A method and system for predicting total primary productivity of ecosystems under drought stress

Abstract

The application provides a method and system for predicting total primary productivity of an ecosystem under drought stress, comprising: coupling nonlinear response under drought stress and hydrothermal carbon physical mechanism to provide important and operable reference for response mechanism of future drought evolution to carbon sink function of an ecosystem under global and regional prediction of climate change scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ecological hydrology prediction, and particularly relates to a method and system for predicting total primary productivity of an ecological system under drought stress. BACKGROUND

[0002] In previous studies, considering the causes of meteorological, hydrological, agricultural and social economy, in order to quantitatively depict the degree of water deficit, scholars at home and abroad have proposed various drought indices such as standardized precipitation index (SPI), standardized precipitation evapotranspiration index, Palmer drought severity index (PDSI), standardized runoff index (SRI) and the like. However, each drought index mainly considers one or more of meteorological and hydrological elements, and cannot comprehensively depict the internal physical characteristics of drought events. With the rapid development of satellite telemetry and data inversion technology in recent years, quantitative observation products based on GRACE and GRACE-FO gravity satellites not only provide wider coverage and higher spatial and temporal resolution, but also effectively make up for the defects of insufficient ground station network, which provides strong data support for drought monitoring in areas without data. Therefore, scholars have begun to use GRACE / GRACE-FO satellite inversion data to evaluate drought, and found that the drought severity index based on terrestrial water storage anomaly (TWSA-DSI) can not only effectively monitor drought events, but also capture large-scale spatiotemporal changes in surface fluxes.

[0003] Drought is one of the most significant stressors affecting the carbon sink function of terrestrial ecosystems at both regional and global scales. Using satellite and flux station data, researchers domestically and internationally have assessed the impact of drought on vegetation carbon budgets and found that extreme droughts significantly reduce gross primary productivity (GPP) and net ecosystem productivity (NEP). Some researchers have suggested that the recent decline in ecosystem diversity and carbon sink capacity in the Amazon rainforest is linked to increased drought severity. Global temperatures and vapor pressure deficits are likely to continue increasing, leading to more frequent and intense heat-drought combined disasters with significant impacts on GPP. Following severe climate disasters, plant recovery is typically delayed due to reduced growth, irreversible damage to hydraulic conductivity, and depletion of carbon storage. This delayed recovery increases vulnerability to subsequent climate disasters, impacting terrestrial ecosystem carbon sinks. The world is currently experiencing climate change dominated by warming, and the response of ecosystem carbon sinks to warming exhibits significant spatial and temporal heterogeneity, influenced by atmospheric water vapor capacity, relative humidity, and atmospheric stability. Especially under drought stress, different vegetation types exhibit seasonal and nonlinear responses to drought under climatic conditions. However, current simulations of global gross primary productivity (GPP) and long-term trends using various methods are subject to significant uncertainty, hindering estimates of the terrestrial carbon budget and understanding of its feedback loop with drought. While scholars worldwide are increasingly focusing on the impact of drought on ecosystem productivity, no literature has comprehensively considered the dynamic responses of terrestrial ecosystem carbon budgets under future drought stress and combined them with emergent constraints to predict ecosystem GPP under future drought stress. Summary of the Invention

[0004] The present invention provides a method and system for predicting the total primary productivity of an ecosystem under drought stress, which is used to solve the defect in the prior art that the nonlinear response characteristics of carbon budget variables under drought stress are not considered in future carbon flux estimation.

[0005] In a first aspect, the present invention provides a method for predicting the total primary productivity of an ecosystem under drought stress, comprising:

[0006] Acquire multi-source data sets including carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data, and Earth system model output in the study area;

[0007] The Clausius-Clapeyron thermodynamic equation was used to determine the saturated vapor pressure deficit, relative humidity, and wet-bulb temperature. The drought characteristics in the study area were extracted based on the run theory and TWSA-DSI:

[0008] Based on the water and heat factors and drought index TWSA-DSI, combined with the convergent cross-mapping function, the causal relationship between the water and heat stress factors, drought indicators and ecosystem GPP is obtained, and the key water and heat factors affecting GPP are extracted;

[0009] Based on the key hydrothermal factors affecting GPP and the drought characteristics, a long-short-term memory machine learning model was constructed using the key factors derived from the ERA5 reanalysis data and the GPP data inverted by MODIS. Combined with the hydrothermal factors under different scenarios output by the Earth System Model, GPP data under different future scenarios were predicted;

[0010] Combining the wet-bulb temperature and terrestrial water storage (TWS) data of each grid point in the study area output by the global climate model, the TWSA-DSI is obtained. The GPP data under different future scenarios are used to construct an emergence constraint model to reduce the uncertainty of GPP prediction under future drought stress.

[0011] According to a method for predicting gross primary productivity of ecosystems under drought stress provided by the present invention, a multi-source data set including carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data, and Earth system model output is obtained in the study area, including:

[0012] The FLUXNET 2015 dataset, which is based on flux tower observations from several locations around the world, was quality controlled and bias corrected, and soil moisture was interpolated using the marginal distribution method.

[0013] Converting the atmospheric reanalysis data and the MODIS satellite inversion data into daily scale data with a preset spatial resolution and monthly scale data with a preset spatial resolution, respectively;

[0014] The three GRACE / GRACE-FO gravity satellite datasets in the gravity satellite inversion data were interpolated to a preset scale spatial grid to make the terrestrial water storage anomaly data consistent with the spatial resolution of the variables extracted by ERA5. The average of each product was taken within each time step to obtain the monthly scale terrestrial water storage anomaly dataset.

