Comparing Time Series of Venµs Surface Reflectance and Flat Reflectance Products for Land Surface Phenology Modeling in Highland Pastures

We compared the performance in land surface phenology (LSP) models using time series of two VENµS products—surface reflectance (SRE) and flat reflectance (FRE)—that were acquired during 2018 over the Naryn site in central Kyrgyzstan. We modeled the LSP as a convex quadratic (CxQ) function of accumulated growing degree-days calculated from MODIS 1 km land surface temperature data. We evaluated LSP models iteratively fitted to time series of five different vegetation indices: three variants of Gitelson’s chlorophyll index using two red-edge bands and the green band, the VENµS variant of the MERIS Terrestrial Chlorophyll Index, and the NDVI. Fitted separately for each of more than 11M pixels, the CxQ model yields two key phenometrics from the fitted parameter coefficients: Peak Height (PH—the maximum model VI) and Thermal Time to Peak (TTP—duration of the modeled green-up phase). Comparing SRE and FRE for the NDVI, we found minimal differences with < 0.5% of FRE models not fitting successfully relative to SRE. Differences in PH occurred over steep slopes (>25°). Differences in TTP were also minimal with most differences <|15 °C|. We will present results for the other VIs at the meeting.