ABSTRACT
R/N Regional or national RCGb Index related to soil weathering RID Re¨exion in¨exion di¦erence RK Regression-kriging RMSEP/RMSECV Root mean square error of prediction and
cross-validation
RS Remote sensing RT Regression tree SAR Synthetic aperture radar SEP/SECV Standard error of prediction or
cross-validation Si Silicon SIC Soil inorganic carbon SL Spectral library SM Soil moisture SMA Spectral mixture analysis SMAP Soil moisture active passive SNR Signal-to-noise ratio SOC Soil organic carbon SOM Soil organic matter Sr Strontium SS Spectral sensing SSS Space-Aerial and Orbital-Spectral
Sensing SVMR Support vector machine regression SWIR Shortwave infrared TC Total carbon tbd to be demonstrated TIR Ÿermal infrared UV Ultraviolet UV-VIS-NIR 250-2500 nm (the exact spectral range in
the individual studies may deviate but will stay within these ranges)
V Vanadium VIS Visible VIS-NIR 350-2500 nm λ Wavelength in nm ν and δ CoKriging Energy levels of fundamental vibration in
microscopic interactions γ Gamma
Industrial and agricultural activities are developing faster than the public policy on the use of soil resources. Ÿe world needs more information about soil for land use planning and interpretative purposes. Spectral sensing (SS) has emerged as a major discipline in remote sensing (RS) science in the past years providing important tools to assist in soil information gathering, mapping, and monitoring. Ÿis chapter aims to discuss the role of SS (covering the visible, infrared, thermal, microwave, and gamma ranges of the spectrum) in soil science based on di¦erent sensors, scales, and platforms (laboratory, ¤eld, aerial, and orbital). We review the state of the art and provide guidance on how to use SS for several purposes, for example, soil classi¤cation and mapping, attribute quanti¤cation, soil management, conservation, and monitoring. Research has shown that SS has the capability to quantify soil attributes, such as clay, sand, soil organic matter (SOM), soil organic carbon (SOC), cation exchangeable capacity (CEC), Fe2O3, carbonates, and mineralogy with reliable and repeatable results. Other soil attributes including pH, Ca, Mg, K, N, P, and heavy metals have also been evaluated with variable outcomes. Laboratory and ¤eld-based
measurements are more accurate than aerial or space-based measurements as they are conducted under more controlled environments that are less a¦ected by external factors, such as mixing in the ¤eld-of-view, vegetation cover, stone cover, water content, and atmospheric conditions. Nevertheless, soil is typically evaluated from space using multispectral sensors on board satellites, which o¦er many options in terms of temporal and spatial coverage and resolution, and are commonly available free of charge. On the other hand, hyperspectral images are less commonly applied due to their more limited choices of temporal and spatial resolutions and di¶culty of processing, despite their great potential to correlate with various soil properties. Other SS techniques, such as passive gamma spectroscopy, provide data for surface and below-surface soil inference, primarily relating to the clay content and types of soil minerals, while microwave (i.e., radar) spectroscopy is mainly used in the study of soil moisture. In soil science, there are promising results and growing interest for visible-near-infrared (VIS-NIR) and middle infrared (MIR) spectroscopy as they allow quick, nondestructive, and cost-e¦ective estimation of soil properties, reducing the need for sample preparation and the use of reagents, minimizing pollution. It has been observed that MIR spectroscopy can quantify properties, such as clay, clay-sized mineralogy, SOC, and inorganic carbon (C), more accurately than VIS-NIR. Both physical (descriptive interpretation of spectral information) and statistical (mathematical approach) methods proved to be useful depending on soil and environmental conditions under study. We observed that the most important limitation of VIS-NIR and MIR spectroscopy for soil classi¤cation are their inability to detect soil morphological properties (e.g., soil structure). In the case of VIS-NIR space SS, the limitation is that the radiation only penetrates a few centimeters into the soil surface. On the other hand, satellite-based VIS-NIR data can be used for delineation of soil boundaries supporting soil survey and mapping. SS applicability is also increasing in precision agriculture (PA), coupled with on-the-go sensors that measure soil properties with high sampling density and in real time. Future advances in SS include (1) extraction of moisture e¦ects from intact and ¤eld moist spectra, allowing a comparison with laboratory measurements; (2) development of local, regional, or global soil spectral libraries and their appropriate use; and (3) combining multiple sources of sensed data for better soil inference. Country-based soil spectral libraries started in the early 1980s and today we are moving toward a global spectral library (SL) with contribution from as many as 90 countries. Soil spectral libraries, from global to local, will be the future of soil analysis carrying both spatial and hyperspectral data to derive soil information. SS has the advantage of providing quantitative data, and thus reducing the subjectivity of soil spatial information for decision making. SS techniques are powerful when combined with geoprocessing, landscape modeling, geology, and geomorphology. Ÿe past and new studies on soil ground SS indicate strong information with a great perspective on all SS platforms, specially for existing hyperspectral aerial and orbital sensors and new ones that are being developed and will be launched soon (2017-2020).
