Spanning eight states from Texas to South Dakota, the Ogallala Aquifer is one of the largest and most stressed aquifer systems in the world. Underlying some of the most dynamic irrigated area in the country, it supports a $20 billion agricultural economy, but groundwater use is unsustainable over a large part of the aquifer.
According to experts, maintaining and even expanding irrigation is required to continue growing economically viable crops given rising temperatures and increasing rainfall variability due to climate change. However, without substantial management and technological interventions, irrigation may become impractical even on currently irrigated fields due to ongoing groundwater depletion. Understanding how irrigated areas change over time can help inform forward-looking water management, but most regions lack detailed records of when and where irrigation occurs.
Thanks to an MSU-led team of scientists, including hydrogeologists Jillian Deines and David Hyndman, a new map dataset made from 34 years of satellite imagery – the longest, highest-resolution record of where and when irrigation occurs to date – is now available to farmers, managers and researchers to inform policies and management decisions. The results of their study were recently published in the journal, “Remote Sensing of the Environment.”
Until recently, efforts to map irrigated lands from satellite images across large regions were limited by expensive high-resolution imagery and high computational costs. Now, with free high-quality Landsat satellite imagery from the United States and new cloud computing tools such as Google Earth Engine, researchers can access and analyze large volumes of data to understand changes to land management as far back as the 1970s.
"To manage limited resources into the future, you need to understand the past,” said Deines, a former doctoral student in Hyndman’s MSU Hydrogeology Lab, now a postdoctoral researcher at Stanford University and the paper’s lead author. “Unfortunately, we often lack the long-term records we need. The Landsat satellite archive provides a continuous stream of data spanning more than 35 years. With cloud-based computing tools such as Google Earth Engine, our work can translate those satellite observations into information that helps us understand where we are, where we've been and, hopefully, better plan for where we're going."
Building upon past efforts that focused on the Republican River Basin in the northern Ogallala for 1999-2016, the team used machine learning to produce an annual, high resolution irrigation map time series from 1984 to 2017 over the entire aquifer. Leveraging Earth Engine's extensive data catalog, they combined Landsat imagery, climate and soil data to classify active irrigation each year. Their approach produced maps with high overall accuracy, 91%, that closely matched existing United States Department of Agriculture, or USDA, county statistics. They were then able to identify regions with declining, expanding and stable irrigated areas.
“High resolution maps of annual irrigation provide a strong basis to quantify the influence of physical and socioeconomic factors on farmers annual decisions to irrigate,” said Hyndman, professor and chair of the Department of Earth and Environmental Sciences in the MSU College of Natural Science and senior author. “This, in turn, can help decisionmakers chart a path toward sustainable water use in areas where pumping for irrigated agriculture is depleting aquifers.”
“These maps also provide the necessary inputs to model the effects of irrigation on downwind climate and regional hydrology,” Hyndman said.
Deines noted that there is still more to learn from this dataset. For example, environmental geosciences graduate student and Hydrogeology Lab team member Ally Brady is using the patterns of irrigation over time to document changes in irrigation technology to better understand how these technological advances affect water use and agricultural production.
“Looking to the future, these maps tell us something we didn't know before – the precise locations of irrigated fields overlying rapidly declining portions of the aquifer,” Deines said. “This information is particularly important because, given the declining aquifer storage, we estimate that up to 24 percent of currently irrigated area may be lost this century. For this reason, I'm currently working with a research team from universities across the region to think through what these areas might transition to when irrigation is no longer possible and what the economic impact of these transitions may be."
In addition to Hyndman and Deines, other scientists who participated in the study are Anthony Kendall and Jeremy Rapp from the MSU Hydrogeology Lab, and Morgan Crowley and Jeffrey Cardille from McGill University Department of Natural Resource Sciences, Quebec, Canada.