A quantitative identification method for vegetation loss and recovery under drought stress

Abstract

The present invention relates to a quantitative identification method for vegetation loss and recovery under drought stress, comprising the following steps: S1: using run theory to identify drought events at a grid scale, specifically including characteristics such as the start time, duration, end time, and intensity of the drought event; S2: eliminating vegetation data from historical drought events and estimating vegetation changes caused by non-drought factors; S3: estimating vegetation changes caused by individual drought events; S4: classifying vegetation changes caused by drought events using a probability distribution method; and S5: estimating the time from vegetation loss to recovery based on a threshold for normal vegetation fluctuations. The quantitative identification method for vegetation loss and recovery under drought stress described in the present invention has the advantages of eliminating the influence of non-drought factors on vegetation changes, effectively classifying vegetation loss levels, and determining the time required for vegetation to recover from loss to normal after a drought.

Description

Technical Field

[0001] The present invention belongs to the technical field of drought impact assessment on vegetation, and in particular relates to a quantitative identification method for vegetation loss and recovery under drought stress. Background Art

[0002] Complex responses between vegetation growth and drought are linked to factors such as vegetation type, regional climate, drought intensity, and duration. Currently, scholars at home and abroad have conducted exploratory research on the relationship between drought and vegetation, primarily using statistical methods such as correlation analysis and probability distribution to establish the relationship between vegetation and drought and further explore the characteristics of vegetation's response to drought.

[0003] Studies of vegetation responses to drought have mostly used correlation analysis and other methods to simply quantify the relationship between drought and vegetation. Vicente-Serrano et al. (2012) conducted a correlation analysis using the Standardized Precipitation Evapotranspiration Index (SPEI) and the Normalized Difference Vegetation Index (NDVI) at different time scales, finding that NDVI correlates well with SPEI at shorter time scales, indicating that vegetation growth is sensitive to water changes in arid regions, while the opposite is true in humid regions. Zhao et al. (2018) conducted a correlation analysis using SPEI and NDVI at different time scales to examine the response characteristics of vegetation to drought at different time scales and analyze the changing trends of vegetation under the influence of dryness and wetness. Zhang and Zhang (2019) studied the relationship between natural vegetation dynamics (NDVI) and drought during the growing season in China and found that the positive (negative) correlation between NDVI and drought index occurred primarily in non-humid (humid) regions, suggesting that water stress is unlikely to be a limiting factor for natural vegetation in humid regions. Fang et al. (2019) systematically assessed the conditional probability of vegetation decline under various drought scenarios (moderate, severe, and extreme) using bivariate correlation modeling based on the NDVI and SPEI indices. Similarly, Yuan et al. (2022) constructed a joint distribution model of SPEI and NDVI based on a probability distribution approach and assessed the conditional probability of vegetation (NDVI) loss under various drought conditions (moderate, severe, and extreme). However, these studies only considered dryness and wetness, ignoring the influence of multiple factors such as season and underlying surface conditions on vegetation change. Therefore, these assessment methods have significant uncertainty and are not very reliable.

[0004] Researchers have also attempted to propose methods for identifying vegetation recovery after droughts, such as comparing vegetation status during drought periods with long-term averages (Jha et al., 2019), remote sensing-based monitoring methods (Pérez-Cabello et al., 2021), and drought impact assessment indicators that characterize vegetation decline (Jiao et al., 2021). However, these methods fail to monitor the dynamic changes in vegetation loss and recovery from the perspective of individual drought events, and are unable to accurately quantify the extent (magnitude) of vegetation loss and its corresponding recovery time associated with different types of drought events. Therefore, a quantitative identification method is needed that can accurately reflect the relationship between drought and vegetation and identify the entire process from vegetation loss to recovery. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide a quantitative identification method for vegetation loss and recovery under drought stress, which can eliminate the influence of non-drought factors on vegetation changes, realize the effective classification of vegetation loss levels and determine the time required for vegetation to recover from loss to normal state after drought occurs.

