The paper presents institutional and management arrangements and contrasts water management strategiesat farm, system and sub-basin level and shows how these have led to water savings and increases in waterproductivity. In the water-rich environment of ZIS, farm and canal management of water is much more precisethan in the water-scarce environment of LIS. Yet both systems are close to their water-saving potential. Bothsystems have experienced remarkable increases in irrigation water productivity over time, largely fromincreases in crop yields, but in the case of ZIS also from changes in management. Controlling supplies andreallocating as much as possible to non-agricultural uses while assuring an adequate supply for agriculture isextremely important in ZIS where water productivity per unit of irrigation supply is the key measurement. AtLIS, because of water resource scarcity, there is evidently scope for reducing evaporation from raisedwatertables. Thus, water productivity per unit of evapotranspiration is the key measurement. We suggest thatdesign improvements at LIS be targeted to reduce any non-beneficial evaporation, a recommendation thatholds across many water-scarce environments globally.

1 International Water Management Institute (IWMI), POBox 2075, Colombo, Sri Lanka.

2 Wuhan University, 4300722 Wuhan, People’s Republic of China.

3 FutureWater. Eksterstraat 7, 68233 DH Arnhem, The Netherlands [formerly at IWMI].

4 International Rice Research Institute, DAPO Box 7777, Metro Manila, The Philippines.


77



78


1 Introduction

Between 1999 and 2005, detailed studies of two irri-gation systems were carried out in China.2 TheZhanghe Irrigation System (ZIS) is located just northof the Yangtze River near the city of Jinmen, about200 km west of Wuhan (Figure 1). The LiuyuankouIrrigation System (LIS) lies south of the YellowRiver, just to the east of Kaifeng City.

The different physical and institutional contextsfor each system provided an excellent opportunity togain valuable insights into water savings, water pro-ductivity, institutions and incentives, irrigation oper-ations and infrastructure. Studies were carried out atdifferent scales — field, farm, household, canal level,system and sub-basin level — providing differentperspectives on each of these issues. This paper com-pares and contrasts the two systems to draw outimportant lessons for stakeholders of the systemsand, more widely, for all those involved in improvingirrigation for enhanced water productivity.

The paper is organised as follows. Sections 2 and 3describe the physical and the institutional context andsettings for the two research sites. Section 4 discussesthe incentives to save or reallocate water, and the sixththrough eighth sections the scope for water saving andgains in water productivity. Section 9 is concernedwith scale issues in water-resource management.Section 10 presents strategies for improved water-

resource management in ZIS and LIS. The final sec-tion, Rethinking irrigation, draws some general con-clusions about irrigation that emerge from this study.

2 Comparing ZIS and LIS: the physical context

ZIS lies in the Yangtze River Basin which, from anannual, basin-wide perspective, has ample water, butlocally and in certain seasons physical scarcity maybe an issue. The basin is ‘open’3 in that not all wateris allocated across uses, one of the reasons that Chinais considering a project for south–north watertransfer. Downstream users will not readily noticewhether or not ZIS depletes more water. On the otherhand, the ZIS storage systems are important in pro-tecting the basin from floods.

LIS is situated within the Yellow River Basin, achronically stressed river. This basin is ‘closed’, in thesense that all water is allocated across uses, and thereis arguably not enough water to meet environmental-flow requirements. If LIS depletes more water, otherusers within the basin will be affected, and it will be acontentious issue in river basin management.

ZIS is situated in hilly terrain that gradually flat-tens to the floodplains of the Yangtze. At ZIS, mostdrainage water readily finds its way back to thenatural drainages and river system where it can becaptured and used or reused, and is classified as anatural recapture zone.4 The soil is clay loam with arelatively low percolation rate. Farmers acting ontheir own and the irrigation authorities in the areahave taken advantage of this situation and built thou-sands of reservoirs and ponds of various sizes tocapture drainage flows. Floods far overshadow waterscarcity as an issue in the area. For safety, reservoirsare often drawn down to low levels in the floodseason, a practice that at times stresses the agricul-tural system. Although not a topic of this study, water

Figure 1. Locations of study sites: ZhangheIrrigation System (ZIS) and LiuyuankouIrrigation System (LIS)

2 This paper is based on the results of the project ‘Morerice less water’ supported by the Australian Centre forInternational Agricultural Research (ACIAR).

3 Open basins are those where useable outflow exists (inexcess of acceptable environmental-flow levels) at theend of the basin and there is additional water forallocation across uses. Closed basins are those where allwater is already allocated to human and environmentaluses (Seckler 1996).

4 Using the hydronomic zone classification system(Molden and Sakthivadivel 2000) in which a zone is anarea where similar strategies can be developed. Anatural recapture zone is an area where drainage flowsby gravity to a river drainage, and can be reused from theriver.


79


quality is of increasing concern, especially pollutionfrom agro-chemicals.

On the flat floodplains of the Yellow River, loamysoil with high percolation rates dominates LIS. Thereare two quite distinct zones within LIS — a naturalrecapture zone upstream of the railway line, and aregulated recapture zone5 downstream of the line(Figure 2). Land use upstream of the line is domi-nated by paddy cultivation, with water in excess ofcrop evapotranspiration (ET) either finding its way todrains or percolating to groundwater. The drainseventually flow to the regulated recapture zonedownstream of the railway line. Farmers use drainagecanals and groundwater as primary sources of water.As in ZIS, reuse is prevalent, except that LIS reliesmore on pumping from drains and groundwater,while ZIS uses gravity and surface storage to captureflows.

The role of groundwater is quite different in thetwo areas. At LIS it is a main source of water belowthe railway line through pumping. At ZIS it is a sig-nificant, but indirect source, with high watertablescontributing directly to crop ET. Much of the ground-water at LIS emanates from recharge from theYellow River and rainfall. Much of the Yellow Riverrecharge is induced by pumping. This undergroundwithdrawal of water apparently goes officially unrec-ognised in Yellow River Basin water allocations. AtZIS, groundwater levels are influenced by topog-raphy and paddy irrigation practices, but it appearsthat these are not actively managed to controlgroundwater levels. Fortunately for both areas,salinity is not a major concern. Before large-scalepumping at LIS in the 1960s, watertable rise andwaterlogging led to salinity build-up because of littlesurface or subsurface drainage at the larger systemscale. Because of installation of pumps, the drainageat LIS is now adequate. Table 1 give a summary com-parison of ZIS and LIS.

5 A regulated recapture zone is where drainage water flowsto drains or groundwater, and its reuse can be regulated bypumps or other hydraulic structures.

Figure 2. Layout of the Liuyuankou Irrigation System (LIS). The rice area served bycanals is north of the railway line, while the diversified cropping systemsupported by groundwater lies south of the line.


80


3 Comparing ZIS and LIS: the institutional context

The institutional context has evolved differently inboth situations. The multi-tiered organisation of irri-gation at ZIS is striking, with several actors — theprovincial authority, the Zhanghe Irrigation Admin-istrative Bureau, the canal management authority(three of the four main canals are managed by ZIS,but one is managed by Jingmen City WaterResources Bureau), and village and farmer groups.

The irrigated area of LIS is smaller than that ofZIS, so the LIS is under the direct control of theKaifeng county and city-level authorities. The irriga-tion department of LIS is the main service provider tofarmers at LIS, delivering bulk supplies of water tovillage groups upstream of the railway line.

ZIS tends to function as a demand system becauseof its built-in flexibility to store water in ponds andreservoirs close to the water users (Loeve et al. 2001).While farmers order water through their water usergroups or village heads, many of the decisions aboutwhen to release water from the ZIS reservoir comefrom higher up in the canal operation hierarchy.Thus, there is a strong element of supply approach inwhich the reservoir operators make decisions basedon the available storage, rainfall and an overall viewof when the crop needs water. Our research tends toshow, for example, that the decline in irrigationreleases from ZIS over time (Figure 3) has put pres-sure on farmers to adopt alternate wetting and drying(AWD) irrigation (Cabangon et al. 2004; Moya et. al.2004), to expand ponds (Mushtaq 2004) and to

recycle water (Loeve et al. 2004a). Furthermore, vol-umetric pricing at the village or farmer group level,adopted in the late 1980s, has provided a furtherincentive to save water at the village and farm level(Mao Zhi and Li 1999).

