| At present, soil salinization and alkalization has become a globally ecological environmental issue. Soil salinization would not only arise soil hardness and decline fertility, but also cause resources destruction and losses in agricultural production,which seriously limit agricultural exploitation and agricultural sustainable development and seriously threaten biosphere and ecological environment.The Yellow River Delta High-Efficiency Eco-Economic Zone at the national level is the important middle-and-low-yielding fields and reserve land resources development zone in our country. Salinized land has a large area and distributed widely in Yellow River Delta,which will seriously restrict the agricultural production activity in this area.Accurately grasping soil salt condition and its dynamic change is of great significance for scientific guidance of salinized soil improvement,effectively prevent the secondary salinization land, reasonable development and utilization of salinized land resource, the improvement of land productivity, and the maintenance of ecological sustainable development.This study chose the typical areas of the Yellow River Delta as the study area.We built soil salinity estimation models by near-ground multispectral data and RS image data, respectively. Firstly, we chose the demonstration area of ‘Bohai Barn'project as the study area, which located in Wudi, Shandong Province, exploring the soil salinity estimation model based on near-ground multispectral in a small region.Several soil salinity estimation models were built by using correlation and regression analysis method conmbined with vegetation indices and actual measured soil salinity,from which the best estimation model for soil salinity estimation was selected.Secondly, in order to explore the applicability of the method, this research ideas and methods were applied to winter wheat planting area in the Yellow River Delta, and wecould get the best estimation model for soil salinity estimation. Then, using RS image in May 2015 and analyzing information features of each band and surface spectral features of research areas, we selected out sensitive bands and built soil salinity estimation model based on remote sensing imagery. And then, we got soil salinity of the study area by decision tree approach based on expert knowledge. Finally, this estimation model was fitted to the RS image of the whole Yellow River Delta winter wheat areas, and we got the spatial distribution map of soil salinity. By applying soil salinity estimation model based on RS image to the RS image of different phase in Kenli, Shandong Province, we got the soil salinity distribution maps and analyzed the dynamic change. Through the study, the main conclusions of this paper as follows:(1) Soil Salinity estimation models based on near-ground multispectral imagesThree vegetation indices(NDVI, SAVI and GNDVI) were used to build 18 models respectively with the actual measured soil salinity, from which the best estimation model for soil salinity estimation was selected. In two different scales, the best estimation models of soil salinity were all the linear models taking SAVI as the dependent variable. Soil salinity estimation model of the small region in the core demonstration area of ‘Bohai Barn' project is better than the model of a big region of winter wheat planting area in Binzhou and Dongying city, Shandong Province. In other words, the correlation coefficient of estimation model in the area of ‘Bohai Barn'project was the highest as 0.901 with a 93.36% accuracy, and the relative error was the smallest as 6.64%. The correlation coefficient and the estimation precision of estimation model in Binzhou and Dongying city is slightly smaller than the estimation model in the core demonstration area of ‘Bohai Barn' project, 0.864 and 81.44%,respectively. However, the relative error of estimation model in Binzhou and Dongying city was larger than the estimation model in the core demonstration area of‘Bohai Barn' project, 18.56%. In general, the accuracy of estimation model would be higher, if the study area was smaller and there are more samples in the same area.(2) Soil salinity estimation models based on RS imageWe extracted soil salinity information by using RS image. By selecting sensitive bands, spectral index was built, and then soil salinity estimation model based onnear-infrared band and middle-infrared band was built. The overall accuracy of estimation model was 93.04% and coefficient of Kappa was 0.7869, it well reflected soil salinity status of the study area.(3) Remote sensing application of estimation modelIn general, the distribution of winter wheat in the study area was significantly decreased from the southwest inland to the northeast coast; soil salinity of the catcher was mainly around 1.5 g · kg-1-3.0 g · kg-1, accounting for 76.90% of the total cultivated area; the catcher with salt content of less than 1.5 g·kg-1 accounted for relatively small, less than 10% of the total area of Winter Wheat; and the catcher with a salt content more than 3.0 g·kg-1 accounted for 14.09%.By applying estimation model based on RS image to RS image of Kenli County in 2000,2005,2011and2015, we got soil salinity distribution map in different periods.The statistical analysis indicated: the distribution of soil salinity in the Kenli County increased from the southwest inland to the east coast. From 2000 to 2015, soil salinity decreased generally in Kenli County. Among them the change of the content of heavy salinized soil and saline soil is the largest. The percentage of moderate saline soil increased from 17.17% in 2000 to 24.44% in 2015 and the percentage of saline soil decreased from 41.54% in 2000 to 32.43% in 2015. The percentage of heavy saline soil always remains at 25% or so. Therefore, the heavy saline soil has the buffer function between them. Soil salt content change trend of moderately saline soil and saline soil dominated the whole change trend of soil salt content within the county. |
PDF Full Text Request