Mapping the Moisture of Surface Soil Using Landsat 8 Imagery (Case study: Suburb of Semnan City)

Monitoring soil surface moisture, as a crucial factor in water and energy cycles, is of great importance in water and soil resource management. This important factor varies dramatically in time and space due to variability in soil characteristics, topography, vegetation, and dynamic nature of the climate. Any change in soil moisture can have an immediate effect on runoff, soil erosion, and plant productivity. Soil moisture is considered as a decisive factor in plant growth. Moreover, a decrease in soil moisture results in an increase in the dust of desert regions. Remote sensing methods can provide continuous soil moisture information on a large scale with acceptable accuracy. In the current study, applying data from Landsat 8 satellite image, different soil surface moisture estimation methods were studied. In order to assess the precision of each method, 80 samples of volumetric soil surface moisture were taken from the vicinity of Semnan province on the exact date the satellite passed over the region.  Some of the applied indices in the study are NDVI, NDTI, NDMI, PSMI, Surface Temperature, and SM_SWIR index. SM_SWIR index, with a correlation coefficient of 0.88 and R² of 0.61 was considered as the suitable index for soil moisture zonation in arid and desert areas. Therefore, SM_SWIR can be an appropriate indicator for soil surface moisture in arid and semi-arid regions. SM_SWIR , NDTI and NDMI indices could be considered as appropriate indicators for soil surface moisture in desert regions with poor vegetation. In the next step, using multivariate regression models, we prepared a soil moisture model using the studied indices. The findings of this study illustrated that Enter regression model has higher accuracy in surface soil moisture mapping.
Extended Abstract
1-Introduction
Monitoring soil surface moisture, as a crucial factor in water and energy cycles, is of great importance in water and soil resource management. Many soil properties such as stability, compressibility, permeability, and ability to transfer are function of moisture content in the soil. This variable is used to calculate the water balance of the area, Plant water requirement, and desertification studies. This important factor varies dramatically in time and space due to variability in soil characteristics, topography, vegetation, and dynamic nature of the climate. Any change in soil moisture can have an immediate effect on runoff, soil erosion, and plant productivity. Soil moisture is considered as a decisive factor in plant growth. Moreover, a decrease in soil moisture results in an increase in dust in desert regions. Continuous measurement of soil moisture is very costly and time-consuming, and the spatial range of measurement is generally limited. Remote sensing methods can provide continuous soil moisture information on a large scale with acceptable accuracy.
2-Materials and Methods
In the current study, applying data from Landsat 8 satellite, different soil surface moisture estimation methods were studied. In order to assess the precision of each method, real time field data were also collected and used. Due to the fact that the soil surface moisture is severely affected by daily evapotranspiration, wind, and airflows, 80 samples of volumetric soil surface moisture were taken from the vicinity of Semnan city on the exact date the satellite passed over the region. Measurement of moisture in the laboratory was carried out by direct method. Some of the applied indices in the study are Normalized differential vegetation index, normalized differential tillage Index, normalized differential moisture index, perpendicular soil moisture index, surface temperature, and SM_SWIR index. Calculating the Pearson coefficient and the coefficient of determination of each of the above indices, the most correlated indices were used to measure soil moisture content. In the next step, the soil moisture estimation functions were calculated using the correlated indices by three methods of Partial least squares regression, Stepwise and Enter method. Considering the high correlation coefficient of the Enter method and the low RMSE compared to the other two methods, regression model of Enter can be used to model the soil moisture zonation in the studied area.
3-Results and Discussion
The results of the study revealed that in desert regions with low vegetation, using short-wavelength infrared bands showed a higher level of correlation with the field data. Among the proposed methods, surface temperature did not show a high correlation with data on moisture. This indicates that thermal bands are affected by other factors besides soil moisture. Thus, perpendicular soil moisture index cannot offer acceptable accuracy as it relies heavily on surface temperature. SMSWIR index, with correlation coefficient of 0.78, presents a slight difference compared to normalized differential tillage indices and normalized differential vegetation index. This index’s square regression value was 0.61, and the root-mean-square error for the regression model results and actual data were estimated to be 3.69. The root-mean-square error for the regression model resulting from normalized differential tillage index, normalized differential moisture index and normalized differential vegetation index and the actual data was 3.77 and 4, respectively.  Therefore, Soil moisture of Short-wavelength infrared bands (SMswir) and normalized differential tillage index and normalized differential moisture index could be an appropriate indicator for soil surface moisture in desert regions with poor vegetation.
In the next step, using normalized differential tillage indices, normalized differential moisture index, normalized differential vegetation index, short-wavelength infrared moisture index and surface temperature index, regression model of these variables were prepared. Regarding the low root mean square error and the high coefficient of explanation of the Enter model among other multivariate regression methods, this model has a good accuracy for soil surface moisture zonation at 5% level.
4-Conclusion
Our study illustrated that the short-wavelength infrared bands of Landsat 8 show more sensitivity to soil moisture changes than near-infrared bands. In this paper, the efficiency of reflective and thermal bands of Landsat 8 for monitoring of surface soil moisture was confirmed. The research was done in a desert region in Semnan. As a result, the reliability of proposed methods in regions with higher vegetation and forests must be evaluated again.