The contrast with LIS is sharp. LIS falls under thelocal administrative jurisdiction, outside of thecommand system of the Yellow River ConservancyCommittee (YRCC), which controls all the divisiongates along the river. Though the LIS has a share ofthe river water, when and how much water can bediverted to the LIS depends on the availability ofwater in the river and the allocation plan drawn up byYRCC and the Provincial Water Resources Bureau.Despite the fact that the Yellow River Basin is shortof water, the institutional structure at LIS, coupledwith a rather poorly developed infrastructure, pro-vides no incentive and facilities for rice farmers northof the railway line (Figure 4) to save water. Becauseof the high watertable and high seepage from irriga-tion canals, practising AWD in rice cultivation is cur-rently out of the question. Below the railway line, thepicture is different, as farmers rely on pumping togrow crops other than rice.

At ZIS, there are multiple needs from the watersector for agriculture, cities and hydropower, andthere are growing environmental concern. Waterresources, initially developed to serve agriculturalpurposes, are being shifted to other uses. Allocatingenough water to these uses, yet meeting agriculturalneeds, is a primary objective of system managers. ZISreservoir managers actively manage the allocation ofwater to different uses. They receive more incomefrom cities and hydropower than from farmers (Table

Table 1. Comparing the Zhanghe Irrigation System (ZIS) and the Liuyuankou Irrigation System (LIS): thephysical context

ZIS LIS

Yangtze Basin – ‘open’ – abundant waterNatural recapture zoneHillyClay loam, low percolation rateDrainage readily (re)captured for reuse – little salinity

Uses surface water storageSurface storage prominentLarge, medium, small reservoirs

Groundwater contributes directly to cropsMainly paddy rice in summer, winter wheat and rapeseeds in winter

Yellow River Basin – ‘closed’ – physically water scarceRegulated recapture zoneFlatLoam, high percolation rateGroundwater pumps recharged by water from rain and other sourcesGroundwater main storage mechanismIrrigation and drainage channels used for rechargePumping from groundwater prevalent, especially in downstream areasGroundwater pumped, and contributes directly to cropsLess paddy rice and a variety of upland crops in summer, mainly winter wheat in winter


81


3). There is a direct incentive to deliver less to agricul-ture (in 2000 water fees per sector were CNY 0.0371for irrigation, 0.068 for cities, and 0.105 for industry).Counteracting this incentive is the role of the Provin-cial Water Resources Bureau which steps into nego-tiations about water allocations and try to ensure thatenough goes to agriculturalists.

At LIS too there is growing competition for limitedsupplies. Similarly, it is important to use water tosupport agriculture, but also to meet other needs. LISirrigation operators regulate the distribution of theirshare of Yellow River water only after the river waterhas passed through the Yellow River diversion gate.In contrast to ZIS, LIS delivers water primarily tofarmers. LIS operators are charged a flat, area-basedfee, and thus have an incentive to irrigate more area.On the other hand, more releases from the YellowRiver would allow the maintenance of a high ground-water table in the rice area through seepage and infil-tration. Moreover, the hydraulic infrastructure at LISaffords such poor water control that more precisedelivery measures are difficult without an overhaulof the physical system, something which systemmanagers have often pointed out.

4 Water savings and reallocation — why save water?

The numerous participants in both systems have dif-ferent reasons to save water. One of the findings ofthe study that was clearly brought home is that theterm water savings is potentially misleading becauseof these different perspectives. It is would be better tounderstand how the flow path of water within thebasin or system changes, and evaluate the trade-offsfor various stakeholders from the basin to the farmerlevel, than to say whether or not an intervention saveswater. To demonstrate this, we discuss the concept ofwater savings from various perspectives.

We have already noted that there is a surplus ofwater in the Yangtze River Basin. Although there hasbeen much debate about the benefits and costs, thereare already plans to move water north from theYangtze River Basin, with the first priority to meetrising non-agricultural demands. At a smaller scalewithin the study area, the Zhanghe Irrigation Admin-istrative Bureau tends to allocate as much of the res-ervoir water as possible to higher-value, non-agricultural uses and therefore benefits from prac-

Table 2. Comparing the Zhanghe (ZIS) and Liuyuankou (LIS) irrigation systems: the institutional context

ZIS LIS

Multi-tiered organisational structure

ZIS reservoir authority serves agriculture, cities, hydropower usesFinancially autonomous reservoir operating authority receives revenue from farm and non-farm sources — financially well-offVolumetric pricing

Good infrastructure, with adequate controlling structuresAlternate wetting and drying irrigation promotedFarm ponds expanded over timeThe fee gai shue (FGS) policy promotes payment directly to irrigation authority

Under the local government system, not the command system of the Yellow River Conservancy CommitteeIrrigation department serves primarily farmers

Finances collected from farmers are insufficient for irrigation department

Flat rate pricing — lack of incentives to promote farm water savings practice for paddy riceInadequate controlling structuresAlternate wetting and drying currently not possibleHeavy groundwater pumping by individuals

Table 3. Sources of income (104 yuan) for Zhanghe Irrigation System ZIS reservoir operation

Appropriation funds from the

provincial government

Gross income from agricultural irrigation

water supply

Net income from the city and industry water supply

Net income from power generation

Other mixed businesses

Total

Average value (1998~2002) 293 315 246 193 23 1070


82


Figure 3. Changes in water released to agriculture and other uses over time in the Zhanghe IrrigationSystem

Figure 4. Trends in water use in the Liuyuankou Irrigation System. While Yellow River diversionshave fallen, pumping from groundwater has increased.


83


tices such as AWD, volumetric pricing, canal liningand pond development that enable them to reducetheir allocations to agriculture without loss in agri-cultural production. The canal managers face a dif-ferent problem. For operating and maintaining thecanals, they rely on payments from water users.

During the years 2002 through 2004, a majorpolicy change took place that affected the way waterwas delivered to farmers from the main reservoir.The policy of fee gai shue (FGS) required thatfarmers get organised to request water from the irri-

gation system and make payments directly to the irri-gation authority, when previously their requests wentthrough the village. Water deliveries, and henceincome from water fees, were sharply down andcanal managers faced a severe budget constraint. Inresponse, farmers relied more on their small storageponds for water supply, and practices such as AWDhelped them to adapt to this policy shift. The farmershave faced both incentives (volumetric pricing) andpressures (reduced deliveries and FGS). As waterdeliveries to irrigation from ZIS have declined

Table 4. Incentives and pressures to save or reallocate water

Group Zhanghe Irrigation System Liuyuankou Irrigation System

Farmer perspective

Action

Rationale

Incentives

Farmers acting out of necessity in light of decreasing agricultural water supplies

Apply less water through alternate wetting and drying strategy, and increase water fees Response to decreasing supplies

Necessity — get enough water to cropsVolumetric pricing to village or farmer groups, cost savings pro-rated to farmers

Understanding the condition of water shortage in the basin and having pressure from water resource managers to ‘use’ less waterApply less water

No good reason for farmers, but there is a view that rice is highly water consuming and a long term habitNo great incentive for paddy rice water savings at present

Irrigation operators

Action

Rationale

Incentives

No external pressure but incentive to deliver more water within the systemReduce ‘losses’ from delivery system

Deliver more water to customersDeliver more water to cities and industriesMore fee collection from farmers, and higher payments from cities and industries.

Great external pressure but little internal incentiveReduce ‘losses’ from delivery system and deliver less water to rice growersDeliver more water to more customers

Expansion of effective irrigation area by canal water

Water resource managers, society

Action

Rationale

Incentives

Obtain higher value from water by responding to increasing demand from other uses, yet keeping agriculture healthyReduce withdrawals from reservoir for agriculture

Maximising the benefits from the reservoir water

More value from water

Reduce water for rice to release water for other uses, yet maintain food production.

