A method and system for estimating the lag time of vegetation growth on terrestrial water storage
By combining Pearson correlation analysis and autocorrelation function, the accuracy and universality of estimation of vegetation growth response lag effect on changes in land water reserves is solved, and efficient and accurate on the long-term sequence cell scale in large areas is achieved Lag time estimation overcomes the limitations of traditional methods and provides important support for the correlation between vegetation growth and land water reserves.
Patent Information
- Application Number
- CN202410499183.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-24
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-04-24
AI Technical Summary
The existing vegetation growth response lag effect estimation methods for changes in land water reserves have limitations in accuracy and universality, especially in large areas, long-term cell scales, and high computational complexity. Overfitting is prone to occur.
Combining Pearson correlation analysis and autocorrelation function, through image preprocessing, seasonal outlier removal and iterative calculation, the correlation coefficients of vegetation index and land water reserve images were constructed, and the lag time of each pixel was gradually determined, and the vegetation growth pairs were quantified. The lag effect of changes in land water reserves.
It achieves efficient and accurate assessment of the correlation and lag between vegetation growth and land water reserves in the target area and long-term time range, overcomes the window setting limitations and overfitting problems of traditional methods, and is suitable for different types of Remote sensing image data has the ability to accurately estimate the lag time.
Smart Images

Figure CN118537722B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of remote sensing data analysis, and in particular to a method and system for estimating the lag time of vegetation growth to terrestrial water storage. Background Art
[0002] Vegetation growth is a slow process that requires the accumulation of sufficient materials and water to cope with environmental changes, so the response to various meteorological factors often has a certain time lag effect. In addition to being affected by current climate factors, vegetation growth is also greatly affected by the cumulative effect of climate conditions in the previous period, and the response time of vegetation to meteorological factors in different regions and vegetation types is not exactly the same. Meteorological condition data such as precipitation, soil moisture, and evapotranspiration are the most commonly used indicators for scholars to study the impact of vegetation growth. In recent years, terrestrial water storage data has been introduced as a new indicator for studying the hydrological impact of vegetation growth. It is composed of canopy water, surface water, soil water, and groundwater, and can fully reflect the changes in terrestrial water resources. It has been widely used in global and regional water storage changes and attribution, drought and flood monitoring, and other studies, and can capture seasonal and interannual changes in vegetation greenness. Water is the main driving force of vegetation change and an important prerequisite for maintaining vegetation growth. It regulates vegetation growth in more than 60% of the world's ecosystems and affects the global water cycle and carbon cycle. Therefore, a clear understanding of the lag effect between changes in terrestrial water storage and vegetation growth is particularly important for predicting future carbon and water cycles.
[0003] In recent years, domestic and foreign scholars have mainly used spatiotemporal statistical methods and machine learning methods to study the lag effect of vegetation growth on the response of terrestrial water storage changes. The first method is based on the principle of spatial analysis, combined with Pearson correlation analysis, window cross-correlation, geographically weighted regression and other methods, selects a time window from the beginning to the end and moves forward, calculates the correlation coefficient between vegetation index and terrestrial water storage, and takes the lag months corresponding to the maximum correlation coefficient as the lag time. This method has great limitations and requires manual setting of the analysis window. The window and time series must be long enough to contain variability, and the accuracy is not high. It is not suitable for lag time estimation at the pixel scale of large areas and long time series; the second method is based on neural networks, support vector machines, decision trees and other methods to establish a relationship model between variables and predict future trends. This method has high computational complexity and is prone to overfitting at the pixel scale, so it is not universal. To address the above issues, this study proposed a method that combines Pearson correlation analysis and autocorrelation function to analyze the lag effect of vegetation growth in response to changes in terrestrial water storage, which can accurately estimate the correlation and lag time between vegetation growth and terrestrial water storage. Summary of the invention
[0004] In order to solve the problem of accurately evaluating the lag effect of vegetation growth at the pixel scale on the change of terrestrial water storage in the study area, the present invention proposes a method and system for estimating the lag time of vegetation growth on terrestrial water storage.
[0005] The technical solution of the method of the present invention is a method for estimating the lag time of vegetation growth on terrestrial water storage, and the specific steps are as follows:
[0006] Step 1: Obtain vegetation index images of the target area at multiple historical moments and terrestrial water storage images of the target area at multiple historical moments;
[0007] Step 2: Through image preprocessing and resampling, a vegetation index image with a uniform temporal and spatial resolution at each historical moment of the target area and a terrestrial water storage image with a uniform temporal and spatial resolution at each historical moment of the target area are obtained respectively;
[0008] Step 3: Through seasonal outlier elimination processing, the vegetation index image after seasonal elimination at each historical moment of the target area and the terrestrial water storage image after seasonal elimination at each historical moment of the target area are obtained respectively;
[0009] Step 4: Calculate the correlation coefficient between the time series of vegetation index image pixels after seasonal elimination and the time series of terrestrial water storage image pixels after seasonal elimination in the target area by using the Pearson correlation coefficient method, and obtain the significance level of each row and column of pixels in the target area by performing a t-test;
[0010] Step 5: Compare the significance level of each row and column of the target area in step 4 with the threshold. If it is less than a certain threshold, the vegetation index image pixel and the terrestrial water storage image pixel have a significant correlation at a certain significance level, and jump to step 6;
[0011] Step 6: Construct a new time series of vegetation index image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of vegetation index image pixels after seasonal elimination corresponding to the iteration number and the new time series of vegetation index image pixels after lagged seasonal elimination through the autocorrelation function;
[0012] Step 7: construct a new time series of terrestrial water storage image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of terrestrial water storage image pixels after seasonal elimination corresponding to the iteration number and the new time series of terrestrial water storage image pixels after lagged seasonal elimination through the autocorrelation function;
[0013] Step 8: Repeat steps 6 and 7 multiple times until the convergence condition is met, and obtain the number of iterations when the convergence condition is met;
[0014] Step 9: Calculate the lag time of each pixel in combination with the number of iterations when the convergence condition is met, which is further used to quantify the lag effect of vegetation growth in the target area on the change of terrestrial water storage;
[0015] Preferably, the vegetation index image with uniform spatiotemporal resolution at each historical moment of the target area is obtained by image preprocessing and resampling in step 2, as follows:
[0016] The vegetation index image of each historical moment in the target area is processed by interpolation to fill missing data, cropping the area of interest, and maximizing the value to obtain the vegetation index image of the specified time resolution at each historical moment in the target area. The vegetation index image of each historical moment in the target area is further combined with the resampling process of the specified spatial resolution to obtain the vegetation index image of the unified temporal and spatial resolution at each historical moment in the target area.
