Salinity modeling by remote sensing in central and southern Iraq

Salinization, leading to a significant loss of cultivated land and crop production, is one of the most active land degradation phenomena in the Mesopotamian region in Iraq. The objectives of this study (under the auspices of ACIAR and Italian Government) are to investigate the possibility to use remote sensing technology to establish salinity-sensitive models which can be further applied to local and regional salinity mapping and assessment. Case studies were conducted in three pilot sites namely Musaib, Dujaila and West Garraf in the central and southern Iraq. Fourteen spring (February - April), seven June and four summer Landsat ETM+ images in the period 2009-2012, RapidEye data (April 2012), and 95 field EM38 measurements undertaken in this spring and summer, 16 relevant soil laboratory analysis result (Dujaila) were employed in this study. The procedure we followed includes: (1) Atmospheric correction using FLAASH model; (2) Multispectral transformation of a set of vegetation and non-vegetation indices such as GDVI (Generalized Difference Vegetation Index), NDVI (Normalized Difference Vegetation Index), EVI (Enhanced Vegetation Index), SAVI (Soil Adjusted Vegetation Index), SARVI (Soil Adjusted and Atmospherically Resistant Vegetation Index), NDII (Normalized Difference Infrared Index), Principal Components and surface temperature (T); (3) Derivation of the spring maximum (Musaib) and annual maximum (Dujaila and West Garraf) value in each pixel of each index of the observed period to avoid problems related to crop rotation (e.g. fallow) and the SLC-Off gaps in ETM+ images; (4) Extraction of the values of each vegetation and non-vegetation index corresponding to the field sampling locations (about 3 to 5 controversial samples very close to the roads or located in fallow were excluded); and (5) Coupling remote sensing indices with the available EM38 and soil electrical conductivity (EC) data using multiple linear least-square regression model at the confidence level of 95% in a stepwise (forward) manner. The results reveal that soil salinity and EM38 readings are negatively correlated with the different vegetation indices, especially, GDVI and NDVI, and positively correlated with T. The models obtained for the pilot sites are presented in Table 1. Although we are still waiting for more laboratory analytical result and satellite imagery for more comprehensive analysis, it is clearly possible to build up salinity models by remote sensing, on which further salinity mapping and assessment can be based. It is also noted that among all the vegetation indices, GDVI is the best salinity indicator followed by NDVI and T. RapidEye image shows lower correlation with EM38 measurements and EC because fallow and crop rotation issue cannot be sorted out by one acquisition image.Table 1: Salinity models obtained from the pilot sitesNote: EMV- Vertical reading of EM38, EC - Electrical conductivity in dS/m