Gap Filling Cloudy Sentinel-2 NDVI and NDWI Pixels with Multi-Frequency Denoised C-Band and L-Band Synthetic Aperture Radar (SAR), Texture, and Shallow Learning Techniques

Multispectral imagery provides unprecedented information on Earth system processes: however, data gaps due to clouds and shadows are a major limitation. Normalized-Difference Vegetation Index (NDVI) Water (NDWI) two spectral indexes employed for monitoring vegetation phenology, land-cover change more. Synthetic Aperture Radar (SAR) with its cloud-penetrating abilities can fill using coincident imagery. In this study, we evaluated C-band Sentinel-1, L-band Uninhabited Aerial Vehicle SAR (UAVSAR) texture gap filling efficient machine learning regression algorithms across three seasons. Multiple models were including Support Vector Machine, Random Forest, Gradient Boosted Trees an ensemble of models. The Gap ability was Sentinel-2 from the same date, 3 days 8 later than both sensors in September. Sentinel-1 winter spring seasons also evaluated. Because contains noise, compared robust de-noising methods performance against refined lee speckle filter. Mean Absolute Error (MAE) rates cloud gap-filling model assessed different dataset combinations land covers. results indicated de-noised UAVSAR GLCM provided highest predictive random forest R2 = 0.91 (±0.014), MAE 0.078 (±0.003) 0.868 (±0.015), 0.094 during errors observed bare ground forest, while lowest herbaceous woody wetland. Results January June without less strong at 0.60 (±0.036), 0.211 (±0.005) (NDVI), 0.61 (±0.043), 0.209 0.72 (±0.018), 0.142 (±0.004) 0.77 (±0.022), 0.125 June. Ultimately, suggest metrics accurately predict NDVI NDWI most These shallow rapidly trained applied faster intensive deep or time series methods.