[0017] In step 2, the land water storage image with uniform spatiotemporal resolution at each historical moment of the target area is obtained through image preprocessing and resampling, as follows:
[0018] The terrestrial water storage images of the target area at each historical moment are interpolated to fill missing data, crop the area of interest, and perform maximum value synthesis in turn to obtain the terrestrial water storage images of the target area at each historical moment with the specified time resolution, and further combined with the resampling process of the specified spatial resolution to obtain the terrestrial water storage images of the target area at each historical moment with the unified temporal and spatial resolution;
[0019] Preferably, the step 3 is as follows:
[0020] The vegetation index image with uniform spatiotemporal resolution at each historical moment of the target area is processed by removing seasonal outliers to obtain the vegetation index image after seasonal removal at each historical moment of the target area;
[0021] The terrestrial water storage images with uniform spatiotemporal resolution at each historical moment of the target area are processed by removing seasonal outliers to obtain the terrestrial water storage images after seasonal removal at each historical moment of the target area;
[0022] Preferably, the time series of vegetation index image pixels after seasonal elimination in the target area in step 4 is constructed as follows:
[0023] Extract the seasonal vegetation index image pixels of each historical moment of the target area in turn to construct the time series X of the seasonal vegetation index image pixels of the target area. i(m,n), which is the seasonally eliminated vegetation index image pixel in the mth row and nth column at the i-th historical moment, where 1≤i≤L, and L is the number of historical moments;
[0024] The construction process of the time series of the terrestrial water storage image pixels after seasonal elimination in the target area in step 4 is as follows:
[0025] Sequentially extract the seasonally eliminated terrestrial water storage image pixels of the target area at each historical moment to construct the time series Y of the seasonally eliminated terrestrial water storage image pixels of the target area. i (m,n), which is the seasonally eliminated land water storage image pixel in the mth row and nth column at the i-th historical moment, where 1≤i≤L, and L is the number of historical moments;
[0026] Step 4 is calculated by the Pearson correlation coefficient method, as follows:
[0027] The Pearson correlation coefficient method is used to calculate the time series X of the vegetation index image pixels after seasonal elimination in the target area. i (m,n) and the time series Y of the terrestrial water storage image pixels after seasonal elimination in the target area i The correlation coefficient between (m,n) is r(m,n), and the formula is as follows:
[0028]
[0029] Where L is the number of historical moments; X i (m,n) is the vegetation index image pixel after seasonal elimination in the mth row and nth column at the i-th historical moment, Y i (m,n) is the seasonally eliminated land water storage image pixel in the mth row and nth column at the i-th historical moment; It is the average value of the seasonally eliminated vegetation index image pixels in the mth row and nth column of L historical moments; is the average value of the seasonally eliminated terrestrial water storage image pixels in the mth row and nth column at L historical moments;
[0030] Preferably, the new time series of vegetation index image pixels after the lagged season removal is constructed in step 6, specifically as follows:
[0031] The time series X of vegetation index image pixels after removing the season of the target area i (m,n), where 1≤i≤L, is lagged with a step length of one month to obtain a new time series X of vegetation index image pixels after seasonal elimination after the lag. j (m,n), where a+1≤j≤L, a is the number of iterations, and L is the number of historical moments;
[0032] Step 6 is calculated by the autocorrelation function, as follows:
[0033] Through the autocorrelation function, the original time series X of the vegetation index image pixels after seasonal elimination is calculated when the number of iterations is a. i (m,n) and the new time series X of vegetation index image pixels after lagged seasonal elimination j Correlation between (m,n)
[0034] The autocorrelation function calculation formula is as follows:
[0035]
[0036] Where L is the number of historical moments; X i (m,n) is the vegetation index image pixel after seasonal elimination in the mth row and nth column at the i-th historical moment, X i-a (m,n) is the vegetation index image pixel after seasonal elimination in the mth row and nth column at the iath historical moment; It is the average value of the seasonally eliminated vegetation index image pixels in the mth row and nth column of L historical moments;
[0037] Preferably, the new time series of terrestrial water storage image pixels after the lagged season removal is constructed in step 7, specifically as follows:
[0038] The time series Y of the terrestrial water storage image pixels after removing the seasons of the target area i (m,n), where 1≤i≤L, is lagged with a step length of one month to form a new time series Y of land water storage image pixels after the lagged season is removed j (m,n), where a+1≤j≤L, a is the number of iterations, and L is the number of historical moments;
[0039] Step 7 is calculated by the autocorrelation function, as follows:
[0040] Calculate the original time series Y of the terrestrial water storage image pixels after seasonal elimination when the number of iterations is a i (m,n) and the new time series Y of the terrestrial water storage image pixels after the lagged seasonal elimination j Correlation between (m,n)
[0041] The specific formula for calculating the autocorrelation function is as follows:
[0042]
[0043] Where L is the number of historical moments; Y i (m,n) is the seasonally eliminated land water storage image pixel in the mth row and nth column at the i-th historical moment, Y i-a (m,n) is the seasonally eliminated land water storage image pixel in the mth row and nth column at the iath historical moment; is the average value of the seasonally eliminated terrestrial water storage image pixels in the mth row and nth column at L historical moments;
[0044] Preferably, the convergence condition in step 8 is as follows:
[0045]
[0046] Where L is the number of historical moments, the iteration is stopped and the number of iterations a* when the convergence condition is met is output;
[0047] Preferably, the lag time of each pixel is calculated in combination with the number of iterations when the convergence condition is met in step 9, as follows:
