A reconfigured precipitation downscaling simulation method based on high-resolution terrain features

Abstract

This invention discloses a reconfigurable precipitation downscaling simulation method based on high-resolution topographic features, comprising: encapsulating precipitation downscaling calculation modes and assigning them numbers; extracting DEM raster data of the target watershed; calculating the slope and aspect of the target watershed based on the DEM raster data; dividing the target watershed into regions according to topographic factors, and matching different downscaling modes in different regions for precipitation downscaling calculation. This invention ensures both the accuracy and reliability of the calculation results, while also solving the precipitation downscaling problem in high-altitude regions with significant topographic spatial heterogeneity. It primarily utilizes a watershed digital elevation model to quickly and automatically identify and match suitable precipitation downscaling methods in the target region, facilitating the direct application of topographic feature-based precipitation downscaling calculation methods, and rapidly obtaining detailed and personalized precipitation data for use as input to hydrological models, which can further promote the in-depth development of hydrological forecasting.

Description

Technical Field

[0001] The present invention relates to the field of hydrological and meteorological data, and in particular to a reconfigured precipitation downscaling simulation method based on high-resolution terrain features. Background Art

[0002] Precipitation is the most important input source for hydrological models and has a significant impact on the forecast accuracy of the models. In addition, precipitation is the most difficult climate variable to measure because it has a large temporal and spatial heterogeneity, which depends on the geomorphic characteristics of the region.

[0003] Downscaling technology is a bridge connecting global climate models and regional climate models or hydrological models, aiming to improve the accuracy of large-scale climate predictions at the local scale. With the deepening of climate change research and the increasing demand for refined understanding of surface hydrological processes, downscaling technology has become a research hotspot in climate science and water resources management. Precipitation downscaling methods mainly include dynamic downscaling and statistical downscaling. Dynamic downscaling usually uses simulations of global climate models to derive regional climate models, which can generate a set of fine-scale information. The resulting simulations are constrained by solid physics principles, but are computationally expensive and susceptible to biases in global climate models. Another method is statistical downscaling, which constructs a statistical model to draw the relationship between historical large-scale and small-scale meteorological data. Statistical downscaling has the advantages of low computational complexity and easy application. However, it relies on long-term, high-quality observations to establish strong statistical relationships.

[0004] In order to solve the impact of terrain spatial heterogeneity on precipitation and improve the accuracy of precipitation data, the terrain factors affecting precipitation are used as the basis to quantify the impact of elevation, slope and aspect on precipitation. Appropriate precipitation downscaling methods are used for calculation in different regions, thereby constructing a reconfigured precipitation downscaling simulation method based on high-resolution terrain features, which is exactly the problem that the present invention needs to solve. Summary of the invention

[0005] The purpose of the present invention is to provide a precipitation downscaling simulation method based on high-resolution terrain features, taking into account the impact of the spatial variability of terrain features on precipitation products, using high-resolution terrain data, dividing different regions according to the spatial distribution of elevation, and using appropriate downscaling methods to perform precipitation downscaling calculations in different regions. It has the advantages of high computational efficiency and objective and reasonable results, is conducive to the rapid calculation of precipitation downscaling calculation methods based on high-resolution terrain features, and is worthy of promotion.

[0006] To achieve the above functions, the present invention designs a reconfigured precipitation downscaling simulation method based on high-resolution terrain features. For the target basin, the following steps S1 to S4 are performed to complete the precipitation downscaling calculation of the target basin:

[0007] Step S1: Collect satellite precipitation data of the target basin, precipitation data of precipitation monitoring stations, and elevation data DEM, and construct various precipitation downscaling calculation models based on different precipitation downscaling calculation methods, and number them respectively;

[0008] Step S2: extracting the raster data of the elevation data DEM of the target watershed;

[0009] Step S3: Calculate the slope and slope direction of the target watershed based on the raster data of the target watershed elevation data DEM;

[0010] Step S4: Based on the terrain characteristics of the target watershed, the target watershed is vertically divided into five regions, namely, V1, V2, V3, V4, and V5, and each region of the target watershed is matched respectively; and different precipitation downscaling calculation modes are matched in different regions to perform precipitation downscaling calculations.

