d3.2.2:guidelines&forpredictive...
TRANSCRIPT
-
D3.2.2: Guidelines for predictive differential irrigation scheduling and
water application WP3.2 – Upscaling VRT for nutrient and water efficiency and yield
optimization
Alfonso Calera, Julio Villodre, Jesús Garrido, José González (UCLM)
First version – M12
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 633945.
Ref. Ares(2016)1002521 - 28/02/2016
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
2 | 16
Document Information Grant Agreement Number 633945 Acronym FATIMA Full Title of Project Farming Tools for external nutrient inputs and water Management Horizon 2020 Call SFS-‐02a-‐2014: External nutrient inputs (Research and innovation Action) Start Date 1 March 2015 Duration 36 months Project website www.fatima-‐h2020.eu Document URL (insert URL if document is publicly available online) REA Project Officer Aneta RYNIAK Project Coordinator Anna Osann Deliverable D2.1.1 FATIMA webGIS conceptual design document Work Package WP3.2 – Upscaling VRT for nutrient and water efficiency and yield Date of Delivery Contractual 1 March 2016 Actual 29 Feb 2016 Nature R -‐ Report Dissemination Level CO Lead Beneficiary 01_UCLM Lead Author Alfonso Calera (UCLM) Email [email protected] Contributions from Internal Reviewer 1 Francesco Vuolo (BOKU) Internal Reviewer 2 Carlo de Michele (Ariespace) Objective of document Rationale and operation of predictive differential irrigation scheduling
and water application Readership/Distribution All FATIMA Regional Teams; All WP leaders and other FATIMA team
members; European Commission / REA
Keywords webGIS, user requirements, conceptual design, co-‐creation Document History
Version Issue Date Stage Changes Contributor Draft v00 22/02/2016 Draft
Disclaimer
Any dissemination of results reflects only the authors’ view and the European Commission is not responsible for any use that may be made of the information it contains.
Copyright
© FATIMA Consortium, 2015 This deliverable contains original unpublished work except where clearly indicated otherwise. Acknowledgement of previously published material and of the work of others has been made through appropriate citation, quotation or
both. Reproduction is authorised provided the source is acknowledged. Creative Commons licensing level
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
3 | 16
Executive summary
Currently operative use of dense time series of multispectral imagery at high spatial resolution is able to monitor crop development across its growing season at a suitable time and spatial scales. Crop growing cycle is described through reflectance and Vegetation Indices providing spatial distribution of crop growth. These time series of images, jointly with meteorological data are able to provide accurate maps of daily transpiration and so crop water requirements by using the remote sensing-‐based approach crop coefficient, Kc, and reference evapotranspiration, ETo, where Kc is derived from spectral reflectances and ETo from meteorological data. A water balance in the root soil layer enables us to calculate irrigation water requirements at appropriate scale for monitoring water management near-‐ real time. This approach could be coupled to the remote sensing-‐based surface energy balance which uses surface temperature as primary input. But according users requirement, what we could call “remote sensing-‐driven crop water management” requires at least two steps more to be placed into the day-‐to-‐day routine on farming irrigation: On the one hand, for planning irrigation the users require the forecasting of crop water requirements for the week ahead; it can be achieved by extrapolating crop coefficient trend and by using weather forecasting for ETo estimation. On the other hand, decision makers in charge of irrigation require access to this information in an easy-‐to-‐use way on real time. It can be achieved through leading edge webGIS tools, which facilitates co-‐creation and collaboration with stakeholders. In this deliverable we describe a modular system based on the integration of Earth observation (EO) remote sensing (from satellites and ultralights), soil wireless sensor networks (WSN), and weather observations and forecasting into a webGIS, in order to monitor plant growth status and to determine their water requirements a week ahead. By using SPIDERwebGIS tool, this information is brought to the users in a timely way for practice. The current version (v01) reflects the status of M12, with the basic methodology available and functioning. The integration of other similar approaches as well as further calibration and validation will lead to updated methodologies, which will be documented in a subsequent document version.
