Search

As a part of the Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA) program in India, durable assets under various work categories are created in rural areas by employing adult members belonging to the marginal communities. Taking the Prakasam district in Andhra Pradesh as an example, we analyzed the spatial clustering of the various work implemented under MGNREGA using kernel density estimation (KDE). We also analyzed the spatial clustering of villages in terms of overall assets and their statistical significance using hotspot analysis (Getis-Ord). The socio-economic factors influencing village-level clustering are analyzed using a machine learning model. Results show significant spatial variations in kernel density and village hotspots, thus, indicating the demand-driven nature of the program as influenced mainly by the marginal worker population female followed by the main other worker population. The study thus offers a methodological framework that may help improve and complement the empirical indicators currently employed in mapping the performance of the MGNREGA and similar programs.

The accurate and complete spatial distribution of soil texture is required for proper land use planning, soil management practices and other interventions linked to environmental protection. In the present study, we spatially predicted sand, silt and clay content in top soil (0–15 cm) in parts of Gadag district in Karnataka, India using satellite-based indices and geostatistical modeling. The study was conducted to evaluate the Landsat 8 Operational Land Imager (OLI) remote sensing data as auxiliary variable along with CartoDEM for spatial prediction of sand, silt and clay content with a total of 57 number of field observations. The bands 2, 3, 4, 5, 6 and 7, Normalized Difference Vegetation Index (NDVI), The Grain Size Index (GSI) and the relationship between band 4 and 3, 4 and 7 and 6 and 7 of Landsat 8 OLI and slope, elevation of CartoDEM were used as covariates for modeling. Among these covariates NDVI and relationship between band 4 and band 7 were not significantly correlated with the sand, silt and clay content. However digital number (DN) of band 2, 3, 4, 5, 6, 7, GSI and relationship between band 4 to band 3 were significantly correlated with the surface soil sand, silt and clay content. The DN of band 2, 3, 4, 5, 6, 7, GSI and relationship between band 4 to band 3 were selected as auxiliary data for the estimation of sand and silt content, whereas the DN of band 5 and 7 explained most of the variability of soil clay. The Regression Kriging (RK) and Multiple Linear Regression (MLR) were employed to predict soil texture and compared for better accuracy of estimation. The greater spatial variability (58.8, 48.2 and 33.2%) of sand, silt and clay content respectively were predicted with lower estimation errors (RMSE = 7.41, 4.68 and 7.64, R2= 0.95, 0.55 and 0.91 respectively) in RK as compared to MLR method (RMSE = 9.18, 5.65 and 11.13, R2= 0.90, 0.38 and 0.80 respectively). The average decrease of 31.3, 19.2 and 17.5% in the prediction error was observed in RK over MLR approach in clay, sand and silt prediction respectively. The uncertainty of the prediction was calculated by 95% confidence intervals, which showed 7.51, 43.5 and 49.6% for sand, silt and clay prediction respectively. The current study showed that RK approach can be useful to predict sand, silt and clay content over MLR approach.