[0048] The lag time of each pixel is obtained by combining the correlation coefficient between the time series of the vegetation index image pixels after seasonal elimination and the time series of the terrestrial water storage image pixels after seasonal elimination in the target area and the number of iterations when the convergence condition is met;
[0049] If r(m,n)>0, the lag time lag(m,n) of the pixel in the mth row and nth column is a*;
[0050] Otherwise, the lag time lag(m,n) of the pixel in the mth row and nth column is 12-a*;
[0051] Step 9 is further used to quantify the hysteresis effect of vegetation growth in the target area on the change of terrestrial water storage, as follows:
[0052] It is further used to generate a lagged time grid map of the vegetation index response to terrestrial water storage in the target area to quantify the lagged effect of vegetation growth in the target area on terrestrial water storage changes.
[0053] The present invention also provides a system for estimating the lag time of vegetation growth on terrestrial water storage, comprising:
[0054] An image acquisition module is used to acquire vegetation index images of a target area at multiple historical moments and terrestrial water storage images of a target area at multiple historical moments;
[0055] A module for unified spatiotemporal resolution is used to obtain vegetation index images with unified spatiotemporal resolution at each historical moment of the target area and terrestrial water storage images with unified spatiotemporal resolution at each historical moment of the target area through image preprocessing and resampling processing;
[0056] A seasonal elimination module is used to obtain a vegetation index image after seasonal elimination at each historical moment of the target area and a terrestrial water storage image after seasonal elimination at each historical moment of the target area through seasonal outlier elimination processing;
[0057] The correlation coefficient calculation module is used to calculate the correlation coefficient between the time series of vegetation index image pixels after seasonal elimination and the time series of terrestrial water storage image pixels after seasonal elimination in the target area by using the Pearson correlation coefficient method, and obtain the significance level of each row and column of pixels in the target area by performing a t-test;
[0058] The significance level judgment module is used to compare the significance level of each row and column of the target area with the threshold. If it is less than a certain threshold, the vegetation index image pixel and the terrestrial water storage image pixel have a significant correlation at a certain significance level, and jump to the autocorrelation calculation module;
[0059] The vegetation index image autocorrelation calculation module is used to construct a new time series of vegetation index image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of vegetation index image pixels after seasonal elimination corresponding to the number of iterations and the new time series of vegetation index image pixels after lagged seasonal elimination through the autocorrelation function;
[0060] The terrestrial water storage autocorrelation calculation module is used to construct a new time series of terrestrial water storage image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of terrestrial water storage image pixels after seasonal elimination corresponding to the number of iterations and the new time series of terrestrial water storage image pixels after lagged seasonal elimination through the autocorrelation function;
[0061] An iterative update module is used to repeatedly call the vegetation index image autocorrelation calculation module and the terrestrial water storage autocorrelation calculation module until the convergence condition is met, and obtain the number of iterations when the convergence condition is met;
[0062] The pixel lag time calculation module is used to calculate the lag time of each pixel in combination with the number of iterations when the convergence condition is met, and is further used to quantify the lag effect of vegetation growth in the target area on the change of terrestrial water storage.
[0063] The present invention can efficiently and accurately evaluate the correlation and hysteresis between vegetation growth and terrestrial water storage for target areas and continuous long time series. After testing, the present invention overcomes the limitations of traditional methods that require manual setting of analysis windows and time series or insufficient window length. The calculation process is simple and efficient, applicable to different types of remote sensing image data, and has the ability to accurately estimate lag time at the pixel scale of large-area long time series. It not only considers the hysteresis between vegetation index and terrestrial water storage, but also considers the periodicity and hysteresis of the two variables themselves, and combines the positive and negative response relationships of pixels to accurately calculate the lag time of each pixel in the area. The lag time estimation method established by the present invention is of great significance for revealing the correlation and temporal correlation between vegetation growth and terrestrial water storage, and provides strong support for the prediction of future carbon and water cycles. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] Figure 1 : A flow chart of a method according to an embodiment of the present invention;
[0065] Figure 2 : Pearson correlation grading diagram of NDVI and TWSA of different vegetation types in the global region according to an embodiment of the present invention;
[0066] Figure 3 : Average autocorrelation function curve diagram of NDVI and TWSA of different vegetation types in the global region according to the embodiment of the present invention;
[0067] Figure 4 : A histogram of the lag time distribution of NDVI of different vegetation types in the global region to TWSA response in an embodiment of the present invention; DETAILED DESCRIPTION
[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0069] In specific implementation, the method proposed in the technical solution of the present invention can be implemented by technical personnel in this field using computer software technology to realize automatic operation process. System devices for implementing the method, such as computer-readable storage media storing the corresponding computer program of the technical solution of the present invention and computer equipment running the corresponding computer program, should also be within the scope of protection of the present invention.
