Method for determining runoff uncertainty key driving factors based on component type hydrological model

Abstract

The invention provides a method for determining runoff uncertainty key driving factors based on a component-type hydrological model. The method comprises the following steps: collecting and arranging observation of a research basin site, and analyzing daily rainfall data of a grid data set; a component type hydrological model is constructed based on the component type modeling framework, multiple model structures are configured for each hydrological sub-process, and the model structures of different hydrological sub-processes can be flexibly coupled; each model structure of different rainfall data and hydrological sub-processes is associated with an integer, discrete sampling is carried out on the rainfall and hydrological sub-processes in a specified integer range, and continuous sampling is carried out on model parameters in a specified parameter range; a dynamic Sobol'sensitivity analysis method is adopted, daily simulation runoff serves as an analysis target variable, daily total order sensitivity indexes of input data, a model structure and model parameters are calculated, and annual trend changes of the daily total order sensitivity indexes of the input data, the model structure and the model parameters are analyzed through a multi-year average sequence analysis method.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological forecasting, and in particular to a method for determining key driving factors of runoff uncertainty based on a component-based hydrological model. Background Art

[0002] Accurate runoff simulation and forecasting can provide effective technical support for flood and drought disaster forecasting and early warning, flood resource development and utilization, and water resource management and allocation. However, due to the dual limitations of human cognition and the complexity of natural systems, the actual precipitation-runoff process is difficult to accurately simulate, resulting in significant uncertainty in runoff results. A deeper understanding of the sources of runoff simulation uncertainty and its impact on the runoff process has important theoretical guidance for improving runoff simulation accuracy.

[0003] The three key links in the runoff simulation chain—input data, model structure, and model parameters—are the primary sources of runoff simulation uncertainty. While research on these three factors has yielded some results, most existing studies have limited their impact on runoff simulation uncertainty to individual assessments, lacking comprehensive comparative analyses of all three. Research is still needed to comprehensively analyze the impact of precipitation data, model structure, and model parameters on runoff simulation uncertainty and identify the dominant factors driving runoff simulation uncertainty. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for determining the key driving factors of runoff uncertainty based on a component hydrological model. The present invention uses a variety of precipitation data as model input, constructs a modular component hydrological model, performs basin hydrological simulation, and adopts a dynamic Sobol' sensitivity analysis method to comprehensively analyze the impact of precipitation data, model structure and model parameters on runoff simulation uncertainty. It can comprehensively consider the sources of uncertainty in hydrological simulation and provide a basis for improving runoff simulation.

[0005] In order to achieve the above technical features, the purpose of the present invention is to provide a method for determining key driving factors of runoff uncertainty based on a component-based hydrological model, comprising: Step 1: Collection and collation of rainfall input data: Collect and organize daily precipitation data from observations and reanalysis grid datasets at study basin sites; Step 2: Hydrological sub-process model structure configuration: A component-based hydrological model is constructed based on a component-based modeling framework. Multiple model structures are configured for each hydrological sub-process, and the model structures of different hydrological sub-processes can be flexibly coupled. Step 3, sampling of integer and continuous variables: Each model structure of different precipitation data and hydrological subprocesses is associated with an integer, the precipitation and hydrological subprocesses are discretely sampled within the specified integer range, and the model parameters are continuously sampled within the specified parameter range; Step 4: Using the dynamic Sobol' sensitivity analysis method, with daily simulated runoff as the analysis target variable, calculate the daily overall order sensitivity index of input data, model structure and model parameters; Step 5: Use the multi-year average sequence analysis method to analyze the intra-year trend changes of the input data, model structure, and daily total order sensitivity index of model parameters.

[0006] The collection and organization of rainfall input data in step 1 specifically includes selecting four sets of precipitation data, including station network measured data and reanalysis grid datasets. The reanalysis grid datasets include the CRUJRA dataset, which combines the reanalysis data of the Climate Research Institute of the University of East Anglia in the UK and Japan, the China Regional Surface Meteorological Element Dataset CMFD developed by the Institute of Tibetan Plateau Research of the Chinese Academy of Sciences, and the fifth-generation European Centre for Medium-Range Weather Forecasts Climate Land Surface Reanalysis Dataset ERA5-Land.

