engleski

Soil salinity assessment at landscape level using difuse reflectance spectroscopy and geostatistics

Soil salinization caused by natural or human-induced processes is a worldwide environmental hazard impacting more than 100 countries. In Europe this problem is most pronounced in the Mediterranean basin region including the coastal region of Croatia where seawater intrudes through porous media into calcareous aquifers and salinizes both ground and surface water. This is especially evident in the Neretva River Valley (with over 6000 ha of irrigated land), located in southeastern Croatia. Agricultural production in this region, which is very important for citrus and leafy vegetable production in Croatia, is becoming more affected by periodical or occasional soil and water salinization. Identification and monitoring of the process of salinization of agricultural soils represent the basic, but also the most expensive part of the saline soil management. Spectral analysis combined with chemometrics can provide low-cost and high-density soil data for predicting various soil properties. However, a relatively weak correlation between the acquired spectra and the measurements of salinized soil properties makes spectroscopy difficult to use in salinity assessment, especially for low and moderately saline soils. It would therefore be useful if reflectance spectroscopy, combined with geostatistics, could be used as an efficient method for monitoring and mapping of soil under salinity risk. The aim of this work was to study the usefulness of a spectral indicator, derived from radiometric data, for mapping spatial distribution of soil salinity at landscape level. Using an intensive soil survey in the Neretva River valley (conducted at 1:25000 map scale), consisting of 245 agricultural topsoil (0-25 cm) samples, we assessed the feasibility of reflectance spectroscopy (RS) for describing topsoil salinity and its spatial variability across the landscape. Reflectance of air-dried and sieved (<2 mm) soil samples was measured in 2 nm bandwidths over 350- 2500 nm range with a UV/VIS spectrophotometer equipped with an integrating sphere (Jasco V-560). To summarize the information content of the spectra, a partial least squares regression analysis (PLSR) was performed using electrical conductivity (EC) as a response variable and the spectra (restricted to 380-2430 nm and transformed into absorbance) as predictors. The EC values showed a large departure from normal and were transformed into Gaussian shaped variables. The PLSR was carried out on the mean centered and variance scaled data of the response and predictor (spectral data) variables. The first significant latent vector of predictors, accounting for more than 85% of the spectral variance, was extracted and used as a synthetic spectral index (SI) for topsoil salinity characterization. A multivariate geostatistical approach was applied to the dataset including SI, EC, cation exchange capacity (CEC) and Ca content. After fitting a Linear Model of Coregionalization (LMC) to the set of experimental variograms, all data were interpolated at the nodes of a 20m x 20m 2D regular grid using point ordinary co-kriging and a spherical neighborhood of 3235.87 m in size. Topsoil samples EC values have a mean of 1.5 dSm–1, ranged from 0.3 to 10.5 dSm–1, and show a highly skewed distribution with 75% of values below 2 dSm–1. CEC in the soil ranged from 0.2 to 42 cmolckg–1 and Ca concentration was from 48 to 1337 ppm indicating high variability of soil properties in the study area. An isotropic LMC was fitted to the data and included three spatial structures: a Nugget effect, a spherical model with range of 835 m, and a spherical model with range of 12, 271 m ; the last two components account for most of the documented spatial variation. The SI was significantly correlated to EC ; a scale dependent correlation coefficient was equal to 0.43 at shorter range and 0.83 at longer range. Comparison between the EC and SI maps revealed some similarities in the spatial dependence patterns. Despite some differences, a common trend in decreasing topsoil salinity with increasing distance from the sea is detectable. The most considerable differences between the two maps are found in the north-east and north-west parts of the area, which might be related to the presence of clayey soils at salinization risk owing to upward capillary movements. This study demonstrates that by integrating spectroscopy with multivariate geostatistics it is possible to estimate a synthetic index, which could then be used as a potential tool for soil salinity risk assessment.

Salinity; Mediterranean; spectroscopy; geostatistic; risk assessment

nije evidentirano