CN116522145B - Drought prediction method considering space-time constraint and vegetation condition - Google Patents

Drought prediction method considering space-time constraint and vegetation condition Download PDF

Info

Publication number
CN116522145B
CN116522145B CN202310526893.2A CN202310526893A CN116522145B CN 116522145 B CN116522145 B CN 116522145B CN 202310526893 A CN202310526893 A CN 202310526893A CN 116522145 B CN116522145 B CN 116522145B
Authority
CN
China
Prior art keywords
vegetation
drought
index
time
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202310526893.2A
Other languages
Chinese (zh)
Other versions
CN116522145A (en
Inventor
常胜
陈虹
谭深
马宗瀚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Aerospace Information Research Institute of CAS
China Aero Geophysical Survey and Remote Sensing Center for Natural Resources
Original Assignee
Aerospace Information Research Institute of CAS
China Aero Geophysical Survey and Remote Sensing Center for Natural Resources
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Aerospace Information Research Institute of CAS, China Aero Geophysical Survey and Remote Sensing Center for Natural Resources filed Critical Aerospace Information Research Institute of CAS
Priority to CN202310526893.2A priority Critical patent/CN116522145B/en
Publication of CN116522145A publication Critical patent/CN116522145A/en
Application granted granted Critical
Publication of CN116522145B publication Critical patent/CN116522145B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/27—Regression, e.g. linear or logistic regression
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Human Resources & Organizations (AREA)
  • Artificial Intelligence (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Agronomy & Crop Science (AREA)
  • Animal Husbandry (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Mining & Mineral Resources (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a drought prediction method considering space-time constraint and vegetation condition, which comprises the following five steps: 1) Acquiring and processing multisource remote sensing data required by model construction; 2) Constructing a drought monitoring model considering space-time constraint, and solving a high-spatial resolution drought index; 3) Constructing a vegetation key character prediction model, and estimating to obtain a prediction result of a low-resolution vegetation key character index; 4) Fusing the high-spatial-resolution drought index and the low-resolution vegetation key character prediction result to construct a vegetation drought prediction coupling model; 5) And obtaining a drought prediction result. The invention provides a general drought prediction method with mechanism constraint, and the drought is predicted by combining vegetation character index simulation results and utilizing long time sequence historical data, so that a high-resolution and high-precision prediction result is obtained.

Description

Drought prediction method considering space-time constraint and vegetation condition
Technical Field
The invention relates to a drought prediction method, in particular to a drought prediction method considering space-time constraint and vegetation condition.
Background
Drought is a natural disaster which has great influence on economy, society and environment, and has the characteristics of high occurrence frequency, long duration and large influence range. According to the statistics of the International disaster database (EM-DAT), the loss caused by global drought reaches 2210 hundred million dollars between 1960 and 2016. With the aggravation of global greenhouse effect, the drought occurrence frequency shows a remarkable rising trend, and the occurrence range, the influence area and the degree are also remarkably increased. Therefore, the drought is effectively monitored in time, the drought and the influence of the drought on regional vegetation and environment are accurately predicted, and the method has important significance in early warning of the drought, making disaster prevention and reduction, agriculture, ecology and water resource management policies.
Traditional vegetation drought monitoring and prediction mainly depend on information such as precipitation, soil moisture content and the like of ground observation points, are limited by the number and spatial distribution of stations, have limitations on data representativeness and timeliness, and cannot meet application requirements of large-area fine drought monitoring and prediction. The rapid development of remote sensing technology enables regional and even global scale near real-time drought monitoring. A great number of researches are carried out by means of multisource satellite remote sensing data, the degree of vegetation suffering from water stress is directly or indirectly diagnosed by monitoring factors such as vegetation greenness, environmental parameters (such as temperature, soil water content, precipitation and the like), and a series of developed drought remote sensing monitoring models and indexes are widely applied to researches of different spatial scales. As an extension of drought monitoring, the existing drought prediction method mainly adopts a mathematical statistics and machine learning/deep learning method, and predicts drought by establishing a numerical relation between a drought index and remote sensing and meteorological interpretation variables. However, current drought prediction research is mostly based on the variation trend and law of historical drought indexes and explanatory variables (such as meteorological data and soil moisture, etc.), and lacks support of vegetation physiological and ecological mechanisms, which results in great influence on accuracy of vegetation drought prediction. Therefore, there is a need to more deeply study the physiological and ecological response mechanism of vegetation to drought to further improve the accuracy and reliability of drought prediction.
