This study investigates the complex relationship between temperature and precipitation—and their joint role in rainfall prediction—across diverse climates worldwide, including North America, Europe, Australia, and Central Asia . The authors first conduct correlation analyses using high-resolution meteorological data from NOAA and NSIDC, followed by development of predictive models leveraging both Multiple Linear Regression (MLR) and Long‑Short‑Term Memory neural networks (LSTM) . Their findings reveal that although MLR can capture broad trends (with R² values up to ~0.99 for temperature predictions), it underperforms in rainfall estimation (87 % error rate). In contrast, the LSTM model achieves near-perfect accuracy (~F1 = 0.998, 0.2 % error) in predicting rainfall events in Sydney’s dataset . The authors conclude that temperature and precipitation exhibit a notable inverse correlation and that advanced machine‑learning techniques, such as LSTM, offer a superior approach for capturing complex, non‑linear interactions essential for accurate rainfall forecasting.

Research carried out in collaboration with Cardiff Metropolitan University and University College Dublin