Detecting soil salinization in arid regions using spectral feature space derived from remote sensing data

Soil salinization,especially secondary soil salinization caused by irrigation activities is one of the primary ecological and environmental concerns in the arid and semi-arid regions of China. A critical research question is to quickly and accurately monitor soil salinization in arid and semi-arid regions so that prevention strategies can be developed quickly and deployed efficiently. Traditional techniques based on field soil sampling and laboratory experiments,though could be rather accurate for the sampling sites and their immediate vicinity,can hardly produce real-time evaluation. Monitoring soil salinization using remotely sensed imageries,however,starts to attract scholarly attention during the past decades due to the almost real-time information collection,vast geographic coverage,and rich information contained in the remotely sensed imageries. The current study is an attempt to employ remote sensing technique to monitor soil salinization in the arid /semiarid Weigan-Kuqa Delta Oasis region in Xinjiang,Western China. Data are collected from Landsat-TM and Landsat-ETM+multiple-spectral remote sensing imageries. Field soil samples at the 0—10 cm depth are obtained for validation purposes aswell. The study intends to establish a statistical relationship between the degrees of soil salinization and surface biophysical reflective characteristics that are captured by the remote sensing imageries. Spectral un-mixing analysis of the multispectral imageries produces three groups of commonly used spectral information for soil salinization monitor and evaluation,i. e.,individual spectra that are sensitive to soil salinization,that can be used to derive vegetation cover,and that can be used to derive soil moisture contents. The study then combines these groups of information establish three two-dimensional and two three-dimensional soil salinization monitoring indices. The three two dimensional indices include: Vegetation fraction and Soil Index( VSSI),Soil water contents and Vegetation fraction Soil Index( SVSI) and Soil water contents and Soil salinization fraction Soil Index( SSSI). The two three-dimensional soil salinization monitoring indices include: Soil salinization fraction-vegetation fraction-Water contents Soil Index( SVWSI) and Soil Distance Index( SDI). Statistical analyses using these obtained two dimensional and three dimensional indices with field soil sample data are conducted as well. The result suggests that all the indices are able to provide sufficient monitoring and evaluating performance of the severity of soil salinization in our designated study region. Three dimensional indices,however,tend to be more sensitive to soil salinization than the two dimensional indices. In particular,SVWSI and SDI are highly correlated with soil salt contents at the 0—10 cm depth,with correlation coefficients of R2= 0. 8325 and R2= 0. 8646,respectively. The result suggests that higher dimensional indices derived from remote sensing imageries might provide more accurate soil salinization monitoring measurements than lower dimensional indices due to enriched information structure. Since obtaining spectral information from remote sensing imageries is relatively straightforward and is often either real-time of near real-time,our suggests that rich information that can be derived from remotely sensed imageries shall be of invaluable importance to provide real-time and accurate evaluation and monitor for soil salinization monitoring and evaluation might provide timely strategies that can mitigate or even prevent further soil salinization in arid and semi-arid regions.