[0006] The present invention is achieved through the following detailed technical solutions:

[0007] A method for quantitatively identifying vegetation loss and recovery under drought stress comprises the following steps:

[0008] S1: Use run theory to identify drought events at the grid scale, including the start time, duration, end time and intensity of drought events;

[0009] S2: Eliminate vegetation data during historical drought periods and estimate vegetation changes caused by non-drought factors;

[0010] S3: Estimation of vegetation changes caused by drought events;

[0011] S4: Use probability distribution method to classify vegetation changes caused by drought events;

[0012] S5: Estimate the time from vegetation loss to recovery based on the threshold of normal vegetation fluctuation.

[0013] Compared to the existing technology, the present invention, in its study of vegetation response to drought, first removes vegetation data (NDVI) from historical drought periods, separates vegetation changes caused by non-drought factors (such as season and underlying surface conditions) during the drought event, and then uses the overall vegetation change minus the vegetation change caused by non-drought factors to obtain the amount of vegetation change caused by drought. The amount of vegetation change caused by drought obtained in this way is more accurate. In addition, based on the statistical vegetation changes (NDVI) corresponding to historical drought events, the present invention effectively divides the vegetation loss level based on probability distribution theory, clarifies the magnitude of vegetation loss caused by different types of drought events, and on this basis, further determines the time required for vegetation to recover from loss to normal after the drought occurs based on the vegetation loss threshold, so that the determined time is more accurate.

[0014] Furthermore, the S3 step is specifically as follows:

[0015] S3: Identify the lag time of vegetation response to drought. This method calculates overall vegetation change from the first month before the drought event to the lag month after the drought ends, as well as vegetation change due to non-drought factors. The difference between the two is the vegetation change due to drought. This method effectively separates vegetation changes due to non-drought factors (such as season and underlying surface conditions), improving the accuracy of estimates of drought-induced vegetation change.

[0016] Furthermore, the S4 step is specifically as follows:

[0017] S4: Fit a probability distribution for vegetation changes caused by historical droughts, select the optimal distribution, convert it into a standard normal distribution, and classify the extent of vegetation loss based on the probability of occurrence. Based on probability distribution theory, a quantitative identification method for vegetation loss thresholds was developed. This method can effectively estimate the magnitude of vegetation loss caused by different types of drought events and achieve a classification of vegetation loss levels.

[0018] Furthermore, the identification of drought events is based on drought indices, which are the Standardized Precipitation Evapotranspiration Index (SPEI) and the Normalized Difference Vegetation Index (NDVI). The use of drought indices can make the identification of drought events more accurate.

[0019] Furthermore, if the SPEI of a drought event is less than -0.5 and the duration is greater than 2 months, and the time interval between adjacent drought events is less than 2 months, the two drought events are merged into one event. This can reduce errors and improve the accuracy of subsequent calculations.

[0020] Furthermore, the step S2 specifically includes: removing the vegetation data during the drought period of the study period, and calculating the multi-year average vegetation index from January to December as the vegetation state not affected by the drought factor. The specific calculation formula is as follows:

[0021]

[0022] in, represents the NDVI for month i when drought conditions were not present (1 ≤ i ≤ 12); n is the number of years during the study period when month i was not affected by drought. This method can further accurately isolate vegetation changes caused by non-drought factors (such as season and underlying surface conditions), significantly improving the accuracy of estimates of vegetation changes caused by drought.

