Integrating water supply constraints into irrigated agricultural simulations of California

Environmental Modelling & Software 96 (2017) 335e346

Contents lists available at ScienceDirect

Environmental Modelling & Software journal homepage: www.elsevier.com/locate/envsoft

Integrating water supply constraints into irrigated agricultural simulations of California Jonathan M. Winter a, *, Charles A. Young b, Vishal K. Mehta b, Alex C. Ruane c, Marzieh Azarderakhsh d, Aaron Davitt e, Kyle McDonald f, Van R. Haden g, Cynthia Rosenzweig c a

Department of Geography, Dartmouth College, Hanover, NH 03755, USA Stockholm Environment Institute, Davis, CA 95616, USA NASA Goddard Institute for Space Studies, New York, NY 10025, USA d Fairleigh Dickinson University, Teaneck, NJ 07666, USA e The Graduate Center, City University of New York, New York, NY, 10016, USA f The City College of New York, New York, NY, 10031, USA g The Ohio State University, Wooster, OH, 44691, USA b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 5 January 2017 Received in revised form 19 June 2017 Accepted 29 June 2017

Simulations of irrigated croplands generally lack key interactions between water demand from plants and water supply from irrigation systems. We coupled the Water Evaluation and Planning system (WEAP) and Decision Support System for Agrotechnology Transfer (DSSAT) to link regional water supplies and management with field-level water demand and crop growth. WEAP-DSSAT was deployed and evaluated over Yolo County in California for corn, rice, and wheat. WEAP-DSSAT is able to reproduce the results of DSSAT under well-watered conditions and reasonably simulate observed mean yields, but has difficulty capturing yield interannual variability. Constraining irrigation supply to surface water alone reduces yields for all three crops during the 1987e1992 drought. Corn yields are reduced proportionally with water allocation, rice yield reductions are more binary based on sufficient water for flooding, and wheat yields are least sensitive to irrigation constraints as winter wheat is grown during the wet season. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Agriculture Irrigation Water resources management Crop model Evapotranspiration United States

Software availability

Program size: 500 MB

Name of software: WEAP-DSSAT Developers: Charles Young, Jonathan Winter, Vishal Mehta, Alex Ruane Contact address: Dartmouth College, 6017 Fairchild Hall, Hanover, NH 03755 Telephone: þ1 603-646-6456 Email: [email protected] Year first available: 2016 Hardware required: PC, Intel with 4 GB RAM recommended Software required: Microsoft Windows Availability and cost: Licensed software Program language: Fortran, Python, and Delphi

1. Introduction

* Corresponding author. 6017 Fairchild Hall, Hanover, NH 03755, USA. E-mail address: [email protected] (J.M. Winter). http://dx.doi.org/10.1016/j.envsoft.2017.06.048 1364-8152/© 2017 Elsevier Ltd. All rights reserved.

Irrigated farms account for 80%e90% of consumptive water use in the United States and $118.5 billion of US agricultural production (Solley et al., 1998; Schaible and Aillery, 2012). Despite the high productivity of irrigated croplands, agriculture is typically the lowest value sector in a water resources system, and, subject to water regulations and rights, vulnerable to reductions during drought. A major challenge for the hydrologic and agricultural communities is assessing the effects of climate change on the sustainability of regional water resources and irrigated agricultural land (Walthall et al., 2013). A key component of this challenge is the fact that most agricultural models that have sophisticated representations of crop physiology, management, and yield, and are thoroughly evaluated at the field scale, lack constraints on irrigation supply (Winter et al., 2017). Many crop models are run with

336

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

scheduled irrigation or unlimited automatic irrigation (e.g., Bondeau et al., 2007; Jones et al., 2003; Raes et al., 2009), each of which has important disadvantages. With scheduled irrigation the dates and amounts of irrigation are prescribed in advance. This is problematic in locations and years where this information is not readily available, or in predictive applications (e.g., seasonal forecasts, climate change projections). Automatic irrigation doesn't require prior knowledge of water applications, instead relying on a rule-based approach (often linked to soil moisture) to irrigate from an unlimited water supply. This leads to inconsistent, and occasionally implausible, biases and errors that muddle the picture for stakeholders and policymakers seeking to understand agricultural sustainability issues. Multiple studies have addressed the impacts of climate on irrigated agriculture at scales ranging from regional to global; however, few have explicitly coupled a crop model to a calibrated hydrologic and water resources allocation model. Elliott et al. (2014a) compared water supply projections from ten global hydrologic models and water demand projections from six global gridded crop models. Results suggested that the effects of reduced irrigation were comparable to the direct impacts of climate change on global production of maize, soybean, wheat, and rice, with some acute regional impacts. However, Deryng et al. (2016) noted discrepancies in projections of agricultural water supply and demand due to varying responses of crop water use to increased CO2 by global gridded crop models, and the lack of response of crop water use to increased CO2 by some global hydrologic models. Piontek et al. (2014) explored climate change impacts on multiple sectors, including water availability and agriculture. Several areas were found to have overlapping risk for severe change both in water availability and climate, including the southern Amazon Basin and regions in South Asia. Wada et al. (2013) used a set of seven global hydrologic models to explore the change in irrigation water demand by the end of the century, finding a considerable increase during summer months in the northern hemisphere. Huntington and Niswonger (2012) focused on the seasonal timing of streamflow and surface and groundwater interactions over the Western United States. Specifically, they used an integrated model of surface water and groundwater inclusive of snowpack and snowmelt forced with twelve general circulation model (GCM) projections. Future climate was shown to decrease summertime flows by more than 30% averaged across the ensemble, with reductions found even in GCM simulations that projected increased annual precipitation. Groundwater is a critical source of water for irrigated agriculture, and groundwater management remains a salient issue for € ll et al., 2012). Taylor et al. sustainable irrigated agriculture (Do (2013) outline the complexity of groundwater response to climate change and human impacts, including annual precipitation and streamflow; timing, intensity, and duration of precipitation and streamflow; land use; snowpack; groundwater pumping; and surface water irrigation. Groundwater pumping has been shown to be unsustainable in the Central Valley of California; however, to date the use of groundwater has been largely unrestricted in California (Famiglietti, 2014). The water-agriculture nexus has been identified as a high priority area within the Agricultural Model Intercomparison and Improvement Project (AgMIP; Rosenzweig et al., 2015, 2013). The importance of irrigated agriculture to global food production, as well as the response of crop water supply and demand to climate, necessitate explicitly simulating effects of water availability on irrigated yields. In the following sections, we describe the development, application, and evaluation of a coupled hydrologic, water resources allocation, and crop model. The objective of creating this coupled model is to more realistically simulate irrigated agricultural yields, and specifically to develop a modeling system in which

water shortages (e.g., decreased precipitation, enhanced evapotranspiration, changes in allocation) directly impact irrigated crop yields, with applications for identifying and testing policy and management approaches. 2. Model description and development To simulate water demand and supply for irrigated agriculture, a model must link information about regional water supplies and management with field-level water demand and crop response as it develops throughout the season. To accomplish this, we coupled the Decision Support System for Agrotechnology Transfer (DSSAT; Hoogenboom et al., 2012; Jones et al., 2003) to the Water Evaluation and Planning system (WEAP; Yates et al., 2005a, 2005b). Below we describe the development of the coupled model, WEAP-DSSAT, as well as our iterative simulation approach, which together add water supply constraints to automatic irrigation in DSSAT and detailed crop water use and yields to WEAP. 2.1. Water Evaluation and Planning system The Water Evaluation and Planning system (WEAP) is used to model water supply and management within WEAP-DSSAT. WEAP is an object-orientated model that solves a time evolving mass balance based on a water allocation objective function. At each time step, hydrologic fluxes from the surface and near surface are passed to appropriate river and groundwater objects, where they are balanced with an objective function that maximizes satisfaction of demand and instream flow requirements, subject to supply priorities based on water rights and regulations, demand site preferences, mass balances, and other constraints (Yates et al., 2005a, 2005b). WEAP has a flexible time step that can range from daily to annual, which also determines the time scale over which water allocation is calculated. WEAP divides study regions into user-defined sub-catchments; groundwater basins; irrigated areas; urban/export uses; environmental requirements; and water system elements such as canals, diversions, and reservoirs. Water supply in WEAP is provided by an embedded hydrologic model forced by an external climate dataset that simulates runoff, groundwater-surface water interactions, and snow processes. Before the addition of DSSAT, agricultural water demand in WEAP could be simulated using a variety of approaches that range in complexity from a simple crop coefficient method to more complex hydrology-based algorithms that incorporate runoff, infiltration, soil moisture storage, deep percolation, and evapotranspiration as a function of soil moisture status. The WEAP framework readily accommodates user specified models, or modules, that can be plugged into and controlled by WEAP's water budget and allocation logic. WEAP has been used for range of applications, including large river basins with substantial irrigation such as California's Central Valley (Mulligan et al., 2011; Sandoval-Solis et al., 2010), village scale modeling of community livelihoods (Varela-Ortega et al., 2011), and the exploration of climate change impacts on hydropower generation (Mehta et al., 2011). WEAP deployed over the Central Valley has been shown to adequately represent both local and regional water balances, and the allocation of groundwater and surface water supplies (Purkey et al., 2008; Yates et al., 2009, 2008). A feature critical to assessing the impacts of climate change on agriculture, and which is notably lacking in the California implementation of WEAP, is a representation of plant physiology in plant water use and yields. WEAP simulates crop water use by assigning a seasonal cycle of agricultural vegetation to every user-defined subcatchment. For each time step, potential evapotranspiration is scaled by a crop coefficient. This approach, while reasonable for

