Improving the monitoring of sugarcane residues in a tropical environment based on laboratory and Sentinel-2 data

Abstract

ABSTRACT Retaining a residue cover (RC) over the soil is the key to ensure its health. Since tropical agriculture systems, mostly in sugarcane, are characterized by intensive management, the discrimination of residue types becomes difficult and aims to be detected. Our goal was to identify features regarding sugarcane residues using hyperspectral and simulated Sentinel-2 dataset. Therefore, we collected green leafs (GL), straw (ST) and burn material (BM), added to two classes of contrasting soils, Ferralsols (S1) and Arenosols (S2). We developed control samples (GL, ST, BM, and soil) and also treatments with a combination of soil and RC with different dosages. We used the FieldSpec (350 to 2500 nm) to obtain the spectral data to each sample. The spectral data were subsequently used for the simulation of Sentinel-2 bands. GL presented features related to water absorption (1400 and 1900 nm) and pigments in the visible region. ST does not present features relating to pigments, neither water absorption. Due to the increase of BM, we observed a decreasing in the intensity of reflectance. Regardless the type of RC, soil features (kaolinite and mineralogy) were lost when more than 70% coverage was reached. The soil background affects the reflectance in all treatments, brighter soils influenced the spectral in the vis region. Sentinel-2 presented a strong positive correlation in the RedEdge (band 5) to assess green vegetation, in the Short-wave infrared (SWIR) to non-photosynthetically active material and in the blue band to BM. The Cellulose Absorption Index (CAI) combined with Normalized Difference Vegetation Index (NDVI) and reflectance in the red could assist in distinction of the tree RC in mixed systems. We provided an interesting spectral analysis that could assist in futures studies to improve indices on the task for soil management.