Sensitivity Analysis of Chlorophyll Indices to Soil Optical Properties Using Ground-Reflectance Data

In precision agriculture, crop nitrogen status could be estimated based on the measurement of leaf chlorophyll content at specific stages of crop development. Over the last decade, several spectral chlorophyll indices have been developed to estimate chlorophyll content both at the leaf and the canopy level using hyperspectral remote sensing data and considering different crop types. For an accurate interpretation of chlorophyll indices derived from hyperspectral data, a "true" chlorophyll content value attributed only to the crop cover signal and free from any non-photosynthetic elements is required. However, in remote sensing, in spite of the correction and the standardization of the various radiometric distortions (topography, atmosphere, sensor drift, BRDF, etc.), the chlorophyll indices remain always sensitive to the artifacts caused by the soil optical properties particularly in an earlier stage of crop growth. This paper focuses on the evaluation and comparison of the sensitivity of several chlorophyll indices (PRI, NDPI, GNDVI, hNDVI, SIPI, SRPI, NPCI, PSSRa, PSNDa, OSAVI, CARI, MCARI and TCARI) to bare soil optical properties variation. In order to achieve the goal of this investigation, spectroradiometric measurements were acquired above 120 bare soil plots with various optical properties and selected from different agricultural lands. The results show that SIP[, SRPI, PSSRa, NDPI, NPCI and GNDVI indices have non-negligible RMSE related to the optical properties of bare soils, and will be very difficult to interpret at low leaf area index (LAI). The PSNDa, OSAVI and hNDVI show an RMSE less than 10%. However, this error remains significant. The PRI, CARI, MCARI and TCARI are basically not sensitive to changes in the soil optical properties (RMSE less than 2%) and permit a better estimation of chlorophyll content in sparse crop cover environment independently from the bare soil background.