Reduce withdrawals from Yellow River and reduce overall evapotranspiration.Maximising the benefits from the Yellow River Conservancy Committee allocation, better equity above and below the railway lineMore value from water (increase water productivity) and better equity


84


(Figure 3) they have adopted AWD, increased thenumber of ponds and recycled water.

The Yellow River Basin is short of water and thereis considerable pressure to reallocate water, espe-cially to the lower reaches of the Basin (Henan andShandung provinces). Water releases from theYellow River to LIS have gradually declined and thistrend might be expected to continue. Meanwhile,pumping of groundwater for both agricultural andnon-agricultural purposes (in nearby Kaifeng City)has increased (Figure 4). Given this situation, the irri-gation operators would benefit by reducing seepagefrom the canal system and delivering less water to therice producers, and delivering more water down-stream of the railway line. A reduction in water torice could lead to a fall in groundwater levels and areduction in evaporation losses. Farmers might beencouraged to adopt AWD if convinced that therewould be no sacrifice in yield. Reallocation of watercould benefit farmers below the railway line.

One of the important and surprising contrastsbetween ZIS and LIS is the physical context andmotivation for saving water. ZIS is located in a phys-ically water-abundant area, while LIS is in a physi-cally scarce area. Yet at ZIS there are more water-saving activities at farm and irrigation-system scalewith farmers practising AWD, and managers andextension agents actively promoting means of watersavings. At LIS, there is no great incentive forfarmers to practice water savings for paddy rice, andthe amount of water delivered to fields in relation toET is high when compared with circumstances in ZIS(Table 5). Again unlike in ZIS, LIS system managersalso have no great incentive to be stricter on waterdeliveries. From a basin perspective, the place thatneeds to practise water savings (LIS) does not payattention to it, while the place where scarcity is not anissue (ZIS) is actively practising water savings. Whatis the reason for this contradiction?

Rice farmers above the railway line at LIS do notseem to have a compelling economic reason to refusehigh rates of water delivery to their fields. At ZIS onthe other hand, farmers feel compelled to adopt water-saving practices. Yield levels with AWD are about thesame as those from traditional practices, and costs arealso about the same (Cabangon et al. 2004; Moya etal. 2004). The main reason that farmers practise AWDat ZIS is apparently a response to a declining supply ofwater to irrigation. The farmer motivation is one ofsurvival — the necessity to cope with falling watersupplies. AWD helps farmers to adapt to a condition

of lower supply and obtain at least the same output fora lower water input. At LIS, water deliveries remainhigh. Farmers have no real reason to practise AWD,and in fact do not.

We can explore the role of pricing in both cases.There is a flat, area-based pricing scheme at LIS, sothere is no incentive from rice farmers to reducedeliveries. Downstream of the railway line, farmerspay the electrical costs of pumping, and employwater-saving practices and technologies (forexample, they use a flexible pipe called a ‘whitedragon’ to carefully deliver water to fields withminimum seepage). At ZIS, in contrast, volumetricpricing at the village or farmer group level was intro-duced in the 1980s (Mao Zhi and Li 1999). Costsavings are pro-rated to individual farmers, providingan incentive to adopt water-saving practices. Onecould argue, however, that reduced deliveries fromthe reservoir provided the primary incentive foradoption of water-saving practices.

5 Basin and system level outcomes

As already noted, there is pressure for water savingsalong the Yellow River. Any wasted water in agricul-ture would readily be used to serve environmental,industrial or urban needs or, for that matter, to betterserve agriculture. There is intense societal pressure tosave water.

At ZIS, in contrast, there is a need to make sure thatvarious sectors are allocated sufficient water. Thisprocess of allocation and re-allocation is done at thesub-basin level by reservoir managers at ZIS. In con-

Table 5. On-farm water application for paddy growthseasona in the Zhanghe (ZIS) andLiuyuankou (LIS) irrigation systems

ZIS (1999–2000)

LIS (2001–2003)

Irrigation application (mm)

417–470 512–590

Rainfall (mm) 407–310 462–360

Total inflow (mm) 824–780 974–950

Rice evapotranspiration (mm)

613 525

Yield (kg/ha) 7925–6500 7636a The growth season for paddy is from about 20 May to 10

September at ZIS, and from 20 June to 20 October at LIS. Thedata are from Lu et al. (2003) and Dong et al. (2004)


85


trast to the Yellow River, along the Yangtze Riverthere is evidently no great pressure to make sure morewater flows in the river.

At both systems, an important way to reallocatewater is to ‘save’ water that is perceived to be wasted(down the drain), and reallocate it to other uses. Thisalready takes place at both systems. The depletionfraction measured at the scale of ZIS shows thatabout 90% of water is depleted by various uses andthat there is therefore little remaining scope for addi-tional savings. In fact, of concern is the need to meetdownstream commitments for human or environ-mental needs. These should be considered beforetrying to recapture further water before it leaves thesystem boundaries.

Reducing evaporation or transpiration fluxes isanother way to ‘save’ water. Evaporation is targetedfirst, because transpiration is directly related to themarketable yield of crops. At LIS, there are evidentlyhigh rates of non-productive evaporation from areasof shallow groundwater. It seems that no attention isbeing given to this at present. A reduction in thiswould free-up water that could be reallocated. AtZIS, crop ET is only 36% of total ET in the area.Major shifts in land use could change ET. A reduc-tion in ET from the area would, however, mean moreflows reaching the Yangtze River, which is poten-tially detrimental given the river’s propensity toflooding, and not necessarily beneficial from theoverall basin perspective. The past strategy has beento limit drainage flows out of the ZIS area, andconvert these into more ET.

Table 6 summarises incentives by different actorsto change water use. At ZIS, the incentive is one ofsurvival in light of reduced supplies, and for system

managers, there is a strong financial incentive. Thereis little incentive for farmers or system managers totarget evaporation losses.

6 The role of secondary storage and rain

Irrigation studies have traditionally considered therole of delivering water from the main sources — res-ervoirs or canals. An important reason for this is thatinitial investments are made in dams, reservoirs,diversion structures and canal systems, and theiroperation is given due importance. In some cases,secondary storage is built within irrigation to helpoperations. In other cases, the importance of sec-ondary storage such as small dams or reservoirsevolves over time.

At ZIS, there are literally thousands of small andmedium reservoirs within the system, some of whichare from the original design in the 1950s, and otherswhich have been added by farmers or local authori-ties. Most farmers receive water from the main irri-gation canals originating from ZIS. In addition, manyfarmers receive water from small reservoirs or ponds,and some farmers from a third surface source —pumping from drainage canals, or simply usinggravity-driven drainage flows from upstream. Thecombined management of these ultimately deter-mines water productivity. As part of the response todeclining water releases from the main reservoir toagriculture, farmers have relied increasingly on thesealternative sources.

Over time, farmers have increased the number ofponds (Mushtaq 2004). The temporary introductionof FGS (2002–2004) forced farmers to rely much lesson the main reservoir at ZIS. Table 7 shows that. eventhough reservoir releases to agriculture fell sharplyduring 2002 and 2003, the overall area and yield didnot suffer, apart from a slight reduction in averageyield in 2002. Farmers were able to rely on rain andfarm ponds for their main water supply. The questionarises as to the significance of these ponds and theirrelation to water-saving irrigation practices such asAWD in enabling the reduction of releases from ZIS.It is generally accepted that the ponds, by providingfarmers with a source of water on demand, have facil-itated the adoption of AWD. Another question iswhether agriculture needs any water from the mainreservoir if local sources can provide the supply.Modelling by N. Roost (formerly IWMI, unpub-

Table 6. A summary of incentives to save wateramong the different stakeholders in theZhanghe (ZIS) and Liuyuankou (LIS)irrigation systems

ZIS LIS

Farmers – reduce application reduce E or ET

MediumNone

LowNone

Irrigation managers – reduce delivery to agriculture High Low

Basin resource management reallocation reduce E

HighLow

HighHigh


86


lished data) shows that in normal years this may bepossible, but in dry years, the reservoir provides alife-saving water source of water.