Key words: crop water management, remote sensing, weather forecasting, webGIS
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
4 | 16
Table of Contents Executive summary ........................................................................................................................................... 3
1 Guidelines for predictive differential irrigation scheduling and water application ................................... 5
2 Remote Sensing for Irrigation Water Management .................................................................................. 6
Monitoring crop development at right spatial and temporal scale .................................................. 6
Remote sensing-‐based estimates of Irrigation Water Requirements ............................................... 7
2.2.1 The remote sensing-‐based crop coefficient .............................................................................. 8
2.2.2 Remote sensing surface energy balance ................................................................................. 11
2.2.3 Summary of remote sensing-‐based evapotranspiration estimates ......................................... 11
3 Predictive differential crop water requirements ..................................................................................... 13
Planning irrigation for the next week .............................................................................................. 13
3.1.1 Extrapolating reflectance-‐based basal crop coefficient .......................................................... 14
3.1.2 Forecasting reference evapotranspiration ETo ....................................................................... 14
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
5 | 16
1 Guidelines for predictive differential irrigation scheduling and water application
This Deliverable D3.2.2 addresses the optimization and fine tuning of irrigation water supply to the demands of growing crops across space and time to obtain the maximum yield, by using time series of spectral imagery. It describes how these time series of images enable determination of spatial variability of irrigation requirements at intra-‐plot scale. Moreover, forecasting of crop water requirements a week ahead is a key for sustainable water management practices according users requirement.
The current version (v01) reflects the status of M12, with the basic methodology available and functioning. The integration of other similar approaches as well as further calibration and validation will lead to updated methodologies, which will be documented in a subsequent document version.
We propose a modular system based on the integration of Earth observation (EO) remote sensing (from satellites and drones), crop and soil wireless sensor networks (WSN), and weather observations and forecasting into a webGIS, in order to monitor plant growth status and to determine their differential water requirements a week ahead. Therefore, irrigation scheduling and water application could be tailored to the crop demands if variable doses rate of water can be applied, or optimized according the available irrigation system. The WSN provides essential in-‐situ information from soil moisture sensors, complementary to the EO-‐based data. By this way farmers can address short-‐term management strategies for irrigation management, what is a central product in FATIMA portfolio, as is showed in Table 1 about basic FATIMA products to be delivered to stakeholders.
Table 1.-‐ Summary of FATIMA products based on the users requirements
Time frame
Decisions to make Products Delivery of products
Short term
Crop Monitoring
How much water next week?
How much nutrients and its spatial distribution?
Time series of images + ground sensors + weather
Maps of crop water requirements next week
Maps of crop nutrients requirements
webGIS
Midterm/
Long term
Cropping system/crop rotation/no tillage/organic and conservation agriculture
Planning the crop: Seed/ Density/ Main labors/ Climate, Water and Nutrients/Manure management
Assessment for optimum yield through economic analysis including policies and environmental constraints
Reports/
Manuals
webGIS
Multispectral time series of images, jointly with meteorological data are able to provide accurate maps of daily transpiration and so crop water requirements by using the remote sensing-‐based approach crop coefficient, Kc, and reference evapotranspiration, ETo. Where Kc is derived from spectral reflectances and ETo from meteorological data. A water balance in the root soil layer enables us to calculate irrigation water
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
6 | 16
requirements at appropriate scale for monitoring water management near-‐ real time. This approach could be coupled to the remote sensing-‐based surface energy balance which uses surface temperature as primary input. But what we could call “remote sensing-‐driven crop water management” requires at least two steps more to be placed into the day-‐to-‐day routine on farming irrigation: On the one hand, for planning irrigation the users require the forecasting of crop water requirements for the week ahead; it can be achieved by extrapolating crop coefficient trend and by using weather forecasting for ETo estimation. On the other hand, decision makers in charge of irrigation require access to this information in an easy-‐to-‐use way on real time. It can be achieved through leading edge webGIS tools, which facilitates co-‐creation and collaboration with stakeholders.
To answer to the question how much water to apply and where, maps of crop water requirements, CWR, for the next week will be used. They will be generated by combining weather forecast and extrapolation of basal crop coefficient from the trend of the plot as is described by VI_EO time series. Predictive CWR under controlled deficit irrigation (case of wine grape is an example) will be performed by using soil water balance. Ground truth from soil moisture sensors will assess the quality of prediction. CWR map will provide the diagnosis tool about the spatial variability at the pixel size scale, although specific VRT devices for differential spatial water application are less developed than for nutrients. Options for the latter will be explored.
By this way we contribute to increasing farmers’ competitiveness through the reduction of production cost and providing a reliable and up-‐to date tool for land and water management and policy-‐making.