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

337

hydrologic applications, does not include climate impacts on agricultural productivity or capture the effects of a changing environment on plant physiology, such as increased carbon assimilation and water use efficiency under enhanced carbon dioxide concentrations or increased transpiration under warmer temperatures (Kimball, 2011; Tubiello et al., 2007). These effects are critical for projections of irrigation demand, crop yields, and water management. Coupling DSSAT, a process-based crop model, to WEAP expands WEAP's capabilities beyond hydrological applications, and makes it ideally suited for exploring water-agriculture interactions. 2.2. Decision Support System for Agrotechnology Transfer Farm-level crop demands and the growth processes leading to yield are handled within WEAP-DSSAT by the Crop Estimation through Resource and Environment Synthesis (CERES) component of the Decision Support System for Agrotechnology Transfer (DSSAT). DSSAT is a set of applications designed to understand, predict, and manage cropping systems at the field scale (Hoogenboom et al., 2012; Jones et al., 2003). DSSAT is essentially a point-based (single hectare) model that runs on a daily time step, with yields dependent on surface meteorology, a detailed soil profile, cultivar genetics, and crop management. DSSAT simulates biological processes that track carbon, nitrogen, water, and energy budgets (as well as associated stresses) through a plant's developmental stages as it interacts with the environment and is managed by a grower. DSSAT simulates a closed water balance that has compared well with soil moisture measurements in the field (Casanova et al., 2006; Sau et al., 2004). DSSAT supports models for 42 unique crop species, including all the major staple crops, and has been used in agricultural systems around the world for a variety of applications, including climate change impacts (Mearns et al., 1996; Parry et al., 2005; Rosenzweig et al., 2014; Ruane et al., 2013; White et al., 2011). In our simulations we run the CERES-Maize, CERESWheat, and CERES-Rice applications of DSSAT, which have been widely used in crop modeling and studies of California's Central Valley (e.g., Brumbelow and Georgakakos, 2001; Greenwald et al., 2006; Lobell and Ortiz-Monasterio, 2007). 2.3. Coupling DSSAT to WEAP The link between WEAP and DSSAT was implemented by editing the source code of WEAP, written in Delphi, to write DSSAT input files, call the DSSAT executable, and read DSSAT output files. Within WEAP, catchment objects are used to represent hydrological processes, including crop water use. Within each catchment climate is assumed uniform. Sub-catchments represent unique combinations of hydrologic parameters (e.g., land cover, soils). It is at the subcatchment scale that hydrologic calculations are performed and DSSAT was integrated into the WEAP framework. For the subcatchments in which DSSAT is used, DSSAT simulates the hydrologic cycle including irrigation surface runoff, infiltration, evapotranspiration, and deep percolation, with surface runoff and deep percolation being passed back to WEAP. WEAP is used to calculate the flow of water in the stream network, storage of water in reservoirs and groundwater, and allocation of water to demands for all sub-catchments, as well as hydrologic processes in the subcatchments DSSAT is not used in. WEAP water allocation is solved at a user-defined time step that can range from daily to yearly, in contrast to the fixed daily time step of DSSAT. If the WEAP time step is longer than a day, DSSAT output is aggregated for input to WEAP. The process for conducting a WEAP-DSSAT simulation is described in Fig. 1. Water allocation in WEAP is solved by a linear program, so for each allocation time step (in this case, monthly) it is necessary to develop linear equations that represent surface runoff

Fig. 1. WEAP-DSSAT irrigation flowchart. Components are color-coded based on the length of time step considered for action e blue is yearly, green is user-defined (in this case monthly), and red is daily.

and deep percolation as a function of percentage irrigation demand fulfilled, or the portion of the required irrigation water that is allocated based on supply, hereafter referred to as allocation. To develop these equations, WEAP runs DSSAT once with unlimited or full irrigation and once with zero irrigation for the entire water year. Linear equations are then generated that relate allocation to surface runoff and deep percolation so that WEAP can calculate surface runoff and deep percolation from irrigation dates and amounts passed by DSSAT. In addition, the full irrigation run transcribes the irrigation dates and amounts from the automatic irrigation algorithm assuming unlimited water to a DSSAT input file for the catchment (blue boxes in Fig. 1). Once the linear relationships are generated, the water allocation problem is solved and the allocation is computed at a user-defined time step (in this case monthly, green boxes in Fig. 1) for every catchment. For each allocation time step and catchment, WEAPDSSAT checks the demand from the DSSAT full irrigation run against the water supply for agriculture (water for other uses is considered unavailable) passed from WEAP. If the allocation is 100% then the full irrigation results are stored for each day (red boxes in Fig. 1), no additional DSSAT runs are conducted, and WEAP-DSSAT moves to the next allocation time step. If the allocation is less than 100%, then DSSAT is rerun from the beginning of the water year with reduced irrigation. This reduction in irrigation is accomplished through the modification of DSSAT input files. For each instance of DSSAT run by WEAP, WEAP creates an input file that contains settings for the DSSAT simulation, including irrigation dates and amounts. As described above, this input file is first created assuming unlimited water. WEAP-DSSAT is then run iteratively, stopping the simulation for any month that the allocation is less than 100%, editing the input file to irrigate on the same days as the water unlimited run but with the irrigation amount scaled by

338

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

allocation, and then rerunning from the beginning of the water year using surface runoff and deep percolation from the reduced allocation simulation. This process is repeated each month until the simulation ends or the harvest date is reached. Because DSSAT cannot run a single simulation with both prescribed and automatic irrigation, the DSSAT algorithm for automatically triggering irrigation was implemented in the coupling of WEAP to DSSAT, allowing WEAPDSSAT simulations to use prescribed irrigation amounts for past monthly time steps and automatic irrigation amounts for the current monthly time step. The result of this iterative approach is a time series of irrigation dates determined using the same automatic irrigation algorithm in DSSAT with irrigation amounts adjusted based on allocation. Once the simulation finishes, some key DSSAT output variables are aggregated by WEAP-DSSAT into a single file for analysis, including soil moisture from the DSSAT soil water file; planting date, harvest date, and season length from the DSSAT plant growth file; evapotranspiration from the DSSAT soil-plantatmosphere file; and agronomic results (yield, biomass, etc.) from the DSSAT summary file. We note that both WEAP and DSSAT are relatively computationally efficient, so our iterative approach over regional spatial scales such as the Central Valley is not limiting. The final component of the WEAP-DSSAT coupling is the specification of soil moisture initial conditions for the next model year. The use of continuously simulated soil moisture and nutrients, as opposed to annually reinitialized simulations, has been shown by Basso et al. (2015) to be extremely important for accurate crop modeling. At the end of each yearly simulation, soil moisture values for each soil layer are read from the soil water file and then used in the next year's input file as the initial soil moisture conditions. 3. WEAP-DSSAT deployment over Yolo County and evaluation datasets We apply WEAP-DSSAT to a county in California's Central Valley as a test case to explore the impacts of climate and water supply on a productive and drought-prone irrigated agricultural system. 3.1. Study region California's $42.5-billion agricultural sector is supported by a complex water management system prone to acute shortages and susceptible to climate change (Garfin et al., 2014; NASS, 2014). The scale of water development in California is among the most substantial in the world, including massive shifts in water from one basin to another over distances of hundreds of kilometers in order to satisfy water demands (National Research Council, 2011). We focus our test case deployment of WEAP-DSSAT on Yolo County, and our agricultural analyses on Yolo County Flood Control and Water Conservation District (YCFCWCD), shown in Fig. 2. Yolo County is a well-studied, intensively managed, and waterlimited agricultural area in the Sacramento River basin in northern Central Valley, with crops covering 57% of the land and agriculture accounting for almost 95% of the County's total water withdrawal of approximately 1.23 billion m3 yr1 (Mehta et al., 2013). YCFCWCD covers 41% of Yolo County's irrigated area and is located in the western-central portion of the county (Mehta et al., 2013). YCFCWCD supplies surface water to its agricultural customers from Cache Creek, primarily from Clear Lake and the Indian Valley Reservoir. Despite this infrastructure, total irrigation demand exceeds what YCFCWCD can supply, and there have been three times in the past 40 years with absolutely no water supplied by the District (1977, 1990, 2014). Additional details on Yolo County and YCFCWCD agriculture can be found in Mehta et al. (2013).

