Underground Water Level Prediction in Remote Sensing Images Using Improved Hydro Index Value with Ensemble Classifier

The economic sustainability of aquifers across the world relies on accurate and rapid estimates groundwater storage changes, but this becomes difficult due to absence in-situ surveys in most areas. By closing water balance, hydrologic remote sensing measures offer a possible method for quantifying changes storage. However, it is uncertain what extent data can provide an assessment these changes. Therefore, new framework implemented work predicting underground level using images. Generally, defined into five levels: Critical, Overexploited, Safe, Saline, Semi-critical, based quantity. In manuscript, images were acquired from At first, Wiener filtering was employed preprocessing. Secondly, Vegetation Indexes (VI) (Normalized Difference Index (NDVI), Normalized Snow (NDSI), Infrared index (IRI), Radar (RVI)), statistical features (entropy, Root Mean Square (RMS), Skewness, Kurtosis) extracted preprocessed Then, combined as novel hydro index, which fed Ensemble Classifier (EC): Neural Networks (NN), Support Vector Machine (SVM), improved Deep Convolutional Network (DCNN) models prediction obtained results prove efficacy proposed by different performance measures. shows that False Positive Rate (FPR) EC model 0.0083, better than existing methods. On other hand, has high accuracy 0.90, superior traditional models: Long Short-Term Memory (LSTM) network, Naïve Bayes (NB), Random Forest (RF), Recurrent (RNN), Bidirectional Gated Unit (Bi-GRU).