湖北强降水频次时空特征及基于GWR模型的地形关系分析

利用1983—2017年湖北省汛期74个国家气象站逐小时降水数据，按长、短历时强降水事件分类研究强降水频次的时空特征，并运用普通最小二乘法(OLS)、地理加权回归(GWR)方法定量探讨强降水频次与地形因子之间的关系。(1) 湖北汛期长、短历时强降水年频次周期变化明显，年代际变化(≥10 a)存在1个主振荡模态，年代际以下尺度(＜10 a)存在2个主振荡模态。(2) 长历时强降水旬频次在梅雨期达到顶峰，盛夏期减少，而短历时则在梅雨结束后的7月中旬出现跃升；长、短历时强降水频次日变化曲线都为单峰结构。(3) 湖北长、短历时强降水高频次站点多出现在地面存在准常定中尺度辐合线或涡旋的特定地形条件下。(4) 地理加权回归较传统普通最小二乘法显著提高了强降水频次与海拔高度、坡度的拟合效果。结合拟合系数显著性检验分析，地理加权回归不适用于样本偏少、站点稀疏的鄂西山地，更适用于多中小尺度地形的湖北中东部。(5) 地理加权回归模型中，海拔高度与长历时强降水频次在大别山东麓西侧正相关最大，在大别山西麓南侧负相关最大，坡度则正相反；海拔高度、坡度对短历时强降水频次的最大影响在大别山东麓西侧以及沿长江干流的低洼城市带武汉-黄石地区，武汉站分别为-0.20次/米、6.43次/度，这里地形坡度影响远超海拔高度。 

SPATIO-TEMPORAL CHARACTERISTICS OF FREQUENCY OF HEAVY RAINFALL EVENTS IN HUBEI PROVINCE AND ANALYSIS OF TOPOGRAPHIC RELATIONSHIP BASED ON GWR

Based on the hourly heavy precipitation data from 74 national meteorological observation stations during the rainy seasons (May to September) from 1983 to 2017 in Hubei Province, the heavy precipitation events were classified into long and short duration heavy precipitation events according to their temporal and spatial characteristics. The methods of ordinary least squares (OLS) and geographically weighted regression (GWR) were used to quantitatively explore the relationship between heavy rainfall frequency and terrain factor. The results showed that the annual frequency of short duration heavy precipitation in extended rainy seasons in Hubei Province had obvious periodic variation, and there was one main oscillation mode in interdecadal variation (≥10 a), and two main oscillation modes in the variation of periods with a timescale shorter than interdecadal (< 10 a). The ten-day frequency of longduration heavy precipitation reached its peak in Meiyu periods, and decreased in midsummer periods, while the frequency of short duration heavy rainfall jumped in the middle of July after the end of Meiyu periods. The diurnal variation curves of long and short duration heavy rainfall were both single peak structures. The high-frequency stations of long and short-duration heavy precipitation in Hubei Province mostly appeared under specific terrain conditions where there were quasi-constant mesoscale convergence lines or vortices on the ground. Compared with the traditional OLS method, GWR significantly improved the fitting effect of heavy rainfall frequency, altitude and slope. Combined with the analysis of the significance of the fitting coefficients, GWR was not suitable for the mountainous areas with small samples and sparse sites in western Hubei Province, and more suitable for the central and eastern parts of Hubei Province which is home to medium and small-scale topography. In the geographically weighted regression model, the positive correlation between the altitude and the frequency of long-duration heavy precipitation was the largest on the west side of the eastern foot of Dabie Mountain, and the negative correlation was the largest on the south side of the western foothill of Dabie Mountain, while it was the opposite with slopes. Altitude and slope had a positive effect on the frequency of short-duration heavy precipitation. The greatest impact was on the west side of the eastern foot of the Dabie Mountains and the low-lying urban belt along the mainstream of the Yangtze River in the Wuhan-Huangshi region. They were -0.20 times/m and 6.43 times/degree respectively at the Wuhan station where the influence of terrain slope was much higher than altitude.