[0007] In step 2, a component-based hydrological model is constructed based on the Raven component-based modeling framework.

[0008] In step 2, a variety of model structures are specifically set for the five hydrological sub-processes of snow accumulation and melting, evapotranspiration, infiltration, soil flow and base flow. The model structures of the five different hydrological sub-processes can be flexibly coupled into a hydrological model.

[0009] The Sobol' sensitivity analysis method in step 4 is a global sensitivity analysis method based on variance, which decomposes the variance of the model output into the variance of the single parameter effect and the variance of the interaction between parameters. The total order sensitivity index ST K The most commonly used Sobol' index quantifies the parameter K The impact of the value of on the model output and the parameters K The effects of interactions with other parameters.

[0010] The overall order sensitivity index ST The specific calculation formula is: ; Where: y is the hydrological model output, for y The total variance of Ignoring the parameters KThe variance term generated by the effect; the analysis target variable of the Sobol' sensitivity analysis method can be either the runoff simulation value or the model performance index. The dynamic Sobol' sensitivity analysis method usually uses the daily simulated runoff output by the model as the analysis target variable to obtain daily sensitivity, reflecting how the impact of input factors on runoff uncertainty changes over time.

[0011] The step 5 specifically includes: The long-term sensitivity series is calculated as the average value of each calendar day to form a 365-day multi-year average daily series. The annual patterns of the daily total-order sensitivity index of input data, model structure and model parameters are analyzed. The annual patterns of the daily total-order sensitivity index of different hydrological sub-processes and hydrological model parameters are further analyzed.

[0012] The calendar days mentioned are: January 1, January 2...December 31.

[0013] The present invention has the following beneficial effects: 1. Existing technologies often conduct separate studies on input data, model structure, and model parameter uncertainties, lacking a comprehensive comparative analysis of the three. The present invention uses a variety of precipitation data as model input, constructs a modular component-based hydrological model, performs basin hydrological simulation, and adopts a dynamic Sobol' sensitivity analysis method to comprehensively analyze the impact of precipitation data, model structure, and model parameters on runoff simulation uncertainty. This method can comprehensively consider the sources of uncertainty in hydrological simulation and provide a basis for improving runoff simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below with reference to the accompanying drawings and examples.

[0015] Figure 1 It is a flow chart of the present invention.

[0016] Figure 2 Schematic diagram of the component hydrological model structure.

[0017] Figure 3 The multi-year daily average Sobol' total order sensitivity index of the input data, model structure and model parameters in the Gangtuo Basin.

[0018] Figure 4 It is the multi-year daily average Sobol' total order sensitivity index of the five hydrological sub-processes in the basin above Gangtuo.

[0019] Figure 5 It is the multi-year daily average Sobol' total order sensitivity index of 36 hydrological model parameters in the basin above Gangtuo. DETAILED DESCRIPTION

[0020] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0021] Example 1: Reference Figure 1 , a method for determining key drivers of runoff uncertainty based on a component-based hydrological model, including: Step 1: Collection and organization of rainfall input data. Four precipitation datasets were selected: station-based data and gridded reanalysis datasets. The gridded reanalysis datasets included the CRUJRA dataset (a combination of data from the Climatic Research Unit (CRU) at the University of East Anglia in the UK and Japanese reanalysis data); the China Meteorological Forcing Dataset (CMFD) developed by the Institute of Tibetan Plateau Research, Chinese Academy of Sciences; and the fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5-Land dataset.

[0022] Step 2: Configure the hydrological sub-process model structure. A component-based hydrological model is constructed based on the Raven component-based modeling framework. Multiple model structures are set for the five hydrological sub-processes: snow accumulation and melt, evapotranspiration, infiltration, subsoil flow, and baseflow. The model structures of different hydrological sub-processes can be flexibly coupled into a hydrological model.

