recent trends in crops water productivity across the contiguous states
TRANSCRIPT
January 26, 2017
Michael MarshallWorld Agroforestry CentreUnited Nations Ave, Gigiri
P.O. Box 30677-00100Nairobi, Kenya
Recent trends in crop water productivity across the
contiguous United States: a call for “more crop per drop”
USGS (2005)IPCC-AR5
Climate Change and Water Use in the U.S.
Williams et al. (2015)
California Drought 2012-2016
North American Drought 2012-2013
Funk and Brown (2009)
Jung et al. (2010)
WARMER
June et al. (2004)
Globally
Blue-Green Revolution▪ Crop type and variety▪ Surface/groundwater coordination▪ Precision agriculture
─ Deficit irrigation─ Drip irrigation─ Irrigation scheduling─ Soil salinity
▪ Integrated assessments▪ Water markets or tax
℘1=Total Dry Matter (kg)
Cumlative Transpiration (m3)
℘2=Grain∨Seed Yield(kg)
Cumlative Transpiration (m3)
Water Productivity (WP)Ali and Talukder (2008) define crop WP in terms of total dry matter (net primary production-NPP) and yield:
Rain-fed water productive crops assimilate more carbon, while losing less water to the atmosphere. Irrigated crops are typically on a deficit schedule, so crop yield is more appropriate.
ObjectivesDetermine green, blue, and overall trends in WP for major crops in the U.S. (alfalfa, corn, cotton, rice, sorghum, and soy):
Parameterize models with high resolution Earth observation and climate geospatial data over the primary growing season
Quantitative assessment of WP1 (2001-2015)
Qualitative assessment of WP2 (2008-2015)
Crop Yield Model DevelopmentLight-Use or Production Efficiency Models (PEMs) are a compromise between simple empirical models and fully mechanistic models (e.g. APSIM)
Process-based
Transferable
Climatic constraints (explicit)
Non-climatic (implicit)
Crop Yield Model DevelopmentPEMs perform best for forests and more poorly for croplands/ecosystems.
Based on Challinor et al. (2009) we optimized the approach for croplands:
Sensitivity Analysis
Rigorous model calibration (light- and water-use literature)
Multi-scale evaluation with MOD17 (Running et al., 2004)
Sensitivity AnalysisOne-parameter at a time approach on model inputs: PAR, NDVI, VPDX, and TX
Parameter Description Equation Citation
GPP MAX Maximum daily gross primary production ᵋMAX * FPAR * FM * FT * FA * PAR
F PAR Fraction of photosynthetically active radiation 1.257 * NDVI - 0.161 Bastiaanssen et al., 2003
F M Short-term moisture stress min (1, 1 / √ VPDX) Zhou et al., 2014
F T Temperature stress 1.1814 / ((1 + e0.2 * (Topt - 10 - Tx)) (1 + e0.3 * (Tx - 10 - Topt))) Potter et al., 1993
F A Long-term (seasonal) moisture stress FPAR / FPAR, MAX Fisher et al., 2008
ᵋ MAX Maximum quantum conversion efficiency C3 crops: 0.08 * (CA - Γ) / (CA + 2Γ) Collatz et al., 1991
C4 crops: 0.06 Collatz et al., 1992
Station Parameter B1 (gCO2 m-2 d-1) B0 (gCO2 m-2 d-1)US NE-1 TX (< 0) 16.65 35.73
TX (> 0) -17.96 36.94(-0.65)
VPDX -9.55 36.84PAR 23.56 34.38NDVI 17.43 25.66
CalibrationThe baseline model was most sensitive to PAR (FA), NDVI (FPAR), VPDX (FM), and TX (FT)
Bastiaanssen et al. (2003), Running et al. (2004), and Potter et al. (2003) led to overall best performance for C3 and C4 crops
Optimized Water-Light use (OWL) model
Marshall, M., Tu, K.P., Brown, J., 2017. Light- and water-use efficiency model optimization for large-areacrop yield estimation. Remote Sens. Environ. (under review)
Eddy Covariance Flux Tower Data (OWL)
Eddy Covariance Flux Tower Data (MOD17)
GPP → NPP → Yield (Regional Assessment)GPP minus respiration costs were used to estimate net photosynthesis (Pn):
Pn was summed over a fixed growing season (mid-May to late-October) to estimate NPP.
NPP was converted to yield (Y) using the harvest index (HI), root-to-shoot ratio (RS), and seed moisture content (MC)
𝐘= ∑𝐢=𝐒𝐎𝐒
𝐧𝐏𝐧𝐢×
𝐇𝐈1+𝐑𝐒 ×
11−𝐌𝐂
Climate Inputs DOE interpolated climate fields
from NCDC and NRCS stations CONUS Mosaics Daily 1km Temperature and precipitation --
IWD weighted by DEM--(Thorton et al., 1997)
SWIN and VPD from DTR and dew point (Thorton et al., 2000)
2001-2015 Monthly Average MODIS Albedo
RN from albedo, SWIN, DEM, T, VPD, latitude (Allen et al., 1998)
eMODIS (January 2001)
1.00
0.00
Vegetation Input USGS-EROS MODIS based
expedited (eMODIS) NDVI CONUS Mosaics Near real-time 7-day 250m MODIS Land Science
Collection 5 Atmospherically Corrected Surface
Optimized Savitsky-Golay filter (Chen et al., 2004)
SAVI approximation (Los et al., 2000)
Cropscape 30m (2015)
OWL MOD17 OWL – MOD17
NPP (Fixed Season = mid-May to late-October)
Mea
nSt
dev
D = 0.57RMSE = 1.45RMSEU = 0.97RMSES = 1.08
D = 0.63RMSE = 1.26RMSEU = 0.66RMSES = 1.07
D = 0.50RMSE = 0.55RMSEU = 0.42RMSES = 0.36
D = 0.33RMSE = 0.70RMSEU = 0.53RMSES = 0.45
Cot
ton
Soyb
eans
OWL MOD17
Major Findings of Optimization Procedure
C3 and C4 partitioning was essential- particularly during green-up and brown-down
FA (soil moisture indicator) was an important improvement- particularly for the C3 pathway
The C4 pathway remains underestimated, but model bias is primarily systematic in nature
Model counters MOD17 bias in non-agroecosystems and should be further explored
Fisher, J.B., Tu, K.P., Baldocchi, D.D., 2008. Global estimates of the land–atmosphere water flux based on monthly AVHRR and ISLSCP-II data, validated at 16 FLUXNET sites. Remote Sens. Environ. 112, 901–919.
NASA PT-JPL model
WP1 Trends (2001-2015)
kg•m-3
32.0
-12.2
Yield
ETC
Alfalfa
Irrigated
Rainfed
Combined
Irrigated
Rainfed
CombinedYield
ETC
Rice
Corn
Irrigated
Rainfed
CombinedYield
ETC
Improvements in Progress
eMODIS Remote Sensing Phenology
MODIS Irrigated Agriculture Dataset for the United States (MIrAD-US)
MODIS Global Food Security-support Analysis Data (GFSAD) crop type 2001-2015
Summary WP (high to low): Alfalfa, Corn, Soybeans, Sorghum,
Cotton, and Rice WP increases
–Mid-West: rain-fed corn and soybeans–Texas: irrigated/rain-fed cotton and sorghum
WP declines Ogallala Aquifer: irrigated/rain-fed corn and wheat Central Valley, California and Mississippi: irrigated
rice 2012-13 North American Drought Next step: 30+ year (1982-2012) global assessment
Thank You