[0070] In order to facilitate ordinary technicians in this field to understand and implement the present invention, Python is used to perform Pearson correlation analysis and ACF autocorrelation function analysis, and ARCGIS is used to map the results. Taking the lag time of NDVI response to TWSA at the pixel scale on a global scale based on terrestrial water storage anomaly data (TWSA) and normalized vegetation index data (NDVI) as an example, the present invention is further described in detail.
[0071] The method embodiment of the present invention is a method for estimating the lag time of vegetation growth on terrestrial water storage, which is specifically as follows:
[0072] Step 1: Obtain vegetation index images of the target area at multiple historical moments and terrestrial water storage images of the target area at multiple historical moments;
[0073] This step obtains the 16-Days 250m Normalized Difference Vegetation Index (NDVI) image of the global area from 2004 to 2023 and the terrestrial water storage anomaly (TWSA) image of the global area with a spatial resolution of 0.25° per month from 2004 to 2023.
[0074] Step 2: Through image preprocessing and resampling, a vegetation index image with a uniform temporal and spatial resolution at each historical moment of the target area and a terrestrial water storage image with a uniform temporal and spatial resolution at each historical moment of the target area are obtained respectively;
[0075] In step 2, the vegetation index image with uniform spatiotemporal resolution at each historical moment of the target area is obtained through image preprocessing and resampling, as follows:
[0076] The NDVI images of every 16 days in the global area from 2004 to 2023 are interpolated to fill missing data, crop the area of interest, and perform maximum value synthesis processing to obtain the monthly NDVI images of the global area from 2004 to 2023. Further combined with the resampling processing of 0.25° spatial resolution, the NDVI images of the global area with a spatial resolution of 0.25° per month from 2004 to 2023 are obtained;
[0077] In step 2, the land water storage image with uniform spatiotemporal resolution at each historical moment of the target area is obtained through image preprocessing and resampling, as follows:
[0078] The monthly TWSA images of the global area from 2004 to 2023 are interpolated to fill missing data, cropped in the area of interest, and synthesized at the maximum value to obtain the monthly TWSA images of the global area from 2004 to 2023. The monthly TWSA images of the global area with a spatial resolution of 0.25° are further obtained by resampling the images at a spatial resolution of 0.25°.
[0079] Step 3: Through seasonal outlier elimination processing, the vegetation index image after seasonal elimination at each historical moment of the target area and the terrestrial water storage image after seasonal elimination at each historical moment of the target area are obtained respectively;
[0080] The step 3 is as follows:
[0081] The NDVI images with a spatial resolution of 0.25° for each month in the global region from 2004 to 2023 are processed by removing seasonal outliers to obtain the NDVI images after seasonal removal for each month in the global region from 2004 to 2023;
[0082] The 0.25° spatial resolution TWSA images of the global area from 2004 to 2023 are processed by removing seasonal outliers to obtain the seasonally removed TWSA images of the global area from 2004 to 2023.
[0083] Step 4: Calculate the correlation coefficient between the time series of vegetation index image pixels after seasonal elimination and the time series of terrestrial water storage image pixels after seasonal elimination in the target area by using the Pearson correlation coefficient method, and obtain the significance level of each row and column of pixels in the target area by performing a t-test;
[0084] The construction process of the time series of vegetation index image pixels after seasonal elimination in the target area in step 4 is as follows:
[0085] Extract the NDVI image pixels of the global region after seasonal elimination every month from 2004 to 2023 in turn to construct the time series X of the NDVI image pixels of the global region after seasonal elimination i (m,n), which is the seasonally eliminated NDVI image pixel in the mth row and nth column of the i-th month, where 1≤i≤L, L is the number of months;
[0086] The construction process of the time series of the terrestrial water storage image pixels after seasonal elimination in the target area in step 4 is as follows:
[0087] Extract the seasonally eliminated TWSA image pixels of each month in the global area from 2004 to 2023 in turn to construct the time series Y of seasonally eliminated TWSA image pixels of the global area i (m,n), which is the seasonally eliminated TWSA image pixel in the mth row and nth column of the i-th month, where 1≤i≤L, L is the number of months;
[0088] Step 4 is calculated by the Pearson correlation coefficient method, as follows:
[0089] Pearson correlation analysis based on Python is used to calculate the time series X of NDVI image pixels after seasonal elimination in the global region i (m,n) and the time series Y of TWSA image pixels after seasonal elimination in the global area i The correlation coefficient between (m,n) is r(m,n), and the formula is as follows:
[0090]
[0091] Where L is the number of months; X i (m,n) is the seasonally eliminated NDVI image pixel in the mth row and nth column of the i-th month, Y i (m,n) is the seasonally eliminated TWSA image pixel in the mth row and nth column of the i-th month; is the average value of the seasonally eliminated NDVI image pixels in the mth row and nth column of month L; is the average value of the seasonally eliminated TWSA image pixels in the mth row and nth column of month L;
[0092] Step 5: Compare the significance level of each row and column of the target area in step 4 with the threshold. If it is less than a certain threshold, the vegetation index image pixel and the terrestrial water storage image pixel have a significant correlation at a certain significance level, and jump to step 6;
[0093] Combined with the significance level of each row and column of the global area in step 4, the significance level is compared with the threshold value of 0.05. If it is less than 0.05, the NDVI image pixel and the TWSA image pixel have a significant correlation at the 95% significance level, and jump to step 6;