[0011] Beneficial effects: Compared with the prior art, the present invention provides a precipitation downscaling simulation method based on high-resolution terrain features, which quantifies the influence of elevation, slope and aspect on precipitation based on terrain factors affecting precipitation, and adopts appropriate precipitation downscaling methods for calculation in different regions, thereby proposing a precipitation downscaling calculation method reconfiguration technology based on high-resolution terrain features; this not only ensures the accuracy and reliability of the calculation results, but also solves the precipitation downscaling problem in high-altitude areas with obvious terrain spatial heterogeneity; the method is mainly based on terrain factors with high temporal and spatial resolution, the data source is stable and reliable, and the functional relationship between variables in the method is clear, which is conducive to quickly and automatically identifying and matching appropriate precipitation downscaling methods in the target area, simplifies the extraction steps through digital watershed technology, and ensures the objectivity and rationality of the results, which is conducive to the direct call of the precipitation downscaling calculation method based on terrain features, and quickly obtains fine and personalized precipitation data, which is used as the input of the hydrological model, and can further promote the in-depth development of hydrological forecasting. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 is a flow chart of a method for reconfiguring precipitation downscaling simulation based on high-resolution terrain features provided according to an embodiment of the present invention;

[0013] Figure 2 is a schematic diagram of bilinear interpolation provided according to an embodiment of the present invention;

[0014] Figure 3 is a DEM schematic diagram of target watershed elevation data provided according to an embodiment of the present invention;

[0015] Figure 4 is a schematic diagram of the target watershed slope calculated according to an embodiment of the present invention;

[0016] Figure 5 is a schematic diagram of the calculated target watershed slope aspect provided by an embodiment of the present invention;

[0017] Figure 6 is a schematic diagram of elevation partitioning of a target watershed provided according to an embodiment of the present invention;

[0018] Figure 7 It is a schematic diagram of the reconfiguration distribution of the precipitation downscaling pattern calculated according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0020] A reconfiguration precipitation downscaling simulation method based on high-resolution terrain features provided by an embodiment of the present invention is referred to Figure 1 The target basin is the Tunxi River Basin in Anhui Province, which has a catchment area of ​​2,696.76 square kilometers and is close to the southeast coast of China. It is located in the subtropical monsoon climate zone, with many frontal rains in spring and early summer, and many typhoons in summer and autumn. The direction of the monsoon circulation is basically orthogonal to the direction of the main mountain ranges, and the mountains play a role in blocking the northern cold current and typhoons. The terrain of the Tunxi River Basin is high in the west and low in the east, with the maximum, minimum and average elevations of 1,398m, 116m and 380m respectively, with a large relative height difference.

[0021] The original data of the target basin elevation data DEM uses the 90m resolution SRTM (Shuttle Radar Topography Mission) data provided by the geospatial data cloud. Perform the following steps S1 to S4 to complete the downscaling calculation of the precipitation in the target basin:

[0022] Step S1: Collect satellite precipitation data of the target basin, precipitation data of precipitation monitoring stations, and elevation data DEM, and construct various precipitation downscaling calculation models based on different precipitation downscaling calculation methods, and number them respectively;

[0023] Step S1 specifically includes the following steps:

[0024] Step S1.1: Based on the weighted average method, construct the first precipitation downscaling model, which is numbered WA;

[0025] The first precipitation downscaling model is as follows:

[0026] For a set of large-scale precipitation data X = {x1, x2, ..., x i …,x n}, where x irepresents the value of the ith large-scale precipitation data point, and n is the total number of large-scale data points; the large-scale precipitation data is converted into a data set B with m small-scale precipitation data points. 2, ...,b i ,…,b m}, where m>n, that is, the number of small-scale precipitation data points is greater than the number of large-scale precipitation data points;