3.2. Model setup WEAP-DSSAT is based on a WEAP model developed for Yolo County (Mehta et al., 2013). In that model the entire area of Yolo County as well as the portions of Lake County that provide water to Yolo County were simulated using catchment objects. The boundaries of the areas represented by each catchment object are shown in Fig. 2. In the upper Cache Creek watershed the model contains reservoir objects for Clear Lake and Indian Valley Reservoir, the main sources of surface water in Yolo County during the summer months. Reservoir operations are controlled by the Solano Decree which has been programmed into WEAP (Mehta et al., 2013). Downstream, on the valley floor, the irrigated area is subdivided into catchments that represent the Capay Valley, Yolo County Flood Control and Water Conservation District, and all other irrigated area in Yolo County. Water preferences are set in the model so that surface water is used first if available, and groundwater is used second. The model was calibrated using crop water use information from the Department of Water Resources, reservoir storage information, and historical diversions. Additional details on WEAP setup and calibration can be found in Mehta et al. (2013). WEAP-DSSAT was initialized October 1, 1980, and run through September 30, 2008, capturing the 1981e2008 water years, with WEAP run at a monthly time step to leverage the calibration and evaluation of WEAP over Yolo County by Mehta et al. (2013) and for computational efficiency, and DSSAT run at the default daily time step. WEAP and DSSAT were forced using surface (2-m) air temperature and precipitation from the Maurer et al. (2002) 1/8 gridded observational dataset, and surface wind speed, surface relative humidity, and insolation from AgMERRA (Ruane et al., 2015). Specifically, WEAP was forced using monthly temperature and precipitation, and DSSAT was forced using daily temperature, precipitation, wind speed, relative humidity, and insolation data, allowing both models to respond consistently to changes in climate. Input climate time series were extracted from gridded datasets using the nearest neighbor of the WEAP catchment centroid. DSSAT has primarily been used for staple crops, so our initial assessment of climate impacts on irrigated agriculture is focused on corn (maize; Zea mays), winter wheat (Triticum aestivum), and rice (Oryza sativa). However, we note that specialty crops (e.g., tomatoes, almonds, grapes) dominate agricultural production by value in Yolo County. Crop management information is described in Table 1. Both the planting and harvesting end date are consistent with ~ athe crop phenology and typical planting dates in the region (Pen n et al., 2011; USDA NASS, 2010), and are expected not to be Barraga met only in exceptional circumstances. Soil properties, including soil moisture initialization, were defined for all crops using a generic deep sandy loam profile. Details on the construction and attributes of the soil profile can be found in the source code (Hoogenboom et al., 2012). We focus on the role of water in limiting crop production, so shut off all nutrient stresses in DSSAT under an assumption of well-fertilized fields. Automatic irrigation for corn and wheat is triggered when the fraction of soil moisture in the top 30 cm of the soil profile falls below 50% of soil saturation. Water is then applied as furrow irrigation until the soil moisture level reaches 100% saturated. Rice is flooded on the first day of the planting window through harvest to a depth of 50 mm. Bund height is set to 100 mm, and percolation rate is assumed to be 4 mm day1. Simulations for each crop were run using the observed atmospheric CO2 concentration time series of Keeling et al. (2001). We note that while planting date, harvesting date, and irrigation applications will vary by year due to weather, all other crop management is static. Crop varieties were selected for each of the three crops based on the field and management characteristics of Yolo County. Field corn

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

339

Fig. 2. Modeled agricultural area (YCFCWCD) and Yolo and Lake County catchments, with inset of California's Central Valley and Yolo County.

Table 1 Crop management for WEAP-DSSAT simulations. Note that the planting for wheat occurs in the year previous to harvest year, and that crops are harvested automatically at maturity. The harvesting end date is included to ensure crops that are planted but do not mature are cleared before the start of the next water year. Crop

Density [m2]

Row Spacing [cm]

Depth [cm]

Planting Window

Harvesting End Date

Corn Rice Wheat

6 200 270

61 25 16

7 2.5 3

3/1e6/1 5/1e5/25 10/20e12/31

9/30 9/30 9/30

in the Sacramento Valley is grown both for grain and silage, with the harvested acreage planted for grain accounting for approximately 70% of total field corn harvested acreage. However, the split in grain and silage production is variable, and often dual-purpose varieties are planted that can be used for either grain or silage based on market conditions (Frate and Schwanki, n.d.). Farmers in this area generally use slow maturing corn varieties to take advantage of the long growing season in the region (Frate and ~ a-Barraga n et al., 2011). We selected a corn Schwanki, n.d.; Pen cultivar in DSSAT that contains the generic attributes of California corn, but does not match a specific variety grown in Yolo County to maintain flexibility for model applications beyond Yolo County but within the Central Valley. The majority of California's wheat production is hard red spring wheat planted in the fall (University of California Division of Agriculture and Natural Resources, 2006).

Important characteristics for California wheat production include short stature, lodging resistance, soil saturation tolerance, shatter resistance, and nitrogen responsiveness (University of California Division of Agriculture and Natural Resources, 2006). A medium grain hard red spring wheat cultivar with the desired properties was not available in DSSAT, so a modified hard red winter wheat available in DSSAT was determined to be the best choice. Rice production in California is dominated by semi-dwarf, mediumgrain varieties, with the majority classified as early maturity (Espino et al., 2013; Geisseler and Horwath, 2013). These varieties provide a high yield and quality of rice, as well as flexibility in planting dates due to the relatively short time to maturity. As with corn, a rice cultivar in DSSAT was selected that contains the key properties of rice currently grown in California, but is not a specific rice variety from Yolo County to ensure that the model is valid for

