Our analysis demonstrates that four core strategies can—in principle— meet future food production needs and environmental challenges if deployed simultaneously. Adding them together, they increase global food availability by 100–180%, meeting projected demands while lowering greenhouse gas emissions, biodiversity losses, water use and water pollution. However, all four strategies are needed to meet our global food production and environmental goals; no single strategy is sufficient. We have described general approaches to solving global agricultural challenges, but much work remains to translate them into action. Specific land use, agricultural and food system tactics must be developed and deployed. Fortunately, many such tactics already exist, including precision agriculture, drip irrigation, organic soil remedies, buffer strips and wetland restoration, new crop varieties that reduce needs for water and fertilizer, perennial grains and tree-cropping systems, and paying farmers for environmental services. However, deploying these tactics effectively around the world requires numerous economic and governance challenges to be overcome. For example, reforming global trade policies, including eliminating price-distorting subsidies and tariffs, will be vital to achieving our strategies. In developing improved land use and agricultural practices, we recommend following these guidelines: Solutions should focus on critical biophysical and economic ‘leverage points’ in agricultural systems,plastic gardening pots where major improvements in food production or environmental performance may be achieved with the least effort and cost. New practices must also increase the resilience of the food system.

High-efficiency, industrialized agriculture has many benefits, but it is vulnerable to disasters, including climatic disturbances, new diseases and economic calamities. Agricultural activities have many costs and benefits, but methods of evaluating the trade-offs are still poorly developed. We need better data and decision support tools to improve management decisions, productivity and environmental stewardship. There are multiple paths to improving the production, food security and environmental performance of agriculture, and we should not be locked into a single approach a priori, whether it be conventional agriculture, genetic modification or organic farming. The challenges facing agriculture today are unlike anything we have experienced before, and they require revolutionary approaches to solving food production and sustainability problems. In short, new agricultural systems must deliver more human value, to those who need it most, with the least environmental harm.Evapotranspiration, also known as crop consumptive water use in agricultural lands, is often the largest yet most uncertain component of the agricultural water balance in semiarid regions . Irrigation accounts for 70% of human freshwater use and supports 45% of the world’s food supply production . Most irrigation water is lost to the atmosphere via evapotranspiration . Information about evapotranspiration can be applied to optimized sustainable regional water allocations and farm-level irrigation management, improving water, and food security in a changing climate to meet the demands of a growing population . As one of the world’s most productive agricultural regions, California is the major producer of various specialty crops including half of the fruits, vegetables, and nuts grown in the United States .

A quarter of the nation’s produce comes from the Central Valley alone, due to its fertile soils and extended growing season, generating $20 billion of agricultural sales in 2017 . More than 7 million acres of the Central Valley are irrigated via a system of reservoirs and canals, accounting for about 75% of the state’s irrigated lands and consuming about 23 trillion liters of water per year . The acreage of high value yet water-intensive tree crops, such as almonds, pistachios, and walnuts, has been expanding in recent decades, increasing water demand in the state’s largest groundwater basin . Groundwater is a major source of water supply in the Central Valley, ranging from 30% during wet years to 70% during extremely dry years . During the recent 2012–2016 drought, groundwater pumping was intensified, which led to the doubling of the reduction in the valley’s groundwater storage to 11 trillion liters per year from the 2007–2009 value . This groundwater overdraft due to excessive pumping accounts for 13% of water sources in the San Joaquin Valley, which is the southern half of the Central Valley . It caused significant land subsidence , and subsequent damage on aqueducts, costing the state “tens of millions of dollars” in repairs to the aqueduct in the last 40 years . The intensifying drought and reduction in water supply during the summer growing season are expected to continue, due to a projected decrease in snow pack and increasing temperature in the coming decades . Consequences of chronic groundwater overdraft and non-point sources pollution motivated the enactment of the Sustainable Groundwater Management Act in 2014. It requires local groundwater agencies in critically over-drafted basins to achieve a sustainable water balance by 2040 . Another management challenge under water scarcity is to balance the competing beneficial water uses by many stakeholders, e.g., maintaining river flows to support estuarine habitat .

Some potential water management alternatives to attain Sustainable Groundwater Management Act requirements include demand management. For example, via reduced irrigated area, in addition to increasing supply via groundwater recharge. Spatial evapotranspiration estimate records could inform stakeholders on historical and present water demand. Time series of accurate evapotranspiration maps can also help water managers better understand the dynamics of groundwater pumping, improve regulation oversight, and evaluate the impacts of the implemented water policy . As 77% of cropland owners in California have fields smaller than 0.4 km2 , remotely monitoring individual land owner’s water use is only possible by using routine satellite observations at a relatively high spatial resolution. A few algorithms of varying complexity have been developed to map crop evapotranspiration using remote sensing data, due to the large spatial coverage and consistent imagery acquisition by satellite instruments . For example, the crop-coefficient-based evapotranspiration approach has been used by California’s growers for irrigation management . This method is relatively easy to implement but does not account for other factors, such as water stress and thus likely overestimates actual evapotranspiration . Energy balance-based approaches, on the other hand, estimate evapotranspiration as a residual between available energy and sensible heat flux , for example, as adopted by the Surface Energy Balance Algorithm for Land  and Mapping EvapoTranspiration at high Resolution with Internalized Calibration . In METRIC,blueberry pot size the sensible heat is often estimated using clusters of hot and cold pixels observed within each Landsat scene . The pixel selection can be fully automated for large-scale applications or manually selected by professional users for more accurate regional estimates. More sophisticated methods involve solving the soil and canopy latent heat components within a process model, constrained by satellite data . The Atmosphere-Land EXchange Inverse model, for example implements the Two-Source Energy Balance in a time differential mode, based on two snapshots of high temporal frequency geostationary satellite thermal observations, to reduce the sensitivity to the errors in absolute land surface temperatures . DisALEXI has been further developed to estimate evapotranspiration at 30-m resolution, by bringing in additional higher spatial resolution thermal data from Landsat . A recent comparative study of crop evapotranspiration estimates over the Sacramento-San Joaquin Delta showed that DisALEXI estimates had an RMSE of 1.43 mm day−1 and mean bias of 0.13 mm day−1 while the calibrated METRIC prepared by the Irrigation Training & Research Center had an RMSE and mean bias of 2.55 and 2.06 mm day−1, respectively . Another biophysical model, Breathing Earth System Simulator , couple surface energy balance, photosynthesis, and stomatal conductance processes, to estimate evapotranspiration, forced by biophysical parameters from Moderate Resolution Imaging Spectroradiometer 8-days 1-km observations. One compromise between the simple crop-coefficient-based approaches and more complex energy balance models described above is the Priestley-Taylor approach. It estimates evapotranspiration over the wet and extensive surface . It is primarily driven by available energy and a Priestley-Taylor coefficient that partitions available energy to latent and sensible heat. This coefficient was found to be constant for estimating potential evapotranspiration.