[0027] For the small-scale precipitation data point b j , whose value is estimated by weighted average of large-scale precipitation data points:

[0028]

[0029] In the formula, w ij is the weight, representing the large-scale precipitation data point x i For small-scale precipitation data point b j The contribution degree, weight w ij It is determined according to the inverse distance method, as follows:

[0030]

[0031] Where, d ij is the distance between point j and point i, α is a positive parameter used to adjust the rate of distance influence. When the α value is large, the influence of distance decreases faster, usually α = 2;

[0032] The first precipitation downscaling model is simple and easy to use. It is used to calculate the average value in the region and can be used when the data in the region changes relatively smoothly and there is no obvious spatial trend.

[0033] Step S1.2: Based on the bilinear interpolation method, construct the second precipitation downscaling model, which is numbered BI;

[0034] The second precipitation downscaling model is as follows:

[0035] Based on the interpolation function f in Q 11 =(x1,y1),Q 12 =(x1,y2),Q 21 =(x2,y1),Q 22 =(x2,y2) The interpolation formula is as follows:

[0036]

[0037] Reference Figure 2 , according to the interpolation formula, obtain the value of the interpolation function f at any point P = f(x, y);

[0038] Bilinear interpolation is an extension of linear interpolation with an interpolation function of two variables. The core idea is to perform linear interpolation in two directions. Compared with simple averaging, bilinear interpolation can better take into account spatial changes and can be used when the spatial distribution is relatively uniform and the data points are relatively sparse.

[0039] Step S1.3: Based on geographically weighted regression, construct the third precipitation downscaling model, which is numbered GW;

[0040] The third precipitation downscaling model is as follows:

[0041]

[0042] In the formula, y i is the precipitation deviation, i.e., the deviation between the satellite precipitation data on grid i and the precipitation data at the precipitation monitoring station; a i is the regression coefficient of the constant term, n is the number of explanatory variables (including elevation, aspect, slope, etc.); x ik is the kth explanatory variable on grid i, a ik is the kth regression coefficient on grid i, s i is a random error;

[0043] The regression coefficient on grid i is as follows:

[0044] a(i)=(X T w(i)X) -1 X T w(i)b i

[0045] Where a(i) is the n×1 regression coefficient matrix on grid i, w(i) is an m×m diagonal matrix, i.e., the spatial weight matrix; X is the m×n spatial variable matrix (elevation, aspect, slope), b i is the precipitation deviation;

[0046] The spatial weight is determined by the Gaussian weight function:

[0047]

[0048] In the formula, w ij is the spatial weight value in the spatial weight matrix, b is the non-negative attenuation parameter describing the functional relationship between spatial weight and distance, d ij is the distance between point j and point i, when d ij When it is 0, the spatial weight value w ij is 1, that is, the spatial weight value w ij Reach the maximum; as the distance d between point j and point i ij The spatial weight value wij Gradually decreases; when the distance d between point j and point i ij When the spatial weight value w reaches the maximum ij Approaching 0.

[0049] Geographically weighted regression adds a spatial dimension to the traditional regression model, allowing model parameters to vary with geographic location, thereby better capturing spatial heterogeneity. It can be used in downscaling problems with strong spatial heterogeneity, that is, when precipitation in different regions responds differently to environmental factors.

[0050] Step S1.4: Based on hybrid geographically weighted regression, the fourth precipitation downscaling model is constructed, and the corresponding number is MG;

[0051] The fourth precipitation downscaling model is as follows:

[0052]

[0053] In the formula, y i is the precipitation deviation, k is the number of local terrain variables, l is the number of global terrain variables; x ik is the kth local terrain variable on grid i, a ik is the kth regression coefficient on grid i; x il is the lth global terrain variable on grid i, a il is the lth regression coefficient on grid i; θ i is a random error;

[0054] make Where X b is the elevation data DEM, slope, aspect local variables, W i is the weight matrix, which is determined by the Gaussian function:

[0055]

[0056] Where, d ij is the distance from point i to point j; b is the bandwidth;

[0057] Then the global coefficient and local coefficient are:

[0058]

[0059] In the formula, a′ i is the global coefficient of grid i, X a is the global terrain variable, I is the unit matrix, y i is the precipitation deviation, X is the m×n spatial variable matrix;

[0060] b′ i =Y(yi -X a a′ i )

[0061] In the formula, b′ i is the local coefficient of grid i, X a is the global terrain variable, y i is the precipitation deviation, a′ i is the global coefficient of grid i.