340

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

applications throughout California. Due to the unsustainable use of groundwater in California (Famiglietti, 2014), we ran two sets of simulations to capture the two extremes of groundwater supply. In the first, we automatically irrigated with no limitations on water amount (WEAP-DSSAT Full). In the second set of simulations (WEAP-DSSAT Surface), we highlight the potential impacts of eliminating groundwater pumping by irrigating exclusively with surface water. In these simulations, crops are automatically irrigated until irrigation water demand exceeds surface water supply, and then irrigation applications are proportionally reduced after that based on the ratio of surface water supply to irrigation water demand. We note that the areas simulated are all within the surface irrigation canal service area of YCFCWCD, so are therefore less sensitive to the elimination of groundwater pumping than Yolo County overall. We further note that the WEAP-DSSAT Surface simulations are hypothetical and proportionally reducing irrigation applications is a simplistic approach. A more practical modeling or scenario-based treatment of groundwater, as well as representations of crop prioritization (e.g., irrigating higher value fields or perennials first), fallowing, and deficit irrigation, would be necessary for more realistic simulations of Yolo County irrigated agriculture. 3.3. Observed yield and evapotranspiration datasets To assess the accuracy of WEAP-DSSAT predictions of grain yield, we compared simulated yields of corn, wheat, and rice to average yields reported for Yolo County by the USDA's National Agricultural Statistical Service (1980e2014) and observed yields from agronomic field trials conducted in Yolo County. Field trial data for irrigated corn (1994e2007) and irrigated wheat (1994e2013) yields were obtained from a long-term experiment conducted at the University of California's Russel Ranch Sustainable Agriculture Facility (Scow, 2009). The corn varieties planted in this experiment were long duration hybrids: Pioneer 3162 (1994e2000) and ST7570RR (2001e2007). Over the period of this field trial three hard red spring wheat varieties were used; Yolo (1994e2002), Summit (2003e2005), and Cal Rojo (2006e2013). Observed rice yields were obtained from a database of annual variety trials conducted by the University of California at three field sites in Yolo County (University of California Cooperative Extension, 2015): Greer Farm (1981e1997, 1999e2003), Erdman Farm (2004e2005), and Webster Farm (2006e2010, 2012e2014). Yield data from the M-202 cultivar was used for all years. We note that while cultivars are known for the field trials, we selected generic cultivars in WEAP-DSSAT to keep our test case consistent with a regional application of WEAP-DSSAT in which cultivars would not be known. Generic, uncalibrated cultivars have been used extensively in regional- to global-scale crop model simulations (e.g., Elliott et al., 2014b; Kucharik, 2003; Rosenzweig et al., 2014); however, we acknowledge that this simplification has consequences for interpretating results. We did conduct an analysis to explore the sensitivity of simulated yields to cultivar parameters. Specifically, for corn and rice we increased P1 and P5 by 10% and decreased P1 and P5 by 10%, and for wheat we increased P5 by 10% and decreased P5 by 10%. Corn yields were most sensitive to changes in P1 and P5, approximately 15% on average, followed by rice, ~4% on average, and with wheat being relatively insensitive to changes in P5, less than 2% (not shown). Some observed yield units needed to be converted for consistency with WEAP-DSSAT, which simulates dry grain in kg ha1. NASS corn was converted from bu ac1 to kg ha1 assuming 56 lb bu1. NASS wheat was converted from bu ac1 to kg ha1 assuming 60 lb bu1. Rice field trial yield data were converted from units of lb ac1 at 14% moisture to kg ha1 dry grain yield. All other

observations either needed no or simple unit conversions. WEAP-DSSAT evapotranspiration in Yolo County was compared to remotely sensed evapotranspiration estimates from the Terrestrial Observation and Prediction System (TOPS) Satellite Irrigation Management Support (SIMS; Melton et al., 2012). While both WEAP and DSSAT simulate ET, we only consider ET output from DSSAT as it should best represent plant water loss over agricultural areas. SIMS integrates satellite, ground observations, and model results to produce near real-time simulations and forecasts of environmental conditions, including potential evapotranspiration (Nemani et al., 2009). Crop evapotranspiration in SIMS is calculated using the California Department of Water Resources Spatial California Irrigation Management Information System (Hart et al., 2009) and a remotely sensed crop coefficient (Kc), described in more detail below. This presents a marked improvement over the use of weather station potential ET with a fixed Kc, which is spatially coarse and does not account for crop rotation, phenology, and land use changes between years. SIMS combines TOPS with Landsat 5 Thematic Mapper (TM), Landsat 7 Enhanced Thematic Mapper (ETMþ) and MODerate resolution Imaging Spectroradiometer (MODIS) satellite imagery to create a map product that covers approximately 4 million ha of agricultural land in California at a 30-m resolution (Melton et al., 2012). Landsat imagery provides the spatial resolution for field scale measurements, while MODIS provides temporal resolution with gap-filling due to the Landsat 7 Scan Line Corrector (SLC) failure, thus providing timely data accessibility and applicability (Melton et al., 2012). SIMS provides information on crop conditions and water use through the TOPS SIMS website (http://ecocast.arc. nasa.gov/dgw/sims/), including basal crop evapotranspiration (ETcb; Fig. 3), basal crop coefficient (Kcb), fractional cover (FC), normalized difference vegetation index (NDVI) from a combination of Landsat and MODIS, and NDVI from Landsat only. We note that ETcb doesn't account for soil evaporation, free water surface evaporation, or reduced transpiration from crop water stress unless the stress affects crop development. We obtained monthly SIMS ETcb for 2011e2014 for Yolo County. Data for 2012 relied more heavily on MODIS as Landsat 5 data were not available, and Landsat 8 was not yet collecting data. Crop boundary information downloaded from Yolo County (http://www. yolocounty.org) was used to identify and extract ETcb for each field of corn, rice, and wheat, and then generate a county-scale mean of ETcb for each crop. We caveat that due to temporal coverage of climate data used to force WEAP-DSSAT, SIMS data does not overlap with WEAP-DSSAT simulations, so seasonal cycles of ET for WEAPDSSAT (1981e2008) and SIMS (2011e2014) were calculated by crop to produce an approximate comparison between remotely sensed and modeled crop ET for corn, rice, and wheat in Yolo County. 4. Results and discussion WEAP-DSSAT produces near-identical (within 1%) yields relative to the original DSSAT model under well-watered conditions for corn, rice, and wheat, with one exception (not shown). DSSAT irrigates on the day of harvest in the wheat simulation for 2001, while WEAP-DSSAT does not because it is unrealistic. This alters the initial soil moisture conditions enough for 2002 to produce a 24% increase in DSSAT wheat yield relative to WEAP-DSSAT. Below we compare yields and evapotranspiration simulated by WEAP-DSSAT to observations for Yolo County. 4.1. Yield evaluation Fig. 4 shows WEAP-DSSAT simulated, NASS, and field trial yields for corn, rice, and wheat, and Table 2 contains the means of and

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

341

Fig. 3. SIMS mean annual ETcb (mm/month) over Yolo County for 2011e2014.

correlations across simulated and observed yields. As described above, we ran WEAP-DSSAT with both a full allocation of water, representing surface water augmented by unlimited groundwater, and then a surface water only allocation, which falls short of irrigation water demand in some years. We primarily analyze WEAPDSSAT simulations with a full allocation (WEAP-DSSAT Full) because we are using irrigated field trial data, and believe full irrigation is more relevant to NASS, and WEAP-DSSAT Full allows us to confirm the consistency of WEAP-DSSAT to the offline version of DSSAT when run with unlimited water. However, we also include the results from WEAP-DSSAT irrigated exclusively with surface water (WEAP-DSSAT Surface) in Fig. 4 and Table 2 for reference. On average WEAP-DSSAT Full corn yields agree with field trials, and are approximately 10% higher than NASS yields. Differences between field trial and NASS yields could be the result of a wide range of management and environmental factors. In addition, field trials can be managed in a way that is not practical at larger scales, and NASS averages yields over the entire county. WEAP-DSSAT Full overestimates rice yields by 13% and 26% relative to field trials and NASS, respectively. Unlike corn and wheat, water stress is probably not a cause of differences across yields as all fields are flooded. WEAP-DSSAT Surface better simulates mean yields of corn when compared to NASS and rice when compared to either observational dataset, but low correlations suggest this accuracy is false. WEAPDSSAT Full simulates wheat well, with yields only slightly lower than both NASS (7%) and field trials (4%). When rounded, wheat yields produced by WEAP-DSSAT Full and WEAP-DSSAT Surface are identical. WEAP-DSSAT Full simulations of corn produce the highest correlation coefficients, both with field trials (0.43) and NASS (0.36). The correlation of WEAP-DSSAT Surface and NASS is poor while the correlation of WEAP-DSSAT Full and WEAP-DSSAT Surface to field trials is identical for corn. The unexpected identical correlations of