[0094] The data were processed in ARCGIS, and the SetNull() function of the "Raster Calculator" was used to set the correlation values of pixels with a significance level greater than or equal to 0.05 to null values, that is, pixels that failed the 95% significance level test were classified as "non-significant", and then the Pearson correlation coefficient was classified into "significant positive" and "significant negative" according to whether the value was greater than 0. The "display of zoning geometric statistics in table" tool was used to calculate the area proportion of each category. TWSA and NDVI in Eurasia, North America and northern South America generally showed a negative correlation, while other regions generally showed a positive correlation; the area with significant positive correlation accounted for 36.85% of the total land area, while the area with significant negative correlation accounted for 36.63%;
[0095] For the convenience of analysis, the global vegetation is classified into seven categories based on land use type: "Treecover", "Shrubland", "Grassland", "Cropland", "Herbaceous wetland", "Mangroves", and "Moss and lichen". Using the "Display Zoning Statistics in Table" tool of ARCGIS, the area proportion of the correlation values of the seven categories is calculated, as follows: Figure 2 As shown, it can be seen that shrubs have the highest positive correlation ratio, which is 84.65%, that is, in most shrub areas, TWSA and NDVI show an obvious positive correlation; grassland, cultivated land and mangroves also show a relatively obvious positive correlation; while woodland, herbaceous wetlands and mosses show an obvious negative correlation, with negative correlation ratios of 53.00%, 57.44% and 68.05% respectively.
[0096] Step 6: Construct a new time series of vegetation index image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of vegetation index image pixels after seasonal elimination corresponding to the iteration number and the new time series of vegetation index image pixels after lagged seasonal elimination through the autocorrelation function;
[0097] Step 6 constructs a new time series of vegetation index image pixels after lagged seasonal elimination, as follows:
[0098] The time series X of NDVI image pixels after removing the seasons in the global area i (m,n), where 1≤i≤L, is lagged with a step length of one month to obtain a new time series X of NDVI image pixels after seasonal elimination after the lag.j (m,n), where a+1≤j≤L, a is the number of iterations, and L is the number of months;
[0099] Step 6 is calculated by the autocorrelation function, as follows:
[0100] Based on Python's acf() function, calculate the original time series X of the NDVI image pixels after seasonal elimination when the number of iterations is a i (m,n) and the new time series X of NDVI image pixels after seasonal elimination after lag j Correlation between (m,n)
[0101] The autocorrelation function calculation formula is as follows:
[0102]
[0103] Where L is the number of months; X i (m,n) is the seasonally eliminated NDVI image pixel in the mth row and nth column of the i-th month, X i-a (m,n) is the seasonally eliminated NDVI image pixel in the mth row and nth column of the iath month; is the average value of the seasonally eliminated NDVI image pixels in the mth row and nth column of month L;
[0104] Step 7: construct a new time series of terrestrial water storage image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of terrestrial water storage image pixels after seasonal elimination corresponding to the iteration number and the new time series of terrestrial water storage image pixels after lagged seasonal elimination through the autocorrelation function;
[0105] Step 7 constructs a new time series of terrestrial water storage image pixels after lagged seasonal elimination, as follows:
[0106] The time series Y of TWSA image pixels after removing the seasons in the global area i (m,n), where 1≤i≤L, is lagged with a step length of one month to form a new time series Y of TWSA image pixels after lagged seasonal elimination j (m,n), where a+1≤j≤L, a is the number of iterations, and L is the number of months;
[0107] Step 7 is calculated by the autocorrelation function, as follows:
[0108] Based on Python's acf() function, the original time series Y of the TWSA image pixels after seasonal elimination is calculated when the number of iterations is a. i(m,n) and the new time series Y of TWSA image pixels after lagged seasonal elimination j Correlation between (m,n)
[0109] The specific formula for calculating the autocorrelation function is as follows:
[0110]
[0111] Where L is the number of months; Y i (m,n) is the seasonally eliminated TWSA image pixel in the mth row and nth column of the i-th month, Y i-a (m,n) is the seasonally eliminated TWSA image pixel in the mth row and nth column of the iath month; is the average value of the seasonally eliminated TWSA image pixels in the mth row and nth column of month L;
[0112] Step 8: Repeat steps 6 and 7 multiple times until the convergence condition is met, and obtain the number of iterations when the convergence condition is met;
[0113] The convergence conditions described in step 8 are as follows:
[0114] and
[0115] Where L is the number of months, the iteration is stopped and the number of iterations a* when the convergence condition is met is output;
[0116] Step 9: Combining the correlation coefficient between the time series of the vegetation index image pixels after seasonal elimination and the time series of the terrestrial water storage image pixels after seasonal elimination in the target area, and the number of iterations when the convergence condition is met, the lag time of each pixel is obtained, which is further used to generate a lag time grid map of the vegetation index response to terrestrial water storage in the target area, so as to quantify the lag effect of vegetation growth in the target area on the change of terrestrial water storage;
[0117] The step 9 is specifically as follows: if r(m,n)>0, the lag time lag(m,n) of the pixel in the mth row and nth column is a*; otherwise, the lag time lag(m,n) of the pixel in the mth row and nth column is 12-a*;
[0118] Take the average autocorrelation function diagram of seven vegetation types as an example. Figure 3 As shown, the horizontal axis represents the number of iterations, and the vertical axis represents the corresponding correlation value. First, it can be seen that the autocorrelation function values of TWSA and NDVI generally reach the maximum value when the lag period is 12 months, so both of them generally have periodicity, and the period is 12 months; combined Figure 2From the proportion of positive and negative responses of vegetation types, it can be seen that shrubs, grasslands, cultivated land and mangroves have positive responses, so the lag time is the number of iterations that first meet the conditions, which are 3 months, 3 months, 3 months and 1 month respectively; while woodlands, herbaceous wetlands and mosses have negative responses, so the lag time is 12 months minus the number of iterations that first meet the conditions, resulting in a lag period of 9 months, 9 months and 10 months respectively.