[0062] Hybrid Geographically Weighted Regression is an extension of Geographically Weighted Regression, which allows each regression parameter to have different spatial scales and captures spatial heterogeneity more flexibly. When the downscaling analysis needs to consider multiple influencing factors and the spatial influence ranges of these factors are different, hybrid Geographically Weighted Regression can provide more accurate local estimates.

[0063] Step S1.5: Based on the Bayesian model average, the fifth precipitation downscaling model is constructed, and the corresponding number is BM.

[0064] The fifth precipitation downscaling model is as follows:

[0065]

[0066] In the formula, Y is the target variable, X is the observed data, and H h is the hth model in the model set, P(Y|X,H h ) is a given model H h and the conditional probability of Y under data X, P(H h |X) is the model H under given data X h The posterior probability of .

[0067] Bayesian model averaging improves the accuracy and reliability of predictions by combining the predictions of multiple models and taking their uncertainties into account. It is suitable for dealing with model selection and downscaling problems with large uncertainties.

[0068] Step S2: extracting the raster data of the elevation data DEM of the target watershed; specifically comprising the following steps:

[0069] Step S2.1: filling the depression;

[0070] Step S2.2: Calculate flow direction;

[0071] Step S2.3: Calculate the confluence flow;

[0072] Step S2.4: Determine the watershed outlet site;

[0073] Step S2.5: extracting the target watershed;

[0074] The DEM diagram of the target watershed elevation data extracted in step S2 is referenced Figure 3 .

[0075] Step S3: Calculate the slope and slope direction of the target watershed based on the raster data of the target watershed elevation data DEM; specifically, the following steps are included:

[0076] Step S3.1: Calculate the 1km basin slope from the target basin elevation data DEM raster data, refer to Figure 4 ;

[0077] Step S3.2: Calculate the 1km watershed slope from the target watershed elevation data DEM raster data, refer to Figure 5 .

[0078] Step S4: Based on the terrain characteristics of the target watershed, the target watershed is vertically divided into five regions, namely, V1, V2, V3, V4, and V5, and each region of the target watershed is matched respectively; and different precipitation downscaling calculation modes are matched in different regions to perform precipitation downscaling calculations.

[0079] Step S4 specifically includes the following steps:

[0080] Step S4.1: Based on the terrain characteristics of the target watershed, the target watershed is vertically divided into five regions: V1, V2, V3, V4, and V5, and each region of the target watershed is matched respectively. The V1, V2, V3, V4, and V5 regions correspond to the following elevation data intervals: <100m; 100-500m; 500-1000m; 1000-3000m; >3000m; see the schematic diagram of the elevation division of the target watershed for reference. Figure 6 ;

[0081] Step 4.2: Match different precipitation downscaling calculation models in each region, referring to Figure 7 , as follows:

[0082] Region V1 contains plains and lowland areas with relatively gentle terrain changes and relatively low spatial heterogeneity, matching the WA precipitation downscaling calculation model, which is simple, fast and suitable for processing relatively uniform data;

[0083] The V2 region contains hilly and low mountainous areas, where the terrain undulation gradually increases and the spatial heterogeneity begins to become significant but is still relatively mild, matching the BI precipitation downscaling calculation model. Because the precipitation in this elevation range does not vary much in space, this model is sufficient to provide reasonable downscaling results.