WEAP-DSSAT Full and WEAP-DSSAT Surface to field trials is a result of field trial yield data availability coinciding with years in which surface water provided a full allocation. For both rice and wheat, WEAP-DSSAT Full yield correlations to NASS and field trial data are lower, averaging approximately 0.30. For comparison, correlations between NASS and field trial yields are about 0.50 across all crops. WEAP-DSSAT Surface correlations for rice and wheat across both observational datasets are much lower, and often close to zero or even negative, underscoring the importance of groundwater extraction for the region. We note that more accurate simulations of yields could likely be achieved by calibrating the model using field trial data, as well as more specific cultivars, planting dates, and water and nutrient management. 4.2. Evapotranspiration analysis Overall WEAP-DSSAT reasonably simulates average growing season evapotranspiration (ET) for corn and wheat, and captures some features of the ET seasonal cycle when compared to SIMS ET; however, there are differences between WEAP-DSSAT and SIMS within season. We again caveat that the seasonal cycles and mean values for WEAP-DSSAT and SIMS are for 1981e2008 and 2011e2014, respectively, and note that California was in a persistent drought during the growing seasons SIMS is available for, which could have led to discrepancies between WEAP-DSSAT and SIMS planting, crop phenology, and harvesting, and consequently ET. While caution should be exercised in interpreting results, we nonetheless feel that comparing WEAP-DSSAT and SIMS ET is useful given the scarcity of comparisons between crop model simulated and observed ET. WEAP-DSSAT Full and Surface capture some features of the SIMS ET corn seasonal cycle, but greatly underestimate ET in August (Fig. 5). WEAP-DSSAT Full is more accurate on average (7%

342

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

overestimates ET early in the growing season (May and June) and underestimates ET late in the growing season (September). While there are a variety of potential reasons for the differences in ET between WEAP-DSSAT and SIMS, of particular note are planting date and the timing and length of flooding, which are likely not uniform across WEAP-DSSAT simulations and SIMS observations. WEAP-DSSAT Full and Surface ET over wheat, which is largely rainfed and grown in the winter, are similar to (7% underestimation) SIMS ET on average (Table 3). WEAP-DSSAT overestimates ET from November through February, and then underestimates ET in April. This could be due to differences in the timing of crop coefficient values in SIMS and crop phenology in WEAP-DSSAT, which tends to harvest in late March or early April and thus reduces ET late in the growing season, or a misclassification of mature wheat in SIMS as a crop with higher ET. 4.3. Groundwater supply sensitivity assessment

Fig. 4. WEAP-DSSAT with unlimited water and surface water only simulated yields, and NASS and field trial observed yields for: a) corn, b) rice, and c) wheat.

underestimation); however, this is partially a result of compensating errors in the seasonal cycle. The SIMS ET seasonal cycle is more consistent with WEAP-DSSAT Surface, particularly from May through July. The underestimation of WEAP-DSSAT ET in August could be due to discrepancies in maturity date either from the cultivar used in WEAP-DSSAT simulations or corn taking longer to mature due to water stress in 2011e2014. WEAP-DSSAT Full and Surface overestimate ET for rice by 29% and 23%, respectively (Table 3). WEAP-DSSAT accurately simulates rice ET in July and August, especially when run with surface water only, but

Assessing exactly how irrigation patterns change under drought in Yolo County is very challenging. While there is some fallowing and deficit irrigation, many growers access groundwater to enable high yields despite reduced surface water allocations. However, California's Sustainable Groundwater Management Act (SGMA), passed in the fall of 2014, mandates the development of Groundwater Sustainability Plans, which are likely to place limits on groundwater pumping by the early 2020s. Fig. 6 shows the water allocation and yield ratios of WEAPDSSAT Surface, in which irrigation is supplied exclusively by surface water mimicking a simplified application of the SGMA, to WEAP-DSSAT Full. Overall, both corn and rice fare relatively well without groundwater, as in most years irrigation demand is met by surface water, with the exception of the middle of the 1987e1992 drought (California Department of Water Resources, 1993). In the early part of the drought (1987 and 1988), water allocations for corn and rice were close to full, and yields were unchanged from wellwatered. In 1989, 1991, both corn and rice received no irrigation, resulting in failed crops. In 1990, fields were allocated roughly 35% of irrigation demand. This resulted in an approximately 75% reduction of corn yield relative to a full water allocation yield and a complete loss for rice. The following year (1991) water allocations were ~70% of irrigation demand, which reduced corn productivity by about 25% but had no impact on rice production relative to wellwatered. The remainder of the time series to 2008 contains a number of small reductions in water allocation. Most are less than 5% and none impact yield with the exception of corn in 2008, where a water allocation of 86% of irrigation demand resulted in corn yields 2% lower than well-watered. We again note that the areas simulated are all within the surface irrigation canal service area of YCFCWCD, so are therefore expected to be less sensitive to the

Table 2 Means of and correlations across simulated and observed yields. Both metrics are calculated over the length of record, shown in Fig. 4. Source and Crop

Mean [tons ha1]

Correlation Field Trial [ ]

Correlation NASS [ ]

WEAP-DSSAT Full Corn WEAP-DSSAT Surface Corn NASS Corn Field Trial Corn WEAP-DSSAT Full Rice WEAP-DSSAT Surface Rice NASS Rice Field Trial Rice WEAP-DSSAT Full Wheat WEAP-DSSAT Surface Wheat NASS Wheat Field Trial Wheat

11.8 10.6 10.7 11.5 10.8 9.6 8.6 9.6 5.1 5.1 5.5 5.3

0.43 0.43 0.51 e 0.33 0.05 0.52 e 0.27 0.01 0.50 e

0.36 0.03 e 0.51 0.30 0.19 e 0.52 0.29 0.18 e 0.50

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

a) Simulated and Observed Corn Evapotranspiration

300

WEAP-DSSAT Full

WEAP-DSSAT Surface

1.8

SIMS

1.0 0.8 0.6

100

0.4

50

0.2

4

5

6

7

0 1980

8

b) Simulated and Observed Rice Evapotranspiration

] 300

1.8

1985

1990

1995

2000

2005

2010

b) Simulated Water Allocation and Rice Yield Ratios

1.6

250

1.4 1.2

200

Ratio [ ]

-1

Evapotranspitation [mm month

Yield

1.2

150

150 100

1.0 0.8 0.6 0.4

50 0

Water

1.4

200

0

a) Simulated Water Allocation and Corn Yield Ratios

1.6

250

343

0.2

5

150

6

7

8

9

c) Simulated and Observed Wheat Evapotranspiration

0 1980

1.8

1985

1990

1995

2000

2005

2010

c) Simulated Water Allocation and Wheat Yield Ratios

1.6 1.4 1.2

100

1.0 0.8 0.6

50

0.4 0.2

0 10

11

12

1

2

3

4

0 1980

1985

1990

1995

2000

2005

2010

Year

Month Fig. 5. WEAP-DSSAT simulated (1981e2008) and SIMS (2011e2014) evapotranspiration growing seasonal cycles for: a) corn, b) rice, and c) wheat. Blue and red lines show irrigation supply derived from unlimited water and surface water only, respectively, while crosses denote overlapping 95% confidence intervals.

Fig. 6. Water allocation and yields ratios (WEAP-DSSAT Surface divided by WEAPDSSAT Full) for: a) corn, b) rice, and c) wheat. Water allocation ratios less than one denote irrigation water demand exceeding surface water supply. Yield ratios less and more than one describe reduced and enhanced yields, respectively.

Table 3 Mean of simulated and observed evapotranspiration. Both metrics are calculated over the length of record for the growing season, as described in Fig. 5.

First, wheat in California is largely grown in the winter when rainfall is more plentiful. When combined with a relatively drought-tolerant crop like wheat, these fields can usually be managed as rainfed. We note that Fig. 6 does not describe the amount of precipitation during the growing season (larger for wheat) or amount of irrigation applied (less for wheat). In 1989, 1991, wheat received no surface water for irrigation and yields were reduced by 18% on average relative to well-watered. As California was emerging from the drought in 1992, wheat received 67% of its full water allocation, and yields were 12% lower than well-watered. In addition to the intuitive differences in wheat yields between WEAP-DSSAT Full and WEAP-DSSAT Surface, we also found unexpected yield increases of 35% and 62% in 1998 and 2003, respectively, despite a full or slight decrease in growing season water allocation. These spikes are a result of irrigation applications in the previous or current water year, which caused relatively small reductions in soil moisture that substantially altered planting. The DSSAT automatic planting algorithm requires soil moisture greater

Source and Crop

Mean [mm]

WEAP-DSSAT WEAP-DSSAT SIMS Corn WEAP-DSSAT WEAP-DSSAT SIMS Rice WEAP-DSSAT WEAP-DSSAT SIMS Wheat

123 114 132 171 164 133 38 38 41

Full Corn Surface Corn Full Rice Surface Rice Full Wheat Surface Wheat

elimination of groundwater pumping than Yolo County and the Central Valley overall. While the drought of 1987e1992 clearly impacts water allocations for wheat, overall wheat yield reductions are small when compared to corn and rice (Fig. 6). There are several reasons for this.

