Runoff information and its dynamics are critical for supporting watershed management; however, spatio-temporal data about runoff is rare or unavailable in data-scarce regions. Information about the performance of remote sensing-based runoff and its potential application is limitedly known. In data-scarce regions, this condition impedes comprehensive watershed assessment especially in the midst of climate change impacts. This study examined the performance of globally available monthly runoff dataset provided by TerraClimate at ~ 4 km spatial resolution and employed them to assess the runoff dynamics in a humid tropic watershed. Monthly TerraClimate data shows a moderate performance with an r of 0.63, RMSE of 57–127 mm/month and NRMSE of 18–30% to the simulated runoff from a well-calibrated model. The upper region of Brantas watershed was found to be the hotspot of high runoff. About 25% of the study area belongs to high runoff (0–33rd percentile). Over the last two decades, runoff has been slightly increased across the study area. Green vegetation fraction (GVF), precipitation, and topography are critical for regulating runoff dynamics. While topography and precipitation impact on runoff are straightforward, the GVF’s role is complex and site-specific. High runoff was found mostly to be associated with high precipitation and steep slope. GVF appears to be less effective in representing ground cover against runoff generation due to high variability of actual ground cover types. Using time-series and change vector analysis (CVA) of runoff and GVF, the dynamics of watershed condition was examined. Long-term CVA analysis also found that the condition in Brantas watershed was fluctuated with slight increase in impaired condition. The study exemplified the potential use of the remote sensing-based runoff data in a tropical data-scarce region. Despite limitation of the runoff data due to its moderate performance, the globally available monthly runoff data from TerraClimate can be used to support regional water resource assessment in data-scare regions. Future improvement that includes downscaling and use of machine learning can be considered to improve the remotely sensed runoff data to deliver the bigger benefits of such data.
Read full abstract