Automated Detection of On-Farm Irrigation Reservoirs in Two Critical Groundwater Regions of Arkansas:A Necessary Precursor for Conjunctive Water Management

Authors: SHULTS, DANIEL - Arkansas State University; NOWLIN, JOHN - Arkansas State University; Reba, Michele; Massey, Joseph; HASHEM, AHMED - Arkansas State University

Submitted to: International Journal of Geospatial and Environmental Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/1/2023
Publication Date: 2/1/2024
Citation: Shults, D., Nowlin, J., Reba, M.L., Massey, J., Hashem, A. 2024. Automated Detection of On-Farm Irrigation Reservoirs in Two Critical Groundwater Regions of Arkansas:A Necessary Precursor for Conjunctive Water Management. International Journal of Geospatial and Environmental Research. 15(1):1-22. https://doi.org/10.4018/IJAGR.2024.15.1.
DOI: https://doi.org/10.4018/IJAGR.2024.15.1

Interpretive Summary: Parts of the Lower Mississippi River Basin are experiencing significant groundwater declines. Conservation practices focused on collecting, storing, and reusing surface water are being implemented in the region. In order to plan such activities, water managers need tools to help document and assess existing surface water infrastructure. Currently, there is not an efficient and effective method for differentiating between irrigation reservoirs and other bodies of water using remotely sensed data. This research developed an automated method for detecting irrigation reservoirs by measuring the spatial elevation characteristics of waterbodies in relation to surrounding terrain. This method was able to accurately locate on-farm reservoirs for a 750 thousand ha (1.8 Mac) area in a matter of hours rather than the weeks required to manually prepare the assessment dataset at a regional basis. This method can be used to augment periodic inventories of agricultural water resources across the region and will be useful for water managers in low relief regions to potentially identify future reservoir sites.

Technical Abstract: Annually, Arkansas accounts for nearly 50% of U.S. rice production; rice is the most water intensive crop grown in Arkansas, consuming about 800 mm (31.5 in) of water during a growing season. It might be expected that surface water would be plentiful in a state that receives ~1,200 mm (47 in) of precipitation each year, however, rainfall is often unpredictable and isolated spatially during the growing season. Therefore, farmers rely on groundwater to irrigate crops. Because of limited recharge rates and excessive pumping, Eastern Arkansas is experiencing significant groundwater declines. Conservation practices focused on collecting, storing, and reusing surface water from precipitation, field runoff and excess streamflow are being implemented in the region. In order to plan such activities, water managers need tools to help document and assess existing surface water infrastructure. Currently, there is not an efficient and effective method for differentiating between irrigation reservoirs and other bodies of water using remotely sensed data. This research developed an automated method for detecting irrigation reservoirs by measuring the spatial elevation characteristics of waterbodies in relation to surrounding terrain. Waterbodies were identified using near-infrared aerial imagery (NAIP 2019). A localized mean elevation statistic was calculated from a digital elevation model (DEM) at two distances beyond the waterbody. These zones were used to classify a subset of waterbodies as reservoirs. Results were compared to an existing curated dataset for irrigation reservoirs. The model successfully detected 93% of these known reservoirs. This method was able to accurately locate on-farm reservoirs for a 750 thousand ha (1.8 Mac) area in a matter of hours rather than the weeks required to manually prepare the assessment dataset.