Mapping Soil Salinity Using a Combined Spectral Response Index for Bare Soil and Vegetation: A Case Study in the Former Lake Texcoco, Mexico.
Norma Fernandez-Buces1, Christina Siebe Grabach1, José Luis Palacio2, and Silke Cram3. (1) Instituto de Geología, Univ Nacional Autónoma de México, Circuito escolar, Ciudad Universitaria, s/n, Mexico City, Mexico, (2) Dirección General de Estudios de Posgrado, Uinv Nacional Autónoma de México, Circuito interior, Ciudad Universitaria, s/n, Mexico City, Mexico, (3) Instituto de Geografía, Univ Nacional Autónoma de México, Circuito exterior, Ciudad Universitaria, s/n, Mexico City, Mexico
Salinization is one of the main causes of soil degradation in arid and semiarid regions around the world. Salt-affected landscapes are highly sensitive to changes in climatic, edaphic and hydrological conditions in time and space, therefore their characterization and mapping is difficult because the salt concentration may vary seasonally and substantially over short distances. Remote sensing is widely used to lower survey costs, but existing studies usually analyze bare soils and make little reference to the halophytic plants and their role as salinity indicators. Our work aims to correlate soil characteristics (Electric Conductivity in saturation extract (ECe) and Sodium Absorption Ratio (SAR) with the spectral response of plant species and bare soils, integrating an algorithm to allow multi-scale mapping using remote sensors. Ground radiance was measured on different plant species and bare soils, using a Milton Multiband Radiometer sensing three bands in the visible spectrum and one in the Near InfraRed (NIR). A Combined Spectral Response Index (COSRI) was calculated for bare soils and vegetation by adjusting the Normalized Difference Vegetation Index (NDVI). ECe and SAR were determined in surface soil samples. Correlation coefficients between COSRI and soil salinity were obtained and a model was adjusted to predict soil salinity. Landsat-ETM and airborne digital images were used to calculate raster maps of COSRI, and ECe and SAR distribution maps were estimated using the adjusted models. Almost the entire study area is saline-alkaline, with very large variation in salt concentrations over short distances (<10 m). The variation on surface soil salinity (0-15 cm) is very large, from extremely large electric conductivities (ECe) (1 319 dS*m-1), due to the presence of saline crusts, to moderate ECe (11.7 dS*m-1). Salt emergences form crusts of different colors and patterns. Saline grassland is the predominant type of vegetation in the study area. Species dominances (D.I.) and their associations correspond to particular levels of salinity, with Distichlis spicata as the most dominant species at sites severely affected by salts, followed by Suaeda torreyana. On the other hand, Eragrostis mexicana (Hor.) Link, Tamarix gallica L, Hordeum jubatum L and Chaenopodium album L are dominant at sites moderately affected; whereas Bidens odorata Cav, Chloris virgata Swartz, Sonchus oleraceus L and Lepidium virginicum L are dominant at sites slightly affected. A Combined Spectral Response Index (COSRI) was calculated based on the analysis of spectral responses of different bare soil patterns and colors and vegetation cover an dominant species presence. Such index correlated significantly with soil ECe and SAR (–0.885 and –0.857, respectively). Dense vegetated areas yield large index values because of their relatively large NIR reflectance and small reflectance in the visible. Less dense vegetated showed lower index values. In contrast, water, clouds and salt affected soil have larger visible reflectance than in the NIR, thus, they yield negative index values. Rocks and slightly salt affected bare soils give similar reflectance in both bands, resulting in COSRI values near cero. Medium (1:70,000) and large (1:10,000) scale maps were obtained using Landsat and airborne images, respectively. COSRI raster maps were calculated with images spectral bands, and two exponential models were used to estimate surface soil EC and SAR values for each pixel. Variance accounted for by exponential models for ECe and SAR was of 0.826 and 0.751, respectively. The accuracy of the results for predicted electric conductivity and SAR were tested against results for an independent set of soil samples. It may be concluded that the method is an easy, low-cost procedure to map salt-affected areas.