Recipient Organization
UNIV OF MARYLAND
(N/A)
COLLEGE PARK,MD 20742
Performing Department
(N/A)
Non Technical Summary
Climate change is already having a profound impact on agriculture around the world. Crop losses due to extreme heat are projected to rise substantially through the end of the century. Whether and how farmers will adapt to high temperatures are therefore crucial policy questions.Irrigation represents a key form of agricultural climate adaptation. For example, hot temperatures cause substantial crop losses in the rainfed Eastern U.S. where irrigation is less common, but not in the irrigation-dependent Western U.S. However, despite its importance for agriculture--and its susceptibility to climate change--the role of groundwater remains understudied. In particular, little is known about the extent to which farmers can use groundwater to adapt to climate change, since measuring groundwater extraction is difficult. Moreover, understanding the role of groundwater as a source of climate adaptation is becoming increasingly urgent, as underground aquifers are being rapidly depleted around the globe. The goal of this project is to empirically assess the importance and sustainability of adaptive irrigation under climate change.We have three objectives that will contribute to our knowledge about the linkages between agriculture, groundwater, and climate change in the United States. Objective 1 is to estimate the link between extreme heat and adaptive irrigation. Objective 2 is to predict groundwater stock dynamics under climate change. Objective 3 is to analyze the relationship between crop insurance and water use. In meeting these three objectives, we will leverage high-resolution data and employ rigorous econometric methods to produce results that are both academically interesting and policy relevant.
Animal Health Component
20%
Research Effort Categories
Basic
80%
Applied
20%
Developmental
0%
Goals / Objectives
Irrigation represents a key form of agricultural climate adaptation. For example, hot temperatures cause substantial crop losses in the rainfed Eastern U.S. where irrigation is less common, but not in the irrigation-dependent Western U.S. However, despite its importance for agriculture--and its susceptibility to climate change--the role of groundwater remains understudied. In particular, little is known about the extent to which farmers can use groundwater to adapt to climate change, since measuring groundwater extraction is difficult. Moreover, understanding the role of groundwater as a source of climate adaptation is becoming increasingly urgent, as underground aquifers are being rapidly depleted around the globe. The goal of this project is to empirically assess the importance and sustainability of adaptive irrigation under climate change.Objective 1: Estimate the link between temperature and adaptive irrigation.Our first objective is to estimate how farmers use irrigation as a form of adaptation to extreme heat. Irrigation is a crucial adaptation tool that helps protect against heat-related crop losses. Understanding how farmers actually use this tool in practice will shed light on the costs of climate change and also the extent to which a hotter climate will likely increase the demand for groundwater for adaptive irrigation.We will leverage electricity consumption data to estimate the causal relationship between temperature and groundwater irrigation. The vast majority of groundwater pumps in the U.S. are powered by electricity, which serves as the sole variable input for pulling water up to the surface. Electricity utilities collect customer-level consumption data at a high temporal frequency, providing accurate measures of when and where groundwater pumping occurs. Using these data, we overcome a major challenge in the empirical literature on groundwater use, since reliable microdata on pumping behavior are rare.We will merge a retrospective panel of temperature and precipitation data with electricity data from two sources: (i) a sample of utility-level data on electricity distribution to irrigation customers compiled by USDA's Rural Utilities Service; and (ii) a sample of over 10,000 groundwater pumps from Pacific Gas & Electric (PGE), California's largest investor-owned utility. We will also combine the latter dataset with detailed data on pump efficiencies and groundwater levels to recover the quantity of groundwater pumped. We will use both datasets to estimate binned temperature-response functions using a panel fixed effects framework (following the climate impacts literature). This estimation strategy isolates idiosyncratic variation in temperature, controlling for average temperature in each location, average groundwater use at each location, and year-specific factors common to all pumps in the sample.Our estimates will recover the marginal effect of an additional hot day on agricultural groundwater use, including both the quantity of groundwater pumped and the extensive-margin decision of whether to pump. Since our PGE data are at high temporal frequency, we will estimate temperature response functions at the annual and daily levels. Daily regressions will be particularly informative of farmers' irrigation behavior--for example, we can test if farmers irrigate in anticipation of a hot day, as well as on the hot day itself. These results will contribute to the climate economics literature by showing how adaptive irrigation works to moderate the temperature-yield relationship.Objective 2: Predict groundwater stock dynamics under climate change.Objective 1 focuses on short-run adaptation behavior, treating irrigation demand as a static problem. That analysis captures the farmer's immediate decision to irrigate when unexpected heat shocks threaten to damage their crops. It also