[0015] Five global climate model datasets released by the sixth phase of the International Coupled Model Intercomparison Project (CMIP6) were used.

[0016] According to a method for predicting the gross primary productivity of an ecosystem under drought stress provided by the present invention, the Clausius-Clapeyron thermodynamic equation is used to determine the saturated water vapor pressure deficit, relative humidity, and wet-bulb temperature. The drought characteristics in the study area are extracted based on run theory and TWSA-DSI, including:

[0017] The saturation water vapor pressure is determined using the Clausius-Clapeyron thermodynamic equation:

[0018]

[0019] in, and are the first and second integral constants, which are taken as 273.16 K and 611 Pa respectively; Lv is the latent heat of vaporization constant, which is taken as 2.5×106 J kg-1; Rv is the water vapor gas constant, which is taken as 461 J kg-1 K-1; is the saturation vapor pressure deficit, T is the air temperature;

[0020] The temperature T 2m and dew point temperature T dew Substituting into the Clausius-Clapeyron thermodynamic equation, we can obtain the near-saturation vapor pressure deficit ;

[0021] The temperature of each grid point ( T 2m ) and dew point temperature ( T dew ) are substituted as input variables into the Clausius-Clapeyron thermodynamic equation to calculate the relative humidity of each grid point ;

[0022] Based on the air temperature and relative humidity data, calculate the wet-bulb temperature at each grid point:

[0023] Where RH is the monthly average relative humidity, T 2m is the monthly mean temperature, T w is the derived wet-bulb temperature, atan is the inverse tangent function;

[0024] Extraction of abnormal land water storage data :

[0025]

[0026] in, is the jth month of the i-th year TWSA data, and The jth month in the study period TWSA The mean and standard deviation of

[0027] According to the run theory, the drought frequency, duration and intensity of the period in which the TWSA-DSI is less than the preset drought threshold are determined.

[0028] According to a method for predicting gross primary productivity of an ecosystem under drought stress provided by the present invention, based on water and heat factors and drought index TWSA-DSI, combined with a convergent cross-mapping function, the causal relationship between the water and heat stress factors, drought index and ecosystem GPP is obtained, and key water and heat factors affecting GPP are extracted, including:

[0029] The hydrothermal factor and the drought index are used as variables X of the convergent cross-mapping function, and GPP is used as variable Y of the convergent cross-mapping function;

[0030] Constructing a coupled logic differential equation of the convergent cross-mapping function, and determining the intrinsic growth rate of variable X, the intrinsic growth rate of variable Y, the mutual dynamic influence factor of the variables, and the time step of the coupled logic differential equation;

[0031] The cross-mapping skill coefficients of the variable X and the variable Y are calculated to meet the cross-mapping skill coefficients corresponding to the preset convergence conditions, and the key hydrothermal factors affecting GPP are extracted.

[0032] According to a method for predicting gross primary productivity of an ecosystem under drought stress provided by the present invention, based on the key water and heat factors affecting GPP and the drought characteristics, a long-short-term memory machine learning model is constructed using the key factors of ERA5 reanalysis data and GPP data inverted by MODIS. The model is combined with the water and heat factors under different scenarios output by the Earth system model to predict GPP data under different future scenarios, including:

[0033] Extract several driving factors from ERA5 reanalysis data, terrestrial water storage data retrieved from GRACE gravity satellites, and GPP data retrieved from MODIS, and use the top number of driving factors as key water and heat factors.

[0034] Determine the key hydrothermal factors corresponding to the current moment and several moments before the current moment as input variables of the long-short-term memory machine learning model, and train to obtain a GPP prediction model;

[0035] The hydrothermal factors, TWSA-DSI and drought characteristics under different scenarios output by the Earth system model are used to drive the GPP prediction model to obtain GPP data under different future scenarios.

[0036] According to a method for predicting gross primary productivity of ecosystems under drought stress provided by the present invention, the future drought index (TWSA-DSI) is obtained by combining the wet-bulb temperature and terrestrial water storage (TWS) data of each grid point within the study area output by a global climate model. An emergent constraint model is constructed based on the GPP data under different future scenarios to reduce the uncertainty of future GPP predictions under drought stress, including:

[0037] Based on the temperature and relative humidity data of each global climate model historical period, combined with the wet-bulb temperature of each grid point in the study area, the linear regression method was used to obtain the trend of annual average wet-bulb temperature;

[0038] A spatial sliding window method is used to obtain the multi-gridpoint annual average wet-bulb temperature change trend data under five global climate models, and a first emergent constraint sub-model is constructed based on the multi-gridpoint annual average wet-bulb temperature change trend data, the GPP data under different future scenarios, the first constraint parameter and the second constraint parameter;

[0039] A spatial sliding window method is used to obtain multi-grid TWSA-DSI under five global climate models, and a second emergence constraint sub-model is constructed based on the multi-grid TWSA-DSI, the GPP data under different future scenarios, the third constraint parameter, and the fourth constraint parameter.

[0040] Solving the first constraint parameter, the second constraint parameter, the third constraint parameter, and the fourth constraint parameter based on a least squares method, and obtaining the emergence constraint model from the first emergence constraint sub-model and the second emergence constraint sub-model;

[0041] The emergence constraint model is driven under different future scenarios to obtain the GPP prediction results under drought stress under different scenarios.

