000			03049nam a22001577a 4500
082			_a629.8
100			_aMunir, Safdar _9130646
245			_aLong-Term Temperature Forecasting using Deep Learning and Extreme Indices Computation / _cSafdar Munir
264			_aIslamabad : _bSMME- NUST; _c2025.
300			_a105p. _bSoft Copy _c30cm
500			_aClimate change has intensified the demand for accurate long-term temperature forecasting, particularly in highly vulnerable regions such as Pakistan, where rising temperatures and extreme events threaten agriculture, water security, public health, and infrastructure. Conventional climate models, including General Circulation Models (GCMs) and Regional Climate Models (RCMs), provide essential large-scale projections but face constraints in spatial resolution, computational cost, and applicability in data-scarce contexts. Recent advances in deep learning offer powerful alternatives, capable of capturing complex nonlinear relationships and producing high-resolution forecasts that can also inform the computation of extreme climate indices. This study integrates DL based forecasting with extreme temperature index computation to generate actionable insights for adaptation planning in Major Cities of Pakistan. Daily temperature data from the EC-EARTH3-VEG climate model (converted to °C) were bias-corrected using quantile mapping under two Shared Socioeconomic Pathways (SSP4.8 and SSP5.85). After data preprocessing multiple deep learning models LSTM, CNN-LSTM, Transformer, Transformer XL and Deep Ensemble with Monte Carlo (MC) Dropout, were trained on 1980–2040 data and evaluated for 2040–2050 and then forecasted till 2100. Model was assessed using MAE, RMSE, R² with MC dropout. Deep Ensemble outperformed all the single model (MAE = 0.4588; RMSE = 0.751; R² = 0.992). Projections indicate a Tmax rise of ~2–4 °C and Tmin rise of ~1.5–3 °C by 2100. According to extreme climates indices calculation there continue to be increases in warming, especially with regards to more heat waves that last longer in addition to increases in the number of hot days and warm nights as well as fewer frost days. The warming of nights more than the warming of days, closing the diurnal temperature range. There is also more total precipitation which is distributed in a growing number of days, and longer periods of rain which means that heavier rains are becoming less common, which indicates the beginnings of a warmer, more humid, but drier and stressed climate. The study highlights that DL Ensemble frameworks can deliver robust, long term temperature projections and calculation of extreme indices for 25 major cities of Pakistan and for sake of brevity only data of Lahore, Karachi, Quetta and Gilgit city shown in paper reset are in appendix.
650			_aMS Robotics and Intelligent Machine Engineering _9119486
700			_aSupervisor : Dr. Zaib Ali _9119496
856			_uhttp://10.250.8.41:8080/xmlui/handle/123456789/55218
942			_2ddc _cTHE
999			_c614826 _d614826