Loading...

Drought Uncertainty Analysis under Climate Change Using Copula

Abbasian, Mohammad Sadegh | 2014

998 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 46076 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Abrishamchi, Ahmad
  7. Abstract:
  8. The classical approaches of drought analysis are performed using historical records and univariate statistical methods. In the use of historical records it is assumed that statistical characteristics of droughts will be exactly repeated in the future (the assumption of stationarity) and in the use of univariate methods it is assumed that drought variables (e.g. duration and severity) are independent; while on the contrary, global warming changes long-term climate patterns, which is called climate change, and also drought variables have significant correlation. Thus, results of classical approaches of drought analysis have uncertainties.The primary objective of this thesis is to propose an appropriate multivariate procedure for quantifying uncertainties of analyses of historical droughts arising from climate change. The study is performed using daily precipitation data of Tabriz synoptic station during the years 1956-2009. To achieve the objective, firstly the A2 (medium-high emissions) and the B2 (medium-low emissions) greenhouse gas emissions scenarios from the global circulation model HadCM3 are used as inputs to the Statistical Downscaling Model (SDSM) to downscale precipitation during 1961-2099. Data of the period 1961-1990 and the period 2070-2099 are considered as representatives of past climate and future climate, respectively. The comparison of past and future climate precipitation shows that the A2 and B2 scenarios predict a mean annual precipitation decrease of 59 mm (18 percent) and 9 mm (3 percent), respectively. Then, bivariate frequency analyses of duration and severity of past and future climate droughts are done using copulas. Finally, uncertainties of the stationarity assumption are quantified by plotting joint return periods of past and future droughts. Actually, differences between identical return periods of past and future droughts are considered as uncertainties of analyses of past droughts under climate change. The comparison of 2-year return period curves shows that climate change has slight effects on droughts with low return periods. Therefore, for low return periods it is reasonable to use results of analyses of historical droughts for long-term planning of water resources. But the comparison of 200-year return period curves shows that climate change effects on drought with high return periods are significant. Hence, for high return periods it is uncertain to use results of analyses of historical droughts for long-term planning of water resources.Another objective of the thesis is to compare univariate and multivariate analyses. To this objective, curves of joint return periods of drought duration and severity and also conditional return periods of duration and severity are plotted in order to be compared with univariate return periods. Results show that since univariate analyses assume variables are independent, they cannot properly describe the drought event and overestimate or underestimate results. Therefore, uncertainties of univariate analyses are significant. Hence, it is recommended to replace multivariate analyses with univariate analyses whenever enough knowledge of multivariate analyses, appropriate computational resources and enough data are available. A secondary objective of the thesis is to examine the importance of choosing an appropriate probability distribution for computing Standardized Precipitation Index (SPI). Results show that inappropriate fit to precipitation data results in significant error in the computation of SPI. In addition, two parameter gamma and Pearson type III distributions provide appropriate fits to precipitation data of Tabriz station. Hence, according to this thesis and previous studies, for comparing SPI it is permitted to use two parameter gamma and Pearson type III distributions by default
  9. Keywords:
  10. Drought ; Uncertainty Analysis ; Variable Infiltration Capacity (VIC)Model ; Copula Functions ; Downscaling

 Digital Object List

 Bookmark

No TOC