Abstract
The standardized precipitation index (SPI) is widely used in drought assessments to monitor drought conditions due to its simple data requirement and multi-scale characteristics. However, there are some uncertainties existing in the process of its calculation cannot be ignored for accurate drought assessments. This study, taking the Heihe River basin in northwest of China as the study area, mainly focuses on investigated the uncertainty issuesies both in SPI calculation and in drought characteristics associated with the probability distributions and parameter estimation errors. based on the precipitation data at 4 9 meteorological stations covering a period of 1960-2015 in Heihe River basin(Northwest China). 10Ten probability distributions (two- and three-parameter Log-logistic and Lognormal, Generalized Extreme Value, Pearson-III, Burr, Gamma, Inverse Gaussian and Weibull) are employed were selected to estimate the SPI.and Maximum Likelihood Estimation isMaximum Likelihood EstimationMLE was used to estimate distribution parameters. Randomly generating parameters based on the normality assumption was is applied to quantify the uncertainty of parameter estimations and their effects toeffects on drought index. Results showed that, Log-Logistic logistic type distribution presents quite close performance with the benchmark Gamma distribution, and thus was is recommended as an alternatives in SPI calculation in fitting the precipitation data over the study area., and resulting in large uncertainty in SPI values although it passed the Kolmogorov-SmirnovK-S and Anderson-DarlingA-D tests. Effects of both uncertainty sources grades are more reflected on extreme droughts (extremely dry or wet). The more extreme the SPI value, the greater uncertainties caused bythese two both sources. Furthermore, the drought characteristics vary a lot from different distributions and parameter errors. These findings highlight the importance of uncertainty analysis of drought assessments, given that most studies in climatology focus on extreme values for drought analysis.
Highlights
Drought is one of the most common natural disasters usually with a high degree of damage and a wide range of influences (Xu et al, 2005; Mishra and Singh, 2010; Wang et al, 2012), which has become a hot topic in the fields of ecology, meteorology, and hydrology (Lynch et al, 2018; Zhou and Liu, 2018)
Guttman (1999) verified that the Pearson type III distribution is a better universal model in America; Sienz et al (2012) concluded that the Weibull-type distributions give distinctly improved fits compared to gamma in Europe; Angelidis et al (2012) found that the log-normal distribution gives almost the same results as gamma in the calculation of standardized precipitation index (SPI) at 12- and 24month scales in Guadiana (Portugal); Gabriel and Monica (2015) demonstrated that the generalized normal distribution presents the best performance in fitting precipitation series in Brazil
This result keeps consistent with the finding of Vergni et al (2017), who concluded that the two-parameter gamma distribution provides less reliable estimates of the precipitation probability than the three-parameter Pearson type III and the generalized normal distribution in their study case
Summary
Drought is one of the most common natural disasters usually with a high degree of damage and a wide range of influences (Xu et al, 2005; Mishra and Singh, 2010; Wang et al, 2012), which has become a hot topic in the fields of ecology, meteorology, and hydrology (Lynch et al, 2018; Zhou and Liu, 2018). Among the drought indices, standardized precipitation index (SPI) is widely used (Moreira, 2015; Zhang et al, 2017; Merabti et al, 2018; Oliveira-Júnior et al, 2018; Tirivarombo et al, 2018) because it can determine drought at different time scales and only requires precipitation data (Ma et al, 2013). (2012), Hong et al (2013), Gabriel and Monica (2015), Wu et al (2016), and Vergni et al (2017) indicated that the applicability of theoretical distributions in describing the cumulative precipitation was inconsistent across different regions. Guttman (1999) verified that the Pearson type III distribution is a better universal model in America; Sienz et al (2012) concluded that the Weibull-type distributions give distinctly improved fits compared to gamma in Europe; Angelidis et al (2012) found that the log-normal distribution gives almost the same results as gamma in the calculation of SPI at 12- and 24month scales in Guadiana (Portugal); Gabriel and Monica (2015) demonstrated that the generalized normal distribution presents the best performance in fitting precipitation series in Brazil
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
