A method and system for predicting sudden drought in soil water

Abstract

The present invention discloses a method and system for predicting sudden soil water droughts. The method includes: obtaining historical index data of each grid point, and obtaining soil water drought situation data corresponding to each grid point at different times according to a time step ΔT; identifying soil water drought events through time series connection based on the soil water drought situation data corresponding to each grid point at different times, and obtaining data on the area covered by each soil water drought event in the study area changing with time; constructing a random forest model for predicting the drought formation time F with the average precipitation P, average temperature T, and distance value G as input variables and the drought formation time F as the output variable according to the data on the area covered by each soil water drought event changing with time; statistically obtaining drought expansion rate data of each grid point based on the soil water drought situation data corresponding to each grid point at different times; and predicting data on the development trend of the drought area. The present invention can improve the accuracy of predicting sudden soil water drought events.

Description

Technical Field

[0001] The present invention relates to a method for predicting drought events, and in particular to a method and system for predicting soil water flash droughts. Background Art

[0002] Global warming has caused a sharp increase in global temperatures, resulting in significant changes in the energy budget balance and material exchange processes among different spheres, bringing irreversible impacts to the climate system. Drought events are greatly affected by rising temperatures, with their evolution speed gradually accelerating and intensity gradually increasing. More severe cases will form extreme flash droughts. Extreme flash droughts are characterized by greater loss degrees and lower prediction accuracies compared to ordinary drought events.

[0003] How to predict extreme flash droughts is an important topic for current researchers facing drought problems. Many studies analyze and evaluate entire drought events by analyzing the characteristics of droughts at individual grid points, ignoring the spatio-temporal characteristics of droughts as three-dimensional phenomena. Additionally, due to the suddenness of flash drought events, how to accurately predict flash drought events is also an issue that needs to be further addressed in related research. Summary of the Invention

[0004] Object of the Invention: Aiming at the above problems, the present invention proposes a method and system for predicting soil water flash droughts, which can improve the accuracy of predicting soil water flash drought events and enhance the early warning level of flash droughts in the region.

[0005] Technical Solution: The technical solution adopted by the present invention is a method for predicting soil water flash droughts, including:

[0006] Dividing the research area into grids, and obtaining data on the soil water drought conditions corresponding to each grid point at different times according to the historical index data of each grid point at a time step of ΔT;

[0007] According to the data on the soil water drought conditions corresponding to each grid point at different times, identifying soil water drought events through time series connection to obtain data on the areas covered by each soil water drought event in the research area changing with time; according to the data on the areas covered by each soil water drought event changing with time, taking the average precipitation P, average temperature T, and distance value G as input variables and the drought formation time F as the output variable, constructing a random forest model for predicting the drought formation time F; the distance value G is the minimum distance between the grid point and the edge of the drought patch;

[0008] According to the data on the soil water drought conditions corresponding to each grid point at different times, statistically obtaining the drought expansion rate data of each grid point;

[0009] Predicting the development trend data of the drought area, including: setting G maxis the maximum value of the distance value G in the identified soil water drought event; for any grid point x, if G ≤ G max , according to the drought expansion rate data of each grid point, the expansion rate η of grid point x is obtained. According to the average precipitation P, average temperature T, and distance value G of grid point x, the drought formation time F of grid point x is output through the random forest model; all grid points with F ≤ ΔT are selected, and grid points with an expansion rate η greater than the probability threshold are determined to be drought-stricken at the next moment; the prediction time T is divided into several segments according to the step size ΔT, and the calculation of this step is repeated until the prediction ends at time T to obtain the drought area development trend data;

[0010] According to the predicted drought area development trend data, calculate the drought cumulative area-time normalization curve, and judge the suddenness degree of drought according to the drought cumulative area-time normalization curve.

[0011] The exponential data uses soil water moisture SM.

[0012] Use soil water moisture SM to judge the soil water drought situation. By calculating the percentile value of soil water moisture SM, if the percentile value of soil water moisture SM is lower than the set threshold, it is judged as soil water drought.

[0013] Using soil water moisture SM to judge the soil water drought situation includes the following process: extract the daily soil water moisture data within the selected time in the study area to obtain the average soil water moisture data for a period of time, and then calculate the marginal distribution function of the average soil water moisture data; use the K-S test method and root mean square error method to fit the best probability distribution function; according to the best probability distribution function, convert the soil water moisture value into a quantile, and if the quantile of soil water moisture SM is lower than the set threshold, it is judged as soil water drought.

[0014] The time-course connection includes the following process: judge the coverage area of drought events in two adjacent time periods. When the overlap degree of the coverage areas of two droughts in adjacent time periods is higher than the set value, it is judged as the same drought event.

