Spatial patterns of water quality and remote sensing indices from UAV-based multispectral imagery across an irrigation pond

Water quality of irrigation water is an essential factor for public safety and farm sustainability. Imaging surface water sources from unmanned aerial vehicles (UAVs) has become an important source of water quality information. Water quality variables (WQVs) in irrigation ponds have been shown to have persistent spatial patterns. The objective of this work was to test the hypothesis that (a) persistent spatial patterns can be found in reflectance and remote sensing indices from UAV-based multispectral imagery of irrigation ponds, and (b) those patterns can significantly correlate with patterns of WQVs. We utilized data from sampling, in-situ sensing, and UAV-based imaging of a commercial 4-ha farm pond in Maryland. Seventeen water quality variables were measured on a permanent grid during the irrigation season concurrently with the imaging of the pond with the MicaSense RedEdge camera at five wavelengths. Twenty-four remote sensing indices were computed. Spatial patterns were determined using the mean relative difference method. The water quality patterns appeared to reflect differences in distances from banks, closeness to the creek meeting the pond, the degree of water stagnancy, dominant wind directions, and a geese congregation site. High (>0.8) Spearman correlation coefficients were found for turbidity, photosynthetic pigments, and organic carbon in water. These variables' patterns had similarities with patterns of remote sensing indices AFAI, TCARI, TCI, and MCARI. Patterns of E. coli strongly correlated with the pattern of reflectance at the red wavelength. Given the high spatiotemporal variability of WQVs in irrigation ponds, determining patterns of remote sensing indices can be useful for the design of surveys or monitoring important aspects of water quality.