Assessment of remotely-sensed canopy measurements for mapping soil nutrient deficiencies in Mexico croplands

Abstract:

Soil fertility plays a critical role in agro-ecosystems by providing the foundation for crop water and nutrient uptake. Nutrient deficient or otherwise degraded soils may result in lower yielding crops and can reduce the effectiveness of applied fertilizers. Adapting management practices to account for local soil deficiencies remains difficult, especially in smallholder systems, where widespread collection of soil and plant tissues is often cost and time prohibitive. However, satellite remote sensing imagery offers the opportunity to collect high resolution data with global coverage. Here we explore satellite-based methods for detecting soil and crop nutrient deficiencies in North-Central Mexico, using 10 – 20 meter resolution data from the ESA’s Sentinel-2 Multi Spectral Instrument Level-1.

Typical vegetation indices used for crop yield prediction, such as the Normalized Difference Vegetation Index (NDVI) or Green Chlorophyll Index (GCI) are not highly sensitive to leaf chlorosis caused by nutrient deficiency in the soils. We therefore assess a set of vegetation indices derived from Sentinel-2’s Red Edge bands and designed for detecting leaf chlorosis for their ability to explain variation in soil fertility and crop yields. These include the Simplified Canopy Chlorophyll Content Index (SCCCI), Triangular Chlorophyll Index / Optimized Soil Adjusted Vegetation Index ratio (TCI/OSAVI), and the Transformed Chlorophyll Absorption Reflectance Index / OSAVI ratio (TCARI/OSAVI).Vegetation index values during the growing season peak were sampled at ground-collected soil samples and stratified by the top and bottom 10% of various soil features. Fields in the upper quantile of organic matter, zinc, and phosphorous showed a consistently higher value of SCCCI, TCI/OSAVI, and TCARI/OSAVI than those in the bottom 10%. We also find the SCCCI shows moderate correlation (r = 0.28) with the residuals from a yield prediction model based on GCI and weather. This relationship indicates the potential of the red edge vegetation indices to add information related to plant nutrient deficiency to satellite-based yield models. Our findings show the use of satellite data to accurately map soil deficiencies in a way that could potentially be applied at a global scale, regardless of the availability of ground-collected samples.