The EO methodology for mapping crop water requirements in a pixel by pixel basis is mature and operational (Calera et al 2013, D’Urso et al 2013) by using FAO56 and soil water balance model, in combination with biophysical crop parameters and ground-‐based meteorological data. Consolidation of this approach will be the first step to do, including the implementation of two-‐source model for separating soil evaporation and canopy transpiration. Exploitation of the improved spectral resolution of new generation of sensors will be exploited to enhance existing methodologies.
The forward step is to produce the map of crop water requirements predictions for the next week, which is a very practical product for users in addition to estimates for the past week. To this aim, distributed short term forecasting of relevant meteorological data (from high-‐resolution or local weather prediction models) will be used as input in the above mentioned models. From the knowledge of crop water requirement, irrigation can be supplied either to satisfy full requirements, either to manage under deficit controlled irrigation. Ground data of soil moisture will provide the required ground truth to verify through the soil water balance the quality of the products.
2 Remote Sensing for Irrigation Water Management
Monitoring crop development at right spatial and temporal scale
Currently monitoring crop development across its growing season at a suitable time and spatial scales is possible throughout use of dense time series of multispectral imagery at high spatial resolution. Given the crops canopy evolves rapidly in mostly cases, single satellite sensors or platforms cannot address these changes at high spatial resolution (5-‐30 m) due to the underlying limitations of data availability, including cloudiness, and tradeoffs(eg in the case of Landsat8 the revisit time is 16 days). Virtual constellations of planned and existing satellite sensors may help to overcome this limitation by combining existing observations to mitigate limitations of any one particular sensor. Virtual constellation is defined by the Committee on Earth Observation Satellites (CEOS) as “a set of space and ground segment capabilities operating together in a coordinated manner, in effect a virtual system that overlaps in coverage in order to
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
7 | 16
meet a combined and common set of Earth observation requirements. The individual satellites and ground segments can belong to a single or multiple owners”.
Deliverables D2.2.1 “Methodology for dense high-‐resolution EO time series, gap filled” and D2.2.2 “Methodology for EO-‐based crop water requirements forecast” describe the basis of virtual constellation used in FATIMA. It principally combines sensors with similar spatial, spectral, temporal, and radiometric characteristics to describe canopy evolution with a frequency approaching one image per week. It takes advantage of free and open access to satellite imagery and value-‐added data products, currently those images acquired by Landsat8 and Sentinel2a and released via web by USGS and ESA respectively, what have revolutionized the role of remote sensing in Earth system science and practice.
Remote sensing-‐based estimates of Irrigation Water Requirements
Irrigation water requirements are usually estimated by applying a water mass balance in the soil explored by roots which accounts for evapotranspiration, precipitation, run-‐off, deep drainage and soil water storage. Evapotranspiration of crop stand, ET, is the exchange of water vapor between the atmosphere and land surface shaped by crop canopy and soil beneath. ET is the crop water consumption and its variability across space is the key to determine variability of irrigation water requirements.
Physics of evapotranspiration from land surfaces is well described by the Penmann-‐Monteith equation which relies on the surface energy balance and the resistances approach for describing transport of water vapor, distinguishing between bulk surface and aerodynamics (Monteith and Unsworth,1990). Bulk surface resistance includes stomatal resistance which drives the transpiration process, and so the physiological control on ET by plants is introduced.
Remote sensing approaches for ET estimation have been following two main models streams based on either surface energy balance or either reflectance-‐based crop coefficient (Allen et al.,AWM, 2011); some attempts to apply directly remote sensing-‐based parameters into Penmann-‐Monteith equation have been done as well. The Figure 1 shows a general overview of remote sensing-‐based different approaches. Spatial and temporal spatial resolution of so elaborated maps of ET and irrigation water requirements is depending on the pixel size of utilized input imagery. For those based on reflectance-‐based crop coefficient, which relies on VIS-‐NIR imagery, the pixel size ranges usually 5-‐30 m, and a lot of commercial [ World View, Rapid Eye, DMC, Deimos,…] and free [L8 and Sentinel2a] sensors are currently in orbit, which warranty their use for the next years. For those based on surface temperatures, the pixel size ranges from 100 m for thermal sensor on board of L8 to 1000 m for MODIS and Sentinel3.
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
8 | 16
Figure 1.-‐ Overview of remote sensing-‐based approaches for estimates of ET and Irrigation Water Requirements. The spatial scale of so elaborated maps is related with the pixel size of utilized imagery.
