Uncertainty in SPI Calculation and its Impact on Drought Assessment in Different Climate Regions over China

View More View Less
  • 1 State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering Sciences, Hohai University, Nanjing, China
  • 2 Institute of Geophysics, Polish Academy of Sciences, Warsaw, Poland
© Get Permissions
Restricted access

Abstract

Uncertainty in the calculation of a Standardized Precipitation Index (SPI) attracted growing concerns in the hydrometeorology research community in the last decade. This issue is addressed in the present study from the perspective of candidate probability distributions, the data record length, the cumulative timescale and the selection of a reference period with the bootstrap and Monte Carlo methods using daily precipitation data observed in four climate regions across China. The impacts of the uncertainty in an SPI calculation on drought assessment are also investigated. Results show that the Gamma distribution is optimal in describing the cumulative precipitation in China; among the four timescales investigated in the present study, the minimal timescale appropriate for SPI calculation is 20 days for the humid region, 30 days for the semi-humid/semi-arid region and Tibetan Plateau (mostly its eastern part), and 90 days for the arid region. The uncertainty in SPI calculation decreases with the increase of timescale and record length, essentially as a consequence of the decrease of the confidence interval width of Gamma distribution parameters with the increase of timescale and record length. But there is little improvement for the parameter estimation with record length longer than 70 years. There is greater uncertainty for high absolute SPI values than for small ones, consequently there is greater uncertainty in assessing extreme droughts than moderate droughts. Reference period selection has large impacts on drought assessment, especially in the context of climate change. The uncertainty of the SPI calculation has large impacts on categorizing droughts, but no impact on assessing the temporal features of drought variation.

Corresponding author: Wen Wang, wangwen@hhu.edu.cn

Abstract

Uncertainty in the calculation of a Standardized Precipitation Index (SPI) attracted growing concerns in the hydrometeorology research community in the last decade. This issue is addressed in the present study from the perspective of candidate probability distributions, the data record length, the cumulative timescale and the selection of a reference period with the bootstrap and Monte Carlo methods using daily precipitation data observed in four climate regions across China. The impacts of the uncertainty in an SPI calculation on drought assessment are also investigated. Results show that the Gamma distribution is optimal in describing the cumulative precipitation in China; among the four timescales investigated in the present study, the minimal timescale appropriate for SPI calculation is 20 days for the humid region, 30 days for the semi-humid/semi-arid region and Tibetan Plateau (mostly its eastern part), and 90 days for the arid region. The uncertainty in SPI calculation decreases with the increase of timescale and record length, essentially as a consequence of the decrease of the confidence interval width of Gamma distribution parameters with the increase of timescale and record length. But there is little improvement for the parameter estimation with record length longer than 70 years. There is greater uncertainty for high absolute SPI values than for small ones, consequently there is greater uncertainty in assessing extreme droughts than moderate droughts. Reference period selection has large impacts on drought assessment, especially in the context of climate change. The uncertainty of the SPI calculation has large impacts on categorizing droughts, but no impact on assessing the temporal features of drought variation.

Corresponding author: Wen Wang, wangwen@hhu.edu.cn
Save