7 Response to reduced supplies

A response at LIS to reduced deliveries from theYellow River has been to increase pumping fromgroundwater. It is doubtful whether the overallsupply of water to crops has fallen significantly overtime, and whether the overall ET has changed.Groundwater, much of which emanates from theYellow River itself, has simply replaced surfacewater diversions into the system. Thus, the ground-water plays a very big role in sustaining agriculturalpractices and productivity at LIS (Table 8).

Management of rain, both at farm and sub-basinscales plays an important role in overall water man-agement at ZIS. At the farm level, again in responseto reduced deliveries, farmers capture as much rain aspossible by building high bunds and practisingAWD. The latter maintains low water levels withinthe fields and storage volume to capture rain.

The ZIS system configuration is very effective atcapturing and using rain at larger scales. Internalcatchments in the system provide water to small and

medium reservoirs that ultimately serve farmers.Many small ponds capture excess flows resultingfrom off-field drainage of rain and irrigation water.At larger scales this capture and recapture of run-offand drainage flows ultimately keeps water within thesystem to meet the needs of various uses.

At LIS, the rain serves as an important source ofrecharge. At large scales at ZIS, almost all rain iseffectively utilised by agriculture, either directly bycrops, or indirectly by providing recharge to ground-water, which is then pumped again for agriculture.

8 What is the scope for water savings and water productivity gains?

Looking at sub-basin scale at ZIS and LIS, thedepleted fraction is already quite high in both sys-tems, and reducing outflow could have adverse con-sequences for downstream uses. (The depletedfraction of gross inflow is the evaporation and tran-spiration by all uses divided by the rain plus irrigationinflow.) At both systems, the process fraction ofdepleted water is not extremely high. (The processfraction of gross inflow is the rice ET divided by rainplus irrigation inflow and indicates the amount ofinflow that is depleted by ET rice.) At ZIS, much

Table 7. The introduction of the fee gai shue (FGS) policy resulted in less water being released from the ZhangheIrrigation System (ZIS) reservoir, but the overall area under paddy and yields were not greatly affected,demonstrating the role of secondary storage in overall water management

Year Water release from the Zhanghe Reservoir

(100 million m3)

Rainfall(mm)

Area irrigated by water from Zhanghe

Reservoir (’000 ha)

Planted area with paddy in whole ZIS

(’000 ha)

Yield(t/ha)

200220032004

0.140.381.35

568590703

9.6447.4463.00

105102113

8.737.528.76

Note: farmers reported that the paddy yield in 2002 fell about 20~30%, but the yield data available from ZIS records show that theaverage yield decline was small. Rainfall figures are from Tuanlin Research Station and other data from ZIS records.

Table 8. Responses to reduction in supplies of water to irrigation in the Zhanghe (ZIS) and Liuyuankou (LIS)irrigation systems

ZIS LIS

Alternate wetting and drying irrigationIncreased ponds and secondary storage

Effective use of rainMore precise deliveryCropping pattern change (from two to one rice crop; from paddy to upland crops)

More pumpsControlled recharge of groundwater through irrigation and drainage systemsRain important to recharge groundwaterReduction of outflow to downstream areasReduction of paddy area


87


non-process depletion provides water for non-cropvegetation. At LIS, however, there is a large amountof non-productive evaporation and apparently scopefor reducing evaporation.

The concept of water productivity incorporateswater savings as well as gains in mass or value. If lesswater is applied or depleted, this is related to savings.The numerator of the water productivity equationincreases when more mass or value is achieved. Soeven if there is little scope for savings, there could bepossibilities for gains in value through improvingyield or value of output (reducing input costs orchanging crops) for the same amount of water, byreallocating irrigation water to higher-value useswithin agriculture (higher-value crops) or betweensectors, or by reducing negative externalities. Yieldlevels are already quite high in both systems. But inboth systems there already is a reallocation acrosssectors that increases the benefit per unit of water.The challenge is to manage this reallocation to ensurethat agriculture is able to maintain productivity. Thebiggest productivity gains may be in reducing exter-nalities such as pollution or damage to other users,but we did not study this aspect in great detail.

9 Scale and water resources management

The studies have amply demonstrated the importanceof considering scale in agricultural water manage-ment. Considering actions at only the field scale andsimply extrapolating up to system or basin level ishighly likely to lead to misunderstanding. Many

other factors come into play when considering waterproductivity at larger scales (Table 9).

At ZIS, our research was aimed at understandingthese cross-scale interactions. Figure 5 shows thedepleted fraction [ET/(surface and subsurface inflowplus rain)] at different scales. At field scale, thedepleted fraction is quite high, because farmers care-fully manage limited supplies including rain. But at ameso scale the depleted fraction drops at the studyarea, because the area contains forests which act as acatchment for downstream areas. Yet at larger scales,

Table 9. Factors that influence water resource use and productivity at various scales in the Zhanghe (ZIS) andLiuyuankou (LIS) irrigation areas

ZIS LIS

Micro (field) scale

Meso scale

System scale

Sub-basin

On-farm water management practicesLocal influences of groundwaterRun-off and capture in small ponds from other land uses

Influence of non-agricultural usesWater delivery practicesUse of water from internal storagePolicies – fee gai shue (FGS) Influence of multiple usesConsideration of downstream needs

On-farm water management practices

Reuse of drainage flows originating upstream of railway lineGroundwater recharge and reusePumping or recharge of drainage water originating from upstream areas

Groundwater interaction with nearby citiesInduced recharge from Yellow River Basin

Figure 5. Depleted fraction [ET/(inflow + rain)]estimated at different scales in theZhanghe Irrigation System (Loeve et al.2004). The depleted fraction (DF)available adjusts for canal inflow minusoutflow across the study domain.Differences in DF across scale are due tofarmer practices, influences of other landuse, capture of internal run-off and reuseof drainage flows.


88


the depleted fraction increases again because ofcapture and use of run-off, and recapture and reuse ofdrainage flows. Ultimately, at system level thedepleted fraction is quite high, showing that the scopefor additional saving in the area is limited.

10 Strategies for water management at ZIS and LIS

Not surprisingly, the strategies employed at ZIS andLIS differ markedly due to the difference in contextdiscussed above. At ZIS the basic approach is to:• Keep as much water as possible in upstream

storage (considering too the need for flood control,which requires low water levels in reservoirs atcertain times of the year)– Reduce releases to agriculture– This promotes the development of internal

secondary storage• Use stored water to control reallocation to different

uses– Water in the reservoir can be better targeted for

city or hydropower use• Promote gains in water productivity per unit of

irrigation supply– Because farmers receive an increasingly smaller

supply, AWD is a means to adapt. The sameyield can be achieved with less water input fromthe reservoir.

– Reduce seepage from conveyance structures toallow a higher proportion of canal water to reachfarms.

An alternative strategy would be to release amplewater from storage, and rely on recycling and reuse ofdrainage flows. A major disadvantage of this strategyis that it is more difficult, or even impossible, forsystem managers to control water once it leaves theirmanagement domain. For example, if water seepsfrom canals, farmers are able to reuse it, but systemmanagers cannot capture this water for delivery toother uses.

At LIS, the prevailing strategy is one of conjunc-tive use. Groundwater provides an important bufferin case Yellow River supplies are further reduced.The results of this strategy have been impressive interms of low water wastage and high water produc-tivity.

The common view at LIS is that, because rice is aheavy user of water, deliveries to rice farmers shouldbe reduced, or the area under the crop should be

reduced. An alternative strategy to the prevailingone is to reduce deliveries to rice, and promotesurface water deliveries below the railway line(Figure 2). This could be done by providing bettercanal control and promoting AWD practices. How-ever, we question whether this will lead to net gainsor just a redistribution of water, lessening the need topump groundwater.

Instead, we propose a shift away from thinkingabout reducing deliveries to reducing any non-pro-ductive evaporation. The strategy is to identifywhere evaporation is occurring and control water toreduce this. Evaporation occurs from shallow water-tables, especially before and after the rice season.This could be reduced by introducing drainage orreducing deep percolation. Applying AWD may alsolower the groundwater depth and therefore reducethe non-beneficial evaporation from fallow landwithin the rice area. Crop ET of rice is higher thanfrom other potential crops such as maize which hasan average ET value of about 420 mm comparedwith about 525 mm for rice.

