| Summary, etc |
Coal mining is extensively practised in the East Jaintia Hills district of Meghalaya. But due to unscientific extraction and disorganized dumping of coal mine spoils have led to soil problems in these regions and are adversely affecting the unmined areas resulting in loss of soil fertility and shows variability of nutrients across the different locations. Direct measurement of soil nutrients in the field and laboratory provides more accurate and site-specific information but these are expansive, laborious, and time-consuming and cannot measure soil nutrients at unsampled locations. In recent years, digital mapping of soil nutrients at unsampled locations has been accomplished by using geospatial tools such as Remote sensing, GIS, GPS and geostatistical interpolation techniques. Thus, the present study was conducted in an area of 108 hectares in Latyrke village of the East Jaintia Hills District of Meghalaya which is a representative coal mine-affected area to understand the spatial variability of soil nutrients using the geostatistical model. This region is situated between latitudes 25°20'31.06"N to 25°20'35.78"N and longitudes 92°26'59.31"E to 92°27'53.75"E, at an altitude of 1172 meters above mean sea level. A land use and land cover (LULC) map of the study area was prepared by using Sentinel 2B data of East Jaintia Hills, Meghalaya which was downloaded from Copernicus open excess hub. Then a 50 m × 50 m grid size map of the study area was prepared in ArcGIS 10.8.2 over the LULC map. Sampling was done according to the grid method of grid size 50 m × 50 m. A total of 172 composite soil samples were collected at 0-15cm depth using a handheld Global Positioning System unit (GARMIN GPSMAP-64) from each grid cell (50 m x 50 m) and only from five selected land use namely Agricultural land, Grassland, Forest, Wasteland and Coal dumping site in May 2023. The collected soil samples were randomly split into two sets, 75% of the total samples i.e., 129 samples for SOC, available NPK and micronutrient prediction at the unsampled location using geostatistical models and 25% of the total samples i.e., 43 samples for model validation. Soil samples were analyzed for pH, electrical conductivity (EC), soil organic carbon (SOC), available NPK and available micronutrients (Fe, Mn, Cu, Zn) using standard analytical procedures. Descriptive statistics and geostatistical analysis were done to assess the spatial variability of the soil nutrients and their prediction at unsampled locations, respectively. The best-fitted model was used for making spatial variability map of the selected soil nutrients. The results showed that the value of soil pH ranged from 3.50 to 5.16; electrical conductivity (EC) ranged from 0.03 to 0.15 dS m-1; soil organic carbon (SOC) ranged from 0.66 to 2.81%; available nitrogen (N) ranged from 133.32 to 314.12 kg ha-1; available phosphorus (P) ranged from 6.12 to 24.53 kg ha-1; available potassium (K) ranged from 72.89 to 368.12 kg ha-1 and available micronutrient iron (Fe) ranged from 17.14 to 84.24 ppm; manganese (Mn) ranged from 6.23 to 14.67 ppm, copper (Cu) ranged from 0.30 to 0.86 ppm and zinc (Zn) ranged from 0.2 to 0.62 ppm. The coefficient of variation (CV) value for soil pH, EC, SOC, available NPK and micronutrients Fe, Mn, Cu and Zn were found 7.76%, 44.44%, 34.12%, 25.48%, 36.55%, 47.48%, 33.59%, 20.27%, 24.56% and 15.91% respectively, indicating medium variability of all the soil properties except pH which shows low variability in the study region. The spatial variability analysis of soil properties viz., pH, available NPK and available Fe, Mn were done by using log transformation model. While spatial variability analysis of soil EC, SOC, available Cu and Zn was done by using none transformation model. The value for nugget/sill of soil EC, available (N, P and Fe) were found <0.25 indicating strong spatial dependence but in the case of soil pH, SOC, available (K, Mn,Cu and Zn) were found between 0.25-0.75 indicating moderate spatial dependence among the variables. The exponential model of ordinary kriging was found as the best-fitted model for the prediction of all the selected soil properties by validation with the highest R2 (0.76, 0.72, 0.87, 0.79, 0.83, 0.72, 0.81, 0.74, 0.83 and 0.75) and the lowest RMSE (0.040, 0.249,0.189, 0.111, 0.142, 0.273, 0.164, 0.093, 0.059 and 0.042) value among all the geostatistical models. The spatial variability map of the selected soil chemical properties was prepared by using the exponential model of ordinary kriging. |