[0015] The ordinate of the drought cumulative area-time normalization curve is the normalized drought coverage area, and the abscissa is the normalized time. Judging the suddenness degree of drought according to the drought cumulative area-time normalization curve includes: calculating the slope of the drought cumulative area-time normalization curve, and judging whether the slope is greater than the threshold. If so, the predicted drought type is sudden drought.

[0016] The formula for the normalized drought coverage area is:

[0017]

[0018] In the formula, N Ais the normalized drought coverage area; n is the number of divided time periods, and A j is the maximum drought coverage area projected in the j-th time period; is the sum of the maximum drought coverage areas projected in each time period;

[0019] The so-called normalized time, the calculation formula is:

[0020]

[0021] In the formula, N T is the normalized time; T j is the j-th time period of the drought; is the total sum of each time period of the drought event.

[0022] The present invention proposes a soil water flash drought prediction system, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the soil water flash drought prediction method described above is implemented.

[0023] The present invention proposes a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the soil water flash drought prediction method described above is implemented.

[0024] The present invention also proposes a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the soil water flash drought prediction method described above is implemented.

[0025] Beneficial effects: Compared with the prior art, the present invention has the following advantages: The present invention identifies the change of the area covered by the drought event over time through time-course connection, constructs a random forest model to predict the drought occurrence time; by introducing the expansion rate, predicts the drought development trend, and combines the prediction results of the random forest model to judge the global drought situation. Compared with the existing drought prediction algorithms, the present invention can synchronously predict the global drought data, greatly improving the accuracy for the suddenness of drought development, and can intuitively predict the future trend of soil water drought and judge the drought type. Brief Description of the Drawings

[0026] Figure 1 is the flowchart of the soil water flash drought prediction method described in the present invention;

[0027] Figure 2 is the schematic diagram of the predicted drought area development trend data described in the present invention. Detailed Embodiment

[0028] The technical solutions of the present invention will be further described below with reference to the drawings and embodiments.

[0029] The soil water sudden drought prediction method described in the present invention has a flowchart as shown in Figure 1 and includes the following steps, which are executed according to the processes described in the steps, and there is no specific order between the steps:

[0030] Step 1: Grid the research area, obtain the index data corresponding to each grid point within a period of time, determine whether it is soil water drought according to the index data, and obtain the soil water drought situation data corresponding to each grid point at different times according to the time step ΔT (two weeks).

[0031] In this embodiment, the soil water moisture SM is used to judge soil water drought. Calculate the percentile value of the soil water moisture SM. If the soil water moisture SM is lower than the set threshold of 0.3, it is judged as soil water drought.

[0032] Another method of using the soil water moisture SM to judge soil water drought is to extract the daily soil water moisture data within the selected time in the research area to obtain the average soil water moisture data within two weeks. Subsequently, calculate its marginal distribution function, and the function type can be selected from exponential distribution, lognormal distribution, generalized extreme value distribution, etc. The best probability distribution function is fitted by the K-S test method and the root mean square error method (RMSE). After obtaining the probability distribution function, the soil water moisture value can be converted into a percentile value to facilitate the identification of sudden drought. In this embodiment, the grid points with percentile values lower than 0.3 are judged as drought. For different regions, its critical value should change according to local standards.

[0033] Step 2: According to the soil water drought situation data corresponding to each grid point at different times, identify soil water drought events through time series connection to obtain the data of the area covered by each soil water drought event in the research area changing with time.

[0034] The time series connection includes the following processes: (1) In the same time period (such as within one week), for a grid point identified as drought, that is, a grid point with a percentile value lower than 0.3, if there are also grid points identified as drought in the surrounding 3×3 area, these grid points are regarded as covered by the same drought event until there are no new drought grid points within the 3×3 range of the edge grid points of the drought event. After filtering out the drought events with too small coverage area, different drought events occurring in the research area at this time period and their covered areas can be divided. (2) In the next time period, perform the same operation to obtain different drought events occurring in the next time period and their covered areas. (3) Judge the drought coverage areas of the two time periods. It is considered that when the overlap degree of the drought coverage areas in two consecutive time periods is higher than a certain value, these two droughts can be judged as the expansion or reduction of the same drought event over time. The minimum drought coverage area A min and the minimum overlap degree A oThe selection of these two parameters is very important. They may have different values in different regions, so parameter calibration is carried out. Different combinations of the minimum drought coverage area and overlap degree parameters are set, and the obtained evolution process of drought events is fitted to the real evolution process of drought events in the study area. The parameter combination with the best fitting effect is selected as the parameter combination for the study area.

[0035] Step 3: According to the soil water drought situation data corresponding to each grid point at different times, the drought expansion rate data of each grid point is statistically obtained.

[0036] The drought expansion rate data refers to the probability that the central grid point in a 3×3 area at a certain moment will become drought after ΔT under the condition of whether the surrounding 8 grid points are drought or not. ΔT is the set time step. That is, when there are 0-8 drought grid points around at the same moment, the probabilities of the central grid point becoming drought under these 9 situations are respectively. The following is an example of the calculation method of the drought expansion rate when there are 0, 1, and 2 drought grid points around.