344

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

than 40% of saturation, and in 1998 and 2003 the reduction in soil moisture delayed planting by a few weeks, shifting the entire growing season and thus changing the temperatures and precipitation experienced by the plants. While there is some precedent for this sensitivity in the literature (Basso et al., 2015), in the Central Valley practical considerations would likely cause growers to plant wheat earlier regardless of soil moisture and then irrigate to ensure the ability to double-crop their fields. 5. Conclusions The utility of crop models that can only be run with scheduled irrigation or unlimited automatic irrigation is diminished for predictive applications over irrigated areas. Growers will adapt irrigation schedules to future weather conditions, which is incompatible with a priori scheduling of timing and amount of irrigation, and the vast majority of irrigated croplands are irrigated because water is scarce locally, making irrigation from an unlimited water supply a flawed assumption. To address the deficiencies in each of these approaches, we coupled a crop model to a calibrated hydrologic and water resources allocation model. DSSAT adds the ability to simulate yields and a more sophisticated representation of crop water use to WEAP, WEAP provides water supply for irrigated agriculture, and an iterative approach to running WEAPDSSAT applies the water supply limitation from WEAP to automatic irrigation in DSSAT. This functionality allows WEAP-DSSAT to explore a variety of issues at the water-agriculture interface, such as future yields under limited water availability, irrigation optimization, and strategies for increasing resiliency to drought. We deploy and test our coupled model in Yolo County, an agriculturally important and water-limited area of California. WEAP-DSSAT is largely able to reproduce DSSAT yields under well-watered conditions. WEAP-DSSAT reasonably simulates mean yields of corn, rice, and wheat, and demonstrates mixed skill in capturing observed yield time series. The means and seasonal cycles of evapotranspiration modeled by WEAP-DSSAT are generally consistent with SIMS remotely sensed ET estimates, especially for corn and rice midgrowing season; however, some significant differences exist. Predicting future water supplies for irrigation in Yolo County is complicated by the unsustainable extraction of groundwater, which is used to support agricultural production even in extreme drought years. However, groundwater regulations recently enacted by the State of California have the potential to substantially reduce current pumping rates. We test the sensitivity of corn, rice, and wheat production in Yolo County Flood Control and Water Conservation District to an elimination of groundwater use for irrigation. We find yield reductions and failures in all crops examined during the 1987e1992 drought. Corn yields are reduced proportionally with water allocation, rice yields resemble more of a step function based on sufficient water for flooding, and wheat yields are least sensitive to irrigation constraints as winter wheat is grown during the wet season so does receive some rainfall. While we believe that WEAP-DSSAT is an important advancement in the simulation of agriculture-water interactions, we note that our treatment of groundwater is simplistic and our analyses are limited to field crops. True groundwater use for agriculture in Yolo County is subject to a complicated decision criteria including crop water demand, cost of pumping, depth to water table, and investment in infrastructure, which are outside of the scope of our work. Yolo County is agriculturally diverse, so while corn, wheat, and rice are important crops, we did not assess other key crops for the region such as tomatoes, almonds, alfalfa, and walnuts, because modules for these specialty crops are largely unavailable in DSSAT. Future work includes expanding both the number of crops and area simulated to enable statewide assessments for California

agriculture from seasonal to climatological timescales. Refinement of both groundwater and how growers practically respond to reduced irrigation allocations (e.g., fallowing, deficit irrigation, groundwater pumping) would also create more realistic assessments of climate impacts on agriculture. In addition, an evaluation of DSSAT to simulate water-limited yields in California is needed to assess the accuracy and constraints of the model. Finally, better management information would provide an opportunity to improve automatic planting, harvesting, and irrigation algorithms in WEAP-DSSAT. Acknowledgments This work was supported by the NASA Applied Sciences Program (NNH11ZDA001N-WATER). We thank Lyndon Estes and Cheryl Porter for sharing methods for compiling and running DSSAT on Linux platforms. We thank Forrest Melton and Alberto Guzman for providing access to SIMS evapotranspiration data for our study area. References Basso, B., Hyndman, D.W., Kendall, A.D., Grace, P.R., Robertson, G.P., 2015. Can impacts of climate change and agricultural adaptation strategies Be accurately quantified if crop models are annually Re-Initialized? PLoS ONE 10, e0127333. http://dx.doi.org/10.1371/journal.pone.0127333. Bondeau, A., Smith, P.C., Zaehle, S., Schaphoff, S., Lucht, W., Cramer, W., Gerten, D., Lotze-Campen, H., Müller, C., Reichstein, M., Smith, B., 2007. Modelling the role of agriculture for the 20th century global terrestrial carbon balance. Glob. Change Biol. 13, 679e706. http://dx.doi.org/10.1111/j.1365-2486.2006.01305.x. Brumbelow, K., Georgakakos, A., 2001. An assessment of irrigation needs and crop yield for the United States under potential climate changes. J. Geophys. Res. Atmos. 106, 27383e27405. http://dx.doi.org/10.1029/2001JD900034. California Department of Water Resources, 1993. California's 1987-92 Drought: a Summary of Six Years of Drought. Casanova, J.J., Judge, J., Jones, J.W., 2006. Calibration of the CERES-Maize model for linkage with a microwave remote sensing model. Trans. ASAE 49, 783e792. Deryng, D., Elliott, J., Folberth, C., Muller, C., Pugh, T.A.M., Boote, K.J., Conway, D., Ruane, A.C., Gerten, D., Jones, J.W., Khabarov, N., Olin, S., Schaphoff, S., Schmid, E., Yang, H., Rosenzweig, C., 2016. Regional disparities in the beneficial effects of rising CO2 concentrations on crop water productivity. Nat. Clim. Change 6, 786e790. €ll, P., Hoffmann-Dobrev, H., Portmann, F.T., Siebert, S., Eicker, A., Rodell, M., Do Strassberg, G., Scanlon, B.R., 2012. Impact of water withdrawals from groundwater and surface water on continental water storage variations. J. Geodyn. 59e60, 143e156. http://dx.doi.org/10.1016/j.jog.2011.05.001. Elliott, J., Deryng, D., Müller, C., Frieler, K., Konzmann, M., Gerten, D., Glotter, M., €rke, M., Wada, Y., Best, N., Eisner, S., Fekete, B.M., Folberth, C., Foster, I., Flo Gosling, S.N., Haddeland, I., Khabarov, N., Ludwig, F., Masaki, Y., Olin, S., Rosenzweig, C., Ruane, A.C., Satoh, Y., Schmid, E., Stacke, T., Tang, Q., Wisser, D., 2014a. Constraints and potentials of future irrigation water availability on agricultural production under climate change. Proc. Natl. Acad. Sci. 111, 3239e3244. Elliott, J., Kelly, D., Chryssanthacopoulos, J., Glotter, M., Jhunjhnuwala, K., Best, N., Wilde, M., Foster, I., 2014b. The parallel system for integrating impact models and sectors (pSIMS). Environ. Model. Softw. 62, 509e516. http://dx.doi.org/ 10.1016/j.envsoft.2014.04.008. Espino, L., Fischer, A., Godfrey, L., Greer, C., Hill, J., Horwath, W., Klonsky, K., Linquist, B., Livingston, P., McKenzie, K., Mutters, C., Ruark, M., Salmon, T., Six, J., Whisson, D., Williams, J., van Kessel, C., 2013. Rice Production Workshop Manual. Famiglietti, J.S., 2014. The global groundwater crisis. Nat. Clim. Change 4, 945e948. Frate, C., Schwanki, L., n.d. Crop Irrigation Strategies: Corn [WWW Document]. URL http://ucmanagedrought.ucdavis.edu/Agriculture/Crop_Irrigation_Strategies/ Corn/ (Accessed 24 November 14) Garfin, G., Franco, G., Blanco, H., Comrie, A., Gonzalez, P., Piechota, T., Smyth, R., Waskom, R., 2014. Ch. 20: southwest. In: Melillo, J.M., Richmond, T.C., Yohe, G.W. (Eds.), Climate Change Impacts in the United States. The Third National Climate Assessment. U.S. Global Change Research Program. Geisseler, D., Horwath, W.R., 2013. Rice production in California. Greenwald, R., Bergin, M.H., Xu, J., Cohan, D., Hoogenboom, G., Chameides, W.L., 2006. The influence of aerosols on crop production: a study using the CERES crop model. Agric. Syst. 89, 390e413. http://dx.doi.org/10.1016/ j.agsy.2005.10.004. Hart, Q.J., Brugnach, M., Temesgen, B., Rueda, C., Ustin, S.L., Frame, K., 2009. Daily reference evapotranspiration for California using satellite imagery and weather station measurement interpolation. Civ. Eng. Environ. Syst. 26, 19e33. http:// dx.doi.org/10.1080/10286600802003500.