2.2.1 The remote sensing-‐based crop coefficient
In the seventies of past century a practical approach for the ET estimation was developed, assuming that ET could be estimated like a product of two factors, so giving the known Penman-‐Monteith “two-‐step” procedure (Allen et al., 1998), usually termed crop-‐coefficient approach. The first factor is the evaporative power of the atmosphere, or reference evapotranspiration, ETo, which is obtained from meteorology. The second factor is the crop coefficient, Kc,adj , which parameterizes the characteristics of the actual canopy relative to that of the an ideal reference surface (ideal grass or alfalfa). The coefficient Kc,adj includes (a) the water stress coefficient, depending on the soil water content in the root soil layer, (b) the basal crop coefficient, Kcb, and (c) the evaporative component of the bare soil fraction, Ke.
The relationship between basal crop coefficient and spectral reflectance through a spectral vegetation index (SVI) is the key to introduce remote sensing in the application of the Penman-‐Monteith equation by the “two-‐step” procedure described above. This relationship is empirically supported in many crops (Bausch 1993); (Hunsaker et al. 2003); (Neale et al. 1989); (Gonzalez-‐Dugo y Mateos, 2008) (Gonzalez-‐Dugo et al., 2009) (Campos et al., 2010) (DÚrso et al., 2010) and physically-‐based (Gonzalez, 2008) (Choudhury et al., 1994). The well-‐known capability of SVI to describe the fraction of absorbed photosynthetic active radiation, fAPAR, at canopy level is related with the physics that underlies the reflectance-‐based basal crop coefficient (Asrar, 1989) (Seller, 1989) (Seller et al., 1997): this ability of SVI enables to describe the photosynthetic size of the canopy (Wiegand and Richardson, 1990) (Calera et al.,2004).
So, these theoretical and empirical bases provide support to consider that reflectance-‐based basal crop coefficient Kcb represents the “potential” or maximum ratio between transpiration and reference ET for the canopy, what happens in the case of an unstressed canopy, such as is defined basal crop coefficient concept. A relationship is
It means that the product Kcb*ETo provides a measure of potential transpiration T of the canopy. Potential transpiration means transpiration of the plant stand without any type of stress, due to water shortage, atmospheric conditions or other causes, that produces stomatal control.
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
9 | 16
When the potential energy of the soil water drops below a crop-‐dependent threshold value, the transpiration rate drops below the maximum value and stoma closure happens, even before external changes on leaf chlorophyll (typical yellowing) appear. The lowering coefficient of maximum ET is called the stress coefficient that ranges between 0 and 1. In the water balance approach, the stress coefficient is calculated usually from the ratio between actual soil water content and a determined threshold value (Allen et al. 1998). When water stress appears, current transpiration is calculated from the product of three factors: reference ET, reflectance–based crop coefficient and stress coefficient, according Eq.1.
VI-‐Kcb relationships doesn’t capture evaporation from bare soil. Then to estimate crop coefficient, Kc, is required to estimate the evaporative coefficient Ke that is depending of cover fraction, irrigation system, and frequency of irrigation, as is described in Allen et al., (1998).
Figure 2.-‐ (a) Basal crop coefficient as was measured by Wright (1982) for wheat and corn against reference evapotranspitration alfalfa-‐based (b) typical NDVI curves for the same crops from selected plots of top yield in La Mancha area. Both figures display the same behavior across time.
2.2.1.1 Integrating reflectance-‐based Kcb into FAO56 procedure
Basal crop coefficient as it is derived from multispectral imagery is a basic input in the widely used FAO56 model (Allen et al., 1998) for crop evapotranspiration calculation, involving root soil water balance. Equation 1 shows the calculation process.
ET= Ks Kc ETo = (Ks Kcb+ Ke) ETo (1) where ET: evapotranpiration of the plant stand Kcb: Basal crop coefficient, derived from multispectral imagery, is defined as the ratio of the unstressed crop transpiration over the reference evapotranspiration. Analogous to a transpiration coefficient (dimensionless) Ks: Water stress coefficient, which is calculated on the basis of soil water balance in the root layer (dimensionless) Ke: Bare soil evaporation component, which is calculated on the basis of soil water balance in the upper soil layer (dimensionless) ETo is the evaporative power of atmosphere, or reference crop evapotranspiration, determined from meteorological data; it is defined [FAO56] like the evapotranspiration of "A hypothetical reference crop with an assumed crop height of 0.12 m, a fixed surface resistance of 70 s m-‐1 and an albedo of 0.23”
According eq. 1, the product Kcb ETo represents the maximum of potential transpiration of an unstressed canopy, the product Ks Kcb ETo, represents the actual transpiration of a canopy, and the product Ke ETo is the evaporation from bare soil fraction.
