Projection and prediction of heat waves for an arid region in the context of climate change

• Main Author: Ullah Khan, Najeeb
• Format: Thesis
• Language: English
• Published: 2020
• Subjects: TA Engineering (General). Civil engineering (General)
• Online Access: http://eprints.utm.my/id/eprint/92181/1/NajeebullahKhanPSKA2019.pdf.pdf; http://eprints.utm.my/id/eprint/92181/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:134437

Forecasting temperature extremes especially heat waves are extremely important for developing preparedness and planning mitigation measures, particularly in the context of climate change. The major objective of the present study was to assess the ongoing changes and possible future changes in heat waves and development of robust statistical model for forecasting heat wave which can adapt with changing climate. Pakistan, which is one of the most affected countries of the world to heat waves in recent years was considered as the study area. Novelties of the study are the methods proposed for defining heat waves, reliable projection of heat waves with associated uncertainties, and development of robust forecasting models which can adapt with climate change. Available in-situ temperature records, gauge-based gridded temperature data and temperature simulations of general circulation data (GCM) of Coupled Model Intercomparison Project Phase 5 (CMIP5) were used for defining heat waves and assessment of historical changes and future projections of temperature extremes and heat waves, while the reanalysis atmospheric data of National Centres for Environmental Prediction (NCEP) was used for the development of heat wave forecasting models. A threshold-based approach which able to demarcate the historical heat wave affected area is proposed for defining heat waves, GCMs were selected based on their capability to simulate different characteristics of heat waves and different state-of-the-art machine learning methods (ML) were used for the development of the seasonal and daily heat wave forecasting models. The study revealed that the daily maximum temperature more than 95-th percentile threshold for consecutive five days or more can well reconstruct the spatial pattern of heat wave in Pakistan. The assessment of trends in heat waves based on the derived definition revealed increase in heat wave duration and affected area in Pakistan at a rate of 0.71 days/decade and 1.36% of the total area of Pakistan per decade respectively. Four GCMs namely, CCSM4, CESM1(BGC), CMCC-CM and NorESM1-M were found to have better ability for the projection of all the characteristics of heat waves. The projection of heat waves using the selected GCMs revealed a high increase in the heat wave indices particularly for representative concentration pathways (RCP) 8.5. Heat wave frequency was projected to increase up to 12 events per year in most parts of the country, while some areas would experience heat waves for more than 100 days in a year. The higher increase in heat waves indices was projected in highly populated eastern and southern coastal regions which are already prone to high occurrence of heat waves. Forecasting models were developed for the prediction of triggering date and seasonal number of heat waves days in order to aid in coping and mitigation capacity revealed that the Quantile Regression Forests (QRF) models were able to forecast the triggering and departure dates of heat waves with an accuracy of up to ±5 days. On the other hand, the Support Vector (SVM) model was found to have the higher skills in forecasting number of heat wave days with a month time-lag (R2: 0.89- 0.9, NRMSE%: 32.6-31.8, rSD: 0.98-0.96, and md: 0.8). The analysis of synoptic patterns revealed that the wind vectors, relative humidity and geopotential height are the most potential indicators of heat waves in Pakistan. The forward-rolling based forecasting model proposed for the prediction of heat waves to accommodate the changing pattern of the atmospheric variables responsible for heat waves due to the changes in climate was found to forecast heat waves reliably.