In summary, the existing drought prediction method mainly has two technical defects: firstly, the existing model mostly assumes that the drought distribution mode is kept unchanged in time and space dimensions, and the space autocorrelation characteristic of drought is not fully considered, so that the accuracy of drought prediction is limited; secondly, the current prediction research on vegetation drought is focused on empirical statistical analysis of logarithmic values, the support of vegetation physiological and biochemical processes is lacking, the evolution rule of the drought cannot be accurately reflected, and a model established based on an empirical relationship changes along with the selection of regions and samples, so that the universality of a prediction method is limited.
Disclosure of Invention
In order to solve the defects of the technology, the invention provides a drought prediction method considering space-time constraint and vegetation condition.
In order to solve the technical problems, the invention adopts the following technical scheme: a drought prediction method taking space-time constraint and vegetation condition into consideration comprises the following five steps:
1) Acquiring and processing multisource remote sensing data required by model construction;
2) Constructing a drought monitoring model considering space-time constraint, and solving a high-spatial resolution drought index;
3) Constructing a vegetation key character prediction model, and estimating to obtain a prediction result of a low-resolution vegetation key character index;
4) Fusing the high-spatial-resolution drought index and the low-resolution vegetation key character prediction result to construct a vegetation drought prediction coupling model;
5) And obtaining a drought prediction result.
Further, considering that the construction of the space-time constraint drought monitoring model requires long-time series of medium-low resolution remote sensing data, including vegetation index NDVI of MODIS daily products 0 And surface temperature LST 0 Near infrared ρ NIR0 And short wave infrared reflectance data ρ SWIR0 ；
Calculating a near infrared band and a short wave infrared band to obtain a vegetation water index NDWI reflecting the water content of a vegetation canopy:
wherein ρ is NIR For near infrared reflectance products ρ SWIR Is a short wave infrared reflectivity product;
calculating a temperature condition index TCI, a vegetation state index VCI and a vegetation water index WCI respectively by using a surface temperature data set LST, a vegetation index data set NDVI and a vegetation water index NDWI, wherein the temperature condition index TCI, the vegetation state index VCI and the vegetation water index WCI are expressed as follows:
in LST max And LST min Respectively, historical maximum and minimum values of LST, NDVI max And NDVI min Respectively, historical maximum and minimum values of NDVI, NDWI max And NDWI min Historical maximum and minimum values for NDWI, respectively; LST (least squares) i 、NDVI i 、NDWI i LST, NDVI and NDWI corresponding to the ith pixel respectively;
the soil moisture index SWCI and precipitation index PCI were calculated as follows:
in SM max And SM min Respectively historical maximum value and minimum value of soil moisture in vegetation root zone, rain max And Rain min Respectively historical maximum and minimum values of precipitation; SM (SM) i 、Rain i SM and Rain corresponding to the ith pixel; using a long-time series vegetation root zone soil moisture and precipitation data set in analyzing the product ERA 5;
thus, the construction of the interpretation variable set required by the construction of the drought monitoring model is completed.
Further, the data required for constructing the vegetation key trait prediction model are as follows: and selecting proper monitoring stations according to the requirements, collecting environmental element data such as temperature, water vapor pressure difference value, radiation and the like of the monitoring stations, collecting FAPAR data sets of MODIS, and analyzing 2 m air temperature, soil water content and short wave radiation SDSR data in the ERA5 product.
Further, the construction process of the drought monitoring model taking space-time constraint into consideration is as follows: based on a drought remote sensing monitoring index model, introducing time and space dimensions, assuming regression coefficients as functions of space positions and observation moments, considering space-time non-stationarity, and constructing a drought monitoring model taking space-time constraints into consideration by combining space autocorrelation, time autocorrelation and geographic position information.