[0023] Furthermore, the specific steps of S3 are as follows:

[0024] S301: First, calculate the lag time of vegetation response to drought, use the maximum Pearson correlation coefficient method to determine the time lag effect of drought on vegetation, calculate the correlation coefficients between SPEI and NDVI with different lag times (1-24 months), and select the maximum correlation coefficient r i The corresponding lag months i is taken as the lag time;

[0025] r i =corr(NDVI,SPEI i )

[0026] r max-lag =max(r i )

[0027] Among them, r i is the Pearson correlation coefficient for time lag i months, r max-lag For r i The maximum value of i in the present invention is 0 to 24 months;

[0028] S302: For a drought event, the overall change in vegetation NDVI after the drought and the change in vegetation NDVI caused by non-drought factors are:

[0029] TC ndvi =NDVI e -NDVI s

[0030] NC ndvi =N_NDVI e -N_NDVI s

[0031] Among them, TC ndvi represents the overall change in vegetation NDVI after a drought event, NCndvi represents the vegetation NDVI changes caused by non-drought factors, the subscript s represents the first month before the drought event begins, and e represents the lagged month after the drought event ends;

[0032] S303: The NDVI change ΔVC caused by the drought event can be expressed as:

[0033] ΔVC=TC ndvi -NC ndvi By determining the time-lag effect of drought on vegetation, the impact of drought on vegetation can be calculated more accurately.

[0034] Furthermore, the specific steps of S4 are as follows:

[0035] S401: Calculate the ΔVC caused by all drought events at the grid scale. This paper uses distribution probability to describe the change of ΔVC. By selecting multiple distributions to fit ΔVC, the optimal distribution is selected using the Kolmogorov-Smirnov test and root mean square error. In order to better quantify the changing characteristics of vegetation loss and growth, the fitted distribution is then converted into a standard normal distribution (ΔVC normal ), wherein the multiple distributions include normal distribution, gamma distribution, generalized extreme value distribution, and lognormal distribution;

[0036] S402: Referring to the idea of ​​Standardized Precipitation Index (SPI) drought classification (Mckee et al., 1993), based on the idea of ​​probability distribution, the present invention converts ΔVC normal <-0.5 (the probability of occurrence is about 30.8%) is defined as vegetation loss, and ΔVC normal >0.5 (occurrence probability is about 30.8%) is defined as vegetation growth;

[0037] S403: Define four vegetation loss levels based on the magnitude of vegetation change: slight loss (-1 < ΔVC normal <-0.5), moderate loss (-1.5<ΔVC normal <-1), severe losses (-2<ΔVC normal <-1.5) and extreme losses (ΔVC normal <-2), with corresponding probabilities of approximately 15%, 9.2%, 4.4%, and 2.3%, respectively. Using probability distribution theory, we can effectively estimate the magnitude of vegetation loss caused by different types of drought events, achieve a classification of vegetation loss, and enable a refined assessment of the impact of drought on vegetation.

[0038] Furthermore, the specific steps of S5 are:

[0039] Identify the time it takes for vegetation to recover from loss to normal fluctuation state: When the vegetation in a certain month after the drought ends is within the normal fluctuation range relative to the month before the drought, the vegetation is considered to have returned to normal growth state. The calculation formula is as follows:

[0040] ΔVC -0.5 ≤TC ndvi [s,e+k] -NC ndvi [s,e+k] ≤ΔVC 0.5

[0041] Where s represents the first month before the drought event begins; e represents the lag month after the drought event ends; k is the time required for vegetation to recover to normal growth state after the drought, in months; TC ndvi [s,e+k] represents the overall change in NDVI from the onset of the drought event to vegetation recovery; NC ndvi [s,e+k] represents the change in NDVI caused by non-drought factors from the onset of the drought event to the recovery of vegetation, ΔVC -0.5 and ΔVC 0.5 The ΔVC values ​​corresponding to the thresholds of -0.5 and 0.5 are shown respectively. The time required for vegetation to recover from loss to normal after a drought is more accurately determined based on the vegetation loss threshold.