[0023] Step 3: Integer and continuous variable sampling method. Each model structure of different precipitation data and hydrological subprocesses is associated with an integer. Precipitation and hydrological subprocesses are discretely sampled within the specified integer range, and model parameters are continuously sampled within the specified parameter range.

[0024] Step 4: Dynamic sensitivity analysis. The dynamic Sobol' sensitivity analysis method is used to calculate the daily total order sensitivity index of input data, model structure and model parameters with daily simulated runoff as the analysis target variable. The Sobol' sensitivity analysis method is a global sensitivity analysis method based on variance, which decomposes the variance of the model output into the variance of the effect of a single parameter and the variance of the interaction between parameters. The total order sensitivity index The most commonly used Sobol' index quantifies the parameter K The impact of the value of on the model output and the parameters K The effects of interactions with other parameters.

[0025] ; Where, y is the hydrological model output, for yThe total variance of Ignoring the parameters K The variance term generated by the effect of the Sobol' sensitivity analysis method can be either a simulated runoff value or a model performance indicator. The dynamic Sobol' sensitivity analysis method usually uses the daily simulated runoff output by the model as the analysis target variable to obtain daily sensitivity, reflecting how the impact of input factors on runoff uncertainty changes over time.

[0026] Step 5: Use multi-year average series analysis to analyze the intra-annual trends in the daily overall-order sensitivity index of the input data, model structure, and model parameters. The long-term sensitivity series is calculated as the average of each calendar day (January 1, January 2, … December 31), forming a 365-day multi-year average daily series. The intra-annual patterns of the daily overall-order sensitivity index of the input data, model structure, and model parameters are analyzed. The intra-annual patterns of the daily overall-order sensitivity index of different hydrological subprocesses and hydrological model parameters are further analyzed.

[0027] Example 2: The upper reaches of the Yangtze River are from its source area to Yichang, Hubei Province. The total length of the upper reaches is about 4,504 km, and the controlled basin area is about 1 million km. 2 The upper Yangtze River basin encompasses major tributaries such as the Jinsha, Yalong, Mintuo, and Jialing Rivers. The upper Yangtze River basin is a strategic water source for my country's water resources allocation system and the world's largest clean energy corridor. The basin's geographical environment, climatic conditions, and underlying surface types are complex and diverse, and precipitation and runoff characteristics vary spatiotemporally across subbasins. To account for the spatiotemporal variability of precipitation runoff processes under these complex underlying surface conditions, a subbasin in the upper Yangtze River basin, located above Gangtuo, was selected as the research object for runoff simulation uncertainty analysis.

[0028] Step 1: Collection and organization of rainfall input data. Daily precipitation data from observations and reanalysis grid datasets at basin sites were collected and organized. The measured data from the telemetry station network in the upper reaches of the Yangtze River covers daily precipitation data from 2011 to the present, with a total of 1,078 measured sites. The basin-averaged rainfall was calculated using the Thiessen polygon method. The CRUJRA dataset, which combines data from the Climatic Research Unit (CRU) of the University of East Anglia in the UK and Japanese reanalysis data, contains global precipitation data from 1901 to 2018 with a temporal resolution of 6 hours and a spatial resolution of 0.5°. The CMFD (China Meteorological Forcing Dataset), a surface meteorological element dataset for China developed by the Institute of Tibetan Plateau Research, Chinese Academy of Sciences, contains precipitation data for China from 1979 to 2018 with a temporal resolution of 3 hours and a spatial resolution of 0.1°. The fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF Reanalysis) climate land surface reanalysis dataset ERA5-Land (ECMWF Reanalysis) v5-Land), which contains global precipitation data from 1950 to the present, with a temporal resolution of 1 hour and a spatial resolution of 9 km. Based on various data sources, daily precipitation data from 2011 to 2018 was selected for this study.