Further, the specific steps of constructing the drought monitoring model taking into consideration space-time constraints are as follows:
first, constructing an initial model: the effect of the spatial and temporal dimensions is considered, expressed as follows:
in the formula, VDI c Is the drought index at position (x, t); beta 0 (x, t) represents a constant coefficient, beta n (X, t) is an independent variable X n Regression coefficients at spatial position x and time t; x is X n (n=1, 2,..k) represents a set of explanatory variables including a temperature condition index TCI, a vegetation state index VCI, a vegetation water index WCI, a soil water index SWCI, and a precipitation index PCI, wherein k represents the number of explanatory variables; epsilon is the error term;
secondly, the model introduces space-time non-stationarity:
to consider spatiotemporal instability, we assume the regression coefficient β n (x, t) is a function of spatial position x and observation time t, fitted using a two-dimensional cubic B-spline function:
wherein beta is n,j (x) Is the regression coefficient, phi, at the spatial position x j (t)) is a B-spline function at time t, j representing the regression coefficient and the index in the spline function;
third, the model introduces spatial autocorrelation:
to take into account spatial autocorrelation, a spatial weight matrix ω (X) is introduced to assign a weight value to each spatial position X, a spatial autocorrelation model is used to calculate ω (X), and ω (X) is then combined with each independent variable X n Multiplying to obtain weighted independent variables:
ω(X n )＝ω(x)×X n (9)
fourth, the model introduces temporal autocorrelation and geographic location information:
introducing a space-time weight matrix v (X, t) into the model to represent the time autocorrelation and geographical position information, namely, simultaneously distributing a weight value to each space position X and observation time t, constructing the space-time weight matrix, and combining v (X, t) with each weighted independent variable omega (X n ) Multiplying to obtain a final weight independent variable:
z(X n ,t)＝ν(x,t)×ω(X n ) (10)
fifthly, constructing a drought monitoring model taking space-time constraint into consideration:
will respond to variable VDI c And a weight argument z (X n Substituting t) into the model to obtain a drought monitoring model taking space-time constraints into consideration as follows:
further, based on the constructed drought monitoring model taking space-time constraint into consideration, a least square method is used for solving the model to obtain regression coefficients of each independent variable on each spatial position and observation time, drought monitoring and prediction are carried out, and a final high-spatial resolution drought index value VDI is obtained c 。
Further, the construction process of the vegetation key character prediction model comprises the following steps: based on the ecological hydrologic optimality principle, the plant function and structural character adaptation mechanism to the environment under the condition of quantitative expression water stress is used for constructing a vegetation key character prediction model under the influence of drought stress.
Further, the specific steps of constructing the vegetation key character prediction model are as follows:
under drought conditions, the assimilation and fixation of carbon in the vegetation photosynthetic process is distributed to the leaves in proportion to participate in subsequent photosynthesis, and the biomass of the leaves can be described by the product of two structural property indexes, namely the leaf weight LMA and the leaf area index LAI, so that the leaf area index LAI of the vegetation structural property indexes can be expressed as:
in the sigma (GPP-R e ) Accumulating net carbon benefits for the current growing season, wherein GPP is total primary productivity, and estimating by a pervasive photosynthesis estimation model Pmodel; f (f) 1 For the ratio assigned to the leaves, a reference value of 0.3 is set for woody plants, and a reference value of 0.4 is set for herbaceous plants; the specific blade weight LMA is positively correlated with the variation trend of the blade age, and the median 100 is used as a reference value for arid and semiarid regions or estimated according to the blade age;
and (3) estimating and obtaining a predicted value of the leaf area index LAI of the low-resolution vegetation structural property index through a formula (12) as a parameter of drought coupling prediction in the next step.
Further, the vegetation drought prediction coupling model is constructed by the following steps:
preparation of high spatial resolution drought index VDI of long time series using drought monitoring model taking into account space-time constraints c Meanwhile, acquiring a vegetation key character leaf area index LAI of the same time sequence according to a vegetation key character prediction model; the vegetation drought prediction model constructed by the quantitative expression of the RF fitting device has the following specific expression:
NVDI c,q+1 ＝F(NVDI c,q ,LAI q+1 )+Bias (25)
wherein q and q+1 are corresponding phases, F is a nonlinear model constructed based on RF, and NVDI of q phase is utilized c,q And phase q+1 LAI q+1 Co-predicting; bias is a residual term driven by related environmental elements and is used for representing an unknown process of a vegetation prediction model and counteracting the influence caused by systematic deviation.
