Based on the daily precipitation data of 27 meteorological stations from 1960 to 2009 in the Huaihe River Basin, spatio-temporal trend and statistical distribution of extreme precipitation events in this area are analyzed. Annual maximum series (AM) and peak over threshold series (POT) are selected to simulate the probability distribution of extreme precipitation. The results show that positive trend of annual maximum precipitation is detected at most of used stations, only a small number of stations are found to depict a negative trend during the past five decades, and none of the positive or negative trend is significant. The maximum precipitation event almost occurred in the flooding period during the 1960s and 1970s. By the L-moments method, the parameters of three extreme distributions, i.e., Generalized extreme value distribution (GEV), Generalized Pareto distribution (GP) and Gamma distribution are estimated. From the results of goodness of fit test and Kolmogorov-Smirnov (K-S) test, AM series can be better fitted by GEV model and POT series can be better fitted by GP model. By the comparison of the precipitation amounts under different return levels, it can be found that the values obtained from POT series are a little larger than the values from AM series, and they can better simulate the observed values in the Huaihe River Basin.
Sensitivity analysis of hydrological model is the key for model uncertainty quantification. However, how to effectively validate model and identify the dominant parameters for distributed hydrological models is a bottle-neck to achieve parameters optimization. For this reason, a new approach was proposed in this paper, in which the support vector machine was used to construct the response surface at first. Then it integrates the SVM-based response surface with the Sobol’ method, i.e. the RSMSobol’ method, to quantify the parameter sensitivities. In this work, the distributed time-variant gain model (DTVGM) was applied to the Huaihe River Basin, which was used as a case to verify its validity and feasibility. We selected three objective functions (i.e. water balance coefficient WB, Nash-Sutcliffe efficiency coefficient NS, and correlation coefficient RC) to assess the model performance as the output responses for sensitivity analysis. The results show that the parameters g1 and g2 are most important for all the objective functions, and they are almost the same to that of the classical approach. Furthermore, the RSMSobol method can not only achieve the quantification of the sensitivity, and also reduce the computational cost, with good accuracy compared to the classical approach. And this approach will be effective and reliable in the global sensitivity analysis for a complex modelling system.
Using the daily temperature data of 95 meteorological stations from Sichuan- Chongqing Region and its surrounding areas, this paper adopted these methods (e.g., linear regression, trend coefficient, geographical statistics, gray relational analysis and spatial analysis functions of GIS) to analyze the relations of temperature variability with topography, latitude and longitude. Moreover, the rank of gray correlation between temperature variability and elevation, longitude, latitude, topographic position and surface roughness also was measured. These results indicated: (1) The elevation affected temperature variability most obviously, followed by latitude, and longitude. The slope of the linear regression between temperature change rate and elevation, latitude and longitude was 0.4142, 0.0293 and -0.3270, respectively. (2) The rank of gray correlation between temperature change rate and geographic factors was elevation > latitude > surface roughness > topographic position > longitude. The gray correlation degree between temperature change rate and elevation was 0.865, followed by latitude with 0.796, and longitude with 0.671. (3) The rate of temperature change enhanced with the increase of elevation. Especially, the warming trend was significant in the plateau and mountain areas of western Sichuan, and mountain and valley areas of southwestern Sichuan (with the warming rate of 0.74℃/10a during the 1990s). However, there was a weak warming trend in Sichuan Basin and its surrounding low mountain and hilly areas. (4) The effects of latitude on temperature change rate presented the specific regulation, which the warming rate of low-latitude areas was more significant than that of high-latitude areas. However, they were consistent with the regulation that the increasing of low temperature controlled most of the warming trend, due to the effects of terrain and elevation on annual mean temperature. (5) Basically, temperature variability along longitude direction resulted from the regular change of elevation along longitude. It was suggested that, in Sichuan-Chongqing Region, special features of temperature variability largely depended on the terrain complexity (e.g., undulations, mutations and roughness). The elevation level controlled only high or low annual mean temperature and the range of temperature change rate in the macro sense.
