skip to main content


Title: IrrMapper: A Machine Learning Approach for High Resolution Mapping of Irrigated Agriculture Across the Western U.S.
High frequency and spatially explicit irrigated land maps are important for understanding the patterns and impacts of consumptive water use by agriculture. We built annual, 30 m resolution irrigation maps using Google Earth Engine for the years 1986–2018 for 11 western states within the conterminous U.S. Our map classifies lands into four classes: irrigated agriculture, dryland agriculture, uncultivated land, and wetlands. We built an extensive geospatial database of land cover from each class, including over 50,000 human-verified irrigated fields, 38,000 dryland fields, and over 500,000 km 2 of uncultivated lands. We used 60,000 point samples from 28 years to extract Landsat satellite imagery, as well as climate, meteorology, and terrain data to train a Random Forest classifier. Using a spatially independent validation dataset of 40,000 points, we found our classifier has an overall binary classification (irrigated vs. unirrigated) accuracy of 97.8%, and a four-class overall accuracy of 90.8%. We compared our results to Census of Agriculture irrigation estimates over the seven years of available data and found good overall agreement between the 2832 county-level estimates (r 2 = 0.90), and high agreement when estimates are aggregated to the state level (r 2 = 0.94). We analyzed trends over the 33-year study period, finding an increase of 15% (15,000 km 2 ) in irrigated area in our study region. We found notable decreases in irrigated area in developing urban areas and in the southern Central Valley of California and increases in the plains of eastern Colorado, the Columbia River Basin, the Snake River Plain, and northern California.  more » « less
Award ID(s):
1633831
NSF-PAR ID:
10192099
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
Remote Sensing
Volume:
12
Issue:
14
ISSN:
2072-4292
Page Range / eLocation ID:
2328
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Agricultural fields in drylands are challenged globally by limited freshwater resources for irrigation and also by elevated soil salinity and sodicity. It is well known that pedogenic carbonate is less soluble than evaporate salts and commonly forms in natural drylands. However, few studies have evaluated how irrigation loads dissolved calcium and bicarbonate to agricultural fields, accelerating formation rates of secondary calcite and simultaneously releasing abiotic CO2to the atmosphere. This study reports one of the first geochemical and isotopic studies of such “anthropogenic” pedogenic carbonates and CO2from irrigated drylands of southwestern United States. A pecan orchard and an alfalfa field, where flood-irrigation using the Rio Grande river is a common practice, were compared to a nearby natural dryland site. Strontium and carbon isotope ratios show that bulk pedogenic carbonates in irrigated soils at the pecan orchard primarily formed due to flood-irrigation, and that approximately 20–50% of soil CO2in these irrigated soils is calcite-derived abiotic CO2instead of soil-respired or atmospheric origins. Multiple variables that control the salt buildup in this region are identified and impact the crop production and soil sustainability regionally and globally. Irrigation intensity and water chemistry (irrigation water quantity and quality) dictate salt loading, and soil texture governs water infiltration and salt leaching. In the study area, agricultural soils have accumulated up to 10 wt% of calcite after just about 100 years of cultivation. These rates will likely increase in the future due to the combined effects of climate variability (reduced rainfall and more intense evaporation), use of more brackish groundwater for irrigation, and reduced porosity in soils. The enhanced accumulation rates of pedogenic carbonate are accompanied by release of large amounts of abiotic CO2from irrigated drylands to atmosphere. Extensive field studies and modelling approaches are needed to further quantify these effluxes at local, regional and global scales.

     
    more » « less
  2. Efficient irrigation technologies, which seem to promise reduced production costs and water consumption in heavily irrigated areas, may instead be driving increased irrigation use in areas that were not traditionally irrigated. As a result, the total dependence on supplemental irrigation for crop production and revenue is steadily increasing across the contiguous United States. Quantifying this dependence has been hampered by a lack of comprehensive irrigated and dryland yield and harvested area data outside of major irrigated regions, despite the importance and long history of irrigation applications in agriculture. This study used a linear regression model to disaggregate lumped agricultural statistics and estimate average irrigated and dryland yields at the state level for five major row crops: corn, cotton, hay, soybeans, and wheat. For 1945–2015, we quantified crop production, irrigation enhancement revenue, and irrigated and dryland areas in both intensively irrigated and marginally-dependent states, where both irrigated and dryland farming practices are implemented. In 2015, we found that irrigating just the five commodity crops enhanced revenue by ~$7 billion across all states with irrigation. In states with both irrigated and dryland practices, 23% of total produced area relied on irrigation, resulting in 7% more production than from dryland practices. There was a clear response to increasing biofuel demand, with the addition of more than 3.6 million ha of irrigated corn and soybeans in the last decade in marginally-dependent states. Since 1945, we estimate that yield enhancement due to irrigation has resulted in over $465 billion in increased revenue across the contiguous United States (CONUS). Example applications of this dataset include estimating historical water use, evaluating the effects of environmental policies, developing new resource management strategies, economic risk analyses, and developing tools for farmer decision making. 
    more » « less
  3. Abstract