[0042] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of a method for quantitatively identifying vegetation loss and recovery under drought stress according to the present invention;

[0044] Figure 2 is the ΔVC and ΔVC of the Pearl River Basin from 1982 to 2020 in the embodiment of the present invention. normal Probability distribution diagram of ;

[0045] Figure 3 The spatial distribution of the probability of vegetation increase, loss, normal fluctuation, and the difference between the probability of increase and loss due to drought events in the Pearl River Basin from 1982 to 2020 in the embodiment of the present invention;

[0046] Figure 4 is the spatial distribution of the probability of mild loss, moderate loss, severe loss, and extreme loss of vegetation due to drought events in the Pearl River Basin from 1982 to 2020 in the embodiment of the present invention;

[0047] Figure 5It is the spatial distribution of the average recovery time after vegetation loss caused by all drought events, mild drought events, moderate drought events and severe drought events in the Pearl River Basin from 1982 to 2020 in the embodiment of the present invention. DETAILED DESCRIPTION

[0048] See also Figure 1 The present invention provides a quantitative identification method for vegetation loss and recovery under drought stress.

[0049] A method for quantitatively identifying vegetation loss and recovery under drought stress comprises the following steps:

[0050] S1: Use run theory to identify drought events at the grid scale, including the start time, duration, end time and intensity of drought events;

[0051] S2: Eliminate vegetation data during historical drought periods and estimate vegetation changes caused by non-drought factors;

[0052] S3: Estimation of vegetation changes caused by drought events;

[0053] S4: Use probability distribution method to classify vegetation changes caused by drought events;

[0054] S5: Estimate the time from vegetation loss to recovery based on the threshold of normal vegetation fluctuation.

[0055] Example

[0056] Experimental equipment and data sources

[0057] The system used was Windows 10 64-bit, an Intel Core i7 processor, and 8GB of RAM. The code was written in Matlab. The experimental subject was the Pearl River Basin. Data from 64 meteorological stations were obtained from the China Meteorological Data Sharing Network (https: / / data.cma.cn / ). These data primarily include monthly temperature, precipitation, relative humidity, sunshine, wind speed, and atmospheric pressure from 1982 to 2020. The station data were interpolated to a 0.25° grid (a total of 651 grid points). The monthly SPEI index (SPEI1) was then calculated at these grid points. The NDVI data used include the AVHRR GIMMS NDVI version 3g.v1 dataset (https: / / ecocast.arc.nasa.gov / data / pub / gimms / 3g.v1, 1982–2015) and the MODIS13 A3 dataset (http: / / ladsweb.modaps.eosdis.nasa.gov.search, February 2000–December 2020). The MODIS13 A3 data are primarily used to extend the AVHRR GIMMS NDVI data from January 2016 to December 2020.

[0058] Taking the Pearl River Basin as an example, we quantitatively identified the thresholds for vegetation loss under drought events and the recovery times for different types of vegetation. The detailed steps are as follows:

[0059] S1: Based on historical observation data (1982-2020), the run theory is used to identify drought events at the grid scale (resolution of 0.25°), including the start time, duration, end time and intensity of drought events;

[0060] Specifically, the one-month SPEI index (SPEI1) was used as the research object. Run-length theory was used at the grid level (with a resolution of 0.25°) to identify drought events within the study period. A drought event was considered to have occurred when the SPEI1 was less than -0.5 and lasted for at least two months. When the interval between two drought events was less than two months, the two drought events were merged into one. Characteristics of each drought event were extracted, including its start time, duration, end time, and intensity.

[0061] S2: Eliminate vegetation data during historical drought periods and estimate vegetation changes caused by non-drought factors;

[0062] Specifically, the multi-year average vegetation index (NDVI) from January to December is calculated at the grid points as the vegetation status not affected by drought factors. The calculation formula is as follows:

[0063]

[0064] in, represents the NDVI in month i when not affected by drought (1≤i≤12); n is the number of years in which month i was not affected by drought from 1982 to 2020.