Further, VDI c The response to drought lags the LAI, requiring a prior lag time correction, i.e., for LAI and VDI c The historical data sequence of (2) is smoothed and filtered to obtain the phase of the periodic variation of the two, the phase difference is the delay time, and the VDI after the delay time correction c Recorded as NVDI c ；
The time correction comprises the following four steps:
firstly, smoothing and filtering; for LAI and VDI c Is smoothed and filtered to obtain smoothed LAI sequences SLAI and VDI c Sequence SVDI c ：
SLAI(q)＝Smoothing(LAI(q)) (20)
SVDIc(q)＝Smoothing(VDIc(q)) (21)
Wherein SLAI (q) is a smoothed LAI sequence; SVDIc (q) is smoothed VDI c A sequence; q represents a time phase; smoothing is a Smoothing function;
secondly, calculating a phase difference; calculating the periodic variation phase difference of SLAI and SVDIc by Fourier transform, etc., and recording asThe phase difference may be determined by calculating the phase difference of the harmonic components of the two signals:
wherein phase_difference is a Phase Difference function;
thirdly, calculating lag time; the lag time can be expressed as follows:
wherein T is VDI c Is a period of (2);for SLAI and SVDI c A periodically varying phase difference of (2);
fourth, lag time correction; with the calculation result, VDI can be obtained c Sequence lag τ time to NVDI c ：
NVDIc(q)＝VDIc(q-τ) (24)
Where τ is the lag time and NVDIc (q) is VDI corrected by the lag time c Sequence.
The invention discloses a drought prediction method taking space-time constraint and vegetation condition into consideration, which is a drought prediction method comprehensively considering space-time non-stationarity and vegetation environment adaptation mechanism, and comprehensively considering space-time non-stationarity and space autocorrelation of model relation to improve the space-time characteristic expression precision of drought. From the physiological ecology theory of vegetation, the adaptation rule of vegetation environment is considered, and a vegetation key character index simulation prediction model is constructed, so that powerful theoretical support is provided for improving the drought prediction capability of vegetation; and finally, combining the vegetation drought remote sensing monitoring index model and vegetation structural character simulation prediction model results, constructing a drought prediction model with an ecological theory basis, breaking through the bottlenecks of insufficient mechanical property, insufficient universality and unstable results of the traditional prediction model, and realizing high-precision drought prediction.
Drawings
FIG. 1 is a schematic diagram of the technical process of the present invention.
Detailed Description
The invention will be described in further detail with reference to the drawings and the detailed description.
The drought prediction method taking space-time constraint and vegetation condition into consideration is a drought prediction method taking space-time non-stationarity and vegetation environment adaptation mechanism into consideration, and firstly, the method starts from improving the existing drought remote sensing monitoring index model, comprehensively considers the space-time non-stationarity characteristics between the drought index and an explanatory variable and the space autocorrelation characteristics of the drought, obtains a drought monitoring model taking space-time constraint into consideration, and further improves the accuracy of drought monitoring; based on an ecological hydrologic optimality principle, quantitatively expressing an adaptation mechanism of plant functions and structural characters under the condition of water stress to the environment, and constructing a vegetation key character prediction model under the influence of drought stress; and fusing the high-spatial-resolution drought remote sensing monitoring index with the low-resolution vegetation key character prediction result to finally realize the prediction of the high-precision vegetation drought with universality.
The general technical flow chart of the invention is shown in fig. 1, and is mainly divided into the following five steps: 1) Acquiring and processing multi-source remote sensing data; 2) Constructing a drought monitoring model taking space-time constraint into consideration; 3) Constructing a vegetation key character prediction model; 4) Constructing a vegetation drought prediction coupling model; 5) And obtaining a drought prediction result.
For the five steps, the specific processing procedures are as follows:
1) Acquisition and processing of multisource remote sensing data
In order to meet the requirements of the spatial resolution and time scale of drought monitoring, long-time series of medium-low resolution remote sensing data are needed, including vegetation index NDVI of MODIS daily products 0 And surface temperature LST 0 Near infrared ρ NIR0 And short wave infrared reflectance data ρ SWIR0 (the spatial resolution is 1 km). By operating the near infrared band (the central wavelength is between 0.77um and 0.90 um) and the short wave infrared band (the central wavelength is between 1.55um and 1.75 um), the vegetation water index NDWI reflecting the water content of the vegetation canopy can be obtained:
wherein ρ is NIR For near infrared reflectance products ρ SWIR Is short wave infrared reflectivityAnd (5) a product.
