Studying Vegetation Salinity: From the Field View to a Satellite-Based Perspective

Abstract

Salinization of irrigated lands in the semi-arid Jezreel Valley, Northern Israel results in soil-structure deterioration and crop damage. We formulated a generic rule for estimating salinity of different vegetation types by studying the relationship between Cl/Na and different spectral slopes in the visible–near infrared–shortwave infrared (VIS–NIR–SWIR) spectral range using both field measurements and satellite imagery (Sentinel-2). For the field study, the slope-based model was integrated with conventional partial least squares (PLS) analyses. Differences in 14 spectral ranges, indicating changes in salinity levels, were identified across the VIS–NIR–SWIR region (350–2500 nm). Next, two different models were run using PLS regression: (i) using spectral slope data across these ranges; and (ii) using preprocessed spectral reflectance. The best model for predicting Cl content was based on continuum removal reflectance (R2 = 0.84). Satisfactory correlations were obtained using the slope-based PLS model (R2 = 0.77 for Cl and R2 = 0.63 for Na). Thus, salinity contents in fresh plants could be estimated, despite masking of some spectral regions by water absorbance. Finally, we estimated the most sensitive spectral channels for monitoring vegetation salinity from a satellite perspective. We evaluated the recently available Sentinel-2 imagery’s ability to distinguish variability in vegetation salinity levels. The best estimate of a Sentinel-2-based vegetation salinity index was generated based on a ratio between calculated slopes: the 490–665 nm and 705–1610 nm. This index was denoted as the Sentinel-2-based vegetation salinity index (SVSI) (band 4 − band 2)/(band 5 + band 11).

Keywords

• reflectance spectroscopy
• spectral slope
• salinity
• fresh vegetation
• tomato
• cotton
• Sentinel-2
• Sentinel-2-based vegetation salinity index (SVSI)