[0042] In a second aspect, the present invention further provides a system for predicting total primary productivity of an ecosystem under drought stress, comprising:

[0043] The acquisition module is used to obtain carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data and multi-source data sets output by the Earth system model in the study area;

[0044] The first extraction module is used to determine the saturated water vapor pressure deficit, relative humidity and wet bulb temperature using the Clausius-Clapeyron thermodynamic equation, and extract the drought characteristics in the study area based on the run theory and TWSA-DSI:

[0045] The second extraction module is used to obtain the causal relationship between the water and heat stress factors, drought indicators and ecosystem gross primary productivity (GPP) based on water and heat factors and drought indicators TWSA-DSI, combined with the convergent cross-mapping function, and extract the key water and heat factors affecting GPP;

[0046] A prediction module is used to construct a long-short-term memory machine learning model based on the key water and heat factors affecting GPP and the drought characteristics, using the key factors from the fifth-generation global climate reanalysis data set ERA5 reanalysis data and the GPP data inverted by MODIS, and combining the water and heat factors under different scenarios output by the Earth system model to predict GPP data under different future scenarios;

[0047] The constraint module is used to combine the wet-bulb temperature and terrestrial water storage (TWS) data of each grid point in the study area output by the global climate model to obtain the future drought index (TWSA-DSI). The emergence constraint model is constructed based on the GPP data under different future scenarios to reduce the uncertainty of GPP prediction under future drought stress.

[0048] In a third aspect, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for predicting the total primary productivity of an ecosystem under drought stress as described above is implemented.

[0049] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the prediction of the total primary productivity of an ecosystem under drought stress as described in any one of the above.

[0050] The method and system for predicting the total primary productivity of ecosystems under drought stress provided by the present invention, by coupling the nonlinear response and hydrothermal carbon physical mechanisms under drought stress, provide an important and highly operational reference basis for global and regional prediction and assessment of the response mechanism of future drought evolution to the carbon sink function of ecosystems under climate change scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 Schematic diagram of the process of the method for predicting total primary productivity of ecosystems under drought stress provided by the present invention;

[0053] Figure 2 It is a schematic diagram of the run theory provided by the present invention;

[0054] Figure 3 is a schematic diagram of the convergent cross-mapping technology provided by the present invention;

[0055] Figure 4 This is a schematic diagram of the empty window sliding technology provided by the present invention;

[0056] Figure 5 This is a schematic diagram of the structure of the method for predicting the total primary productivity of an ecosystem under drought stress provided by the present invention;

[0057] Figure 6It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0058] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0059] Figure 1 : is a flow chart of a method for predicting total primary productivity of an ecosystem under drought stress provided by an embodiment of the present invention, such as Figure 1 Shown, including:

[0060] Step 100: Acquire a multi-source dataset of carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data, and Earth system model output within the study area;

[0061] Step 200: Using the Clausius-Clapeyron thermodynamic equation to determine the saturated water vapor pressure deficit, relative humidity, and wet-bulb temperature, the drought characteristics in the study area are extracted based on the run theory and TWSA-DSI:

[0062] Step 300: Based on the water and heat factors and the drought index TWSA-DSI, combined with the convergent cross-mapping function, the causal relationship between the water and heat stress factors, drought index and ecosystem GPP is obtained, and the key water and heat factors affecting GPP are extracted;

[0063] Step 400: Based on the key water and heat factors affecting GPP and the drought characteristics, a long-short-term memory machine learning model is constructed using the key factors derived from the ERA5 reanalysis data and the GPP data inverted by MODIS. The model is combined with the water and heat factors under different scenarios output by the Earth System Model to predict GPP data under different future scenarios.

[0064] Step 500: Combine the wet-bulb temperature and terrestrial water storage (TWS) data of each grid point in the study area output by the global climate model to obtain TWSA-DSI, and construct an emergence constraint model based on the GPP data under different future scenarios to reduce the uncertainty of GPP prediction under future drought stress.

[0065] Specifically, the method for predicting the total primary productivity of an ecosystem under drought stress proposed in an embodiment of the present invention includes the following steps:

[0066] Step 1: Obtain a multi-source dataset within the selected study area, including carbon flux observation data, atmospheric reanalysis data, MODIS satellite retrieval data, gravity satellite retrieval data, and Earth system model output, including gridded terrestrial water storage, gross primary productivity (GPP), and water and heat stress factors (temperature, soil moisture, radiation, precipitation, etc.);

[0067] Step 2: The Clausius-Clapeyron thermodynamic equation is used to determine the saturated water vapor pressure deficit, relative humidity, and wet-bulb temperature. Based on the run theory and the drought index TWSA-DSI, the drought characteristics in the study area are extracted, including frequency, duration, and intensity.

[0068] Step 3: Based on the hydrothermal factors and drought index TWSA-DSI extracted in Steps 1 and 2, the convergent cross mapping function (CCM) is used to evaluate the causal relationship between these long series of hydrothermal factors and drought indicators and ecosystem gross primary productivity (GPP), and then extract the key hydrothermal factors that affect GPP.

[0069] Step 4: Based on the key hydrothermal factors and drought characteristics determined in Step 3, a long-short-term memory machine learning model was constructed using the key factors from the ERA5 reanalysis data since 2000 and the GPP data retrieved from MODIS. The model was combined with the hydrothermal factors under different scenarios output by the Earth System Model to predict GPP data under different future scenarios.

[0070] Step 5: Combine the wet-bulb temperature and terrestrial water storage (TWS) data of each grid point in the study area output by the global climate model to derive the future drought index (TWSA-DSI); based on the GPP dataset predicted in step 4, construct an emergence constraint model to reduce the uncertainty of the gross primary productivity (GPP) prediction under future drought stress.

[0071] By coupling the nonlinear response under drought stress and the hydrothermal carbon physical mechanism, this invention provides an important and highly operational reference basis for global and regional prediction and assessment of the response mechanism of future drought evolution to the carbon sink function of ecosystems under climate change scenarios.