[0084] The V3 region contains mid-mountain to high-mountain areas with dramatic topographic changes, complex climatic conditions, and large spatial heterogeneity, matching the GW precipitation downscaling calculation model. The relationship between environmental factors and precipitation in this elevation range will vary depending on the geographical location, and a more detailed method is needed to capture this spatial variability;

[0085] The V4 region includes high mountains and some plateau areas, with extreme climate, sensitive ecological environment, and extremely high spatial heterogeneity. It matches the MG precipitation downscaling calculation model. The precipitation in different geographical locations in this elevation range is affected by a variety of environmental factors, and the scale of the impact varies. This model can better handle this complexity.

[0086] The V5 region contains plateaus and mountain peaks, where data are scarce, the environment is extreme and highly variable, and traditional methods may be difficult to apply. To match the BM precipitation downscaling model, considering the scarcity and uncertainty of climate data in this elevation range, it may be necessary to integrate multiple models to more accurately reflect complex climate change. This model can improve the accuracy and reliability of predictions by integrating the advantages of multiple models.

[0087] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims

1. A precipitation downscaling simulation method based on high-resolution terrain features, characterized in that: For the target basin, perform the following steps S1 to S4 to complete the downscaling calculation of precipitation in the target basin: Step S1: Collect satellite precipitation data of the target basin, precipitation data of precipitation monitoring stations, and elevation data DEM, and construct various precipitation downscaling calculation models based on different precipitation downscaling calculation methods, and number them respectively; Step S1 specifically includes the following steps: Step S1.1: Based on the weighted average method, construct the first precipitation downscaling model, which is numbered WA; Step S1.2: Based on the bilinear interpolation method, construct the second precipitation downscaling model, which is numbered BI; Step S1.3: Based on geographically weighted regression, construct the third precipitation downscaling model, which is numbered GW; Step S1.4: Based on hybrid geographically weighted regression, the fourth precipitation downscaling model is constructed, and the corresponding number is MG; Step S1.5: Based on the Bayesian model average, the fifth precipitation downscaling model is constructed, and the corresponding number is BM; Step S2: extracting the raster data of the elevation data DEM of the target watershed; Step S3: Calculate the slope and slope direction of the target watershed based on the raster data of the target watershed elevation data DEM; Step S3 specifically includes the following steps: Step S3.1: Calculate the 1km watershed slope from the target watershed elevation data DEM raster data; Step S3.2: Calculate the 1km watershed slope from the target watershed elevation data DEM raster data; Step S4: Based on the terrain characteristics of the target watershed, the target watershed is vertically divided into five regions: V1, V2, V3, V4, and V5, and each region of the target watershed is matched respectively; and different precipitation downscaling calculation modes are matched in different regions to perform precipitation downscaling calculation; Step S4 specifically includes the following steps: Step S4.1: Based on the terrain characteristics of the target watershed, the target watershed is vertically divided into five regions: V1, V2, V3, V4, and V5, and each region of the target watershed is matched respectively, where the V1, V2, V3, V4, and V5 regions correspond to the elevation data intervals: <100m; 100-500m; 500-1000m; 1000-3000m; >3000m; Step 4.2, in each region, match different precipitation downscaling calculation modes, as follows: Region V1 includes plains and lowland areas, matching the WA precipitation downscaling calculation model; Region V2 contains hilly and low mountainous areas, matching the BI precipitation downscaling calculation model; Region V3 includes mid-mountain to high-mountain areas and matches the GW precipitation downscaling calculation model; Region V4 includes high mountains and some plateau areas, matching the MG precipitation downscaling calculation model; The V5 region includes the plateau and high mountain peaks, matching the BM precipitation downscaling pattern.

2. The method for reconfiguration precipitation downscaling simulation based on high-resolution terrain features according to claim 1, characterized in that: The first precipitation downscaling model described in step S1.1 is as follows: For a set of large-scale precipitation data X = {x1, x2, ..., x i …,x n }, where x i represents the value of the ith large-scale precipitation data point, and n is the total number of large-scale data points; the large-scale precipitation data is converted into a data set B with m small-scale precipitation data points. 2, ...,b i ,…,b m }, where m>n; For the small-scale precipitation data point b j , whose value is estimated by weighted average of large-scale precipitation data points: In the formula, w ij is the weight, representing the large-scale precipitation data point x i For small-scale precipitation data point b j The contribution degree, weight w ij It is determined according to the inverse distance method, as follows: Where, d ij is the distance between point j and point i, and α is a positive parameter.