There is often confusion and debate about whetherwe should be thinking in terms of depletion (ET) ordeliveries of water. Obviously, both are important,but they carry different levels of importance in dif-ferent contexts. In the highly stressed Yellow RiverBasin, which is closed and over-committed, waterproductivity analysis is better focused on ET. Onlyby reducing ET will more water be made available,yet maintaining levels of transpiration is importantfor crop production. In fact, efforts should first befocused on decreasing non-productive evaporation.Manipulating deliveries is a means to achieve this,and reducing deliveries is not the end in itself.

In contrast, in the water-abundant Zhanghe Basin,deliveries are far more important than ET from a waterresource perspective (from a service perspective ET isimportant because it defines crop water demand).More or less ET will not be noticeable in the basin con-text. On the other hand, water control, keeping water instorage in the upstream areas of the system, and deliv-ering less water, are all means of reallocating water todifferent uses and of reducing outflows from thesystem. Thus, considering water productivity in termsof deliveries is entirely appropriate.

11 Rethinking irrigation

In this final section we attempt to identify the generallessons from our research in China. How can the


89


lessons learned from our two sites be interpreted on abroader scale? The prevailing notion that guidesmany decisions is that irrigation, especially irrigationfor rice, wastes a lot of water. Thus, the focus hasbeen on reducing deliveries and losses from deliv-eries by lining canals and introducing drip and sprin-kler irrigation. The focus is on irrigation supply, andthe classical concept of irrigation efficiency (typi-cally estimated at 40% in many Asian systems)doesn’t consider return flows and leaves rain out ofthe analysis. There is also the notion that farmers arethe only customers and that irrigation managers areserving farmer needs. It is commonly felt that inter-ventions at the system level (more control infrastruc-ture, better organisation and management) are themain means to change practices. Therefore, iffarmers would pay the real cost (full cost) of water,saving water would take place. In light of scarcityand competition, an expanded view is required whenin comes to developing a strategy for increasing theproductivity of water.

Everyone will agree on the importance of watersavings to make most effective use of water. But wehave shown that there can be several perspectives(farmer, irrigation manager and society), with dif-ferent and competing objectives (save water so thatmore area can be irrigated, save water to save money,save water so that it can be reallocated to cities).Rather than using water savings as an operationalterm, it would be better to follow paths of water fromsource, to delivery to a use, to evaporation, run-off anddeep percolation flows, then to the fate of these flowsincluding reuse. Changes in management strategieswill affect flow paths, incurring costs for some, andproducing benefits for others. Decisions should beguided based on these changes in flow paths, and anunderstanding of who gains, and who bears the cost.

Quite often, farmers rely on multiple sources ofwater including both ground- and surface-waterstorage, yet much effort by irrigation authorities isplaced on managing reservoir and canal water. Rainrepresents a significant source that is often over-looked. A challenge is managing rain by capture inthe field and harnessing run-off generated within irri-gation systems as is done in ZIS. Strategies shouldbetter take into account that farmer investments inconstructing sources and tapping sources, such as wehave seen in ponds at ZIS, can mitigate problems ofscarcity and affect what happens at a larger scale.

Irrigation must be a responsible user of water in abasin context, as demands for non-irrigation uses of

water grow. In spite of calls for integration, irrigationis still dealt with in isolation. In reality, irrigatorsoften have no choice but to adapt to decreasing sup-plies due to reallocation to other uses, as happened inboth case studies here. Not only is it important tounderstand these cross-sectoral interactions, but alsoto engage in negotiations across sectors.

An understanding of context and scale considera-tions will help to identify opportunities and avoid pit-falls. It is vital to consider the system- and basin-levelconsequences of actions taken at farm and field scale.Similarly, basin actions such as reallocation affectfarm actions. The concept of open and closed basinsprovides an initial insight on context. Strategiesappropriate for open basins — managing deliveriesfor high-value productivity while sustaining agricul-tural production — may have to shift when basinsclose due to increased development and competitionfor water resources. In closed basins, typically foundin the semi-arid regions, there appears to be an oppor-tunity for water productivity gains through reducedevaporation. This has not yet been a focus of manywater-savings activities. With increasing populationand demands on water, basins will become moreclosed, and there will be a need to shift our thinkingto the use (evaporation and transpiration) side of theequation, rather than the supply side.

Especially in closed basins, it is important to recog-nise that a change in use will affect other uses of water.Strategies to enhance water productivity should firsttarget flow paths where the use of water is generatingnegative or low values (recognising the values gener-ated by other ecosystems) — for example, evaporationfrom shallow watertables. If a change is suggested, it isimportant to evaluate what happens to the water flowpaths, then consider the trade-offs, who wins, and wholoses. For example, a reduction in drainage flow mayaffect a downstream user. Is or should the downstreamuser be compensated?

Following this logic, reducing evaporation inclosed basins such as the Yellow River Basin bringsan opportunity to free-up water with minimal impacton other uses. This is a much different approach andrequires different analysis than approaches that targetreducing deliveries (e.g. sprinkler irrigation) orseepage (canal lining). The approach is to identifyand quantify non-productive evaporation fluxes, thendevelop strategies on how to reduce these.

Our studies and experience have clearly demon-strated that there are multiple actors (farmers, irriga-tion managers, basin managers, broader society) who


90


influence effective use of water, yet who typicallyhave quite different outlooks and objectives on wateruse. Policies and strategies for changed water use andmanagement must aim at aligning these objectivesand incentives for all actors to obtain wider goals ofimproved water use.

Acknowledgments

We gratefully acknowledge the support provided byACIAR for the project (LWR/2000/030) ‘Growingmore rice with less water: increasing water produc-tivity in rice-based cropping systems’. The projectwas carried out in collaboration with Wuhan Univer-sity, the National Centre for Irrigation and DrainageDevelopment in China, the International RiceResearch Institute, the International Water Manage-ment Institute, and the Commonwealth Scientific andIndustrial Research Organisation in Australia.

References

Cabangon, R.J., Tuong, T.P., Castillo, E.G., Bao, L.X., Lu,G., Wang, G., Cui, Y.L., Bouman, B.A.M., Li, Y.H.,Chen, C.D. and Wang, J.Z. 2004. Effect of irrigationmethod and N-fertilizer management on rice yield, waterproductivity and nutrient use efficiencies in typicallowland rice conditions in China. Paddy and WaterEnvironment, 2(4), 195–206.

Dong, B. Molden, D., Loeve, R., Li, Y.H., Chen, C.D. andWang, J.Z. 2004. Farm level practices and waterproductivity in the Zhanghe Irrigation System. Paddy andWater Environment, 2(4), 217–226.

Loeve, R., Dong, B., Zhao, J.H., Zhang, S.J. and Molden, D.2001. Operation of the Zhanghe Irrigation System. In:Barker, R., Loeve, R., Li, Y.H. and Tuong, T.P., ed.,Water saving irrigation for rice. Colombo, Sri Lanka,International Water Management Institute.

Loeve, R., Hong, L., Dong, B., Man, G., Chen, C.D., Dawe,D. and Barker, R. 2004a. Long term trends in intersectoral

water allocation and crop productivity in Zhanghe andKaifeng, China. Paddy and Water Environment, 2(4),237–245.

Loeve, R., Molden, D., Dong, B., Li, Y.H., Chen, C.D. andWang, J.Z. 2004b. Issues of scale in water productivity inthe Zhang He irrigation system: implications forirrigation in the basin context. Paddy and WaterEnvironment, 2(4), 227–236.

Lu, G., Cabangon, R., Tuong, T.P., Belder, P., Bouman,B.A.M. and Castillo, E. 2003. The effect of irrigationmanagement on yield and water productivity of inbred,hybrid and aerobic rice varieties. In: Bouman, B.A.M.,Hengsdijk, H., Hardy, B., Bindraban, P.S., Tuong, T.P.and Ladha, J.K., ed., Water-wise rice production.Proceedings of the international workshop on water-wiserice production, 8–11 April 2002, Los Baños, Philippines.Los Baños, International Rice Research Institute, 15–28.

