ABSTRACT India is vulnerable to the impacts of climate change because of its large population, diverse landscapes and varied climate conditions. Changes in diurnal temperature range (DTR) caused by climate change can significantly affect heat island effects, increase heat stress, and lead to extreme events, impacting human health, animal behaviour, plant growth and crop yields. Reliable high‐resolution projections of DTR are required for climate risk assessment, agricultural planning and regional adaptation strategies; however, they are limited due to the coarse spatial resolution and inherent biases of global climate models (GCMs). To address this gap, this study employs sequential deep learning (DL) techniques, bidirectional long short‐term memory (BiLSTM) and gated recurrent unit (GRU) to downscale daily maximum ( T max ) and minimum ( T min ) temperatures from 11 CMIP6 models across 14 agro‐climatic zones of India and subsequently calculated DTR. Using historical (1951–2010), mid‐future (2040–2069), and late‐century (2070–2100) datasets under two shared socioeconomic pathways (SSPs), SSP2‐4.5 and SSP5‐8.5 scenarios, the DL models were trained and validated at each grid‐point against observed data. Performance assessment shows that both DL models effectively reproduce historical temperature, with BiLSTM outperforming GRU in terms of lower RMSE and MAE and higher correlation, demonstrating its suitability for fine‐scale climate downscaling. Future projections indicate widespread warming across India, with T min increasing (up to 5°C) more rapidly than T max , resulting in a significant decline in DTR. Under SSP5‐8.5 by the late century, DTR decreases by 2°C–4°C over the Western Himalayas, Indo‐Gangetic Plains, Western Dry Region and Central India; hence, these regions are identified as emerging major hotspots. Seasonal analysis also reveals strong DTR reductions (up to 4°C) during pre‐monsoon and post‐monsoon periods. Overall, this study emphasises the added value of DL‐based downscaling in improving high‐resolution temperature projections and identifies regions at increased risk from declining DTR, underscoring the need for targeted adaptation strategies.
Chaturvedi et al. (Wed,) studied this question.