[0065] S3: Estimation of vegetation changes caused by drought events;

[0066] The specific steps for calculating vegetation changes caused by drought events are as follows:

[0067] S301: First, calculate the lag time of vegetation response to drought. Use the maximum Pearson correlation coefficient method to calculate the correlation coefficient between SPEI1 and NDVI with different lag times (0-24 months) at each grid point. Select the one with the maximum correlation coefficient r i The corresponding lag months i is taken as the lag time (unit: month).

[0068] r i =corr(NDVI,SPEI i )0≤i≤24

[0069] r max-lag =max(r i )0≤i≤24

[0070] Among them, r i is the Pearson correlation coefficient of the time lag i months, where i ranges from 0 to 24 months. NDVI represents the monthly NDVI time series (1982-2020), SPEI represents the SPEI1 time series (1980-2020), and r max-lag For r i For example, for a 1-month lag, the correlation analysis is performed using the monthly SPEI1 from December 1981 to November 2020 and the NDVI from January 1982 to December 2020, and so on up to a 24-month lag.

[0071] S302: For a drought event, the overall change in vegetation (NDVI) after the drought event and the change in vegetation (NDVI) caused by non-drought factors are expressed as:

[0072] TC ndvi =NDVI e -NDVI s

[0073] NC ndvi =N_NDVI e -N_NDVI s

[0074] Among them, TC ndvi Represents the overall change in vegetation (NDVI) after a drought event; NC ndvi The NDVI represents changes in vegetation (NDVI) due to non-drought factors (such as seasonal factors and human activity). The subscript s represents the first month before the drought event begins, and e represents the month after the drought event ends.

[0075] S303: The NDVI change ΔVC caused by the drought event can be expressed as:

[0076] ΔVC=TC ndvi -NC ndvi

[0077] S4: Use probability distribution method to classify vegetation changes caused by drought events;

[0078] The specific steps are as follows:

[0079] S401: Calculate the ΔVC caused by all drought events at the grid scale and use probability distribution to describe the change of ΔVC. ΔVC is fitted by selecting multiple distributions (normal distribution, gamma distribution, generalized extreme value distribution, lognormal distribution), and the optimal distribution is selected using the Kolmogorov-Smirnov test and root mean square error. Figure 2 , the results show that ΔVC follows a normal distribution with a mean of 0.0032 and a standard deviation of 0.0793. In order to better quantify the changing characteristics of vegetation, the fitted normal distribution is converted into a standard normal distribution ΔVC normal .

[0080] S402: Referring to the idea of ​​drought classification based on the Standardized Precipitation Index (SPI), based on the idea of ​​probability distribution, ΔVC normal <-0.5 (the probability of occurrence is about 30.8%) is defined as vegetation loss, and ΔVC normal >0.5 (occurrence probability is about 30.8%) is defined as vegetation growth. Four vegetation loss levels are defined according to vegetation changes: mild loss (-1 < ΔVC normal <-0.5), moderate loss (-1.5<ΔVC normal <-1), severe losses (-2<ΔVC normal <-1.5) and extreme losses (ΔVC normal <-2), the corresponding probability of occurrence is approximately 15%, 9.2%, 4.4% and 2.3% respectively (Table 1). As shown in Table 1, ΔVC normalThe ΔVC corresponding to the threshold of -0.5 is -0.0365, which means that vegetation loss will occur only when ΔVC is less than -0.0365. Since the mean value of ΔVC (0.0032) is much larger than the ΔVC corresponding to the threshold of -0.5 (-0.0365), even if the ΔVC in the range of (0, 0.0032) corresponds to normal It is a negative value, but it is far from meeting the vegetation loss condition and strictly falls within the normal fluctuation range of vegetation (-0.5, 0.5), which also shows that the vegetation loss classification method adopted in this study is reasonable.

[0081] S5: Estimate the time from vegetation loss to recovery based on the threshold of normal vegetation fluctuation.