[0029] Step 2: Configure the hydrological sub-process model structure. Figure 2 As shown in the figure, based on the Raven component modeling framework, 4, 2, 3, 3, and 3 different model structure options are set for the five hydrological subprocesses: snow accumulation and melt, evapotranspiration, infiltration, subsoil flow, and baseflow. The model structures of different hydrological subprocesses can be flexibly coupled into a hydrological model. The component hydrological model can provide a total of 4 × 2 × 3 × 3 × 3 = 216 model structures. The component hydrological model has 36 corresponding model parameters.

[0030] Step 3: Integer and continuous variable sampling method. Each model structure of different precipitation data and hydrological subprocesses is associated with an integer. Precipitation and hydrological subprocesses are discretely sampled within the specified integer range, and model parameters are continuously sampled within the specified parameter range.

[0031] Step 4: Use the dynamic Sobol' sensitivity analysis method, taking daily simulated runoff as the analysis target variable, to calculate the daily total order sensitivity index of input data, model structure and model parameters. The Sobol' sensitivity analysis method is a variance-based global sensitivity analysis method that decomposes the variance of the model output into the variance of the effect of a single parameter and the variance of the interaction between parameters. The total order sensitivity index ST K The most commonly used Sobol' index quantifies the parameterK The impact of the value of on the model output and the parameters K The effects of interactions with other parameters.

[0032] ; Where, y is the hydrological model output, for y The total variance of Ignoring the parameters K The variance term generated by the effect of the Sobol' sensitivity analysis method can be either a simulated runoff value or a model performance indicator. The dynamic Sobol' sensitivity analysis method usually uses the daily simulated runoff output by the model as the analysis target variable to obtain daily sensitivity, reflecting how the impact of input factors on runoff uncertainty changes over time.

[0033] Step 5: Use multi-year average series analysis to analyze the intra-annual trends in the daily overall-order sensitivity index of the input data, model structure, and model parameters. The long-term sensitivity series is calculated as the average of each calendar day (January 1, January 2, … December 31), forming a 365-day multi-year average daily series. The intra-annual patterns of the daily overall-order sensitivity index of the input data, model structure, and model parameters are analyzed. The intra-annual patterns of the daily overall-order sensitivity index of different hydrological subprocesses and hydrological model parameters are further analyzed.

[0034] Figure 3 This is the multi-year daily average Sobol' total-order sensitivity index for input data, model structure, and model parameters in the Gangtuo Basin. For the Gangtuo Basin, model structure is the most sensitive, followed by model parameters, and input data is the least sensitive, indicating that model structure is the primary factor driving runoff simulation uncertainty in the Gangtuo Basin.

[0035] Figure 4 The Sobol' index represents the multi-year daily average of the total sensitivity index for the five hydrological subprocesses in the Gangtuo Basin. In the Gangtuo Basin, the sensitivity of the winter and spring snow accumulation and melt processes exhibits significant intra-annual variability, with the intra-annual trend being opposite to the temperature trend, showing higher sensitivity in winter and lower sensitivity in summer. Soil flow is generally a relatively sensitive process, with slightly higher sensitivity in summer than in winter. The sensitivity of the evapotranspiration process exhibits lower intra-annual variability, with the choice of evapotranspiration model structure having a lower impact on runoff simulation in winter and spring than in summer and autumn. Similar to the snow accumulation and melt processes, the sensitivity of the baseflow process is higher in winter than in summer. The sensitivity of the infiltration process remains relatively low.

[0036] Figure 5The most sensitive parameters include the upper soil outflow index ( x 6) and the potential evapotranspiration correction factor for the evapotranspiration process ( x 8). The secondary sensitive parameters mainly include the maximum outflow rate of the upper soil layer during the intersoil flow process ( x 5) and soil thickness parameters ( x 29). The stage-sensitive parameters are mainly concentrated in the processes related to snowfall, snow accumulation and melting, including the critical temperature of snow melting ( x 26) Rain and snow critical temperature ( x 31). The above analysis shows that the parameters related to the soil flow process, evapotranspiration process, and snow accumulation and melting process are the main parameters affecting the uncertainty of runoff simulation.

