Benchmarking ConvLSTM for One-Day-Ahead IMDAA Rainfall-Field Prediction across Four Indian Cities

2026-07-29

AI Summary

This study evaluates the capability of ConvLSTM to forecast next-day precipitation fields using low-frequency, fine-scale daily reanalysis data over the Indian monsoon region, focusing on IMDAA rainfall records from four cities during the monsoon seasons of 1998–2020. It presents the first systematic comparison of ConvLSTM against fully connected LSTM (FC-LSTM), persistence forecasting, statistical models, and tree-based methods across multiple evaluation dimensions—including full-field prediction, regional averages, spatial anomalies, and extreme rainfall events. Results indicate that ConvLSTM does not significantly outperform FC-LSTM or persistence models; neural networks consistently underestimate intense rainfall magnitudes, whereas persistence demonstrates superior skill in detecting high-rainfall days, particularly in regions with strong spatial coherence such as Mumbai. The findings highlight the critical influence of input structure and spatial continuity on model selection for precipitation forecasting.

Abstract

Convolutional long short-term memory networks (ConvLSTMs) are widely used for precipitation forecasting, but most evidence for their performance comes from dense, high-frequency radar sequences. This study tests whether convolutional recurrence improves one-day-ahead rainfall-field prediction on small daily reanalysis grids. Indian Monsoon Data Assimilation and Analysis (IMDAA) fields for June-September 1998-2020 were analysed for Bengaluru, Delhi, Kolkata and Mumbai. Ten naive, statistical, tree-based and neural approaches were compared using atmospheric-only and rainfall-history-plus-atmospheric inputs. Performance was assessed for complete fields, domain-mean rainfall, spatial anomalies and high-rainfall days. ConvLSTM did not consistently outperform simpler alternatives. FC-LSTM produced the numerically lowest domain-mean rainfall error in Bengaluru, Kolkata and Mumbai, whereas persistence performed best in Delhi. ConvLSTM produced the numerically lowest spatial-anomaly error only in Mumbai, where rainfall fields showed greater short-term spatial continuity and rainfall-history inputs improved all three neural architectures. The difference between ConvLSTM and FC-LSTM was nevertheless small. Neural models underestimated rainfall magnitude and predicted too few threshold exceedances on high-rainfall days, while persistence achieved the highest detection performance in every city. Post-hoc analyses showed that the selected models were most sensitive to the latest input day, with broader recent-lag sensitivity in Mumbai. These findings show that gridded inputs alone do not justify ConvLSTM and that architecture choice should follow strong benchmarking across average, spatial and high-rainfall performance.

Problem

• ConvLSTM
• rainfall prediction
• IMDAA
• daily reanalysis
• benchmarking

Innovation

• ConvLSTM
• rainfall prediction
• benchmarking
• IMDAA
• spatial anomaly