[0082] Specifically, for the scenario of vegetation loss caused by drought events, when the change in vegetation (NDVI) relative to the pre-drought period after the drought ends is less than the threshold of the normal fluctuation range of vegetation (NDVI) (i.e., -0.5 < ΔVC normal <0.5), the vegetation is considered to have returned to normal, and its expression is as follows:

[0083] ΔVC -0.5 ≤TC ndvi [s,e+k] -NC ndvi [s,e+k] ≤ΔVC 0.5

[0084] Where s represents the month before the drought event begins; e represents the lag month of vegetation response to drought; k is the time required for vegetation to recover to normal growth state (months); TC ndvi [s,e+k] represents the overall change in NDVI from the onset of the drought event to vegetation recovery; NC ndvi [s,e+k] It represents the change in NDVI caused by non-drought factors (such as seasonal factors and human activity interference) from the occurrence of drought events to vegetation recovery. -0.5 and ΔVC 0.5 They represent the ΔVC values ​​corresponding to thresholds of -0.5 and 0.5 (-0.0365 and 0.0429, respectively).

[0085] The results and analysis obtained according to the above method are as follows:

[0086] Figure 3Figure 3a shows the spatial distribution of the probability of normal vegetation conditions, vegetation growth, and vegetation loss after the drought event in the Pearl River Basin from 1982 to 2020. It also shows the difference between the probabilities of vegetation growth and loss. 3a is the spatial distribution of the probability of vegetation increase, 3b is the spatial distribution of the probability of vegetation loss, 3c is the spatial distribution of the probability of normal vegetation fluctuation, and 3d is the spatial distribution of the difference between the probability of vegetation growth and loss. Figure 3 In d, positive (negative) values ​​indicate that the probability of vegetation increase is greater (less) than the probability of vegetation loss. Figure 3 As shown in Figure a, in most areas of the Pearl River Basin, the probability of vegetation increase after drought is between 25% and 35%. The probability of vegetation increase is higher in the northwest, northeastern and central parts of the Xijiang Basin and the northern part of the Dongjiang Basin, while it is lower in the southwest of the Xijiang Basin and the southern part of the Pearl River Delta. Figure 3 As shown in b, the area with a high probability of vegetation loss (up to 50%) appears in the southwestern part of the Xijiang River Basin, while the probability of vegetation loss in most areas of the eastern Pearl River Basin is low (<25%). Figure 3 As shown in Figure c, the probability of normal vegetation conditions is highest (up to 60%) in the Dongjiang River Basin, the Pearl River Delta, and the eastern Xijiang River Basin, indicating that the drought has had limited impacts on vegetation in these areas. Overall, the probability of vegetation loss is higher than the probability of gain in the middle and upper reaches of the Xijiang River Basin and the Pearl River Delta. The probability of vegetation gain is higher than the probability of loss in the northeastern Xijiang River Basin and most areas of the Beijiang and Dongjiang River Basins.

[0087] The statistical results of the four types of vegetation loss (mild, moderate, severe and extreme loss) are shown in Table 1 below. Based on the data in Table 1, the probability of vegetation loss of different degrees after the drought event is statistically analyzed, and the spatial distribution results are shown in Table 1. Figure 4 As shown. Figure 4 It can be seen that the smaller the degree of vegetation loss, the higher the probability of occurrence. In general, the probability of occurrence is from high to low as mild loss, moderate loss, severe loss and extreme loss. From a spatial perspective, the Dongjiang River Basin, Beijiang River Basin, Pearl River Delta and the eastern part of the Xijiang River Basin are more likely to experience mild vegetation loss (up to 70%). Figure 4 a) The probability of moderate losses is relatively high in the central part of the Xijiang River Basin (30% to 40%) and low in some areas of the northeastern part of the Xijiang River Basin (<20%, Figure 4 b) The probability of severe and extreme losses is mostly within 25%, and even less than 10% in the Dongjiang River Basin and the Pearl River Delta, so the possibility of occurrence is relatively small ( Figure 4 c and d).