[0072] Based on the above embodiment, step 100 includes:

[0073] In step 1.1, carbon flux observation data for the study area were collected using M global flux tower observations provided by the FLUXNET 2015 dataset. The variables used included soil moisture (including all soil layers), air temperature, precipitation, radiation, relative humidity, and GPP. These data were rigorously quality-controlled and bias-corrected. Missing values ​​for soil moisture were interpolated using the marginal distribution (MD) method. In this example, M is 2¹².

[0074] For each grid point in the study area, daily-scale precipitation, near-surface 2m air temperature, 2m dew point temperature, soil moisture content, shortwave radiation and other water and heat factors in the study area from 1985 to 2020 were collected from the fifth-generation atmospheric reanalysis dataset (ERA5) of the European Centre for Medium-Range Weather Forecasts, and the above variables were converted into daily-scale data with a spatial resolution of 0.5°×0.5°.

[0075] This example uses a global gridded GPP product for the study area from 2000 to 2019, based on MODIS data and theoretical inversion of light use efficiency. This dataset utilizes state-of-the-art vegetation index gap-filling and smoothing algorithms, as well as separate processing for the C3 / C4 photosynthesis pathways, resulting in superior performance when validated against in situ GPP estimates. This example converts these variables to monthly data with a spatial resolution of 0.5° × 0.5°.

[0076] In step 1.2, the three GRACE / GRACE-FO gravity satellite datasets were interpolated to a 0.5° × 0.5° spatial grid to make the spatial resolution of the variables extracted from the TWSA data consistent with that of the ERA5 reanalysis data. The average of each product was taken within each time step to obtain the TWSA monthly dataset.

[0077] Earth's gravity field information from the GRACE / GRACE-FO satellites can be used to infer terrestrial water storage. Data interpretation is conducted by institutions including the Potsdam Geoscience Center in Germany, the Jet Propulsion Laboratory (JPL) at the California Institute of Technology, the Center for Space Research at the University of Texas, Austin (CSR), and NASA's Goddard Space Flight Center (GSFC). These institutions publish monthly-scale global gravity field outputs. In this implementation, the latest mascon (mass concentration) products from JPL, CSR, and GSFC are used to generate a long-series TWSA (terrestrial water storage anomaly) dataset. To account for potential uncertainties arising from different data products, the three GRACE gravity satellite data sets are interpolated to a 0.5° × 0.5° spatial grid. The average of each product is taken at each time step, resulting in a monthly-scale TWSA dataset covering the study area from 2002 to the present.

[0078] Step 1.3, to project future climate scenarios, uses the five most recently released global climate models from the Coupled Model Inter-comparison Project Phase 6 (CMIP6): GFDL-ESM4, IPSL-CM6A-LR, MPI-ESM1-2-HR, MRI-ESM2-0, and UKESM1-0-LL. Compared to its earlier version (CMIP5), CMIP6 uses a matrix framework based on Shared Socioeconomic Pathways (SSPs) and Representative Concentration Pathways (RCPs). This example uses meteorological data output from three emission scenarios: SSP1-2.6, SSP3-7.0, and SSP5-8.5. The output variables of the five global climate models include monthly temperature, relative humidity, precipitation, soil moisture, radiation and carbon flux datasets. The historical period was selected from 1985 to 2014 (a total of 30 years). In order to predict the meteorological characteristics under future climate change scenarios, the last 30-year window of this century (2071-2100) was set as the future period.

[0079] The five global climate models described above were further driven using the global hydrological model H08 to obtain monthly terrestrial water storage data, covering both historical and future periods. The H08 model and related driving methods are conventional techniques in the art.

[0080] Based on the above embodiment, step 200 includes:

[0081] Step 2.1, determine the saturation vapor pressure deficit, relative humidity, and wet-bulb temperature using the Clausius-Clapeyron thermodynamic equation;

[0082] The Clausius-Clapeyron thermodynamic equation can quantitatively describe the saturation water vapor pressure Nonlinear relationship with temperature T:

[0083] (1)

[0084] in, and are the first and second integration constants, which are 273.16 K and 611 Pa respectively; L v is the latent heat of vaporization constant, which is taken as 2.5×106 J kg-1; R v is the water vapor gas constant, which is taken as 461 J kg-1 K-1; is the saturation vapor pressure deficit, and T is the air temperature.

[0085] Dew point temperature represents the temperature at which air is cooled to saturation with water vapor under the conditions of constant water vapor content and air pressure. Substituting it into the Clausius-Clapeyron equation can measure the actual water vapor pressure. T 2m ) and dew point temperature ( T dew ) are substituted into formula (1) to deduce the near-saturation vapor pressure deficit .

[0086] Furthermore, the temperature of each grid point ( T 2m ) and dew point temperature ( T dew ) are substituted into the Clausius-Clapeyron thermodynamic equation as input variables to calculate the relative humidity of each grid point ( RH ), specifically .

[0087] Furthermore, based on the air temperature and relative humidity data, the wet-bulb temperature of each grid point is calculated using the following formula: (2)

[0088] Where: RH is the monthly average relative humidity,T 2m is the monthly mean temperature, T w is the wet-bulb temperature to be derived, and atan is the inverse tangent function.

[0089] In step 2.2, the terrestrial water storage (TWSA) data retrieved from gravity satellites are used to extract the drought index (TWSA-DSI). Based on the run theory, the drought characteristics in the study area are further extracted, including frequency, duration, and intensity.

[0090] Drought Index TWSA-DSI It is a dimensionless normalized water storage anomaly index that is spatially comparable across different hydroclimatic regions. TWSA-DSI The calculation formula is as follows:

[0091] (3)

[0092] Where: Represents the jth month of the i-th year TWSA data, and The jth month in the study period TWSA The mean and standard deviation of .