Mao Zhi and Li Yuanhua 1999. Effect of reform on watercharge in water saving in irrigation scheme in China. In:Proceedings of 2nd Inter-regional Conference onEnvironment – Water, 1–3 September 1999. Lausanne,Switzerland, Presses Polytechniques et UniversitairesRomandes.

Molden, D.J., Sakthivadivel, R. and Keller, J. 2001.Hydronomic zones for developing basin waterconservation strategies. Colombo, Sri Lanka,International Water Management Institute, IWMIResearch Report 56.

Moya, P., Hong, L., Dawe, D. and Chen, C. 2004. Theimpact of on-farm water saving irrigation techniques onrice productivity and profitability in the ZhangheIrrigation System, Hubei, China. Paddy and WaterEnvironment, 2(4), 207–215.

Mushtaq, S. 2004, An assessment of the role of ponds onsustaining crop production and adoption of water savingirrigation practices in China. PhD thesis, University of thePhilippines, Los Baños, Philippines.

Seckler, D. 1996. The new era of water resourcesmanagement: from ‘dry’ to ‘wet’ water savings.Colombo, Sri Lanka, International Water ManagementInstitute, IWMI Research Report 1, 17p.


91


Evaluating system-level impacts of alternative water-saving options

Shahbaz Khan1, Jianxin Mu2, Yaqiong Hu2, Tariq Rana1 and Zhanyi Gao2

Abstract

Improving water-use efficiency is crucial to both Australia and China. In the case of China, development in thelast half century has resulted in the irrigated area growing to 53 million ha, which accounts for 40% of thefarmland. The total water used for irrigation in China gradually increased from approximately 100 billion m3

in 1949 to 358 billion m3 in 1980, since when it has stabilised. However, competition for water from othersectors is increasing. Industrial and municipal water use, for example, have increased rapidly and reduced theproportion used in irrigation from 92% in 1949, to 80% in 1980 and 65% in 1997. This situation demandsmajor improvement in water use-efficiency for irrigated agriculture if current production levels are to bemaintained or enhanced.

Irrigated agriculture makes up 70% of Australia’s consumptive water use. With the water resources inirrigation areas being close to fully allocated, or even over-allocated in some catchments, there is increasedcompetition for water. It is generally accepted that there will be less water available for irrigated agriculturein future and the only way to provide enough water for that purpose will be to use the available resource moreefficiently at both farm and catchment scales.

In both China and Australia, since it is almost impossible to withdraw more water from existing resources,the present irrigation practices and future irrigation developments should focus on improvement of water-useefficiency in all sectors at all scales. However, savings from one part of the system may lead to higher wateruse in another part of the system and the overall improvement may be negligible. Some measures that mayimprove the water productivity in agriculture are canal lining, irrigation scheduling, advanced irrigationtechnologies, improved cropping patterns and conversion to crops with higher economic returns. The key toachieving real and substantial water savings lies in the assessment and hydrologic ranking of water savingoptions in a whole-of-system context.

This paper describes results of a major water-use efficiency study in the Murrumbidgee Valley, Australia andsimilar basin-level studies in China. Benefits of a systems approach are summarised through a hydrologicevaluation of water saving interventions at the field, irrigation area and catchments levels for the case studies.Use of a multi-scale, whole-of-system approach can lead to true water savings, improved socioeconomicconditions and future environmental sustainability.

1 CSIRO Land and Water and Charles Sturt University,School of Science and Technology, Locked Bag 588,Wagga Wagga, NSW 2678, Australia. Email: [email protected]; [email protected]

2 China Institute of Water Resources and HydropowerResearch, 20 West Chegongzhuang Road, Beijing100044, People's Republic of China.


92



93


Introduction to a multi-scale hydrologic systems approach

Improving water-use efficiency is crucial for Aus-tralia and China. In both countries, since it would bealmost impossible to withdraw more water fromexisting resources, the present irrigation practicesand future irrigation developments should focus onimprovement of water-use efficiency at both fieldand catchment scales. The key to achieving real andsubstantial water savings lies in the assessment andhydrologic ranking of water-saving options in awhole-of-system context.

Figure 1 shows the water cycle in an irrigatedcatchment at different spatial scales. Key interven-tion points for improving the sustainability of irriga-tion systems are shown with numbers in circles. Thefactors that can be acted on at these interventionpoints are described below:1. the volume and regime of water extraction from

the river; water rights definition; trading and

regulation of use of water rights; improveddistribution and control of water delivery to farms

2. the volume and regime of water extraction fromgroundwater — extraction must be matched bycatchment and river recharge; improved deliveryto farms

3. the volume and regime of subsurface drainage;improved management to reduce leaching anddrainage to groundwater; reduction of salt load togroundwater through soil storage; improvedinterception of subsurface drainage water andreuse through bio-concentration and extraction;salt management schemes for subsurface drainageand groundwater

4. reduction of water extraction through greater on-farm water-use efficiency

5. improved management of surface water drainage;improved reuse; reduction of contaminants

6. land-use management to control water yield andthe amounts of salt and pollutants carried to riversand groundwater

Figure 1. Schematic of an irrigated catchment, showing with key points of intervention for water saving


94


7. adaptive irrigation management undercircumstances of climatic variability and change.This paper describes results of water-use effi-

ciency studies focusing on intervention points 1–5 inFigure 1, for catchments in Australia and China.

Application of systems approach in Australia

To identify ‘true’ water-saving options it is importantto adopt a multi-scale systems approach foraccounting of all surface water and groundwater useand losses and devise interactions at the catchment,irrigation area and farm levels. An example of asystems analysis — for the Murrumbidgee River inAustralia — is given in the following sections.

Catchment scale

The Murrumbidgee River (Figure 2) has a catch-ment area of around 84,000 km2 and a length of 1600km from its source in the Snowy Mountains to itsjunction with the Murray River. The Murrumbidgee(MIA) and Coleambally (CIA) irrigation areas aremajor irrigation areas situated along the Mur-rumbidgee Catchment. The geographic boundaries ofthe Murrumbidgee catchment include the GreatDividing Range in the east, the Lachlan River Valleyto the north and the Murray River Valley to the south.

The 100-year flow averages show that the totalwater resources of the Murrumbidgee catchment aremade up of average annual flow downstream of Burr-injuck and Blowering dams of around 2900 GL.From dam walls to Wagga Wagga, there is a net gainof around 1460 GL/yr. Between Wagga Wagga andDarlington Point there is an apparent loss of around170 GL/yr. From Darlington Point to Hay the riverrecharges the aquifers of the lower Murrumbidgeesystem with an apparent loss of around 120 GL/yr.Similarly, between Hay and Balranald there is anapparent loss of around 190 GL/yr. An example of asystem’s water balance in shown in Figure 2, for theMurrumbidgee catchment in Australia in 1991.

Under the present cropping regime and irrigationpractices in the Murrumbidgee catchment, theaverage true water losses in the system are:• evaporation from the river ~ 70 GL• supply and storage losses (seepage and

evaporation) in the MIA ~100 GL• on-farm losses in the MIA ~ 90 GL• supply losses (seepage and evaporation) in the CIA

~30 GL• on-farm losses in the CIA ~ 45 GL.

This analysis has shown true losses of greater than300 GL (Khan et al. 2004b), indicating the potentialfor real water savings and better environmental man-agement by investments in management and infra-structure.

