Linear spectral unmixing to monitor crop growth in typical organic and inorganic amended arid soil

Abstract

The soils of the GCC countries are dominantly sandy which is typical of arid regions such as the Arabian Peninsula. Such soils are low in nutrients and have a poor water holding capacity associated with a high infiltration rate. Soil amendments may rehabilitate these soils by restoring essential soil properties and hence enable site revegetation and revitalization for crop production, especially in a region where food security is a priority. In this study, two inorganic amendments; AustraHort and Zeoplant pellet, and one organic locally produced compost were tested as soil amendments at the experimental field of the International Center for Biosaline Agriculture in Dubai, UAE. The main objective is to assess the remote sensing ability to monitor crop growth, for instance Okra (Abelmoschus esculentus), having these amendments, as background with the soil. Three biomass spectral vegetation indices were used namely; NDVI, TDVI and SAVI. Pure spectral signatures of the soil and the three amendments were collected, using a field spectroradiometer, in addition to the spectral signatures of Okra in two growing stages (vegetative and flowering) in the field with a mixed F.O.V of the plant and amended soil during March and May 2015. The spectral signatures were all collected using the FieldSpec® HandHeld 2 (HH2) in the spectral range 325 nm - 1075 nm over 12 plots. A set of 4 plots were assigned for each of the three amendments as follow: three replicates of a 1.5 by 1.5 meter plot with 3kg/m2 of each amendment and 54 plants, one plot as control and all plots were given irrigation treatments at 100% based on ETc. Spectra collected over the plots were inversed in the range of 400-900 nm via a Linear Mixture Model using pure soil and amendments spectral signatures as reference. Field pictures were used to determine the vegetation fraction (in term of area of the F.O.V). Hence, the Okra spectral signatures were isolated for all plots with the three types of amendments. The three vegetation indices were then calculated and compared in the vegetation fraction of the entire F.O.V. The key outcome of this analysis was that a considerable bias was induced when using a mixed F.O.V. In fact, the compost as an organic soil amendment containing dead vegetation affects similarly the sensitivity of the three vegetation indices in the vegetative stage of Okra compared to AustraHort and Zeoplant pellet amended plots. However, the TDVI is very sensitive to vegetation presence even with unmixed crop spectra. AustraHort amendment led to better status of Okra both in March and May with values of 0.19 and 0.28 respectively. Bias induced by some soil amendments can be misleading when upscaling spectral information to satellite imagery with low spectral and spatial resolutions. The results obtained are encouraging for further use of spectral information for crop monitoring in soil containing amendments.