Development of soil moisture maps using image fusion of SCATSAT-1 and MODIS Dataset

Real-time monitoring of the soil moisture level is a challenging task, especially on a large scale. Although remote sensing is a significant way for global-level monitoring, the presence of atmospheric effects and coarse resolution are the two major problems associated with remote sensing. Hyperautomation in remote sensing allows the significant advancement of the automation process through artificial intelligence models in the collection, analysis, and application the geospatial information for various scientific domains. Hyperautomation enables the fusion of satellite data acquired from multisensory sources, especially with optical and radar sensors to generate more detailed and cloud-free soil moisture images. In this article, nearest-neighbor image fusion and neural net change detection have been performed using multisensory datasets from scatterometer satellite sensor, that is, SCATSAT-1 and optical sensor, that is, moderate resolution imaging spectroradiometer (MODIS) to generate the soil moisture maps. With such sensors, global data can be acquired daily without any cost. This model has been demonstrated over a part of India, that is, Haryana by taking into consideration both horizontal transmit and horizontal receive (HH) and vertical transmit and vertical receive (VV) polarization of the SCATSAT-1 dataset. All the outcomes, that is, fused, classified, and changed images have been evaluated at each stage of output. It has been analyzed that the proposed model effectively generated the soil moisture maps with more than 90% accuracy. In the fused images, the root mean square error was varying from 0.0001 to 0.0193. As far as the comparative analysis of SCATSAT-1 (HH) and SCATSAT-1 (VV) is concerned, SCATSAT-1 with HH achieved marginally better (~1%) accuracy as compared to S1 with VV. Hyperautomation facilitates the continuous monitoring of the soil moisture level with the incorporation of neural net-based classification and change detection methods. Additionally, it also allows spontaneous data processing, decision-making procedures, and prediction services