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346 Hoogenboom, G., Jones, J.W., Wilkens, P.W., Porter, C.H., Boote, K.J., Hunt, L.A., Singh, U., Lizaso, J.L., White, J.W., Uryasev, O., 2012. Decision Support System for Agrotechnology Transfer (DSSAT) Ver. 4.5 [CD-ROM]. University of Hawaii, Honolulu, Hawaii. Huntington, J.L., Niswonger, R.G., 2012. Role of surface-water and groundwater interactions on projected summertime streamflow in snow dominated regions: an integrated modeling approach. Water Resour. Res. 48, W11524. http:// dx.doi.org/10.1029/2012WR012319. Jones, J., Hoogenboom, G., Porter, C., Boote, K., Batchelor, W., Hunt, L., Wilkens, P., Singh, U., Gijsman, A., Ritchie, J., 2003. The DSSAT cropping system model. Eur. J. Agron. 18, 235e265. http://dx.doi.org/10.1016/S1161-0301(02)00107-7. Keeling, C.D., Piper, S.C., Bacastow, R.B., Wahlen, M., Whorf, T.P., Heimann, M., Meijer, H.A., 2001. Exchanges of Atmospheric CO2 and 13CO2 with the Terrestrial Biosphere and Oceans from 1978 to 2000. I. Global Aspects. Kimball, B.A., 2011. Lessons from FACE: CO2 Effects and Interactions with Water, Nitrogen and Temperature. Handbook of Climate Change and Agroecosystems: Impacts, Adaptation, and Mitigation. Imperial College Press, London, pp. 87e107. Kucharik, C.J., 2003. Evaluation of a process-based agro-ecosystem model (AgroIBIS) across the U.S. Corn belt: simulations of the interannual variability in maize yield. Earth Interact. 7, 1e33. http://dx.doi.org/10.1175/1087-3562(2003) 007<0001:EOAPAM>2.0.CO;2. Lobell, D.B., Ortiz-Monasterio, J.I., 2007. Impacts of day versus night temperatures on spring wheat yields. Agron. J. 99, 469e477. http://dx.doi.org/10.2134/ agronj2006.0209. Maurer, E.P., Wood, A.W., Adam, J.C., Lettenmaier, D.P., Nijssen, B., 2002. A long-term hydrologically based dataset of land surface fluxes and States for the conterminous United States. J. Clim. 15, 3237e3251. Mearns, L., Rosenzweig, C., Goldberg, R., 1996. The effect of changes in daily and interannual climatic variability on CERES-Wheat: a sensitivity study. Clim. Change 32, 257e292. http://dx.doi.org/10.1007/BF00142465. Mehta, V.K., Haden, V.R., Joyce, B.A., Purkey, D.R., Jackson, L.E., 2013. Irrigation demand and supply, given projections of climate and land-use change, in Yolo County, California. Agric. Water Manag. 117, 70e82. http://dx.doi.org/10.1016/ j.agwat.2012.10.021. Mehta, V., Rheinheimer, D., Yates, D., Purkey, D., Viers, J., Young, C., Mount, J., 2011. Potential impacts on hydrology and hydropower production under climate warming of the Sierra Nevada. J. Water Clim. Change 2, 29e43. Melton, F.S., Johnson, L.F., Lund, C.P., Pierce, L.L., Michaelis, A.R., Hiatt, S.H., Guzman, A., Adhikari, D., Purdy, A.J., Rosevelt, C., Votava, P., Trout, T.J., Temesgen, B., Frame, K., Sheffner, E.J., Nemani, R.R., 2012. Satellite irrigation management support with the terrestrial observation and prediction system: a framework for integration of satellite and surface observations to support improvements in agricultural water resource management. Selected topics in applied earth observations and remote sensing. IEEE J. 5, 1709e1721. http:// dx.doi.org/10.1109/JSTARS.2012.2214474. , G., Jarvis, A., Mulligan, M., Fisher, M., Sharma, B., Xu, Z.X., Ringler, C., Mahe Ramírez, J., Clanet, J.-C., Ogilvie, A., Ahmad, M.-D., 2011. The nature and impact of climate change in the Challenge Program on Water and Food (CPWF) basins. Water Int. 36, 96e124. http://dx.doi.org/10.1080/02508060.2011.543408. National Research Council, 2011. A Review of the Use of Science and Adaptive Management in California's Draft Bay Delta Conservation Plan. National Academies Press. Nemani, R., Hashimoto, H., Votava, P., Melton, F., Wang, W., Michaelis, A., Mutch, L., Milesi, C., Hiatt, S., White, M., 2009. Monitoring and forecasting ecosystem dynamics using the terrestrial observation and prediction system (TOPS). Remote Sens. Environ. 113, 1497e1509. http://dx.doi.org/10.1016/ j.rse.2008.06.017. Parry, M., Rosenzweig, C., Livermore, M., 2005. Climate change, global food supply and risk of hunger. Philosophical Trans. R. Soc. B Biol. Sci. 360, 2125e2138. http://dx.doi.org/10.1098/rstb.2005.1751. ~ a-Barraga n, J.M., Ngugi, M.K., Plant, R.E., Six, J., 2011. Object-based crop identiPen fication using multiple vegetation indices, textural features and crop phenology. Remote Sens. Environ. 115, 1301e1316. http://dx.doi.org/10.1016/ j.rse.2011.01.009. n Piontek, F., Müller, C., Pugh, T.A.M., Clark, D.B., Deryng, D., Elliott, J., Colo lez, F., de, J., Flo €rke, M., Folberth, C., Franssen, W., Frieler, K., Friend, A.D., Gonza Gosling, S.N., Hemming, D., Khabarov, N., Kim, H., Lomas, M.R., Masaki, Y., Mengel, M., Morse, A., Neumann, K., Nishina, K., Ostberg, S., Pavlick, R., Ruane, A.C., Schewe, J., Schmid, E., Stacke, T., Tang, Q., Tessler, Z.D., Tompkins, A.M., Warszawski, L., Wisser, D., Schellnhuber, H.J., 2014. Multisectoral climate impact hotspots in a warming world. In: Proceedings of the National Academy of Sciences 111, 3233e3238. Purkey, D.R., Joyce, B., Vicuna, S., Hanemann, M.W., Dale, L.L., Yates, D., Dracup, J.A., 2008. Robust analysis of future climate change impacts on water for agriculture and other sectors: a case study in the Sacramento Valley. Clim. Change 87, 109e122. http://dx.doi.org/10.1007/s10584-007-9375-8. Raes, D., Steduto, P., Hsiao, T.C., Fereres, E., 2009. AquaCropdthe FAO Crop Model to Simulate Yield Response to Water: II. Main Algorithms and Software Description All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Agron. J. 101, 438e447. http://dx.doi.org/10.2134/agronj2008.0140s. Rosenzweig, C., Elliott, J., Deryng, D., Ruane, A.C., Müller, C., Arneth, A., Boote, K.J.,