3. The method for reconfiguration precipitation downscaling simulation based on high-resolution terrain features according to claim 1, characterized in that: The second precipitation downscaling model described in step S1.2 is as follows: Based on the interpolation function f in Q 11 =(x1,y1),Q 12 =(x1,y2),Q 21 =(x2,y1),Q 22 =(x2,y2) The interpolation formula is as follows: According to the interpolation formula, the value of the interpolation function f at any point P = f(x, y) is obtained.

4. The method for reconfiguration precipitation downscaling simulation based on high-resolution terrain features according to claim 1, characterized in that: The third precipitation downscaling model described in step S1.3 is as follows: In the formula, y i is the precipitation deviation, i.e., the deviation between the satellite precipitation data on grid i and the precipitation data at the precipitation monitoring station; a i is the regression coefficient of the constant term, n is the number of explanatory variables; x ik is the kth explanatory variable on grid i, a ik is the kth regression coefficient on grid i, ε i is a random error; The regression coefficient on grid i is as follows: a(i)=(X T w(i)X) -1 X T w(i)b i Where a(i) is the n×1 regression coefficient matrix on grid i, w(i) is an m×m diagonal matrix, i.e., the spatial weight matrix; X is the m×n spatial variable matrix, b i is the precipitation deviation; The spatial weight is determined by the Gaussian weight function: In the formula, w ij is the spatial weight value in the spatial weight matrix, b is the non-negative attenuation parameter describing the functional relationship between spatial weight and distance, d ij is the distance between point j and point i, when d ij When it is 0, the spatial weight value w ij is 1, that is, the spatial weight value w ij reach the maximum; As the distance d between point j and point i ij The spatial weight value w ij Gradually decreases; when the distance d between point j and point i ij When the spatial weight value w reaches the maximum ij Approaching 0.

5. The method for reconfiguration precipitation downscaling simulation based on high-resolution terrain features according to claim 1, characterized in that: The fourth precipitation downscaling model described in step S1.4 is as follows: In the formula, y i is the precipitation deviation, k is the number of local terrain variables, l is the number of global terrain variables; x ik is the kth local terrain variable on grid i, a ik is the kth regression coefficient on grid i; x il is the lth global terrain variable on grid i, a il is the lth regression coefficient on grid i; θ i is a random error; make Where X b is the elevation data DEM, slope, aspect local variables, W i is the weight matrix, which is determined by the Gaussian function: Where, d ij is the distance from point i to point j; b is the bandwidth; Then the global coefficient and local coefficient are: In the formula, a′ i is the global coefficient of grid i, X a is the global terrain variable, I is the unit matrix, y i is the precipitation deviation, X is the m×n spatial variable matrix; b′ i =And(and i -X a to' i ) In the formula, b′ i is the local coefficient of grid i, X a is the global terrain variable, y i is the precipitation deviation, a′ i is the global coefficient of grid i.

6. The method for reconfiguration precipitation downscaling simulation based on high-resolution terrain features according to claim 1, characterized in that: The fifth precipitation downscaling model described in step S1.5 is as follows: In the formula, Y is the target variable, X is the observed data, and H h is the hth model in the model set, P(Y|X,H h ) is a given model H h and the conditional probability of Y under data X, P(H h |X) is the model H under given data X h The posterior probability of .

7. The method for reconfiguration precipitation downscaling simulation based on high-resolution terrain features according to claim 1, characterized in that: Step S2 extracts the raster data of the elevation data DEM of the target watershed, which specifically includes the following steps: Step S2.1: filling the depression; Step S2.2: Calculate flow direction; Step S2.3: Calculate the confluence flow; Step S2.4: Determine the watershed outlet site; Step S2.5: Extract the target watershed.

Images