[0088] Table 1

[0089]

[0090] The drought events that caused vegetation loss in the historical period were counted at the grid points, and the average recovery time after vegetation loss caused by drought events of different intensities was further counted. The spatial distribution is as follows: Figure 5 Overall, the average recovery time for vegetation loss under all drought conditions in most areas of the Pearl River Basin is within 6 months. In the northern part of the Dongjiang River Basin, the recovery time for vegetation loss may exceed 10 months ( Figure 5 Specifically, vegetation loss caused by mild drought took longer to recover (>10 months) in a few areas, such as the western part of the Xijiang River Basin and the upper reaches of the Beijiang and Dongjiang River Basins. Figure 5 b); Under moderate drought stress ( Figure 5 c) The vegetation recovery time in the Dongjiang River Basin is longer than that in other areas (>8 months); and under severe drought stress ( Figure 5 d) The vegetation recovery time in the upper and lower reaches of the Xijiang River Basin is longer than in other areas.

[0091] This study further calculated the average recovery time required for wetland, forest, grassland, and cultivated land losses caused by different types of drought. The specific data are shown in Table 2 below. Comparing the four types of vegetation, it can be found that under all drought conditions, wetlands have the longest average recovery time (6.4 months), followed by forests (5.1 months) and cultivated land (4.8 months). Grassland has the shortest recovery time (4.2 months).

[0092] Table 2

[0093]

[0094] Compared with the existing methods, the present invention has the following advantages and beneficial effects:

[0095] (1) The present invention can effectively separate vegetation changes caused by non-drought factors (such as season and underlying surface conditions), improve the accuracy of estimating vegetation changes caused by drought, and effectively reduce the uncertainty of drought impact assessment;

[0096] (2) Based on probability distribution theory, a quantitative identification method for vegetation loss thresholds was developed, which can effectively estimate the magnitude of vegetation loss caused by different types of drought events and realize the classification of vegetation loss levels;

[0097] (3) From the perspective of drought event monitoring, a method for calculating the time required for vegetation to recover from loss to normal state was proposed based on the vegetation loss threshold, which effectively solved the problem of estimating vegetation recovery time under different types of drought conditions.

[0098] The above-described embodiments merely represent several implementations of the present invention. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous modifications and improvements without departing from the spirit of the present invention, and the present invention is intended to encompass such modifications and variations.

Claims

1. A quantitative identification method for vegetation loss and recovery under drought stress, characterized in that: The following steps are involved: S1: Using run theory to identify drought events at the grid scale, specifically including characteristics such as the start time, duration, end time, and intensity of the drought event; the identification of the drought event is based on the drought index, which is the Standardized Precipitation Evaporation Index (SPEI) and the Normalized Difference Vegetation Index (NDVI); S2: Eliminate vegetation data during historical drought periods and estimate vegetation changes caused by non-drought factors; S3: Estimation of vegetation changes caused by drought events; S4: Classify vegetation changes caused by drought events using a probability distribution method; the specific steps of S4 are as follows: S401: Calculate the ΔVC caused by all drought events at the grid scale. This paper uses distribution probability to describe the change of ΔVC. By selecting multiple distributions to fit ΔVC, the Kolmogorov-Smirnov test and root mean square error are used to screen the optimal distribution. In order to better quantify the changing characteristics of vegetation loss and growth, the fitted distribution is then converted into a standard normal distribution ΔVC. normal , the multiple distributions include normal distribution, gamma distribution, generalized extreme value distribution, and lognormal distribution; S402: Referring to the idea of ​​​​standardized precipitation index SPI drought level classification, based on the idea of ​​probability distribution, the present invention divides ΔVC normal <-0.5 is defined as vegetation loss, with a probability of 30.8%. normal >0.5 is defined as vegetation growth, with a probability of 30.8%; S403: Define four vegetation loss levels based on the magnitude of vegetation change: Mild loss -1 < ΔVC normal <-0.5, moderate loss -1.5<ΔVC normal <-1, severe loss -2<ΔVC normal <-1.5 and extreme loss ΔVC normal <-2, corresponding to the probability of occurrence of 15%, 9.2%, 4.4% and 2.3% respectively; S5: Based on the threshold of normal vegetation fluctuation, the time from vegetation loss to recovery is estimated; the specific steps of S5 are: Identify the time it takes for vegetation to recover from loss to normal fluctuation state: When the vegetation in a certain month after the drought ends is within the normal fluctuation range relative to the month before the drought, the vegetation is considered to have returned to normal growth state. The calculation formula is as follows: ΔVC -0.5 ≤TC ndvi [s,e+k] -NC ndvi [s,e+k] ≤ΔVC 0.5 Where s represents the first month before the drought event begins; e represents the lag month after the drought event ends; k is the time required for vegetation to return to normal growth after the drought, in months; TC ndvi [s,e+k] represents the overall change in NDVI from the onset of the drought event to vegetation recovery; NC ndvi [s,e+k] represents the change in NDVI caused by non-drought factors from the onset of the drought event to vegetation recovery, ΔVC -0.5 and ΔVC 0.5 They represent the ΔVC values ​​corresponding to thresholds of -0.5 and 0.5, respectively.

