We propose a new architecture based on a fully connected feed-forward Artificial Neural Network (ANN) model to estimate surface soil moisture from satellite images on a large alluvial fan of the Kosi River in the Himalayan Foreland. We have extracted nine different features from Sentinel-1 (dual-polarised radar backscatter), Sentinel-2 (red and near-infrared bands), and Shuttle Radar Topographic Mission (digital elevation model) satellite products by leveraging the linear data fusion and graphical indicators. We performed a feature importance analysis by using the regression ensemble tree approach and also feature sensitivity to evaluate the impact of each feature on the response variable. For training and assessing the model performance, we conducted two field campaigns on the Kosi Fan in December 11–19, 2019 and March 01–06, 2022. We used a calibrated TDR probe to measure surface soil moisture at 224 different locations distributed throughout the fan surface. We used input features to train, validate, and test the performance of the feed-forward ANN model in a 60:10:30 ratio, respectively. We compared the performance of ANN model with ten different machine learning algorithms [i.e., Generalised Regression Neural Network (GRNN), Radial Basis Network (RBN), Exact RBN (ERBN), Gaussian Process Regression (GPR), Support Vector Regression (SVR), Random Forest (RF), Boosting Ensemble Learning (Boosting EL), Recurrent Neural Network (RNN), Binary Decision Tree (BDT), and Automated Machine Learning (AutoML)]. We observed that the ANN model accurately predicts the soil moisture and outperforms all the benchmark algorithms with correlation coefficient (R = 0.80), Root Mean Square Error (RMSE = 0.040 $$\mathrm {m^3/m^3}$$m3/m3 ), and bias = 0.004 $$\mathrm {m^3/m^3}$$m3/m3 . Finally, for an unbiased and robust conclusion, we performed spatial distribution analysis by creating thirty different sets of training-validation-testing datasets. We observed that the performance remains consistent in all thirty scenarios. The outcomes of this study will foster new and existing applications of soil moisture.

This review provides a detailed synthesis of various in-situ, remote sensing, and machine learning approaches to estimate soil moisture. Bibliometric analysis of the published literature on soil moisture shows that Time-Domain Reflectometry (TDR) is the most widely used in-situ instrument, while remote sensing is the most preferred application, and the random forest is the widely applied algorithm to simulate surface soil moisture. We have applied ten most widely used machine learning models on a publicly available dataset (in-situ soil moisture measurement and satellite images) to predict soil moisture and compared their results. We have briefly discussed the potential of using the upcoming NASA-ISRO Synthetic Aperture Radar (NISAR) mission images to estimate soil moisture. Finally, this review discusses the capabilities of physics-informed and automated machine learning (AutoML) models to predict surface soil moisture at higher spatial and temporal resolutions. This review will assist researchers in investigating the applications of soil moisture in the broad domain of earth sciences.

Surface soil moisture has a wide application in climate change, agronomy, water resources, and in many other domain of science and engineering. Measurement of soil moisture at high spatial and temporal resolution at regional and global scale is needed for the prediction of flood, drought, planning and management of agricultural productivity to ensure food security. Recent advancement in microwave remote sensing, especially after the launch of Sentinel operational satellites has enabled the scientific community to estimate soil moisture at higher spatial and temporal resolution with greater accuracy. This study evaluates the potential of Sentinel-1A satellite images to estimate soil moisture in a semi-arid region. Exactly at the time when satellite passes over the study area, we have collected soil samples at 37 different locations and measured the soil moisture from 5 cm below the ground surface using ML3 theta probe. We processed the soil samples in laboratory to obtain volumetric soil moisture using the oven dry method. We found soil moisture measured from calibrated theta probe and oven dry method are in good agreement with Root Mean Square Error (RMSE) 0.025 m ${}^3$ /m ${}^3$ and coefficient of determination (R ${}^2$ ) 0.85. We then processed Sentinel-1A images and applied modified Dubois model to calculate relative permittivity of the soil from the backscatter values ( ${\sigma }^{\circ }$ ). The volumetric soil moisture at each pixel is then calculated by applying the universal Topp’s model. Finally, we masked the pixels whose Normalised Difference Vegetation Index (NDVI) value is greater than 0.4 to generate soil moisture map as per the Dubois NDVI criterion. Our modelled soil moisture accord with the measured values with RMSE = 0.035 and R ${}^2$ = 0.75. We found a small bias in the modelled soil moisture ( $0.02{\mathrm{m}}^3/{\mathrm{m}}^3$ ). However, this has reduced significantly ( $0.001{\mathrm{m}}^3/{\mathrm{m}}^3$ ) after applying a bias correction based on Cumulative Distribution Function (CDF) matching. Our approach provides a first-order estimate of soil moisture from Sentinel-1A images in sparsely vegetated agricultural land.

