Seasonal variation in vegetation water content estimated from proximal sensing and MODIS time series in a Mediterranean Fluxnet site

Abstract

Abstract. This study evaluates three different metrics of vegetation water content estimated from proximal sensing and MODIS satellite imagery: Fuel Moisture Content (FMC), Equivalent Water Thickness (EWT) and Canopy Water Content (CWC). Dry matter (Dm) and Leaf area Index (LAI) were also analyzed in order to connect FMC with EWT and EWT with CWC, respectively. This research took place in a Fluxnet site located in Mediterranean wooded grassland (dehesa) ecosystem in Las Majadas del Tietar (Spain). Results indicated that FMC and EWT showed lower spatial variation than CWC. The spatial variation within the MODIS pixel was not as critical as its temporal trend, so to capture better the variability, fewer plots should be sampled but more times. Due to the high seasonal Dm variability, a constant annual value would not work to predict EWT from FMC. Relative root mean square error (RRMSE) evaluated the performance of nine spectral indices to compute each variable. VARI provided the worst results in all cases. For proximal sensing, GEMI worked best for both FMC (RRMSE = 34.5%) and EWT (RRMSE = 27.43%) while NDII and GVMI performed best for CWC (RRMSE =30.27% and 31.58% respectively). For MODIS data, results were a bit better with EVI as the best predictor for FMC (RRMSE = 33.81%) and CWC (RRMSE = 27.56%) and GEMI for EWT (RRMSE = 24.6%). To explain these differences, proximal sensing measures only grasslands at nadir view angle, but MODIS includes also trees, their shades, and other artifacts at up to 20° view angle. CWC was better predicted than the other two water content variables, probably because CWC depends on LAI, which is highly correlated to the spectral indices. Finally, these empirical methods outperformed FMC and CWC products based on radiative transfer model inversion.