[0037] When there are 0 drought grid points around:

[0038]

[0039] N0 represents the number of times that when the central grid point becomes drought, none of the surrounding eight grid points become drought within ΔT (two weeks), or become drought at the same time as the central grid point; N represents the total number of times that the central grid point becomes drought.

[0040] When there is 1 drought grid point i around:

[0041]

[0042] M i represents the number of times that grid point i becomes drought in the total time series; N i represents the number of times that the central grid point changes from non-drought to drought within ΔT (two weeks) after drought grid point i becomes drought (excluding the number of times when the two grid points become drought at the same time).

[0043] When there are 2 drought grid points i, j around:

[0044]

[0045] M j represents the number of times that grid point j becomes drought in the total time series; N j represents the number of times that the central grid point changes from non-drought to drought within ΔT (two weeks) after drought grid point j becomes drought (excluding the number of times when the two grid points become drought at the same time); A ij represents the number of times that within ΔT two weeks, both grid points i and j become drought, and the central grid point becomes drought after the two grid points; B ijDenote the number of times that droughts occurred in both grid points \(i\) and \(j\) within two weeks of \(\Delta T\).

[0046] Step 4: According to the data on the change of the area covered by each soil water drought event over time, introduce a random forest model to predict the drought formation time \(F\).

[0047] Obtain the average values of precipitation and temperature of all grid points that have experienced droughts outside the drought patch \(S\) among the identified three-dimensional drought events. Using the average precipitation \(P\), average temperature \(T\), and distance value \(G\) as independent variables, and the drought formation time \(F\) as the dependent variable, import them into the random forest model. Randomly select 70% of the samples as the training set and 30% of the samples as the test set, and a prediction model for the drought formation time can be obtained. The distance value \(G\) is the minimum distance from the grid point to the edge of the drought patch \(S\). The average precipitation \(P\) and average temperature \(T\) refer to the average precipitation and average temperature within \(\Delta T\) of each grid point. The drought formation time \(F\) is the time required from the start of the drought patch \(S\) of an identified three-dimensional drought event until the grid point itself becomes drought.

[0048] It should be noted that if a grid point itself is within the drought patch, it is generally not included in the training set and the test set. If, after the area of the drought event changes, this grid point (the grid point within the drought patch) becomes non-drought and then re-enters the drought before the end of this drought event, the drought formation time of this grid point is the time when it changes from non-drought to drought; if there is a grid point that repeatedly changes from non-drought to drought and then to non-drought within a drought event, the drought formation time takes the minimum value between the time required from the formation of the drought patch until it itself becomes drought and the time required to change from non-drought to drought.

[0049] Step 5: Predict the data on the development trend of the drought area.

[0050] For a drought patch that appears in the study area, when its area exceeds the set minimum drought coverage area \(A\) min , the development trend of its area can be predicted.

[0051] Set \(G\) max , as the maximum value of the distance value \(G\) in the identified historical drought events. When the distance of any grid point \(G > G\) max , this grid point is not included in the subsequent calculations. Use the expansion rate \(\eta\) and formation time \(F\) of each grid point to jointly determine whether the grid point will experience drought at the next moment. Divide the prediction time \(T\) into several segments according to the time step \(\Delta T\) (two weeks), set the probability threshold at 60%, screen out all grid points with \(F\leq\Delta T\), and determine that the grid points with an expansion rate \(\eta\) greater than or equal to 60% will experience drought at the next moment. Repeat the calculation of this step until the prediction ends at time \(T\). As Figure 2 shown, it is the prediction situation of the regional drought in a \(\Delta T\) period. The predicted area of the drought in this period is the combined part of green and red, and it is also the initial area for the calculation in the next period.

[0052] Step 6: According to the predicted drought area development trend data, calculate the drought cumulative area-time normalization curve to judge the sudden onset degree of drought.

[0053] The area where drought may develop at a selected time. Therefore, by continuously selecting times, the drought areas for 1, 2, …, n weeks can be obtained, that is, the development of the drought area over time. Through the normalization curve, the development speed of drought under the threshold η is described, and different possible drought development situations can be obtained by changing the threshold η. The drought cumulative area-time normalization curve is used to describe how fast the drought cumulative area changes over time. The calculation formulas for the vertical and horizontal coordinate values are as follows:

[0054]

[0055] In the formula, the total duration of drought is p time periods, and j is the current drought time period. N A is the normalized drought coverage area; A i is the maximum drought projection area in the i-th time period; is the sum of the maximum drought projection areas of each time period, is the sum of the maximum drought projection areas up to the j-th time period.