aligns with the broader economics literature that estimates static temperature response functions using panel fixed effects techniques. This line of research often invokes the Envelope Theorem to translate marginal responses to weather shocks into expected future climate impacts.However, groundwater irrigation is inherently dynamic: irrigation today depletes the stock of groundwater available for irrigation in future periods. This raises future pumping costs as farmers will need to expend more energy to extract groundwater from greater depths. This form of state dependence alters farmers' profit functions and breaks the weather-climate mapping provided by the Envelope Theorem. Intuitively, if adaptive irrigation becomes increasingly expensive in the future, farmers will demand less irrigation on hot days.In Objective 2 we will project how adaptive irrigation is likely to impact groundwater scarcity under climate change while directly incorporating state dependence in groundwater depths. Our key insight is that this state dependence affects irrigation costs, while our estimates from Objective 1 capture the benefits of adaptive irrigation. Since the optimizing farmer should treat these benefits and costs separately, we can project the net impact of climate change on groundwater depths using separate estimates of the benefits (i.e., temperature) response and the cost response.We will parameterize a stylized dynamic model of adaptive irrigation using three sets of econometric estimates from the same sample of California farmers: (i) temperature-response estimates from Objective 1, (ii) groundwater cost elasticity estimates, and (iii) new estimates of the physical relationship between current-year groundwater extraction and next-year groundwater depths. Using state-of-the-art climate projections, we will step forward year-by-year to forecast adaptive irrigation under hotter temperatures (using (i)), net out cost-based demand response (using (ii)), and project the impact on next-period groundwater stocks (using (iii)).Objective 3: Analyze the relationship between crop insurance and water use.Objectives 1 and 2 explore how farmers use irrigation as a means of adaptation to hotter temperatures and a changing climate. However, irrigation decisions are also influenced by other incentives that farmers face. A key source of these incentives is the Federal Crop Insurance Program, which is administered by the USDA's Risk Management Agency (USDARMA). USDA-RMA offers varying subsidies for crop-specific insurance policies. As a result, the crop insurance program influences what farmers choose to grow and the inputs they use--both of which impact water use. Objective 3 aims to empirically assess the interaction between crop insurance and irrigation decisions, in order to understand how subsidized insurance policies promote or crowd out adaptation to climate change.Our empirical approach will leverage a unique aspect of USDA-RMA's insurance program: crop insurance prices are set annually at the county-crop level, to reflect a county's average risk profile. In many cases, farmers in two adjacent counties face different net premiums for the same insurance product. This lets us use a regression discontinuity (RD) design that compares agricultural land in the neighborhood of these county borders, where insurance prices change discontinuously but risk is smooth. We will build a comprehensive geospatial dataset on crop insurance, crop choice, and irrigation to estimate the causal effect of crop insurance subsidies on agricultural water use. We plan to supplement this RD design with a panel fixed effects approach, leveraging variation in insurance prices across counties and across years.
Project Methods
Objective 1Previous research has documented how extreme heat causes crop losses in the rainfed Eastern U.S., but not in the irrigated Western U.S. While this striking difference between rainfed vs. irrigated temperature effects has been attributed to the presence of adaptive irrigation, there is limited direct evidence that farmers irrigate crops more on hot days. This is likely due to the inherent difficulty of observing farmers' irrigation behavior.In order to measure this irrigation behavior, we will use customer-level electricity consumption and billing data from Pacific Gas & Electric (PGE). PGE serves the majority of farmers in California's Central Valley, the nation's most economically valuable crop-producing region. We have access to these confidential data through a non-disclosure agreement, which includes the universe of PGE's agricultural customers. We also observe audit data for nearly 12,000 unique groundwater pumps, which we can match to electricity meters in our sample. For this subset of pumps we know that all electricity consumption goes towards powering vertical groundwater pumps. By combining pump-specific efficiency data (from PGE audits) with publicly available measurements of groundwater depth, we can convert kilowatt-hours of electricity use into acre-feet of groundwater consumption for each pump in our sample. We will also use a nationwide utility-level sample of electricity distributed to irrigation customers from rural electric cooperatives. These data come from the USDA's Rural Utilities Service, and they will allow us to estimate the temperature-irrigation relationship more broadly at the national level.We will spatially merge these two electricity datasets with gridded temperature and precipitation data from the PRISM Climate Group to estimate a panel fixed effects model of temperature response using electricity and groundwater consumption as outcome variables. Since our PGE data report the dollar value of customers' electricity bills, we can also estimate farmers' average cost of adaptive