Background and Objectives: Drought and wetness are complex phenomena in which many factors have different effects. The introduced drought indices have tried to formulate the effect of these factors on drought severity as if they reflect real conditions. Therefore, each index has advantages and disadvantages that a single examination of the index cannot provide comprehensive information in this regard. On the other hand, combining and using several indices with multivariate statistical methods provides useful information. Therefore, in this study, the structural equation modeling approach was used to evaluate the relative contribution of drought drivers to evaluate the weight of precipitation and temperature in drought severity.
Materials and Methods: Structural equation modeling (SEM) was used to analyze the relationship between precipitation and temperature with drought. The data used were precipitation and temperature from 10 synoptic stations in the provincial centers of Iran over a period of 60 years. For each station, annual values of four drought indices (SPI, SPEI, RDI and PDI) were calculated. Before using the indices in structural equation modeling, the values of the indices were cleaned and direction of the indices was corrected (PDI values were multiplied by -1 to align with other indices). In the next step, its structural equation model was designed and implemented. The parameters were estimated using maximum likelihood (ML) method and CFI, RMSEA and SRMR criteria were used to evaluate the model fit. The model fit was accepted when CFI>0.90, RMSEA<0.08 and SRMR<0.08.
Results: The structural equation model showed that the proposed model had a good fit. All four indices (SPI, SPEI, RDI and PDI) had significant factor loadings with drought. The highest loading factor belonged to the SPEI index (λ = 0.86) and the lowest to the SPI (λ = 0.59). The paths between exogenous variables and drought were both significant. Precipitation had a strong and negative effect on drought (β = –0.71, p < 0.001), while temperature showed a positive and significant effect (β = 0.48, p < 0.01). In general, based on these four indices and constructs, 64% of the variance of the latent variable of drought was explained by precipitation and temperature. The relative contribution of precipitation and temperature to drought was calculated as 69 and 31%, respectively, based on the squared coefficients of the structural equation model.
Conclusion: Structural equation modeling showed that the modified combination of SPI, SPEI, RDI, and PDI indices can well-represent the latent concept of drought. Based on these indices, precipitation had the greatest reducing effect on drought severity, while increasing temperature significantly intensified drought. Also, different factor loadings of the indices showed that SPI and RDI reflect precipitation changes more than other indices, while SPEI and PDI are more sensitive to temperature fluctuations. This separation of the indices' functions highlights the value of using structural equation models in analyzing drought dynamics, because it allows for the simultaneous examination of direct and indirect effects of climate variables. Overall, this study confirms that SEM models are a powerful tool for explaining multivariate climate relationships and can be used to assess drought trends and risk at a regional scale. This research was able to address some of these foundations in the field of drought. To complete this process, extensive researches are needed in this field. By using other indicators, it is possible to obtain more accurate results, create a suitable hybrid model for assessing the severity of drought, and simultaneously understand the contribution of variables to the occurrence of the phenomenon.