[0093] This embodiment uses TWSA-DSI<-0.8 to identify droughts, and uses run theory to determine when the abnormal land water storage drought index TWSA-DSI is less than the preset drought threshold, and extracts the annual drought frequency, duration and intensity of the corresponding period. Figure 2 As shown, a schematic diagram of run theory is given.

[0094] Based on the above embodiment, step 300 includes:

[0095] Eight hydrothermal factors, including soil moisture (including all soil layers), air temperature, precipitation, evaporation, radiation, relative humidity, saturated water vapor pressure deficit, and the drought index TWSA-DSI obtained in step 2, were selected as independent variables based on the flux station observation data obtained in step 1, and carbon flux GPP was used as the dependent variable. The convergent cross-mapping technique was used to detect the causal relationship path between hydrothermal factors and carbon flux, and to identify the key hydrothermal influencing factors of carbon flux GPP.

[0096] CCM is derived from the phase space reconstruction theory and is an effective method for analyzing causal relationships and false correlations in multivariate time series of nonlinear dynamic systems. It is suitable for time series data analysis of complex ecosystems and can identify whether there is a nonlinear causal relationship between variables. Based on Takens theory and its extension theorem, the shadow manifold M is determined using observable time series. X and M Y Since variables X and Y are dynamically coupled, MX and M Y The points on the graph correspond to each other in time, which means that X is related to Y. The main algorithm is based on the nearest neighbor prediction method, which involves tracking the forward trajectory of nearby points in the lagged coordinate embedding. X Point M on Y The predictions are made for points on the time series, reflecting causal relationships based on predictive power. Then, by varying the length of the time series, L, different cross-mapping skills, ρ, are calculated. A trend chart of their variation with increasing L is obtained to observe convergence. The convergence of the CCM refers to the property that the correlation coefficient (cross-mapping skill ρ) between the predicted and observed values ​​increases with the length of the time series L used to construct the sample, and remains stable after reaching an extreme value. The model can be represented by two coupled logical difference equations:

[0097] (4)

[0098] (5)

[0099] in and is the time step, and represent the intrinsic growth rate of the variable, and Variables For variables Dynamic impact and variables For variables dynamic impact.

[0100] In this example, variable X represents water and heat factors (air temperature, soil moisture, radiation, precipitation, saturated water vapor pressure deficit calculated in step 2, relative humidity, wet bulb temperature, and drought index TWSA-DSI observed by FLUX flux), and variable Y represents gross primary productivity (GPP) based on FLUX flux observations. The cross-mapping skill ρ is defined as the correlation coefficient between variables X and Y. If ρ increases with the length of the time series and converges, a bidirectional causal effect between X and Y can be inferred. The main influencing factors of X or Y can then be determined by the high or low ρ value that remains stable after converging to an extreme value, thereby extracting the top eight key water and heat factors affecting GPP. Figure 3 As shown, a schematic diagram of the convergent cross-mapping technique is given.

[0101] Based on the above embodiment, step 400 includes:

[0102] Step 4.1: Based on step 3, determine the key hydrothermal factors and drought characteristics that affect GPP, and use the key factors derived from ERA5 reanalysis data since 2000 and MODIS inverted GPP data to construct a relationship model between the driving factors and GPP.

[0103] The key hydrothermal factors affecting GPP determined in step 3 can be used as explanatory variables. ERA5 reanalysis data, terrestrial water storage data retrieved from the GRACE gravity satellite, and GPP data retrieved from MODIS are further used to extract driving factors, including temperature, saturated vapor pressure deficit, precipitation, soil moisture, shortwave radiation, relative humidity, evaporation, drought frequency, duration, and intensity (a total of 10 driving factors). The annual drought characteristics are matched with the monthly key hydrothermal factors, that is, the drought characteristics of each month in the same year are set to the same value, and the top 8 key hydrothermal factors are selected as input variables. The input data of the machine learning model are all MODIS satellite retrieval period (2000-2019), and the time lag effect of each driving factor on GPP is taken into account. For each month, the driving factors of that month and the previous 1-3 months are selected. For this embodiment, a total of 8×4=32 hydrothermal factors are used as model inputs for training the long-short-term memory machine learning model.

[0104] The constructed long short-term memory machine learning model is used to simulate the GPP data since 2000, which is expressed as:

[0105] (6)

[0106] in, represents the total primary productivity simulated by machine learning at time t, represents the input variable at time t, represents the input variable at time t-1, represents the input variable at time t-2, represents the input variable at time t-3; F represents the long short-term memory machine learning model.

[0107] In step 4.2, based on the key factors selected in step 3 and the GPP prediction model constructed in step 4.1, the long series of key hydrothermal factors and terrestrial water reserves under different scenarios output by the Earth system model are used, combined with step 2 to extract the drought index TWSA-DSI for different future scenarios. The run theory is further combined to obtain the frequency, duration, and intensity of droughts under different future scenarios, driving the long-term and short-term machine learning model established in step 4.1 to predict the long series of GPP data sets under different future scenarios.

[0108] Based on the above embodiment, step 500 includes:

[0109] In step 5.1, based on the temperature and relative humidity data of each global climate model historical period, the wet-bulb temperature of each grid point in the study area is calculated in combination with step 2, and the trend of the annual average wet-bulb temperature is inferred using the linear regression method.

[0110] Furthermore, the air temperature of the ERA5 dataset and the relative humidity data derived in step 2 are used to derive the wet-bulb temperature of each grid point during the period 1985-2014, and its changing trend is calculated.

