Time series analysis of remotely sensed data as a tool for vegetation evaluation and soil-water interactions: Implications for digital soil mapping

Abstract

Soil, as a synthetic body, is the result of complex environmental interactions occurring across time and different geographic scales. Its functions are critical to maintaining ecosystem services, such as water redistribution and water storage, nutrient cycling, food production, carbon storage and sequestration, and climate regulation. The knowledge of soil geographic distribution and its relation with landscape dynamics are of paramount importance to sustaining such functions. Digital soil mapping (DSM) methods are intended to solve spatial association models that relate geographic occurrence of soil to soil-forming factors, namely: parent material, topography, climate, organisms, and time. Remotely sensed data are essential to parametrize such models, since they offer measurements of land surface features, both in time and geographic frames. However, despite the good performance of the assimilation of data from optical remote sensors with laboratory analysis and field data into DSM, accessing soil properties under vegetation cover via remote sensing methods still represents a challenging task. Therefore, this research intended to evaluate a method to assess the response of vegetation greenness (VG) to soil condition based on the following hypotheses: a) Modulation of surface reflectance of vegetation is controlled by soil condition and properties related to water dynamics, b) this feature can be measured based on remotely sensed data and a time-spectral signature of vegetation response associated to mentioned soil properties can, therefore, be retrieved, c) elucidation of this relationship can be applied to produce time-synthetic covariates for DSM. Although the initial hypothesis was partially verified statistically, VG temporal signal may not reflect exclusively the effects of water availability and other factors can act as vegetation ‘stressors’ affecting its spectral properties, such as the interaction of soil fertility, toxicity, and taxonomic class. Results of this approach demonstrated that the addition of seasonal variability of vegetation greenness can be applied to access soil subsurface processes, as well as their use as covariates in DSM.