北京三百年降水序列的成分分析及其随机模拟

全球气候变暖背景下,降水量时域分布不均导致局部时段水资源短缺等问题日益严重,合理准确分析区域长序列降水变化规律和演变趋势尤为重要。基于北京地区1724—2019年降水时序资料构建不同长度的降水序列,采用BS-Pettitt耦合模型和极点对称模态分解算法对降水序列组成成分进行分析,并利用随机森林算法对其进行随机模拟。通过近三百年的降水时序检验分析,结果表明：北京地区降水量变化整体呈“上升—平稳—下降”趋势,降水序列在时域中存在1770年、1813年、1871年、1893年、1947年和1999年6个突变点;在频域中存在2.5～4年、7～15年、25～35年、74年左右和95～100年的准周期波动规律,且长时间序列在降水序列成分分析和随机模拟中均表现出较强优势。

Component analysis and stochastic simulation of precipitation series in Beijing during the last 300 years

Under the background of global warming,the uneven spatial and temporal distribution of precipitation has led to increasingly serious problems such as water scarcity in some areas,so it is particularly important to analyze the regional long-term precipitation change patterns and evolution trends reasonably and accurately.The precipitation series of different lengths were constructed based on the precipitation time series data of Beijing from 1724 to 2019,the BS-Pettitt coupling model and the pole symmetry modal decomposition algorithm were used to analyze the components of the precipitation series,and the random forest algorithm was used to simulate them stochastically.Based on the ultra-long precipitation time series,the overall annual precipitation variation in Beijing shows an "increasing-stable-decreasing" trend,and there are six abrupt change points in the time domain,namely 1770,1813,1871,1893,1947 and 1999.In terms of cyclical variation,and there is a quasi-cyclical fluctuation pattern of 2.5-4 years,7-15 years,25-35 years,74 years and 95-100 years in the frequency domain,and the long time series show a strong advantage in both component analysis and stochastic simulation of precipitation series.