For drying environmental conditions, Priestley and Taylor demonstrated that a factor could be added to multiply the Priestley-Taylor evapotranspiration estimates to actual evapotranspiration. They found that the factor remains constant when the cumulative evapotranspiration over a drying surface was below a certain threshold, but the threshold varies from one site to another. After the threshold was exceeded, the scaling factor follows a linear decline to zero after a further 5 cm of evapotranspiration. Similar to Priestley and Taylor ‘s concept of scaling Priestley-Taylor evapotranspiration estimates on wet surfaces to actual evapotranspiration over drying surfaces, remote sensing scientists introduced biophysical controls, such as water stress, to down-regulate the Priestley-Taylor coefficient, thus reducing the fraction of available energy used for latent heat. Fisher et al. , for example, parameterized the Priestley-Taylor coefficient separately for soil evaporation, canopy transpiration, and interception using remotely sensed vegetation index and meteorological data. It has been applied to estimate monthly evapotranspiration globally at 5 km and 1° resolutions using Advanced Very High Resolution Radiometer data , regionally at 1 km using MODIS data and most recently at 70 m using ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station data . Jin et al. optimized Priestley-Taylor coefficient, a a E or PT , as a function of Leaf Area Index and soil moisture for each plant function type, using the eddy covariance tower measurements from AmeriFlux sites, and estimated monthly evapotranspiration at 1 km for the entire continental US, from primarily MODIS data. In this study, our primary objective is to improve our understanding of the agricultural water use patterns in California’s Central Valley, to facilitate water resources planning, in the context of the Sustainable Groundwater Management Act. We first modified, calibrated, and evaluated the refined semiempirical Priestley-Taylor method over major California crops, using the Landsat Analysis Ready Data . This approach further built upon an earlier version adapted for evapotranspiration estimation in Sacramento-San Joaquin Delta . The method was chosen for this study based on its relatively good performance and computational efficiency for large-scale applications . The automated workflow for regional applications was applied to assess the spatial patterns of evapotranspiration across the Central Valley and among dominant crop types and to analyze water use changes between 2014 and 2016 water years. To provide insights for water management agencies in their efforts to prepare for droughts and implement Sustainable Groundwater Management Act, we further provided a comprehensive assessment on agricultural evapotranspiration by boundaries of Groundwater Sustainability Agencies .We focused on the major agricultural production area, about 25 thousand km2 , in the Central Valley of California . The top eight crop types include almond, grapes, corn, rice, alfalfa, walnuts, pistachios, and tomatoes are based on the updated crop layers from the California Department of Water Resources. Together, these account for 68% and 73% of agricultural land use in 2014 and 2016, respectively. This semiarid region has a Mediterranean climate, with mild winter and hot, dry summers. Its mean annual precipitation ranges from 51 cm yr−1 in the north to 13 cm yr−1 in the far south, and the majority of rainfall occurs from November to March. The Central Valley is therefore highly dependent on irrigation and vulnerable to water scarcity. Its groundwater storage depleted by approximately 16 trillion liters between spring 2005 and 2010 . Groundwater depletion during the recent 2012–2016 drought is expected to be even worse, since Xiao et al. estimated that the Central Valley’s groundwater storage depletion doubled to 11 trillion liters per year from the 2007 to 2009 value.GSAs were established under the Sustainable Groundwater Management Act to manage California’s groundwater resources at the local scale. Over 260 GSAs were initially formed in the State’s high and medium priority and some are currently under reorganization and consolidation. Figure 1c shows 134 GSAs in the Central Valley that have >1 km2 of farmland in both 2014 and 2016. The GSAs boundaries are defined by the most recently available DWR’s Exclusive GSA map ; GSAs boundaries were sometimes split into parts by other political boundaries, so we merged the polygon parts that share an identical GSA name, for example, Byron-Bethany Irrigation District GSA-East Contra Costa and Byron-Bethany Irrigation District GSA—Tracy were merged into a single GSA.A total of 26 evapotranspiration measurement sites over agricultural areas were available in the study region, including five agricultural sites from the AmeriFlux network and two from cropland sites established for the shorter-term study of California’s specialty crops .