[0119] At the pixel scale, a lag time grid map of the global NDVI response to TWSA is output. The lag time of the global vegetation NDVI response to TWSA is divided into 0-13 months, and histogram statistics are performed for each vegetation type. Figure 4 As shown, it can be seen that the lag time of forest land is concentrated in 8-9 months, the lag time of shrubs is concentrated in 1 month, the lag time of grassland is concentrated in 4 months, the lag time of cultivated land is concentrated in 1-4 months, and the lag time of herbaceous wetlands, mangroves and mosses is generally concentrated in 8 months.
[0120] The technical solution of the system embodiment of the present invention is a system for estimating the lag time of vegetation growth on terrestrial water storage, comprising:
[0121] All modules of the system are deployed on the server;
[0122] An image acquisition module is used to acquire vegetation index images of a target area at multiple historical moments and terrestrial water storage images of a target area at multiple historical moments;
[0123] A module for unified spatiotemporal resolution is used to obtain vegetation index images with unified spatiotemporal resolution at each historical moment of the target area and terrestrial water storage images with unified spatiotemporal resolution at each historical moment of the target area through image preprocessing and resampling processing;
[0124] A seasonal elimination module is used to obtain a vegetation index image after seasonal elimination at each historical moment of the target area and a terrestrial water storage image after seasonal elimination at each historical moment of the target area through seasonal outlier elimination processing;
[0125] The significance level judgment module is used to calculate the correlation coefficient between the time series of vegetation index image pixels after seasonal elimination and the time series of terrestrial water storage image pixels after seasonal elimination in the target area by using the Pearson correlation coefficient method, and obtain the significance level of each row and column of pixels in the target area by performing a t-test;
[0126] The correlation coefficient calculation module is used to compare the significance level of each row and column of the target area with the threshold. If it is less than a certain threshold, the vegetation index image pixel and the terrestrial water storage image pixel have a significant correlation at a certain significance level, and jump to the autocorrelation calculation module;
[0127] The vegetation index image autocorrelation calculation module is used to construct a new time series of vegetation index image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of vegetation index image pixels after seasonal elimination corresponding to the number of iterations and the new time series of vegetation index image pixels after lagged seasonal elimination through the autocorrelation function;
[0128] The terrestrial water storage autocorrelation calculation module is used to construct a new time series of terrestrial water storage image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of terrestrial water storage image pixels after seasonal elimination corresponding to the number of iterations and the new time series of terrestrial water storage image pixels after lagged seasonal elimination through the autocorrelation function;
[0129] An iterative update module is used to repeatedly call the vegetation index image autocorrelation calculation module and the terrestrial water storage autocorrelation calculation module until the convergence condition is met, and obtain the number of iterations when the convergence condition is met;
[0130] The pixel lag time calculation module is used to combine the correlation coefficient between the time series of the vegetation index image pixels after seasonal elimination of the target area and the time series of the terrestrial water storage image pixels after seasonal elimination of the target area, and the number of iterations when the convergence condition is met, to obtain the lag time of each pixel, which is further used to generate a lag time grid map of the vegetation index response to terrestrial water storage in the target area, so as to quantify the lag effect of vegetation growth in the target area on the change of terrestrial water storage;
[0131] In specific implementation, the method proposed in the technical solution of the present invention can be implemented by technical personnel in this field using computer software technology to realize automatic operation process. System devices for implementing the method, such as computer-readable storage media storing the corresponding computer program of the technical solution of the present invention and computer equipment running the corresponding computer program, should also be within the scope of protection of the present invention.
[0132] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific implementation cases described, or replace them with similar methods, without departing from the spirit of the present invention or exceeding the scope defined by the appended claims.
