ATHYS模型及其在巢湖流域丰乐河水文过程模拟中的应用

ATHYS model and its application on simulation of rainfall-runoff on Fengle catchment of Lake Chao

以巢湖典型支流丰乐河流域为例,利用分布式水文模型ATHYS对流域降雨—径流过程进行模拟,以期为丰乐河流域水土保持、防洪、水资源利用和非点源污染防治等提供科学依据。ATHYS模型将流域分为大小相同的空间小单元,以库模型进行产流计算,以滞后演算法(lag and route)方法进行汇流计算,并以Nash and Sutcliffe指数作为检验模拟结果的标准,模拟流域出口断面日径流量,初步分析流域径流实测值与模拟值的年际变化。模型模拟结果表明,在率定期和验证期模拟与实测径流量的总Nash and Sutcliffe值(Nash值)都可以达到0.80以上,相关系数均可达0.89以上。除了部分峰值和基流,模型能较好地模拟丰乐河的径流过程。模型模拟径流总量的相对误差在率定期为23.1%,验证期为34.6%。流域降雨量、径流量的年内、年际变化均比较明显,径流实测值和模拟值的年际变化也都比较大。综合来看,模型比较适用于所研究流域的日径流模拟。

The rainfall-runoff on the Fengle catchment in the Lake Chao basin was simulated using ATHYS model to provide a scientific basis for the control of erosion, flooding, non-point source pollution, and water utili- zation in the Fengle catchment area. The model is based on Digital Elevation Model （DEM）, land use, soil and climate data of the catchment and the daily discharge data from the outlet Taoxi station. ATHYS model is a dis- tributed hydrological model. The catchment was divided into the same size （90×90 m） meshes that describe DEM, landuse and soil properties. The rainfall at the location of the mesh was interpolated using the Thiessen method and the rainfall data was based on the daily rainfall measurement from eight rain gauge stations. Fengle basin is one of the biggest tributaries of Lake Chao. The basin area is 1500 km^2, The mainstream of the river is about 50 km long and its elevation is from 6 to 463 m. Fengle catchment belongs to the humid subtropical monsoon climate with an average of annual rainfall about 1000 meters. ATHYS model was developed based on a variety of produc- tion and transfer computing models. The runoff from each mesh was calculated using a reservoir model and the runoff volume was routed to the outlet using a lag and route method to get the hydrograph and to obtain the dis- charge volume. STO, INF, 09, ds, the primary parameters in production model, represent the maximum volume of the reservoir （mm）, coefficient of infiltration （mm·h^-1）, coefficient of subsurface runoff （no dimension）, coefficient of soil infiltration （1·d^-1）, respectively. There were two parameters in the transfer function： Vo and Ko. Vo is the speed of propagation （m·s^-1）, and K0 is a diffusion coefficient without dimension. V0 and K0 are assumed here to be identical for each mesh and must be calibrated from rainfall and discharge data. There are various types of lan- duse in the catchment, such as flee, agricultural, nature, urban and so on. Two landuse types, rice and other, were used in this model. The model was run in a continuous way at the daily timescale. The discharge at the outlet was available for ten years duration in which five years＇s discharge data were randomly selected for ＂the calibration purpose. The comparison observed and simulated discharge was based on Nash and Sutcliffe index. Data from other five years were used for validation. The parameter was calibrated using trial and error method. The model demonstrated a good.calibration and validation result with the total Nash coefficients being 0.86 and 0.80, respec- tively. The maximum Nash coefficient in calibration was 0.92, while the minimum value was 0.68. In validation, the maximum and minimum of Nash coefficient were 0.92 and 0.68. Except for one value, the correlation coeffi- cient between observed and simulated discharge was over 0.9 for both calibration and validation periods. The model can generally well reproduce the hydrograph except for some peaks and some baseflow periods. The rela- tive error of the annul runoff volumes between observed and simulated runoff was about 23.1% in calibration pe- riod and 34.6% in validation period. The rainfall and runoff both had a great variation by year, while a corre- sponding relationship was observed between them. The eigenvalues of rainfall and runoff volume had a great in- ter-annual variation and the average value of annual rainfall was 1238 m and the variation coefficient was 20.2%. The average value of annual runoff depth was 0.52 and the variation coefficient of annual runoff depth reached to 47.0%. There is no big difference in the average of annual runoff volume and daily peak runoff volume between observation and simulation values, but the observed and simulated runoff volume both showed a great inter-annual variation in total runoff volume and daily peak runoff volume. ATHYS model has proven to be able to simulate runoff from Fengle catchment in a continuous way.