Evaluating the Utility of Bivariate Copula-Statistical Models for Forecasting Autumn Precipitation (Case study: Northwest of Iran)

Abstract

Teleconnection patterns are one of the causes of precipitation fluctuations in various regions of the world, including Iran. This study aims to develop bivariate models for forecasting autumn precipitation in the north-western region of Iran based on teleconnection indices. Copula functions were selected for this assignment due to the nonlinear relationship between precipitation and the teleconnection indices, as well as the fact that the assumption of normal distribution for precipitation data is not met, rendering Pearson correlation inappropriate. The dependence of Pacific and Atlantic Ocean teleconnection indices with precipitation for the period 1991-2020 was calculated using Kendall's and Pearson's rank correlation coefficients for moving average of one to six months. Appropriate copulas and marginal distributions were then used to model precipitation, and the performance of the developed models was evaluated. The results showed that the strongest correlations were obtained between the precipitation and the NINO3.4, SOI, and MEI indices. Consequently, the bivariate models using these indices demonstrated higher efficiency in simulating precipitation anomalies. Among these models, the one with the NINO3.4 predictor provided the best estimate of precipitation anomaly for 2021 and 2022, with values of -4.2 mm and -5.8 mm, respectively.

Keywords

• anomaly
• bivariate models
• copula
• predictor