[0111] Step 5.2, such as Figure 4 As shown in the figure, for each grid point, the spatial sliding window method is used to search the grid point and its neighboring grid points, a total of 25 grid points. For each grid point, the trend of the annual average wet-bulb temperature in the historical period under each global climate model is calculated (denoted as Twet_Trend ). Further, the data of 25 grid points in 5 global climate models were integrated to obtain 125 sets of historical annual average wet-bulb temperature change trends and the future period predictions in step 4. The pairing combination of , constructs the constraint model 1 as follows:

[0112] (7)

[0113] Where: a1 and b1 represent the parameters of the emergence constraint model 1, Twet_Trend is the trend of wet-bulb temperature;

[0114] In addition, considering the stress effect of drought on the gross primary productivity of ecosystems, the drought index TWSA-DSI for historical and future periods is obtained based on the long series of terrestrial water storage sequences output by the Earth system model. For each grid point, the trend of the annual average drought index TWSA-DSI in the historical period under each global climate model is calculated (denoted as TWSADSI_Trend Furthermore, the data of 25 grid points in 5 global climate models were integrated to obtain 125 sets of historical annual average drought index TWSA-DSI change trends and the future period predicted by step 4. The pairing combination of , the second constraint model is constructed as follows:

[0115] (8)

[0116] Where: a2 and b2 represent the parameters of the emergence constraint model 2, TWSADSI-Trend is the trend of drought index TWSA-DSI. Figure 4 As shown, a schematic diagram of the empty window sliding technology is given.

[0117] Step 5.3, using the least squares method to solve the parameters a1, a2, b1 and b2 of the emergence constraint model; based on the future terrestrial water storage data predicted by the H08 model, combined with step 2, the drought index TWSA-DSI is extracted. In this example, TWSA-DSI < -0.8 is used to identify drought, and then the GPP anomaly of the future drought month is extracted, which is recorded as the predicted result of the total primary productivity under drought stress in the future period. .

[0118] Based on the reanalysis data of the ERA5 dataset, the changing trend of GPP anomalies under drought stress conditions during 1985-2014 was inferred, and the trend term was substituted into the constructed emergence constraint model as follows:

[0119] (9)

[0120] Where: represents the predicted total primary productivity under drought stress in the future period after correction; Twet_Trend is the changing trend of the annual average wet-bulb temperature at this grid point obtained based on step 2; TWSADSI-Trend is the trend of the annual average drought index TWSA-DSI obtained in step 2 at this grid point; the parameters a1, a2, b1 and b2 are derived from the parameters determined by the emergence constraint models 1 and 2.

[0121] In step 5.4, the emergence constraint model under different future scenarios is constructed, and finally the prediction results of total primary productivity under drought stress under three different future scenarios are obtained.

[0122] The following describes the total primary productivity prediction system of an ecosystem under drought stress provided by the present invention. The total primary productivity prediction system of an ecosystem under drought stress described below and the total primary productivity prediction method of an ecosystem under drought stress described above can be referenced to each other.

[0123] Figure 5 : is a schematic diagram of the structure of the method for predicting the total primary productivity of an ecosystem under drought stress provided by an embodiment of the present invention. Figure 5 As shown, it includes: an acquisition module 51, a first extraction module 52, a second extraction module 53, a prediction module 54 and a constraint module 55, wherein:

[0124] The acquisition module 51 is used to obtain carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data and multi-source data sets output by the Earth system model in the study area; the first extraction module 52 is used to determine the saturated water vapor pressure deficit, relative humidity and wet bulb temperature using the Clausius-Clapeyron thermodynamic equation, and extract the drought characteristics in the study area based on the run theory and TWSA-DSI: the second extraction module 53 is used to obtain the causal relationship between the water and heat stress factors, drought indicators and the gross primary productivity (GPP) of the ecosystem based on water and heat factors and drought indicators TWSA-DSI, combined with the convergent cross-mapping function, and extract the key water and heat factors affecting GPP. Factors; the prediction module 54 is used to construct a long-short-term memory machine learning model based on the key water and heat factors affecting GPP, using the key factors derived from the fifth-generation global climate reanalysis data set ERA5 reanalysis data and the GPP data inverted by MODIS, and combining the water and heat factors under different scenarios output by the earth system model to predict the GPP data under different scenarios in the future; the constraint module 55 is used to combine the wet-bulb temperature and terrestrial water storage TWS data of each grid point in the study area output by the global climate model to obtain the future drought index TWSA-DSI, and construct an emergent constraint model based on the GPP data under different future scenarios to reduce the uncertainty of GPP prediction under future drought stress.

[0125] Figure 6 An example of a physical structure diagram of an electronic device is shown below. Figure 6As shown, the electronic device may include: a processor (processor) 610, a communication interface (Communications Interface) 620, a memory (memory) 630 and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call the logic instructions in the memory 630 to execute the method for predicting the total primary productivity of the ecosystem under drought stress, the method comprising: obtaining carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data and a multi-source data set output by the earth system model in the study area; using the Clausius-Clapeyron thermodynamic equation to determine the saturated water vapor pressure deficit, relative humidity and wet-bulb temperature, extracting the drought characteristics in the study area based on the run theory and TWSA-DSI: based on the water and heat factor and the drought index TWSA-DSI, combined with the convergent cross-mapping function, obtaining the water and heat stress factor, The causal relationship between drought indicators and ecosystem GPP is established, and key water and heat factors affecting GPP are extracted; based on the key water and heat factors affecting GPP, the key factors derived from ERA5 reanalysis data and GPP data inverted by MODIS are used to construct a long-short-term memory machine learning model, and the water and heat factors under different scenarios output by the Earth system model are combined to predict GPP data under different scenarios in the future; the TWSA-DSI is obtained by combining the wet-bulb temperature and terrestrial water storage TWS data of each grid point in the study area output by the global climate model, and an emergence constraint model is constructed based on the GPP data under different future scenarios to reduce the uncertainty of GPP prediction under future drought stress.

