Soil Salt Content Retrieval from ALI Multispectral Image Based on GRNN

Soil salinization has been a major progress of soil degradation due to environment evolution and anthropological activities.And therefore,it is of great significance to carry out timely detection of soil salinization,aiding the reclamation and prevention of saline soils.In this paper,a typical subregion of Yellow River Delta was studied for soil salt content(SSC) retrieval.For this purpose,generalized regression neural network(GRNN) were performed based on soil samplings and corresponding laboratory physico-chemical analysis.The model shows a good relationship between the measured and predicted SSC with determination coefficient(R2) of 0.855,and root mean square error(RMSE) of 0.119.Subsequently,the model was applied to ALI reflectance image to yield the SSC map.A comparison shows that GRNN-derived SSC is more likely to accord with ground observations.