Based on China homogenized land surface air temperature and the National Centers for Environmental Prediction/Department of Energy (NCEP/DOE) Atmospheric Model Intercomparison Project (AMIP)-Ⅱ Reanalysis data (R-2), the main contributors to surface air temperature increase in Southeast China were investigated by comparing trends of urban and rural temperature series, as well as observed and R-2 data, covering two periods of 1954-2005 and 1979-2005. Results from urban-rural comparison indicate that urban heat island (UHI) effects on regional annual and autumn minimum temperature increases account for 10.5% and 12.0% since 1954, but with smaller warming attribution of 6.2% and 10.6% since 1979. The results by comparing observations with R-2 surface temperature data suggest that land use change accounts for 32.9% and 28.8% in regional annual and autumn minimum temperature increases since 1979. Accordingly, the influence of land use change on regional temperature increase in Southeast China is much more noticeable during the last 30 years. However, it indicates that UHI effect, overwhelmed by the warming change of background climate, does not play a significant role in regional warming over Southeast China during the last 50 years.
By decomposing and reconstructing the runoff information from 1965 to 2007 of the hydrologic stations of Tuotuo River and Zhimenda in the source region of the Yangtze River, and Jimai and Tangnaihai in the source region of the Yellow River with db3 wavelet, runoff of different hydrologic stations tends to be declining in the seasons of spring flood, summer flood and dry ones except for that in Tuotuo River. The declining flood/dry seasons series was summer > spring > dry; while runoff of Tuotuo River was always increasing in different stages from 1965 to 2007 with a higher increase rate in summer flood seasons than that in spring ones. Complex Morlet wavelet was selected to detect runoff periodicity of the four hydrologic stations mentioned above. Over all seasons the periodicity was 11-12 years in the source region of the Yellow River. For the source region of the Yangtze River the periodicity was 4-6 years in the spring flood seasons and 13-14 years in the summer flood seasons. The differences of variations of flow periodicity between the upper catchment areas of the Yellow River and the Yangtze River and between seasons were considered in relation to glacial melt and annual snowfall and rainfall as providers of water for runoff.
Drought is one of the most destructive disasters in the Lancang River Basin, which is an ungauged basin with strong heterogeneity on terrain and climate. Our validation suggested the version-6 monthly TRMM multi-satellite precipitation analysis (TMPA; 3B43 V.6) product during the period 1998 to 2009 is an alternative precipitation data source with good accuracy. By using the standard precipitation index (SPI), at the grid point (0.25°×0.25°) and sub-basin spatial scales, this work assessed the effectiveness of TMPA in drought monitoring during the period 1998 to 2009 at the 1-month scale and 3-months scale; validated the monitoring accuracy of TMPA for two severe droughts happened in 2006 and 2009, respectively. Some conclusions are drawn as follows. (1) At the grid point spatial scale, in comparison with the monitoring results between rain gauges (SPI1g) and TMPA grid (SPI1s), both agreed well at the 1-month scale for most of the grid points and those grid points with the lowest critical success index (CSI) are distributed in the middle stream of the Lancang River Basin. (2) The same as SPI1s, the consistency between SPI3s and SPI3g is good for most of the grid points at the 3-months scale, those grid points with the lowest were concentrated in the middle stream and downstream of the Lancang River Basin. (3) At the 1-month scale and 3-months scale, CSI ranged from 50% to 76% for most of the grid points, which demonstrated high accuracy of TMPA in drought monitoring. (4) At the 3-months scale, based on TMPA basin-wide precipitation estimates, though we tended to overestimate (underestimate) the peaks of dry or wet events, SPI3s detected successfully the occurrence of them over the five sub-basins at the most time and captured the occurrence and development of the two severe droughts happened in 2006 and 2009. This analysis shows that TMPA has the potential for drought monitoring in data-sparse regions.