[0126] Furthermore, the logic instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0127] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the processor executes the method for predicting the total primary productivity of ecosystems under drought stress provided by the above methods, the method comprising: obtaining carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data and a multi-source data set output by an earth system model in the study area; using the Clausius-Clapeyron thermodynamic equation to determine the saturated water vapor pressure deficit, relative humidity and wet-bulb temperature, extracting the drought characteristics in the study area based on the run theory and TWSA-DSI; based on the water and heat factors and the drought index TWSA-DSI, combined with the convergence A cross-mapping function is used to obtain the causal relationship between the water and heat stress factors, drought indicators and ecosystem GPP, and extract the key water and heat factors affecting GPP; based on the key water and heat factors affecting GPP, the key factors derived from the ERA5 reanalysis data and the GPP data inverted by MODIS are used to construct a long-short-term memory machine learning model, and the water and heat factors under different scenarios output by the Earth system model are combined to predict the GPP data under different scenarios in the future; the wet-bulb temperature and terrestrial water storage TWS data of each grid point in the study area output by the global climate model are combined to obtain the TWSA-DSI, and the GPP data under different future scenarios are used to construct an emergence constraint model to reduce the uncertainty of GPP prediction under future drought stress.

[0128] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0129] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for predicting the total primary productivity of an ecosystem under drought stress, characterized in that: include: Acquire multi-source data sets including carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data, and Earth system model output in the study area; The Clausius-Clapeyron thermodynamic equation was used to determine the saturated vapor pressure deficit, relative humidity, and wet-bulb temperature. The drought characteristics in the study area were extracted based on the run theory and the Terrestrial Water Storage Abnormal Drought Index (TWSA-DSI): Based on water and heat factors and drought index TWSA-DSI, combined with convergent cross-mapping function, the causal relationship between water and heat stress factors, drought indicators and ecosystem gross primary productivity (GPP) was obtained, and the key water and heat factors affecting GPP were extracted. Based on the key water and heat factors affecting GPP and the drought characteristics, a long-short-term memory machine learning model was constructed using the key factors from the fifth-generation global climate reanalysis data set ERA5 reanalysis data and the GPP data inverted by MODIS. Combined with the water and heat factors under different scenarios output by the Earth system model, GPP data under different future scenarios were predicted; The future drought index (TWSA-DSI) is derived by combining the wet-bulb temperature and terrestrial water storage (TWS) data of each grid point in the study area output by the global climate model. An emergent constraint model is constructed based on the GPP data under different future scenarios to reduce the uncertainty of GPP prediction under future drought stress. Based on the key hydrothermal factors affecting GPP and the drought characteristics, a long-short-term memory machine learning model was constructed using the key factors of the ERA5 reanalysis data and the GPP data inverted by MODIS. Combined with the hydrothermal factors under different scenarios output by the Earth System Model, GPP data under different future scenarios were predicted, including: Extract several driving factors from ERA5 reanalysis data, terrestrial water storage data retrieved from GRACE gravity satellites, and GPP data retrieved from MODIS, and use the top number of driving factors as key water and heat factors. Determine the key hydrothermal factors corresponding to the current moment and several moments before the current moment as input variables of the long-short-term memory machine learning model, and train to obtain a GPP prediction model; The hydrothermal factors, TWSA-DSI and drought characteristics under different scenarios output by the Earth System Model are used to drive the GPP prediction model and obtain GPP data under different future scenarios: in, represents the total primary productivity simulated by machine learning at time t, represents the input variable at time t, represents the input variable at time t-1, represents the input variable at time t-2, represents the input variable at time t-3; F represents the long short-term memory machine learning model; The future drought index (TWSA-DSI) is derived by combining the wet-bulb temperature and terrestrial water storage (TWS) data of each grid point in the study area output by the global climate model. An emergent constraint model is constructed based on the GPP data under different future scenarios to reduce the uncertainty of GPP prediction under future drought stress, including: Based on the temperature and relative humidity data of each global climate model historical period, combined with the wet-bulb temperature of each grid point in the study area, the linear regression method was used to obtain the trend of annual average wet-bulb temperature; The spatial sliding window method is used to obtain the multi-gridpoint annual average wet-bulb temperature change trend data under five global climate models. The first emergence constraint sub-model is constructed based on the multi-gridpoint annual average wet-bulb temperature change trend data, the GPP data under different future scenarios, the first constraint parameter and the second constraint parameter: Where: a1 and b1 represent the parameters of the emergence constraint model 1, Twet_Trend is the trend of wet-bulb temperature; The spatial sliding window method is used to obtain multi-grid TWSA-DSI under five global climate models. The second emergence constraint sub-model is constructed based on the multi-grid TWSA-DSI, the GPP data under different future scenarios, the third constraint parameter and the fourth constraint parameter: Where: a2 and b2 represent the parameters of the emergence constraint model 2, TWSADSI-Trend is the trend of drought index TWSA-DSI; Solving the first constraint parameter, the second constraint parameter, the third constraint parameter, and the fourth constraint parameter based on a least squares method, and obtaining the emergence constraint model from the first emergence constraint sub-model and the second emergence constraint sub-model; Where: represents the predicted total primary productivity under drought stress in the future period after correction; Twet_ Trend is the changing trend of the annual average wet-bulb temperature obtained at this grid point; TWSADSI-Trend is the trend of the annual drought index TWSA-DSI obtained at this grid point; the parameters a1, a2, b1 and b2 are derived from the parameters determined by the first and second emergence constraint sub-models; The emergence constraint model is driven under different future scenarios to obtain the GPP prediction results under drought stress under different scenarios.