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
10 | 16
Calculation of reflectance-‐based basal crop coefficient can be done using the eq. 2 (Campos et al., 2010). This equation was derived according TM and ETM+ spectral bands. Values of NDVI need to be atmospherically corrected. Updating required to use atmospherically corrected L8-‐OLI and Sentinel2a data is on going.
Kcb*= 1.44·∙NDVI – 0.1 where Kcb* reflectance-‐based basal crop coefficient [0.15 – 1.15], NDVI, calculated from surface reflectances on TM and ETM+ bands. [Typical range values: bare soil 0.12-‐0.16; maximum NDVI value for very dense green vegetation, 0.91] Components of mass soil water balance and main elements of soil water content available by plants are showed in Figure 3. Running a daily soil water balance where evapotranspiration is calculated using as input the reflectance-‐based basal crop coefficient, and assuming other components of water balance are known, we are able to calculate soil water content. Ground information about soil hydraulics characteristics, precipitation, irrigation system and frequency of water application increases the modelling accuracy. The water stress coefficient is calculated from the ratio between actual soil water content and a determined threshold value crop dependent. A similar procedure is made in the top soil layer, typically the first 10-‐15 cm of soil, to calculate the bare soil evaporation component Ke. Detailed Excel spreadsheets incorporating this procedure could be accessed through the URL http://extension.uidaho.edu/kimberly/2013/04/spreadsheets-‐supporting-‐fao-‐56-‐example-‐calculations/. To perform the satellite-‐driven FAO56 soil water balance it is needed to introduce as input the value of reflectance-‐based Kcb obtained from multispectral imagery. So, the satellite-‐driven FAO56 soil water balance enables to calculate irrigation water requirements in a pixel by pixel basis. In many cases the amount of irrigation is calculated avoiding stress. It is also possible to calculate irrigation water requirements under water stress, as is commonly desired either in controlled deficit irrigation either in supplemental irrigation. Knowledge of desired water stress degree is required, which needs local calibration. The software Hidromore+, available from UCLM, is able to perform the described the satellite-‐driven FAO56 soil water balance in a spatially distributed way. In this case, the required inputs for calculation are maps. So, Hidromore+ incorporates directly the maps of reflectance-‐based basal crop coefficient from multispectral imagery.
Figure 3.-‐ Components of mass soil water balance and main elements of soil water content available by plants to perform the FAO56 soil water balance. Input of evapotranspiration at pixel scale by using the reflectance-‐based basal crop coefficient enables calculation of Irrigation water requirements in a spatially distributed way.
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
11 | 16
2.2.2 Remote sensing surface energy balance
This approach computes land surface evapotranspiration directly. It partitions available energy in the land surface using the radiometric surface temperature (TR), derived from thermal band imagery, to constrain the sensible heat flux, computing latent heat as a residual to the surface energy balance (e.g., Moran et al., 1994; Kustas and Norman, 1996; Gillies et al, 1997; Bastiaanssen et al., 1998).
Successful applications of the surface temperature approach must address the fact that TR differs from the aerodynamic temperature, To, needed to compute sensible heat, particularly for partial vegetation covered surfaces (Kustas, 1990). TR and To are clearly related (Norman and Becker, 1995), but its relationship is highly complex, since TR depends on the temperature of the different elements that occupied the radiometer view, while To depends on surface aerodynamic roughness, wind speed and the coupling of soil and canopy elements to the atmosphere. Several schemes of varying levels of complexity and input requirements have been formulated to deal with this difference. Some employ empirical/semi-‐empirical methods for adjusting TR to To, tuned to account for spatial variability in the roughness lengths for heat and momentum transport (eg. Kustas et al., 1989; Lhomme et al., 1994; Chehbouni et al., 1996; Mahrt and Vickers, 2004). When calibrated with field data, empirical relationships have provided accurate results (Chavez et al., 2005); however, such relationships are typically crop or vegetation specific and are not likely to function correctly when applied to different crop types or landscapes.