Figure 2. A system’s water balance: the Murrumbidgee River, Australia at catchment level 1991. All valuesare in gigalitres (1 GL = 106 litres = 106 m3)


95


Irrigation area scale

Systems analysis at the irrigation area scale providesindications of water savings at the whole-of-irrigation-area level. Figure 3 gives a system water balance forthe Coleambally Irrigation Area (CIA), which spansover 80,000 ha of land. This water balance provides apossible water-use efficiency scenario for the CIA(using 2000–2001 water allocations). The water-useefficiency at various points within the system isexpressed in terms of water delivered versus the watersupplied and net water use through evapotranspirationand the tonnes of produce/GL. Key water-use effi-ciency indicators for the CIA show that irrigation effi-ciency in terms of root-zone storage to the waterdiverted from the source is 70%. Unless there is aninvestment in irrigation infrastructure to improvemeasuring, monitoring and loss reduction, this effi-ciency indicator will remain low. The overall water-use efficiency of the CIA is 77% due to capillary wateruse by the crops. Production efficiency of the CIA is343 tonnes/GL. Further analysis of the whole of theCIA water savings shows (Khan et al. 2004a) that it ispossible to increase economic water-use efficiency

from $91,000/GL to $97,500/GL, and total water useefficiency from 77% to 84% under the current crop-ping and irrigation regimes.

Farm scale

The current state of water use and water produc-tivity in the MIA is summarised in Table 1, whichprovides an overview of the net crop-water require-ments (NCWR), current irrigation levels and yieldsin the MIA. In all cases NCWR are well below themaximum reported irrigation application levels.There are major differences between minimum andmaximum crop yields, as well as the overall amountof water consumed and the NCWR. These dataclearly illustrate that there is a potential to increasefarm profitability through:• better matching of soils and groundwater

conditions with cropping systems• improving irrigation efficiency by 1–5 ML/ha• increasing crop yields by 20–50% by removing the

management, nutrient and salinity constraints.The possible water savings can be summarised in

the form of steps of increasing on-farm and off-farm

=77%

Groundwater Irrigation

Surface Water Irrigation

Conveyance Efficiency

Farm Edge (494GL)= 80%

Water Source (615GL)

Farm Efficiency

Field Edge (464GL)= 94%

Farm Edge (494GL)

Field Efficiency

Root zone storage (715GL)= 92%

Field Edge (774GL)

Effective Rainfall

Conveyance losses• Evaporation • Seepage• Operational losses• Leakages

Distribution Losses• Evaporation • Seepage• Operational losses• Leakages

On Farm Storage

Application Losses• off-target• Evaporation

IrrigationEfficiency

Root zone storage (435GL)=70%

Water Source (620GL)

Water UseEfficiency

ET actual (736GL)

Water Supply (945GL)

WaterProductivity

EconomicReturn

Yield (324401t)


Profit(86112213$)


Surface andGroundwater

Drainage

AbatementCost ($/ML)

• Shallow pumping• Tile drains • Evaporation Basins• SBC

Groundwater Storage

Fallow ET

Regional Groundwater Flow

15 GL

5GL

Shallow Groundwater

Capillary Uptake

310 GL

59 GL

34 GL

CIA-BASE

= 91,000$/GL

= 343t/GL

• deep percolation• non recycled surface runoff

Figure 3. Base case water-use efficiency of the Colleambally Irrigation Area, Australia


96


water savings and water benefits. It is important torecognise that some steps are prerequisites for thenext water-use efficiency level. For example, torealise on-farm water savings it is crucial to imple-ment soil management and groundwater and flowmonitoring programs, to ensure irrigation levels arebeing matched with the crop water requirement, atthe same time considering conversion to advancedirrigation technologies. Similarly, to realise off-farmwater saving options it is vital to know how muchwater is being delivered in space and time beforepiping/lining of channels. It is important to reduce theconveyance difference and narrow the wide gapbetween the gross diversions from rivers to deliveriesto farms by installing state-of-the-art monitoring anddelivery systems as a part of the modern irrigationinfrastructure.

Considering a range of soil, water and groundwaterconditions, Khan et al. (2004b) concluded that on-farmirrigation technology conversions can provide potentialwater savings ranging from 0.1 to 2.2 ML/ha for dif-ferent broadacre crops. For example, for citrus crops,they quote savings of 1.0–2.0 ML/ha for changing from

flood to sprinkler irrigation and 2.0–3.0 ML/ha fromflood to drip; for vineyards, the savings are 1.0–1.5ML/ha for the change for flood to sprinkler and up to4.0 ML/ha from flood to drip irrigation; for vegetables,they quote savings of 0.5–1.0 ML/ha. Modelling simu-lations show water-saving potential of 7% for maize,15% for soybean, 17% for wheat, 35% for barley, 17%for sunflower and 38% for faba bean, if on-farm surfaceirrigation methods can be replaced with pressurisedirrigation systems.

Figure 4 summarises the on-farm and off-farmwater savings and environmental benefits for theMurrumbidgee Irrigation Area in Australia.

Based on recent work by Khan et al. (2004a), thepotential savings for converting from good surfacewater to pressurised irrigation systems (travellingirrigators or centre pivots or equivalent) are shown inTable 2.

A study of on-farm conveyance losses on ninefarms showed that seepage losses vary from 1 to 4%of the total water supplied, which can amount to morethan 60 ML/year.

Figure 4. Stairs to possible water savings in an irrigation area


97


Tab

le 1

. N

et c

rop

wat

er r

equi

rem

ents

(N

CW

R),

rep

orte

d w

ater

use

and

yie

lds

in t

he M

urru

mbi

dgee

Irr

igat

ion

Are

a, A

ustr

alia

. C

rop

area

s ar

e as

rep

orte

d fo

r20

00–2

001

Cro

pC

rop

area

(h

a)N

CW

RR

epor

ted

irri

gatio

na (

ML

/ha)

Rep

orte

d yi

eld

(t/h

a)

(ML

)(M

L/h

a)M

edia

nL

owH

igh

Med

ian

Low

Hig

h

Ric

e46

,120

506,

562

11.0

14.0

12.0

16.0

9.5

6.0

12.0

Whe

at39

,215

111,

835

2.9

2.0

1.0

3.0

5.0

3.0

7.0

Oat

s2,

896

7,51

22.

62.

01.

03.

03.

52.

06.

0

Bar

ley

3,03

48,

615

2.8

2.0

1.0

3.0

5.0

2.5

7.0

Mai

ze2,

924

18,8

136.

48.

56.

012

.09.

56.

015

.0

Can

ola

2,68

54,

643

1.7

2.5

1.0

4.0

2.5

1.8

3.0

Soyb

ean

2,88

118

,383

6.4

8.0

6.0

9.0

2.6

1.5

3.8

Sum

mer

pas

ture

3,92

945

,154

11.5

7.5

7.5

8.0

Win

ter

past

ure

24,1

8450

,403

2.1

5.5

5.5

6.0

Luc

erne

(un

cut)

2,46

843

,291

17.5

10.0

7.0

14.0

7.3

5.0

15.0

Vin

es13

,635

77,5

085.

75.

03.

07.

515

.09.

025

.0

Citr

us8,

700

68,8

617.

97.

04.

510

.038

.020

.060

.0

Ston

e fr

uits

934

9,07

19.

79.

07.

512

.018

.015

.020

.0

Win

ter

vege

tabl

esb

1,50

092

10.

65.

04.

06.

060

.050

.070

.0

Sum

mer

veg

etab

lesc

1,50

08,

906

5.9

7.0

6.0

10.0

90.0

60.0

120.

0

TO

TA

L15

6,60

598

0,47

7d

Sour

ces:

Bee

cher

(19

95),

MD

BC

(19

97),

MIA

&D

LW

MPW

G (

1997

), H

ope

and

Wri

ght (

2003

).

a R

epor

ted

irri

gatio

ns le

vels

are

sub

ject

to a

djus

tmen

t for

mea

sure

men

t err

or; e

.g. 1

4% is

the

acce

pted

und

eres

timat

ion

fact

or f

or D

ethr

idge

whe

els

.b

Irri

gatio

n re

quir

emen

t and

yie

ld is

for

oni

ons.

For

sal

ad c

rops

(le

ttuce

) ir

riga

tion

requ

irem

ent i

s 2.

0–4.

0 an

d yi

eld

is 3

0.0–

40.0

.c

Irri

gatio

n re

quir

emen

t and

yie

ld is

for

tom

atoe

s. F

or m

elon

s, th

e ir

riga

tion

requ

irem

ent i

s 4.