2. The method for predicting total primary productivity of an ecosystem under drought stress according to claim 1, wherein: The multi-source datasets obtained in the study area include carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data, and Earth system model output, including: The FLUXNET 2015 dataset, which contains observations from several global flux towers, was quality controlled and bias corrected, and soil moisture was interpolated using the marginal distribution method. Converting the atmospheric reanalysis data and the MODIS satellite inversion data into daily scale data and monthly scale data with a preset spatial resolution, respectively; The three GRACE / GRACE-FO gravity satellite datasets in the gravity satellite inversion data were interpolated to a preset scale spatial grid to make the terrestrial water storage anomaly data consistent with the spatial resolution of the variables extracted by ERA5. The average of each product was taken within each time step to obtain the terrestrial water storage anomaly monthly dataset. Five global climate model datasets released by the sixth phase of the International Coupled Model Intercomparison Project (CMIP6) were used.

3. The method for predicting total primary productivity of an ecosystem under drought stress according to claim 1, wherein: The Clausius-Clapeyron thermodynamic equation was used to determine the saturated vapor pressure deficit, relative humidity, and wet-bulb temperature. Drought characteristics within the study area were extracted based on run theory and TWSA-DSI, including: The saturation water vapor pressure is determined using the Clausius-Clapeyron thermodynamic equation: in, and are the first and second integral constants, which are taken as 273.16 K and 611 Pa respectively; Lv is the latent heat of vaporization constant, which is taken as 2.5×106 J kg-1; Rv is the water vapor gas constant, which is taken as 461 J kg-1 K-1; is the saturation vapor pressure deficit, T is the air temperature; The temperature T 2m and dew point temperature T dew Substituting into the Clausius-Clapeyron thermodynamic equation, we can obtain the near-saturation vapor pressure deficit ; The temperature of each grid point ( T 2m ) and dew point temperature ( T dew ) are substituted as input variables into the Clausius-Clapeyron thermodynamic equation to calculate the relative humidity of each grid point ; Based on the air temperature and relative humidity data, calculate the wet-bulb temperature at each grid point: Where RH is the monthly average relative humidity, T 2m is the monthly mean temperature, T w is the derived wet-bulb temperature, atan is the inverse tangent function; Extraction of abnormal land water storage data : in, is the jth month of the i-th year TWSA data, and The jth month in the study period TWSA The mean and standard deviation of According to the run theory, the drought frequency, duration and intensity of the period corresponding to the TWSA-DSI being less than the preset drought threshold are determined.

4. The method for predicting total primary productivity of an ecosystem under drought stress according to claim 1, wherein: Based on water and heat factors and drought index TWSA-DSI, combined with the convergent cross-mapping function, the causal relationship between water and heat stress factors, drought indicators and ecosystem GPP is obtained, and the key water and heat factors affecting GPP are extracted, including: The hydrothermal factor and the drought index TWSA-DSI are used as variables X of the convergent cross-mapping function, and GPP is used as variable Y of the convergent cross-mapping function; Constructing a coupled logic differential equation of the convergent cross-mapping function, and determining the intrinsic growth rate of variable X, the intrinsic growth rate of variable Y, the mutual dynamic influence factor of the variables, and the time step of the coupled logic differential equation; The cross-mapping skill coefficients of the variable X and the variable Y are calculated to meet the cross-mapping skill coefficients corresponding to the preset convergence conditions, and the key hydrothermal factors affecting GPP are extracted.

5. A system for predicting the total primary productivity of an ecosystem under drought stress, based on the method for predicting the total primary productivity of an ecosystem under drought stress according to any one of claims 1 to 4, characterized in that: include: The acquisition module is used to obtain carbon flux observation data, atmospheric reanalysis data, MODIS satellite inversion data, gravity satellite inversion data and multi-source data sets output by the Earth system model in the study area; The first extraction module is used to determine the saturated water vapor pressure deficit, relative humidity and wet bulb temperature using the Clausius-Clapeyron thermodynamic equation, and extract the drought characteristics in the study area based on the run theory and TWSA-DSI: The second extraction module is used to obtain the causal relationship between water and heat stress factors, drought indicators and ecosystem gross primary productivity (GPP) based on water and heat factors and drought indicators TWSA-DSI, combined with the convergent cross-mapping function, and extract the key water and heat factors affecting GPP; A prediction module is used to construct a long-short-term memory machine learning model based on the key water and heat factors affecting GPP and the drought characteristics, using the key factors from the fifth-generation global climate reanalysis data set ERA5 reanalysis data and the GPP data inverted by MODIS, and combining the water and heat factors under different scenarios output by the Earth system model to predict GPP data under different future scenarios; The constraint module is used to combine the wet-bulb temperature and terrestrial water storage (TWS) data of each grid point in the study area output by the global climate model to obtain the future drought index (TWSA-DSI). The emergence constraint model is constructed based on the GPP data under different future scenarios to reduce the uncertainty of GPP prediction under future drought stress.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for predicting the total primary productivity of an ecosystem under drought stress as claimed in any one of claims 1 to 4 is implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting the total primary productivity of an ecosystem under drought stress as claimed in any one of claims 1 to 4 is implemented.

8. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for predicting the total primary productivity of an ecosystem under drought stress as claimed in any one of claims 1 to 4 is implemented.

Images