[0056]

[0057] In the formula, N T is the normalized time; T j is the j-th time period of drought; is the sum of each time period of the drought event.

[0058] Taking N A as the vertical coordinate of the curve and N T as the horizontal coordinate of the curve, the cumulative area-time normalization curve of a drought event is obtained. It is considered that when the curve is a straight line with a slope of 1, it is a special state where the area increases uniformly with time, that is, the drought type is not a sudden onset drought. Generally, it is considered that if there is a time period where the slope of the curve is greater than 1, the drought in this time period can be judged as a sudden onset drought. The greater the slope, the greater the sudden onset degree of the drought.

[0059] In one embodiment, a sudden onset drought prediction system for soil water is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-mentioned sudden onset drought prediction method for soil water is implemented.

[0060] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned soil water flash drought prediction method is implemented.

[0061] In one embodiment, a computer program product is provided, including a computer program / instructions, and when the computer program / instructions are executed by a processor, the above-mentioned soil water flash drought prediction method is implemented.

[0062] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0063] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 box or multiple boxes.

[0064] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 box or multiple boxes.

[0065] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 box or multiple boxes.

Claims

1. A method for predicting sudden drought of soil water, characterized in that, Including: Grid the research area, and according to the historical index data of each grid point, obtain the data of soil water drought conditions corresponding to each grid point at different times according to the time step ΔT; According to the data of soil water drought conditions corresponding to each grid point at different times, identify soil water drought events through time series connection, and obtain the data of the area covered by each soil water drought event in the research area changing with time; According to the data of the area covered by each soil water drought event changing with time, taking the average precipitation P, average temperature T, and distance value G as input variables and the drought formation time F as the output variable, construct a random forest model for predicting the drought formation time F; the distance value G is the minimum distance between the grid point and the edge of the drought patch; According to the data of soil water drought conditions corresponding to each grid point at different times, statistically obtain the drought expansion rate data of each grid point; Data for predicting the development trend of drought area, including: setting G max as the maximum value of the distance value G in the identified soil water drought events; for any grid point x, if G ≤ G max , the expansion rate η of grid point x is obtained according to the drought expansion rate data of each grid point, and according to the average precipitation P, average temperature T, and distance value G of grid point x, the drought formation time F of grid point x is output through the random forest model; all grid points with F ≤ ΔT are screened out, and it is determined that the grid points with the expansion rate η greater than the probability threshold will be drought-stricken at the next moment; the prediction time T is divided into several segments according to the step size ΔT, and the calculation of this step is repeated until the end of the prediction at time T to obtain the data of the development trend of the drought area; According to the predicted drought area development trend data, calculate the drought cumulative area-time normalization curve, and judge the suddenness degree of the drought according to the drought cumulative area-time normalization curve; The time series connection includes the following process: judge the coverage area of drought events in two adjacent time periods. When the overlap degree of the coverage areas of two consecutive droughts in adjacent time periods is higher than the set value, it is judged as the same drought event; The ordinate of the drought cumulative area-time normalization curve is the normalized drought coverage area, and the abscissa is the normalized time. Judging the suddenness degree of the drought according to the drought cumulative area-time normalization curve includes: calculating the slope of the drought cumulative area-time normalization curve, and judging whether the slope is greater than the threshold. If so, predict the drought type as sudden drought.

2. The soil water flash drought prediction method according to claim 1, wherein: The index data uses soil water moisture SM.

3. The soil water sudden drought prediction method according to claim 2, characterized in that: Use soil water moisture SM to judge the soil water drought situation. By calculating the percentile value of soil water moisture SM, if the percentile value of soil water moisture SM is lower than the set threshold, it is judged as soil water drought.

4. The soil water sudden drought prediction method according to claim 2, characterized in that: Using soil water moisture SM to judge the soil water drought situation includes the following process: extract the daily soil water moisture data within the selected time in the research area to obtain the average soil water moisture data for a period of time, and then calculate the marginal distribution function of the average soil water moisture data; use the K-S test method and the root mean square error method to fit the best probability distribution function; according to the best probability distribution function, convert the soil water moisture value into a quantile, and if the quantile of soil water moisture SM is lower than the set threshold, it is judged as soil water drought.

5. The soil water sudden drought prediction method according to claim 1, characterized in that, The formula for the normalized drought coverage area is: Where N A is the normalized drought coverage area; n is the number of divided time periods, and A j is the maximum drought coverage area projected in the j-th time period; is the sum of the maximum drought coverage areas projected in each time period; The formula for the normalized time is: where N T is the normalized time; T j is the j-th period of drought; is the sum of each period of the drought event.

6. A soil water sudden drought prediction system, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the soil water sudden drought prediction method described in any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the soil water sudden drought prediction method described in any one of claims 1 to 5.

8. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, it implements the soil water sudden drought prediction method described in any one of claims 1 to 5.

Images