Claims
1. A method for estimating the lag time of vegetation growth on terrestrial water storage, characterized in that: The following steps are involved: Step 1: Obtain vegetation index images of the target area at multiple historical moments and terrestrial water storage images of the target area at multiple historical moments; Step 2: Through image preprocessing and resampling, a vegetation index image with a uniform temporal and spatial resolution at each historical moment of the target area and a terrestrial water storage image with a uniform temporal and spatial resolution at each historical moment of the target area are obtained respectively; Step 3: Through seasonal outlier elimination processing, the vegetation index image after seasonal elimination at each historical moment of the target area and the terrestrial water storage image after seasonal elimination at each historical moment of the target area are obtained respectively; Step 4: Calculate the correlation coefficient between the time series of vegetation index image pixels after seasonal elimination and the time series of terrestrial water storage image pixels after seasonal elimination in the target area by using the Pearson correlation coefficient method, and perform a t-test to obtain the significance level of each row and column of pixels in the target area; Step 5: Compare the significance level of each row and column of the target area in step 4 with the threshold. If it is less than a certain threshold, the vegetation index image pixel and the terrestrial water storage image pixel have a significant correlation at a certain significance level, and jump to step 6; Step 6: Calculate the correlation coefficient between the original time series of vegetation index image pixels after seasonal elimination corresponding to the iteration number and the new time series of vegetation index image pixels after seasonal elimination after lag through the autocorrelation function; Step 7: Calculate the correlation coefficient between the original time series of the seasonally eliminated terrestrial water storage image pixels corresponding to the iteration number and the new time series of the seasonally eliminated terrestrial water storage image pixels after the lag through the autocorrelation function; Step 8: Repeat steps 6 and 7 multiple times until the convergence condition is met, and obtain the number of iterations when the convergence condition is met; Step 9: The lag time of each pixel is calculated in combination with the number of iterations when the convergence condition is met, which is further used to quantify the lag effect of vegetation growth in the target area on the change of terrestrial water storage.
2. The method for estimating the lag time of vegetation growth on terrestrial water storage according to claim 1, characterized in that: In step 2, the vegetation index image with uniform spatiotemporal resolution at each historical moment of the target area is obtained through image preprocessing and resampling, as follows: The vegetation index image of each historical moment in the target area is processed by interpolation to fill missing data, cropping the area of interest, and maximizing the value to obtain the vegetation index image of the specified time resolution at each historical moment in the target area. The vegetation index image of each historical moment in the target area is further combined with the resampling process of the specified spatial resolution to obtain the vegetation index image of the unified temporal and spatial resolution at each historical moment in the target area. In step 2, the land water storage image with uniform spatiotemporal resolution at each historical moment of the target area is obtained through image preprocessing and resampling, as follows: The terrestrial water storage images of the target area at each historical moment are interpolated to fill missing data, crop the area of interest, and perform maximum value synthesis to obtain the terrestrial water storage images of the target area at each historical moment with a specified time resolution. They are further combined with the resampling processing of the specified spatial resolution to obtain the terrestrial water storage images of the target area at each historical moment with a unified temporal and spatial resolution.
3. The method for estimating the lag time of vegetation growth on terrestrial water storage according to claim 2, characterized in that: The step 3 is as follows: The vegetation index image with uniform spatiotemporal resolution at each historical moment of the target area is processed by removing seasonal outliers to obtain the vegetation index image after seasonal removal at each historical moment of the target area; The terrestrial water storage images with uniform spatiotemporal resolution at each historical moment of the target area are processed by removing seasonal outliers to obtain the terrestrial water storage images after seasonal removal at each historical moment of the target area.
4. The method for estimating the lag time of vegetation growth on terrestrial water storage according to claim 3, characterized in that: The construction process of the time series of vegetation index image pixels after seasonal culling in the target area in step 4 is as follows: Extract the seasonal vegetation index image pixels of each historical moment of the target area in turn to construct the time series X of the seasonal vegetation index image pixels of the target area. i (m,n), which is the seasonally eliminated vegetation index image pixel in the mth row and nth column at the i-th historical moment, where 1≤i≤L, and L is the number of historical moments; The construction process of the time series of the terrestrial water storage image pixels after seasonal elimination in the target area in step 4 is as follows: Sequentially extract the seasonally eliminated terrestrial water storage image pixels of the target area at each historical moment to construct the time series Y of the seasonally eliminated terrestrial water storage image pixels of the target area. i (m,n), which is the seasonally eliminated land water storage image pixel in the mth row and nth column at the i-th historical moment, where 1≤i≤L, and L is the number of historical moments; Step 4 is calculated by the Pearson correlation coefficient method, as follows: The Pearson correlation coefficient method is used to calculate the time series X of the vegetation index image pixels after seasonal elimination in the target area. i (m,n) and the time series Y of the terrestrial water storage image pixels after seasonal elimination in the target area i The correlation coefficient between (m,n) is r(m,n), and the formula is as follows: Where L is the number of historical moments; X i (m,n) is the vegetation index image pixel after seasonal elimination in the mth row and nth column at the i-th historical moment, Y i (m,n) is the seasonally eliminated land water storage image pixel in the mth row and nth column at the i-th historical moment; It is the average value of the seasonally eliminated vegetation index image pixels in the mth row and nth column of L historical moments; It is the average value of the seasonally eliminated terrestrial water storage image pixels in the mth row and nth column at L historical moments.