345

Folberth, C., Glotter, M., Khabarov, N., Neumann, K., Piontek, F., Pugh, T.A.M., Schmid, E., Stehfest, E., Yang, H., Jones, J.W., 2014. Assessing agricultural risks of climate change in the 21st century in a global gridded crop model intercomparison. Proc. Natl. Acad. Sci. 111, 3268e3273. http://dx.doi.org/10.1073/ pnas.1222463110. Rosenzweig, C., Jones, J.W., Hatfield, J.L., Antle, J.M., Ruane, A.C., Mutter, C.Z., 2015. The Agricultural Model Intercomparison and Improvement Project: Phase I Activities by a Global Community of Science. Handbook of Climate Change and Agroecosystems: the Agricultural Model Intercomparison and Improvement Project (AgMIP) Integrated Crop and Economic Assessments 3. Rosenzweig, C., Jones, J.W., Hatfield, J.L., Ruane, A.C., Boote, K.J., Thorburn, P., Antle, J.M., Nelson, G.C., Porter, C., Janssen, S., Asseng, S., Basso, B., Ewert, F., Wallach, D., Baigorria, G., Winter, J.M., 2013. The agricultural model intercomparison and improvement Project (AgMIP): protocols and pilot studies. Agric. For. Meteorology 170, 166e182. http://dx.doi.org/10.1016/ j.agrformet.2012.09.011. Ruane, A.C., Goldberg, R., Chryssanthacopoulos, J., 2015. Climate forcing datasets for agricultural modeling: merged products for gap-filling and historical climate series estimation. Agric. For. Meteorology 200, 233e248. http://dx.doi.org/ 10.1016/j.agrformet.2014.09.016. Ruane, A.C., Major, D.C., Yu, W.H., Alam, M., Hussain, S.G., Khan, A.S., Hassan, A., Hossain, B.M.T.A., Goldberg, R., Horton, R.M., Rosenzweig, C., 2013. Multi-factor impact analysis of agricultural production in Bangladesh with climate change. Glob. Environ. Change 23, 338e350. http://dx.doi.org/10.1016/ j.gloenvcha.2012.09.001. Sandoval-Solis, S., McKinney, D., Loucks, D., 2010. Sustainability index for water resources planning and management. J. Water Resour. Plann. Manage 137, 381e390. http://dx.doi.org/10.1061/(ASCE)WR.1943-5452.0000134. Sau, F., Boote, K.J., Bostick, W.M., Jones, J.W., Mínguez, M.I., 2004. Testing and improving evapotranspiration and soil water balance of the DSSAT crop models. Agron. J. 96, 1243e1257. Schaible, G.D., Aillery, M.P., 2012. Water Conservation in Irrigated Agriculture: Trends and Challenges in the Face of Emerging Demands. USDA Economic Research Service. Scow, K., 2009. Cover Crop Biomass Yields. Davis, CA: Russell Ranch Electronic Data Archives: CC-001 (Accessed 23 April 15). http://russellranch.ucdavis.edu/data. Solley, W.B., Pierce, R.R., Perlman, H.A., 1998. Estimated Use of Water in the United States in 1995 (Report No. 1200), Circular. Taylor, R.G., Scanlon, B., Doll, P., Rodell, M., van Beek, R., Wada, Y., Longuevergne, L., Leblanc, M., Famiglietti, J.S., Edmunds, M., Konikow, L., Green, T.R., Chen, J., Taniguchi, M., Bierkens, M.F.P., MacDonald, A., Fan, Y., Maxwell, R.M., Yechieli, Y., Gurdak, J.J., Allen, D.M., Shamsudduha, M., Hiscock, K., Yeh, P.J.-F., Holman, I., Treidel, H., 2013. Ground water and climate change. Nat. Clim. Change 3, 322e329. http://dx.doi.org/10.1038/nclimate1744. Tubiello, F.N., Soussana, J.-F., Howden, S.M., 2007. Crop and pasture response to climate change. Proc. Natl. Acad. Sci. 104, 19686e19690. http://dx.doi.org/ 10.1073/pnas.0701728104. University of California Cooperative Extension, 2015. Delta Crops Resource Management [WWW Document] (Accessed 13 April 15). http://ucanr.edu/sites/ deltacrops/Rice/. University of California Division of Agriculture and Natural Resources, 2006. Small Grain Production Manual (No. 8208). USDA NASS, 2014. Census of Agriculture, p. 2012. USDA NASS, 2010. Field Crops Usual Planting and Harvesting Dates (No. 628). rrez, I., Swartz, C.H., Downing, T.E., 2011. Balancing Varela-Ortega, C., Blanco-Gutie groundwater conservation and rural livelihoods under water and climate uncertainties: an integrated hydro-economic modeling framework. Glob. Environ. Change 21, 604e619. http://dx.doi.org/10.1016/j.gloenvcha.2010.12.001. €rke, M., Gerten, D., Haddeland, I., Hanasaki, N., Wada, Y., Wisser, D., Eisner, S., Flo Masaki, Y., Portmann, F.T., Stacke, T., Tessler, Z., Schewe, J., 2013. Multimodel projections and uncertainties of irrigation water demand under climate change. Geophys. Res. Lett. 40, 4626e4632. http://dx.doi.org/10.1002/grl.50686. Walthall, C.L., Hatfield, J., Backlund, P., Lengnick, L., Marshall, E., Walsh, M., Adkins, S., Aillery, M., Ainsworth, E.A., Ammann, C., Anderson, C.J., Bartomeus, I., Baumgard, L.H., Booker, F., Bradley, B., Blumenthal, D.M., Bunce, J., Burkey, K., Dabney, S.M., Delgado, J.A., Dukes, J., Funk, A., Garrett, K., Glenn, M., Grantz, D.A., Goodrich, D., Hu, S., Izaurralde, R.C., Jones, R.A.C., Kim, S.-H., Leaky, A.D.B., Lewers, K., Mader, T.L., McClung, A., Morgan, J., Muth, D.J., Nearing, M., Oosterhuis, D.M., Ort, D., Parmesan, C., Pettigrew, W.T., Polley, W., Rader, R., Rice, C., Rivington, M., Rosskopf, E., Salas, W.A., Sollenberger, L.E., €ckle, C., Takle, E.S., Timlin, D., White, J.W., Winfree, R., WrightSrygley, R., Sto Morton, L., Ziska, L.H., 2013. Climate Change and Agriculture in the United States: Effects and Adaptation. (Technical Bulletin No. 1935). USDA, Washington, DC. White, J.W., Hoogenboom, G., Kimball, B.A., Wall, G.W., 2011. Methodologies for simulating impacts of climate change on crop production. Field Crops Res. 124, 357e368. http://dx.doi.org/10.1016/j.fcr.2011.07.001. Winter, J.M., Lopez, J.R., Ruane, A.C., Young, C.A., Scanlon, B.R., Rosenzweig, C., 2017. Representing water scarcity in future agricultural assessments. Anthropocene 18, 15e26. http://dx.doi.org/10.1016/j.ancene.2017.05.002. Yates, D., Galbraith, H., Purkey, D., Huber-Lee, A., Sieber, J., West, J., Herrod-Julius, S., Joyce, B., 2008. Climate warming, water storage, and Chinook salmon in California's Sacramento Valley. Clim. Change 91, 335e350. http://dx.doi.org/ 10.1007/s10584-008-9427-8. Yates, D., Purkey, D., Sieber, J., Huber-Lee, A., Galbraith, H., 2005a. WEAP21da

346

J.M. Winter et al. / Environmental Modelling & Software 96 (2017) 335e346

demand-, priority-, and preference-driven water planning model: Part 2: aiding freshwater ecosystem service evaluation. Water Int. 30, 501e512. http:// dx.doi.org/10.1080/02508060508691894. Yates, D., Purkey, D., Sieber, J., Huber-Lee, A., Galbraith, H., West, J., Herrod-Julius, S., Young, C., Joyce, B., Rayej, M., 2009. Climate driven water resources model of the

Sacramento basin. Calif. J. Water Resour. Plann. Manage 135, 303e313. http:// dx.doi.org/10.1061/(ASCE)0733-9496(2009)135:5303. Yates, D., Sieber, J., Purkey, D., Huber-Lee, A., 2005b. WEAP21da demand-, priority-, and preference-driven water planning model: Part 1: model characteristics. Water Int. 30, 487e500. http://dx.doi.org/10.1080/02508060508691893.