Disease prediction of a human means predicting the probability of a patient’s disease after examining the combinations of the patient’s symptoms. Monitoring a patient's condition and health information at the initial examination can help doctors to treat a patient's condition effectively. This analysis in the medical industry would lead to a streamlined and expedited treatment of patients. The previous researchers have primarily emphasized machine learning models mainly Support Vector Machine (SVM), K-nearest neighbors (KNN), and RUSboost for the detection of diseases with the symptoms as parameters. However, the data used by the prior researchers for training the model is not transformed and the model is completely dependent on the symptoms, while their accuracy is poor. Nevertheless, there is a need to design a modified model for better accuracy and early prediction of human disease. The proposed model has improved the efficacy and accuracy model, by resolving the issue of the earlier researcher’s models. The proposed model is using the medical dataset from Kaggle and transforms the data by assigning the weights based on their rarity. This dataset is then trained using a combination of machine learning algorithms: Random Forest, Long Short-Term Memory (LSTM), and SVM. Parallel to this, the history of the patient can be analyzed using LSTM Algorithm. SVM is then used to conclude, the possible disease. The proposed model has achieved better accuracy and reliability as compared to state-of-the-art methods. The proposed model is useful to contribute towards development in the automation of the healthcare industries.

We apply the Support Vector Regression (SVR) machine learning model to estimate surface roughness on a large alluvial fan of the Kosi River in the Himalayan Foreland from satellite images. To train the model, we used input features such as radar backscatter values in Vertical–Vertical (VV) and Vertical–Horizontal (VH) polarisation, incidence angle from Sentinel-1, Normalised Difference Vegetation Index (NDVI) from Sentinel-2, and surface elevation from Shuttle Radar Topographic Mission (SRTM). We generated additional features (VH/VV and VH–VV) through a linear data fusion of the existing features. For the training and validation of our model, we conducted a field campaign during 11–20 December 2019. We measured surface roughness at 78 different locations over the entire fan surface using an in-house-developed mechanical pin-profiler. We used the regression tree ensemble approach to assess the relative importance of individual input feature to predict the surface soil roughness from SVR model. We eliminated the irrelevant input features using an iterative backward elimination approach. We then performed feature sensitivity to evaluate the riskiness of the selected features. Finally, we applied the dimension reduction and scaling to minimise the data redundancy and bring them to a similar level. Based on these, we proposed five SVR methods (PCA-NS-SVR, PCA-CM-SVR, PCA-ZM-SVR, PCA-MM-SVR, and PCA-S-SVR). We trained and evaluated the performance of all variants of SVR with a 60:40 ratio using the input features and the in-situ surface roughness. We compared the performance of SVR models with six different benchmark machine learning models (i.e., Gaussian Process Regression (GPR), Generalised Regression Neural Network (GRNN), Binary Decision Tree (BDT), Bragging Ensemble Learning, Boosting Ensemble Learning, and Automated Machine Learning (AutoML)). We observed that the PCA-MM-SVR perform better with a coefficient of correlation (R = 0.74), Root Mean Square Error (RMSE = 0.16 cm), and Mean Square Error (MSE = 0.025 ${\mathrm{cm}}^2$). To ensure a fair selection of the machine learning model, we evaluated the Akaike’s Information Criterion (AIC), corrected AIC (AICc), and Bayesian Information Criterion (BIC). We observed that SVR exhibits the lowest values of AIC, corrected AIC, and BIC of all the other methods; this indicates the best goodness-of-fit. Eventually, we also compared the result of PCA-MM-SVR with the surface roughness estimated from different empirical and semi-empirical radar backscatter models. The accuracy of the PCA-MM-SVR model is better than the backscatter models. This study provides a robust approach to measure surface roughness at high spatial and temporal resolutions solely from the satellite data.