    In recent decades, irrigated agriculture has expanded dramatically over the Southeastern United States (SEUS). The trend is more likely to continue in future given the need to further improve crop productivity and its resilience against droughts, however, the impact of these SEUS land cover changes remains unknown. This study investigates how and to what extent rain-fed to irrigation-fed (RFtoIF) transition in the SEUS region modulates precipitation spatially and temporally under a severe drought meteorological condition. In this study, we perform three Weather Research Forecasting model simulations with varying degrees of irrigated crop areas with meteorological boundary conditions of a record-breaking 2007 drought in the SEUS region. Results show that the SEUS irrigation expansion reduces both the convective triggering potential and low-level humidity index through land-atmospheric interaction. This is accompanied by reduction in the height of atmospheric boundary layer (ABL)-lifting condensation level crossing and increase in the convective available potential energy. These modulations within the ABL provide a favorable condition for strong deep convection during the drought period. However, the impact on precipitation is heterogeneous, with crop areas undergoing RFtoIF transition experiencing an overall reduction in precipitation while other landcovers experiencing an increase. The reduction in precipitation over RFtoIF transitioned croplands is in part due to moisture redistribution aided by generation of an anomalous high-pressure system. The results highlight the complexity of response of precipitation to irrigation expansion in the SEUS, and underscore the need to perform spatially-explicit analysis for mitigating risks to water resources and food security.

     
    more » « less
  4. Abstract

    The historical rise of irrigation has profoundly mitigated the effect of drought on agriculture in many parts of the United States. While irrigation directly alters soil moisture, meteorological drought indices ignore the effects of irrigation, since they are often based on simple water balance models that neglect the irrigation input. Reanalyses also largely neglect irrigation. Other approaches estimate the evaporative fraction (EF), which is correlated with soil moisture under water-limited conditions typical of droughts, with lower values corresponding to drier soils. However, those approaches require satellite observations of land surface temperature, meaning they cannot be used to study droughts prior to the satellite era. Here, we use a recent theory of land–atmosphere coupling—surface flux equilibrium (SFE) theory—to estimate EF from readily available observations of near-surface air temperature and specific humidity with long historical records. In contrast to EF estimated from a reanalysis that largely neglects irrigation, the SFE-predicted EF is greater at irrigated sites than at nonirrigated sites during droughts, and its historical trends are typically consistent with the spatial distribution of irrigation growth. Two sites at which SFE-predicted EF unexpectedly rises in the absence of changes in irrigation can be explained by increased flooding due to human interventions unrelated to irrigation (river engineering and the expansion of fish hatcheries). This work introduces a new method for quantifying agricultural drought prior to the satellite era. It can be used to provide insight into the role of irrigation in mitigating drought in the United States over the twentieth century.

    Significance Statement

    Irrigation grew profoundly in the United States over the twentieth century, increasing the resilience of American agriculture to drought. Yet observational records of agricultural drought, and its response to irrigation, are limited to the satellite era. Here, we show that a common measure of agricultural drought (the evaporative fraction, EF) can be estimated using widespread weather data, extending the agricultural drought record decades further back in time. We show that EF estimated using our approach is both sensitive and specific to the occurrence of irrigation, unlike an alternative derived from a reanalysis.

     
    more » « less
  5. High-resolution mapping of irrigated fields is needed to better estimate water and nutrient fluxes in the landscape, food production, and local to regional climate. However, this remains a challenge in humid to subhumid regions, where irrigation has been expanding into what was largely rainfed agriculture due to trends in climate, crop prices, technologies and practices. One such region is southwestern Michigan, USA, where groundwater is the main source of irrigation water for row crops (primarily corn and soybeans). Remote sensing of irrigated areas can be difficult in these regions as rainfed areas have similar characteristics. We present methods to address this challenge and enhance the contrast between neighboring rainfed and irrigated areas, including weather-sensitive scene selection, applying recently developed composite indices and calculating spatial anomalies. We create annual, 30m-resolution maps of irrigated corn and soybeans for southwestern Michigan from 2001 to 2016 using a machine learning method (random forest). The irrigation maps reasonably capture the spatial and temporal pattern of irrigation, with accuracies that exceed available products. Analysis of the irrigation maps showed that the irrigated area in southwestern Michigan tripled in the last 16 years. We also discuss the remaining challenges for irrigation mapping in humid to subhumid areas. 
    more » « less