LB 36-307 - Prediction of Soil Moisture on a Restored Wet Meadow Using In-situ and Remotely-sensed Data

The Monitoring of soil moisture (SM) on restored wetlands is an important metric in assessing the effectiveness of restoration on ecosystem resilience to climate change. However, due to the spatial and heterogeneous properties of SM, using in-situ methods for monitoring SM can be difficult. The use of remotely-sensed data acquired by multispectral images can be used to monitor SM at higher spatial resolutions and across a larger land cover. The application of remote sensing tools to estimate SM has been primarily used in agriculture systems, limiting our understanding of their accuracy in some other ecosystem types. In this study, data acquired from Landsat-9 and Blackswift E2 Unmanned Aerial System were used to predict SM on a restored wetland in the Gunnison Basin of Colorado. In-situ SM data including gravimetric SM (GSM) and volumetric SM (VSM) was collected at two time periods (August & September), and at different soil depths (GWC: 0 to 5 cm, 0 to 10 cm, and 0 to 20 cm; VWC: 0 to 7.6 cm and 0 to 20 cm). The prediction of SM was done using the linear regression method, with in-situ and remotely-sensed data as the response and predictor variables, respectively. Prediction accuracy was determined using the coefficient of determination values (R2) and the root mean square error values (RMSE). Between the two in-situ data used, GSM had the best fit for SM prediction compared to VSM. Among the Landsat-9 indices used as predictor variables, the best prediction of SM was by NSDSI2 at the depth of 0 to 5 cm (R2 = 0.77, RMSE = 0.18). Good predictions of SM by Landsat-9 indices (NSDSI2 &NDVI) were also made at the depths of 0 to 10 cm and 0 to 20 cm (R2 = 0.65, RMSE = 0.15; R2 = 0.73, RMSE = 4.09). The overall best prediction of SM was by the UAS-based index (NDREI) at the depth of 0 to 5 cm (R2 = 0.86, RMSE = 7.41). A good prediction of SM was also made at the depth of 0 to 10 cm by UAS-based NDVI (R2 = 0.70, RMSE = 0.14). The conclusions drawn from this study are that remotely-sensed data, particularly UAS vegetation indices, are potential and relevant tools for predicting SM on restored wetlands. This study provides land managers with information on the effectiveness of using RS tools to monitor the success of restoration initiatives in the Gunnison Basin.