Estimation of return levels and associated uncertainties of extreme temperatures using a time-varying framework: a case study in Iran

In recent decades, Iran has seen unprecedented extreme temperatures (ETs) in different climatic zones, resulting in significant shifts and inconsistencies in their distributions. So, estimating ETs and associated uncertainties within a non-stationary (NS) context becomes a crucial step in modeling of hydro-climatic events like floods, droughts etc. This study examines the time-varying evaluation of extreme hot and cold temperatures (EHTs and ECTs) at 12 weather stations in Kerman province, Iran. Moreover, two recently proposed methodologies are investigated: conditional and integrated (unconditional), for estimating return levels (RLs) and their corresponding confidence intervals (CIs) within a NS framework. Analyses were conducted using Generalized Extreme Value (GEV) distribution under two assumptions: stationary (S-GEV) and non-stationary (NS-GEV). The EHTs and ECTs time series from 1979 to 2019 underwent testing for trends, homogeneity, and stationarity. The maximum likelihood estimator (MLE) was adopted to estimate the distribution parameters. The NS impacts of EHTs and ECTs were quantified by calculating the difference between stationary and non-stationary RLs, denoted as SRL and NSRL, respectively. Analysis of trends and stationarity indicated that the EHTs and ECTs time series were non-stationary. The Akaike information criterion (AIC) favored the NS-GEV model over the S-GEV model. Our results demonstrated that NS-GEV frequency analyses have a growing impact on the RL for both EHTs and ECTs. One finding was that the visualization of conditional RL plots turned out to be a valuable approach to assess uncertainties in future scenarios; another that climatology (e.g. arid and excessive arid areas across Kerman) seems to influence shapes and features of RL in future outcomes. Our findings can significantly contribute to policy-making and strategic planning in water resource management, particularly in areas such as infrastructure development and risk assessment.