A class of internally calibrated surface temperature schemes avoids the problem of specifying To by instead modeling the vertical near-‐surface air gradient TA-‐To. These methods are based on selecting pixels in the satellite image representing the extreme heat and moisture exchanging surfaces (i.e., a dry non-‐transpiring surface where ET=0 and a wet surface where ET is at potential) and calculating the spatially distributed sensible heat flux assuming a linear relationship between TR and the near-‐surface air temperature gradient across the image (Bastiaanssen et al., 1998) (Allen et al., 2007). This approach reduces the need for atmospheric correction of TR, which is a cumbersome and error-‐prone process, but it arises new uncertainties associated with “choosing the end point” and assuming the linear relation between TR and TA-‐To.
Other TR-‐based approaches model the effects of partial vegetation cover on To using two-‐source model parameterizations (Shuttleworth and Wallace, 1985; Norman et al., 1995), which partition surface fluxes between the soil and canopy components of the scene. This more physically based approach does not require in-‐situ calibration, although most implementations do require accurate radiometric temperature retrievals. A comparison between a two-‐source model and an internally calibrated model over herbaceous crops (Gonzalez-‐Dugo et al., 2009) showed a reasonable agreement with tower measurements; however Timmermans et al. (2007) found significant discrepancies in the heat flux maps generated by the two approaches, particularly for bare soil and sparse canopy covered areas.
Anderson et al (1997) proposed an improvement of a two-‐source scheme by incorporating a simple description of planetary boundary layer dynamics. The resulting Atmosphere-‐Land Exchange Inverse (ALEXI) and an associated flux disaggregation technique (DisALEXI) is a multi-‐sensor TIR approach to ET mapping that reduces the need of ancillary data input and is able to deal with errors in TR remote estimation by partially working in time-‐differencing mode (Anderson et al. 2010).
2.2.3 Summary of remote sensing-‐based evapotranspiration estimates
A summary of ET estimates by using remote sensing approaches in an operative way could be:
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
12 | 16
Ø Time series of NDVI maps, or other VI indices, can be converted, either through a linear relationship either through more complex models, into maps of basal crop coefficient. Gap filling to obtain daily Kcb maps can be applied to dense time series of multispectral imagery to avoid cloudiness and lacking images. This time series would be provided by virtual multisensor constellation.
Ø The product of daily basal crop coefficient maps and reference evapotranspiration maps provides
directly the daily potential transpiration in a pixel by pixel basis. Reference evapotranspiration ETo maps would be obtained by predictive meteorological models or from agro-‐meteorological station.
Ø Basal crop coefficient as it is derived from multispectral imagery is a basic input in the widely used FAO56 model for crop evapotranspiration calculation, involving soil water balance. Ground information about soil hydraulics characteristics, precipitation, irrigation system and frequency of water application increases the modelling accuracy.
Ø The EO-‐driven soil water balance, that a scheme is shown in Figure 5, enables to calculate irrigation water requirements in a pixel by pixel basis. According FAO56 procedures it is possible to calculate irrigation water requirements under water stress, as is used either in controlled deficit irrigation or in supplemental irrigation. Knowledge of desired water stress degree is required, which needs local calibration.
Ø Remote sensing of evapotranspiration can also be calculated from temperature images by using other techniques like those based on surface energy balance. One of the main constraint of its use is the spatial scale of temperature surface data. The best spatial resolution has the value of 100 m pixel size, provided by Landsat8 sensor with a revisit time of 16 days. Other surface temperature maps have lower spatial resolution, like those provided daily by sensors onboard of MODIS, Sentinel3,… They are too coarse for typical agricultural plots management. Therefore this procedure is complementary with that previously described, providing an independent quality control in the suitable areas.
Figure 4.-‐ Scheme of FATIMA modular system based on the integration of Earth observation (EO), soil wireless sensor networks (WSN), and weather observations and forecasting into a webGIS, to provide users about differential irrigation scheduling, matching water supply to crop water demands.
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
13 | 16
3 Predictive differential crop water requirements Previous sections have described the way we can use remote sensing multispectral imagery to estimate spatially distributed crop water requirements and irrigation requirements. But according users requirement, what we could call “remote sensing-‐driven crop water management” requires two steps more to be placed into the day-‐to-‐day routine on farming irrigation: On the one hand, for planning irrigation the users require the forecasting of crop water requirements for the week ahead; On the other hand, decision makers in charge of irrigation require access to this information in an easy-‐to-‐use way on real time. It can be achieved through leading edge webGIS tools, which facilitates co-‐creation and collaboration with stakeholders.