5. The method for estimating the lag time of vegetation growth on terrestrial water storage according to claim 4, characterized in that: Step 6 is calculated by the autocorrelation function, as follows: Construct a new time series of vegetation index image pixels after lagged seasonal elimination, as follows: The time series X of vegetation index image pixels after removing the season of the target area i (m,n), where 1≤i≤L, is lagged with a step length of one month to obtain a new time series X of vegetation index image pixels after seasonal elimination after the lag. j (m,n), where a+1≤j≤L, a is the number of iterations, and L is the number of historical moments; Through the autocorrelation function, the original time series X of the vegetation index image pixels after seasonal elimination is calculated when the number of iterations is a. i (m,n) and the new time series X of vegetation index image pixels after lagged seasonal elimination j Correlation between (m,n) 6. The method for estimating the lag time of vegetation growth on terrestrial water storage according to claim 5, characterized in that: The autocorrelation function calculation formula is as follows: Where L is the number of historical moments; X i (m,n) is the vegetation index image pixel after seasonal elimination in the mth row and nth column at the i-th historical moment, X i-a (m,n) is the seasonally eliminated vegetation index image pixel in the mth row and nth column at the iath historical moment; It is the average value of the seasonally eliminated vegetation index image pixels in the mth row and nth column of L historical moments.
7. The method for estimating the lag time of vegetation growth on terrestrial water storage according to claim 6, characterized in that: Step 7 is calculated by the autocorrelation function, as follows: Construct a new time series of terrestrial water storage image pixels after lagged seasonal elimination, as follows: The time series Y of the terrestrial water storage image pixels after removing the seasons of the target area i (m,n), where 1≤i≤L, is lagged with a step length of one month to form a new time series Y of land water storage image pixels after the lagged season is removed j (m,n), where a+1≤j≤L, a is the number of iterations, and L is the number of historical moments; Calculate the original time series Y of the terrestrial water storage image pixels after seasonal elimination when the number of iterations is a i (m,n) and the new time series Y of the terrestrial water storage image pixels after the lagged seasonal elimination j Correlation between (m,n) The specific formula for calculating the autocorrelation function is as follows: Where L is the number of historical moments; Y i (m,n) is the seasonally eliminated land water storage image pixel in the mth row and nth column at the i-th historical moment, Y i-a (m,n) is the seasonally eliminated land water storage image pixel in the mth row and nth column at the iath historical moment; It is the average value of the seasonally eliminated terrestrial water storage image pixels in the mth row and nth column at L historical moments.
8. The method for estimating the lag time of vegetation growth on terrestrial water storage according to claim 7, characterized in that: The convergence conditions described in step 8 are as follows: Where L is the number of historical moments, the iteration is stopped and the number of iterations a* when the convergence condition is met is output.
9. The method for estimating the lag time of vegetation growth on terrestrial water storage according to claim 8, characterized in that: When calculating the hysteresis of each pixel in combination with the number of iterations when the convergence condition is met as described in step 9, the details are as follows: The lag time of each pixel is obtained by combining the correlation coefficient between the time series of the vegetation index image pixels after seasonal elimination and the time series of the terrestrial water storage image pixels after seasonal elimination in the target area and the number of iterations when the convergence condition is met; If r(m,n)>0, the lag time lag(m,n) of the pixel in the mth row and nth column is a*; Otherwise, the lag time lag(m,n) of the pixel in the mth row and nth column is 12-a*; Step 9 is further used to quantify the hysteresis effect of vegetation growth in the target area on the change of terrestrial water storage, as follows: It is further used to generate a lagged time grid map of the vegetation index response to terrestrial water storage in the target area to quantify the lagged effect of vegetation growth in the target area on terrestrial water storage changes.
10. A system for estimating the lag time of vegetation growth on terrestrial water storage, characterized in that: include: An image acquisition module is used to acquire vegetation index images of a target area at multiple historical moments and terrestrial water storage images of a target area at multiple historical moments; A module for unified spatiotemporal resolution is used to obtain vegetation index images with unified spatiotemporal resolution at each historical moment of the target area and terrestrial water storage images with unified spatiotemporal resolution at each historical moment of the target area through image preprocessing and resampling processing; A seasonal elimination module is used to obtain a vegetation index image after seasonal elimination at each historical moment of the target area and a terrestrial water storage image after seasonal elimination at each historical moment of the target area through seasonal outlier elimination processing; The correlation coefficient calculation module is used to calculate the correlation coefficient between the time series of vegetation index image pixels after seasonal elimination and the time series of terrestrial water storage image pixels after seasonal elimination in the target area by using the Pearson correlation coefficient method, and to perform a t-test to obtain the significance level of each row and column of pixels in the target area; The significance level judgment module is used to compare the significance level of each row and column of the target area with the threshold. If it is less than a certain threshold, the vegetation index image pixel and the terrestrial water storage image pixel have a significant correlation at a certain significance level, and jump to the autocorrelation calculation module; The vegetation index image autocorrelation calculation module is used to construct a new time series of vegetation index image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of vegetation index image pixels after seasonal elimination corresponding to the number of iterations and the new time series of vegetation index image pixels after lagged seasonal elimination through the autocorrelation function; The terrestrial water storage autocorrelation calculation module is used to construct a new time series of terrestrial water storage image pixels after lagged seasonal elimination, and calculate the correlation coefficient between the original time series of terrestrial water storage image pixels after seasonal elimination corresponding to the number of iterations and the new time series of terrestrial water storage image pixels after lagged seasonal elimination through the autocorrelation function; An iterative update module is used to repeatedly call the vegetation index image autocorrelation calculation module and the terrestrial water storage autocorrelation calculation module until the convergence condition is met, and obtain the number of iterations when the convergence condition is met; The pixel lag time calculation module is used to calculate the lag time of each pixel in combination with the number of iterations when the convergence condition is met, and is further used to quantify the lag effect of vegetation growth in the target area on the change of terrestrial water storage.
