EGU24-16245, updated on 09 Mar 2024
EGU General Assembly 2024
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Stochastic simulation of high space-time resolution precipitation fields in Beijing

Tinghui Li1, Shuiqing Yin1, Zeqi Li1, Maoqing Wang1, and Nadav Peleg2
Tinghui Li et al.
  • 1State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China
  • 2Institute of Earth Surface Dynamics, University of Lausanne, Géopolis, 1015 Lausanne, Switzerland

Precipitation is closely related to many earth surface processes, for some of them, such as urban flooding, high-resolution precipitation fields data are required. However, those high-resolution precipitation fields are often not available for a long enough period to be used for flood estimates. Stochastic models attempting to simulate precipitation at single or multiple sites face challenges in capturing the high spatial heterogeneity inherent in precipitation. We calibrated the Advanced WEather GENerator for a two-dimensional grid (AWE-GEN-2d) to simulate continuous 2-D precipitation fields and evaluated its performance based on CMA Multi-source merged Precipitation Analysis System Product (CMPAS) for the period from 2015 to 2020, with a spatial resolution of 0.01°×0.01° and a temporal resolution of hourly. Characteristics of spatiotemporal precipitation fields for 486 events were analyzed and monthly parameters in AWE-GEN-2d were obtained. AWE-GEN-2d was utilized to stochastically simulate hourly spatiotemporal precipitation fields at a resolution of 0.01°×0.01° for 30 years and its simulation accuracy was subsequently assessed by comparing with the observations. The results showed precipitation fields simulated by AWE-GEN-2d demonstrated consistency with the observed fields in terms of annual and monthly precipitation, the number and duration of precipitation events, and the average hourly precipitation intensity. For extreme hourly precipitation, the 95th and 99th percentiles of hourly precipitation were underestimated by 12.6% and 11.2%, respectively, compared to the observations. In terms of spatial pattern, we calculated the spatial autocorrelation function and spatial variation coefficient of the precipitation fields. The AWE-GEN-2d captured the general pattern but the spatial coefficient of variation was underestimated (spring to winter observations were 0.81, 1.16, 1.05, and 0.70; while the simulated were 0.57, 0.81, 0.74, and 0.49). The temporal autocorrelations were also underestimated, resulting in discontinuity jumps in rainfall centers. Future research work will focus on collecting sub-hourly observation interval data, such as 5 min or 10 min, and improve the simulation of the evolution of precipitation events, especially those with short durations and heavy intensities, which may bring high risks in urban flooding.

How to cite: Li, T., Yin, S., Li, Z., Wang, M., and Peleg, N.: Stochastic simulation of high space-time resolution precipitation fields in Beijing, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-16245,, 2024.