Accurate weather prediction is crucial for effective flood prevention, water
resource management, and agricultural planning, particularly in regions prone to
extreme weather events such as the lower basin of the Mahanadi River. While
traditional weather forecasting methods are useful, they often struggle to account for
the complexity and variability inherent in local weather patterns. Recent years have
seen significant advancements in weather prediction accuracy, thanks to AI-driven
methods, especially those utilizing Machine Learning (ML) and Deep Learning (DL)
techniques. This chapter proposes a Bidirectional Long Short-term Memory (BiLSTM)
deep learning-based design for the rain gauge stations of the lower Mahanadi River
Basin, aiming to predict average temperatures. The most effective Rain Gauge (RG)
network in the Mahanadi basin was obtained through three different cluster analyses,
like K-means, Hierarchical Clustering (HC) Partitioning, and Partitioning Around
Medoids (PAM), using the selected characteristics, which are extracted from the
principal component analysis of seventeen rain gauge stations. The four rain gauge
stations of the lower Mahanadi river basin that were found effective from 15 years’
data are Kantamal, Kesinga, Salebhata, and Sundergarh, with a random split of 70/30
for train/test using Root Mean Squared Error (RMSE), Mean Absolute Error (MAE),
and Normalized RMSE (NRMSE) for performance evaluation of the different models.
The efficiency of the key Rain Gauge (RG) stations identified through cluster analysis
was evaluated using an Artificial Neural Network (ANN) with a single hidden layer,
along with four different Deep Neural Network (DNN) and Bidirectional Long ShortTerm Memory (BiLSTM) models. It is found that the BiLSTM model yields lower
RMSE, MAE, and NRMSE results. Overall, this study demonstrates the potential
benefits of adopting AI-driven approaches in regions and contributes to the growing body of literature on the application of deep learning in meteorology, where accurately
predicting average temperatures is both challenging and critically important.
Keywords: AI-driven, BiLSTM model, Everage temperature, Lower Mahanadi River basin, Neural networks, Rain gauge station.
