On-farm reservoir monitoring using Landsat inundation datasets

Authors: PERIN, VINICIUS - North Carolina State University; TALBURE, MIRELA - North Carolina State University; GAINES, MOLLIE - North Carolina State University; Reba, Michele; YAEGER, MARY - University Of Memphis

Submitted to: Agricultural Water Management
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/7/2020
Publication Date: 1/4/2021
Citation: Perin, V., Talbure, M., Gaines, M.D., Reba, M.L., Yaeger, M. 2021. On-farm reservoir monitoring using Landsat inundation datasets. Remote Sensing of Environment. 246. https://doi.org/10.1016/j.agwat.2020.106694.
DOI: https://doi.org/10.1016/j.agwat.2020.106694

Interpretive Summary: On-farm reservoirs (OFRs) enable farmers to store water during the wet season for crop irrigation during the dry season. However, monitoring the inter- and intra-annual change of these water bodies remains a challenging task because they are typically small and occur in high numbers. Therefore, we used two existing remote sensing inundation datasets to assess surface water area change of OFRs located in Eastern Arkansas. We found the highest mean percent errors (MPE) in size (~20%) for OFRs between 0–5 ha, the smallest size class in our study. Both inundation datasets enabled us to estimate the seasonality in surface area change of OFRs, with the highest surface water extent between March-May, the months when the region receives most of the annual precipitation. This work can be used to enhance hydrological assessments in poorly monitored basins that have a concentration of OFRs.

Technical Abstract: On-farm reservoirs (OFRs)—artificial water impoundments that retain water from rainfall and run-off—enable farmers to store water during the wet season for crop irrigation during the dry season. However, monitoring the inter- and intra-annual change of these water bodies remains a challenging task because they are typically small (<10 ha) and occur in high numbers. Therefore, we used two existing Landsat inundation datasets—the U.S. Geological Survey Dynamic Surface Water Extent (DSWE) and the European Commission’s Joint Research Centre (JRC) Global Monthly Water History—to assess surface water area change of OFRs located in Eastern Arkansas, the third most irrigated state in the U.S. that has seen a rapid increase of OFRs occurrence. We used an existent OFRs dataset as ground truth. We aimed (i) to compare the performance of the DSWE and the JRC when characterizing OFRs of varied sizes and (ii) to assess the impact of climate variables (i.e. precipitation and temperature) on surface water area of OFRs. We found the highest mean percent errors (MPE) in size (~20%) for OFRs between 0–5 ha, the smallest size class in our study. The DSWE had a smaller MPE and higher agreement with our ground truth dataset when compared to the JRC for OFRs smaller than 5 ha (p-value <0.05). Both inundation datasets enabled us to estimate the seasonality in surface area change of OFRs, with the highest surface water extent between March-May, the months when the region receives most of the annual precipitation. Our results showed that both DSWE and JRC can be used to enhance hydrological assessments in poorly monitored basins that have a concentration of OFRs.