Planning irrigation for the next week
Users need to know the crop water requirement CWR for the next week to planning water application in order to fulfill the water requirements of crop, according their irrigation system, electric rates, water availability, precipitation if happens, …, and their own personal availability to do the task. CWR for the next week can be achieved by extrapolating crop coefficient trend and by using weather forecasting for ETo estimation. This approach has been tested last years in the pilot area of La Mancha and is currently operational. Differential CWR maps have been elaborated week to week. Figure 5 shows a CWR maps. Usually the irrigation system doesn’t allow to apply differential irrigation, then an aggregated value for the entire plot has been produced; Figure 6 shows one of the case study followed last campaign. Gained experience indicates that prediction need to be corrected with observation, both Kcb and ETo. Then a simplified water balance is running in parallel to correct the bias that prediction could introduce.
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
14 | 16
Figure 5.-‐ Aspect of Crop Water Requirement map for the next week as can be seen in SPIDER webGIS platform. Color is associated with weekly CWR according the displayed legend. Vector lines delimit plot under monitoring
Figure 6.-‐ Prediction of crop water requirements for the next week. Case study for barley, 2015 campaign. Time trajectory of NDVI describes temporal evolution of crop, showing a typical smooth curve.
3.1.1 Extrapolating reflectance-‐based basal crop coefficient
Extrapolation of reflectance-‐based basal crop coefficient Kcb take advantage of that Kcb time trajectories for crops, derived from NDVI, are usually smooth curves, see Figure 2. Therefore, the Kcb time trajectories are suitable to be extrapolated with the scope of one week by using previous dates to perform the prediction. Currently we are performing a lineal extrapolation based on values of at least two previous images without clouds. Accuracy of this extrapolation depends strongly of the closeness of previous images to the weekly forecasting window. Turning points are critical in the linear extrapolation; hence the limit of value 1.2 is imposed in the maximum values that Kcb can reach.
3.1.2 Forecasting reference evapotranspiration ETo
Reference evapotranspiración ETo is calculated from meteorological data, therefore is suitable to be calculated from meteorological models utilized for forecasting. Two complementary methods with different spatial scope and accuracy have been tested. The first one is to use the full power of meteorological models for forecasting the variables required to compute ETo according FAO56 for the next week. The second one is based in daily temperature forecasting by using it as input into Hargreaves&Samani equation to calculate ETo (Allen et al., 1998). The last method provide low accuracy and it should be restricted to areas where Hargreaves&Samani equation woks well (no windy, no coastal areas) and no forecasting of other variables than temperature are available. Computing ETo according FAO56 from weekly forecasting models is the selected option. The Spanish Meteorological Agency, AEMET (Agencia Estatal de Meteorología) provides routinely the map forecasted ETo product calculated from the HIRLAM model. The spatial scope of this product is the Iberian Peninsula, as it is
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
15 | 16
showed in Figure 7, and the spatial resolution of the raster map is 5 km pixel size. Moreover, AEMET also provides the observed ETo. Then, the observed ETo is also displayed in SPIDER. Figure 8 shows the comparison ETo_forecast and ETo_observ from the last months. Comparison of both products is displayed in Figure 8; comparison against local meteo station also has been made in the Spanish pilot area with good results.
Figure 7.-‐ Weekly reference evapotranspiration ETo forecasting performed by AEMET and displayed by the system SPIDER webGIS.
Figure 8._ Comparison of weekly ETo_observ and ETo_forecast, both time series provided by AEMET for the last months and displayed by the system SPIDERwebGIS.
-
D3.2.2 Guidelines for predictive irrigation scheduling 26/02/2016 v01
16 | 16
References
Michael A.Wulder, Thomas Hilker, Joanne C.White, Nicholas C. Coops, Jeffrey G. Masek, Dirk Pflugmacher, Yves Crevier. Virtual constellations for global terrestrial monitoring Remote Sensing of Environment 170 (2015) 62–76; http://dx.doi.org/10.1016/j.rse.2015.09.001
CEOS (2013). The CEOS virtual constellation concept. Virtual constellations process paper (updated 2013). http://old.ceos.org/index.php?option=com_content&view=category&id=347&Itemid=480;Accessed October, 30th, 2015,