Claims

1. A method for determining key drivers of runoff uncertainty based on a component-based hydrological model, characterized by: include: Step 1: Collection and collation of rainfall input data: Collect and organize daily precipitation data from observations and reanalysis grid datasets at study basin sites; Step 2: Hydrological sub-process model structure configuration: A component-based hydrological model is constructed based on a component-based modeling framework. Multiple model structures are configured for each hydrological sub-process, and the model structures of different hydrological sub-processes can be flexibly coupled. Step 3, sampling of integer and continuous variables: Each model structure of different precipitation data and hydrological subprocesses is associated with an integer, the precipitation and hydrological subprocesses are discretely sampled within the specified integer range, and the model parameters are continuously sampled within the specified parameter range; Step 4: Using the dynamic Sobol' sensitivity analysis method, with daily simulated runoff as the analysis target variable, calculate the daily overall order sensitivity index of input data, model structure and model parameters; Step 5: Use the multi-year average sequence analysis method to analyze the intra-year trend changes of the input data, model structure, and daily total order sensitivity index of model parameters.

2. The method for determining key driving factors of runoff uncertainty based on a component-based hydrological model according to claim 1, characterized in that: The collection and organization of rainfall input data in step 1 specifically includes selecting four sets of precipitation data, including station network measured data and reanalysis grid datasets. The reanalysis grid datasets include the CRUJRA dataset, which combines the reanalysis data of the Climate Research Institute of the University of East Anglia in the UK and Japan, the China Regional Surface Meteorological Element Dataset CMFD developed by the Institute of Tibetan Plateau Research of the Chinese Academy of Sciences, and the fifth-generation European Centre for Medium-Range Weather Forecasts Climate Land Surface Reanalysis Dataset ERA5-Land.

3. The method for determining key driving factors of runoff uncertainty based on a component-based hydrological model according to claim 1, characterized in that: In step 2, a component-based hydrological model is constructed based on the Raven component-based modeling framework.

4. The method for determining key drivers of runoff uncertainty based on a component-based hydrological model according to claim 3, characterized in that: In step 2, a variety of model structures are specifically set for the five hydrological sub-processes of snow accumulation and melting, evapotranspiration, infiltration, soil flow and base flow. The model structures of the five different hydrological sub-processes can be flexibly coupled into a hydrological model.

5. The method for determining key driving factors of runoff uncertainty based on a component-based hydrological model according to claim 1, characterized in that: The Sobol' sensitivity analysis method in step 4 is a global sensitivity analysis method based on variance, which decomposes the variance of the model output into the variance of the single parameter effect and the variance of the interaction between parameters. The total order sensitivity index ST K The most commonly used Sobol' index quantifies the parameter K The impact of the value of on the model output and the parameters K The effects of interactions with other parameters.

6. The method for determining key drivers of runoff uncertainty based on a component-based hydrological model according to claim 5, characterized in that: The overall order sensitivity index ST The specific calculation formula is: ； Where: y is the hydrological model output, for y The total variance of Ignoring the parameters K The variance term generated by the effect; the analysis target variable of the Sobol' sensitivity analysis method can be either the runoff simulation value or the model performance index. The dynamic Sobol' sensitivity analysis method usually uses the daily simulated runoff output by the model as the analysis target variable to obtain daily sensitivity, reflecting how the impact of input factors on runoff uncertainty changes over time.

7. The method for determining key driving factors of runoff uncertainty based on a component-based hydrological model according to claim 4, characterized in that: The step 5 specifically includes: The long-term sensitivity series is calculated as the average value of each calendar day to form a 365-day multi-year average daily series. The annual patterns of the daily total-order sensitivity index of input data, model structure and model parameters are analyzed. The annual patterns of the daily total-order sensitivity index of different hydrological sub-processes and hydrological model parameters are further analyzed.

8. The method for determining key driving factors of runoff uncertainty based on a component-based hydrological model according to claim 6, characterized in that: The calendar days mentioned are: January 1, January 2...December 31.

Images