0–7.

0 an

d th

e yi

eld

is 3

0.0–

40.0

.d

Rep

orte

d gr

oss

dive

rsio

ns f

or 2

000–

01 a

re 1

,048

,000

ML

and

on-

farm

del

iver

ies

857,

000

ML

.


98


Seepage losses measured for over 700 km ofchannel length in the MIA showed that seepagelosses were over 40,000 ML/year and evaporationlosses over 12,500 ML/year. The total losses in givenchannel reaches vary widely and can be 1–30% of thewater supplied and 0.2–9% per km.

Application of systems approach in China

Another example of a systems approach is basin-wide holistic integrated water assessment (BHIWA)for catchment water balance analysis in China. TheBHIWA model was developed by the InternationalCommission on Irrigation and Drainage (ICID) in2003 to specifically address future water scenariosfor food and rural development — water for people aswell as for the environment — in order to use waterresources in a way that achieves sustainable develop-ment. The model was designed to be simple, flexibleand effective. On the premise that precipitation con-stitutes the primary resource, management of eva-potranspiration to increase the flows in rivers andaquifers is considered as a potential developmentstrategy that could be changed through policy inter-vention.

The model is able to account for the whole landphase of the hydrologic sector, including the consid-eration of hydrologic changes due to changes in landuse and agricultural activities. The model is capableof depicting surface and groundwater balances sepa-rately and allowing depiction of interaction betweenthem, as well as impacts of storage and depletionthrough withdrawals. Figure 5 is a schematic depic-tion of the model.

The BHIWA model gives a definition of the wateruse and yield circumstances for surface water andgroundwater, respectively. The model considers thereturns as an additional resource added to the naturalrun-off. In other words, the model considers the ‘netconsumptive use’ rather than withdrawals. Four indi-cators are selected to depict the water situation in termsof quantity and quality. Indicators 1 and 3 depict thelevel of withdrawals as a fraction of total availablewater in the surface and groundwater systems, whileindicators 2 and 4 depict the potential hazard to sur-face- and groundwater quality, respectively.• Indicator 1 Withdrawals/total run-off for surface

water• Indicator 2 Returns/total run-off for surface water• Indicator 3 Withdrawals/total recharge for

groundwater• Indicator 4 Returns/total recharge for

groundwaterWhen 0.4 < Indicator 1 < 0.8, it can be concluded

that the surface water quantity is highly stressed,while 0.4 < Indicator 3 < 0.8 means groundwaterquantity is highly stressed; 0.1 < Indicator 2 < 0.2suggests that surface water quality is under highthreat, while 0.4 < Indicator 4 < 0.8 indicates thatgroundwater quality is under high threat.

Table 4 gives a comparison of water situation indi-cators of Chinese river basins in 2000. Based on theBHIWA model, the total inputs to groundwater are thesum of groundwater resources and return flow fromwell irrigation. The total inputs to surface water (rivers)are the sum of surface water resources and returns. Thereturns to surface and groundwater were estimatedfrom the surpluses of agriculture, industry anddomestic water use (water use minus consumption).

Table 2. Water use and savings (ML/ha) for selected crops under different irrigation technologies

Crop Irrigation method

Surface Sprinkler Water savings

High Low Average High Low Average High Low Average

MaizeSoybeanWheatBarleySunflowerFaba beans

10.66.64.24.37.04.9

4.33.60.50.73.51.5

8.35.42.41.74.63.2

9.25.62.82.44.83.3

4.03.20.50.73.11.4

7.74.62.01.13.82.0

1.41.01.41.92.21.6

0.30.40.00.00.40.1

0.60.80.40.60.81.2


99


Figure 5. Schematic of the basin-wide holistic integrated water assessment model

Table 4. Water Indicators for Chinese rivers, derived from the basin-wide holistic integrated water assessmentmodel

Class description Value of indicator Basin

Very highly stressed through surface withdrawalHighly stressed, through surface withdrawalModerately stressed, through surface withdrawalLow stress, in regard to surface withdrawalSurface water quality under high threat Surface water quality under moderate threatSurface water quality under low threat

Groundwater very highly stressed through withdrawalsGroundwater highly stressed through withdrawalsGroundwater moderately stressedGroundwater low stressed

Groundwater quality under very high threatGroundwater quality under high threatGroundwater quality under moderate threatGroundwater quality under low threat

Indicator 1 >0.80.4 < Indicator 1 < 0.80.2 < Indicator 1 < 0.4Indicator 1 < 0.20.1 < Indicator 2 < 0.20.05 < Indicator 2 < 0.1Indicator 2 < 0.05

Indicator 3 > 0.8

0.4 < Indicator 3 < 0.80.2 < Indicator 3 < 0.4Indicator 3 < 0.2,

Indicator 4 > 0.80.4 < Indicator 4 < 0.80.2 < Indicator 4 < 0.4Indicator 4 < 0.2

NoneHaihe, Huaihe, inland riversSongliao, YellowYangtze, Pearl, southeast, southwestHaihe, Huaihe Songliao, YellowYangtze, Pearl, southeast, southwest, inlandHaihe

Songliao, HuaiheYellowYangtze, Pearl, southeast, southwest, inlandNoneHaiheSongliao, Yellow, HuaiheYangtze, Pearl, southeast, southwest, inland


100


The information in Table 4 shows that systemmodels such as BHIWA can provide a good, system-level indication of surface and groundwater use andthe availability of water resources. However, thisapproach fails to show spatial points of interventionat farm and sub-regional levels. In order to deviseactions for improving water-use efficiency at farm,irrigation district and catchment levels it is necessaryto carry out a multi-scale water-use efficiency anal-ysis, as described for the Murrumbidgee Catchmentin Australia earlier in this paper.

Conclusions

A multi-scale, top-down, systems approach can helpassess relative magnitudes of potential water savings.Multi-scale water-balance studies have highlightedthe need for accurate measuring and monitoringsystems for realising true water savings at the farm,irrigation area and catchment levels. Whereas single-scale, whole-of-system modelling approaches suchas BHIWA can provide a good, system-level over-view of surface and groundwater use, they fail toshow points of intervention at farm and sub-regionallevel due to the lumped nature of the analysis.

Acknowledgments

Funding support from the Australian Centre for Inter-national Agriculture Research (ACIAR), the PrattWater Group and CSIRO’s Water for a HealthyCountry Flagship is appreciated.

References

Beecher, G., McLeod, G.D., Pritchard, K.E. and Russell, K.1995. Final report, benchmarks and best managementpractices for irrigated cropping industries in the southernMurray-Darling Basin. Natural Resources ManagementStrategy project 1 5045.

Hope, M. and Wright, M. 2003. Murrumbidgee catchmentirrigation profile. Written and compiled for the Water UseEfficiency Advisory Unit, Dubbo, NSW, Australia, ajoint initiative of NSW Agriculture and the Department ofSustainable Natural Resources, 13p.

Khan, S., Akbar, S., Rana, Y., Abbas, A., Robinson, D.,Dassanayke, D., Hirsi, I., Blackwell, J. and Xevi, E.,Carmichael A. 2004a. Hydrologic economic ranking ofwater saving options. A consultancy report to the PrattWater Group.

Khan, S., Rana, T., Beddek, R., Blackwell, J., Paydar, Z. andCarroll, J. 2004b. Whole of catchment water and saltbalance to identify potential water saving options in theMurrumbidgee Catchment. A consultancy report to thePratt Water Group.

MDBC (Murray–Darling Basin Commission) 1997. Inlandagriculture, best management practices andbenchmarking study. Inland Agriculture Pty Ltd, inassociation with Hutchins Agronomic Services,Darlington Point.

MIA&D LWMPWG (Murrumbidgee Irrigation Area andDistricts Land and Water Management Plan WorkingGroup) 1997. MIA & Districts Land and WaterManagement Plan, Griffith.


impacts of improved irrigation and drainage systems of the yinchuan … · yinchuan plain it was...

Documents