irrigation. This estimate can be interpreted as a lower bound on farmers' marginal willingness-to-pay for adaptive irrigation, or their marginal willingness-to-pay to avoid a hot day--provided that spending on adaptive irrigation is not offset by other cost reductions. We plan to test for such offsetting cost reductions using data from the Census of Agriculture on other crop inputs, such as fertilizer and pesticides.Objective 2We will use a simple economic model to illustrate how farmers trade off the benefits vs. costs of adaptive groundwater irrigation. On a given hot day, an optimizing farmer will choose to irrigate with groundwater if the benefits of avoided crop losses outweigh the costs of pumping that groundwater. A day of extreme heat should inflict the same damages to crops regardless of whether the hot temperatures are anomalous (i.e., prior to the onset of severe climate change) or more common (i.e., after a rightward shift in the temperature distribution). This suggests that our static estimates from Objective 1 can be interpreted as farmers' marginal benefits of adaptive irrigation under climate change.On the other hand, the costs of adaptive groundwater irrigation are inherently dynamic. As increased extraction further depletes groundwater resources, farmers' pumping costs increase--since more energy is needed to pull groundwater up to the surface from greater depths. These higher pumping costs reduce farmers' groundwater input demand.Importantly, from the farmer's private perspective, we can model the benefits and costs of adaptive irrigation as additively separable: a higher cost per acre-foot of groundwater does not alter the avoided heat damages to crops, and hotter temperatures do not contemporaneously impact equilibrium groundwater depths. With this insight, we can sum our static estimates of adaptive irrigation from Objective 1 with our existing dynamic cost elasticity estimates to characterize the net climate impacts on irrigation and groundwater resource scarcity.We will use these well-identified econometric estimates to parameterize a dynamic model of groundwater extraction under climate change. In each year, hotter temperatures will cause farmers to extract more groundwater for the purposes of adaptive irrigation, which we can forecast using our static estimates from Objective 1. This increase in extraction across all farmers translates into an aggregate lowering of groundwater depths, raising pumping costs in the ensuing year. Using the dynamic elasticity estimates, we will project how these higher costs will reduce next-year groundwater demand, moderating the effect of hotter temperatures. Both the adaptive response and the cost response are likely to substantially alter the current trajectory of groundwater depletion, and it is not obvious ex ante which effect will dominate.We will start by predicting adaptive irrigation under future perturbations of the temperature distribution using our fitted regression model from Objective 1. This will let us predict adaptive irrigation under hotter temperatures. We will also model the dynamics of how this will likely impact groundwater depths, accounting for cost-based demand response. This will let us forward-simulate groundwater levels of California aquifers 20 years into the future.Objective 3We will leverage a unique feature of the Federal Crop Insurance Program, run by USDA-RMA: insurance prices (premiums net of subsidies) are set at the county level, changing differently over time for different crops and locations. Though agricultural risk is smooth across space, USDA-RMA's rules create discontinuities in insurance prices at county borders. We will use this variation in conjunction with high-resolution spatial data to estimate both regression discontinuity (RD) and difference-in-differences (DD) designs, recovering the causal effect of insurance premiums on insurance take-up, crop choice, irrigation, and groundwater levels.We will obtain crop insurance prices from USDA-RMA's Summary of Business data from 1989 to the present. These publicly-available data provide information on each insurance product offered by USDA-RMA, including the premium--designed to be actuarily fair--and the subsidy amount for each commodity-county-year. Net insurance prices are set at the commodity-county-year level, which is key for identification.Our analysis will focus on three sets of outcomes. First, we will use confidential field level data on insurance take-up from USDA-RMA. At present, we have access to these data in Colorado, Nebraska, and Kansas for 2007-2013. These three states sit atop the High Plains (Ogalalla) Aquifer--a key region for our analysis. We have requested access to these data for all states and for a longer time period. The Summary of Business data also report total covered acres for each county-crop-year.Second, we will use two data sources on farmers' crop choices. Our main dataset will come from the USDA-NASS's Cropland Data Layer (CDL). This satellite-derived dataset provides information on what crop is being grown in every 30-by-30 meter pixel in the United States from 1997 to 2022. The confidential USDA-RMA data discussed above also report crop choice for the subset of farmers who purchase insurance.Third, we will estimate how insurance prices impact water use and groundwater levels.To measure the presence of irrigation, we will use the Landsat-based Irrigation Dataset (LANID-US). To capture the intensive margin, we will collect data on groundwater extraction wherever it is available, including electricity consumption dataand water well withdrawal data from Kansas, Colorado, and Nebraska. Finally, we will estimate effects on groundwater levels using publicly available data from monitoring wells from USGS and state water agencies.