Mapping soil salinity using a combined spectral and topographical indices with artificial neural network

Area of Science:

• Environmental Science
• Remote Sensing
• Soil Science

Background:

• Effective monitoring of natural resources, particularly soil, is crucial for conservation.
• Soil salinity poses a significant threat to agricultural productivity and ecosystem health.

Purpose of the Study:

• To monitor and model soil salinity in Qom plain using remote sensing and machine learning.
• To identify key environmental factors influencing soil salinity distribution.

Main Methods:

• Utilized Landsat 8 satellite imagery and Artificial Neural Network (ANN) for soil salinity mapping.
• Incorporated spectral indices (salinity, vegetation, topography, drainage) alongside soil samples for model calibration.
• Employed the Latin hypercube method for precise geographical sampling of 72 surface soil samples.

Main Results:

• The GFF algorithm demonstrated superior performance in modeling soil salinity.
• Topographic Wetness Index (TWI), Salinity Index (SI5), and Normalized Difference Vegetation Index (NDVI) were the most influential factors.
• Areas such as floodplains and lowlands exhibited the highest levels of soil salinity accumulation.

Conclusions:

• Remote sensing combined with ANN provides an effective approach for monitoring soil salinity.
• Understanding the influence of topography and vegetation is key to predicting soil salinity patterns.
• Identifying high-salinity zones can guide targeted land management and conservation strategies.