Deriving seasonal dynamics in ecosystem properties of semi-arid savanna grasslands from in situ-based hyperspectral reflectance

摘要

This paper investigates how hyperspectral reflectance (between 350 and 1800 nm) can be used to infer ecosystem properties for a semi-arid savannagrassland in West Africa using a unique in situ-based multi-angular data setof hemispherical conical reflectance factor (HCRF) measurements.Relationships between seasonal dynamics in hyperspectral HCRF and ecosystemproperties (biomass, gross primary productivity (GPP), light use efficiency(LUE), and fraction of photosynthetically active radiation absorbed byvegetation (FAPAR)) were analysed. HCRF data (ρ) were used to study therelationship between normalised difference spectral indices (NDSIs) and themeasured ecosystem properties. Finally, the effects of variable sun sensorviewing geometry on different NDSI wavelength combinations were analysed. Thewavelengths with the strongest correlation to seasonal dynamics in ecosystemproperties were shortwave infrared (biomass), the peak absorption band forchlorophyll a and b (at 682 nm) (GPP), the oxygen A band at 761 nm usedfor estimating chlorophyll fluorescence (GPP and LUE), and blue wavelengths(ρ₄₁₂) (FAPAR). The NDSI with the strongest correlation to (i)biomass combined red-edge HCRF (ρ₇₀₅) with green HCRF (ρ₅₈₇),(ii) GPP combined wavelengths at the peak of green reflection (ρ₅₁₈,ρ₅₅₆), (iii) LUE combined red (ρ₆₈₈) with blue HCRF(ρ₄₃₆), and (iv) FAPAR combined blue (ρ₃₉₉) and near-infrared(ρ₁₂₉₅) wavelengths. NDSIs combining near infrared and shortwaveinfrared were strongly affected by solar zenith angles and sensor viewinggeometry, as were many combinations of visible wavelengths. This studyprovides analyses based upon novel multi-angular hyperspectral data forvalidation of Earth-observation-based properties of semi-arid ecosystems, aswell as insights for designing spectral characteristics of future sensors forecosystem monitoring.