2. The method for quantitatively identifying vegetation loss and recovery under drought stress according to claim 1, characterized in that: The S3 step is specifically as follows: S3: Identify the lag time of vegetation response to drought. Calculate the overall vegetation change from the first month before the drought event to the vegetation lag month after the drought ends, as well as the vegetation change caused by non-drought factors. The difference between the two is the vegetation change caused by drought factors.

3. The method for quantitatively identifying vegetation loss and recovery under drought stress according to claim 2, characterized in that: The S4 step is specifically as follows: S4: Fit the probability distribution of vegetation changes caused by drought in historical periods, select the optimal distribution, transform it into a standard normal distribution, and classify the degree of vegetation loss according to the probability of occurrence.

4. The method for quantitatively identifying vegetation loss and recovery under drought stress according to claim 3, characterized in that: The SPEI of the drought event is less than -0.5 and the duration is greater than 2 months. When the time interval between adjacent drought events is less than 2 months, the two drought events are merged into one event.

5. The method for quantitatively identifying vegetation loss and recovery under drought stress according to claim 3, characterized in that: The S2 step is specifically as follows: The vegetation data during the drought period of the study period were eliminated, and the multi-year average vegetation index from January to December was calculated as the vegetation state not affected by drought factors. The specific calculation formula is as follows: Among them, NDVI i st represents the multi-year average of NDVI in month i when it is not affected by drought, 1≤i≤12; n is the number of years in which month i is not affected by drought during the study period.

6. The method for quantitatively identifying vegetation loss and recovery under drought stress according to claim 3, characterized in that: The specific steps of S3 are as follows: S301: First, calculate the lag time of vegetation response to drought, use the maximum Pearson correlation coefficient method to determine the time lag effect of drought on vegetation, calculate the correlation coefficients between SPEI and NDVI with different lag times of 1-24 months, and select the maximum correlation coefficient r i The corresponding lag months i is taken as the lag time; r i =corr(NDVI,SPEI i ) r max-lag =max(r i ) Among them, r i is the Pearson correlation coefficient for time lag i months, r max-lag For r i The maximum value of i in the present invention is 0 to 24 months; S302: For a drought event, the overall change in vegetation NDVI after the drought and the change in vegetation NDVI caused by non-drought factors are: TC ndvi =NDVI e -NDVI s NC ndvi =N_NDVI e -N_NDVI s Among them, TC ndvi represents the overall change in vegetation NDVI after a drought event, NC ndvi represents the vegetation NDVI changes caused by non-drought factors, the subscript s represents the first month before the drought event begins, and e represents the lagged month after the drought event ends; S303: The NDVI change ΔVC caused by the drought event can be expressed as: ΔVC=TC ndvi -NC ndvi 。

Images