Removing the influence of cloud by using an S-G smoothing method, and generating a daily vegetation index, surface temperature and vegetation water index sequence data set; then, according to the data quality mark and the limitation of available days in each ten days, carrying out average synthesis on the cloud-removed daily reflectivity data (near infrared and short wave infrared), vegetation indexes, surface temperature and vegetation moisture according to the ten days to obtain a data set of the ten-day reflectivity data, the ten-day vegetation indexes, the ten-day surface temperature and the ten-day vegetation moisture indexes; calculating the maximum value NDVI of each ten days max 、LST max And NDWI max And a ten-day-by-ten-day minimum value NDVI min 、LST min And NDWI min And the like. Calculating a temperature condition index TCI, a vegetation state index VCI and a vegetation water index WCI respectively by using a surface temperature data set LST, a vegetation index data set NDVI and a vegetation water index NDWI, wherein the temperature condition index TCI, the vegetation state index VCI and the vegetation water index WCI are expressed as follows:
in LST max And LST min Respectively, historical maximum and minimum values of LST, NDVI max And NDVI min Respectively, historical maximum and minimum values of NDVI, NDWI max And NDWI min Historical maximum and minimum values for NDWI, respectively; LST (least squares) i 、NDVI i 、NDWI i LST, NDVI and NDWI corresponding to the ith pixel respectively;
the soil moisture index SWCI and precipitation index PCI were calculated as follows:
in SM max And SM min Respectively historical maximum value and minimum value of soil moisture in vegetation root zone, rain max And Rain min Respectively historical maximum and minimum values of precipitation; SM (SM) i 、Rain i SM and Rain corresponding to the ith pixel; the long-time series vegetation root zone soil moisture and precipitation data set used in analyzing product ERA5 is used herein.
Thus, the construction of the interpretation variable set required by the construction of the drought monitoring model is completed.
Furthermore, proper monitoring stations are selected according to the needs, and environmental element data such as temperature, water vapor pressure difference value, radiation and the like of the monitoring stations are collected, and the data can be obtained from meteorological stations or national meteorological bureau. In addition, FAPAR data sets of MODIS are also collected, and 2 m air temperature, soil moisture content and short wave radiation SDSR data in the product ERA5 are analyzed for construction of vegetation key trait predictive model under water stress conditions.
2) Construction of drought monitoring model taking space-time constraint into consideration
The method comprises the steps of simultaneously introducing time and space dimension changes into a drought remote sensing monitoring index model, assuming regression coefficients to be functions of space positions and observation moments, simultaneously taking account of space-time non-stationarity, and constructing a space-time geographic weighted regression drought monitoring model by combining space autocorrelation, time autocorrelation and geographic position information, namely the drought monitoring model taking space-time constraints into consideration, wherein the method comprises the following specific steps of:
first, constructing an initial model: taking into account the influence of the spatial and temporal dimensions, this can be expressed as follows:
in the formula, VDI c Is the drought index at position (x, t); beta 0 (x, t) represents a constant coefficient, beta n (X, t) is an independent variable X n Regression coefficients at spatial position x and time t; x is X n (n=1, 2,..k) represents a set of explanatory variables, where the explanatory variables include a temperature condition index TCI, a vegetation state index VCI, a vegetation water index WCI, a soil water index SWCI, and a precipitation index PCI, where k represents the number of explanatory variables; epsilon is the error term.
Secondly, the model introduces space-time non-stationarity:
to consider spatiotemporal instability, we assume the regression coefficient β n (x, t) is a function of the spatial position x and the observation time t; here a two-dimensional cubic B-spline function can be used to fit:
wherein beta is n,j (x) Is the regression coefficient, phi, at the spatial position x j (t) is a B-spline function at time t, j represents the regression coefficient and the index in the spline function.
Third, the model introduces spatial autocorrelation:
to take into account the spatial autocorrelation, a spatial weight matrix ω (X) is introduced to assign a weight value to each spatial position X, where ω (X) can be calculated using some spatial autocorrelation model such as a spherical autocorrelation model, an exponential autocorrelation model, or the like, and then ω (X) is combined with each independent variable X n Multiplying to obtain weighted independent variables:
ω(X n )＝ω(x)×X n (9)
fourth, the model introduces temporal autocorrelation and geographic location information:
subsequently, a space-time weight matrix v (X, t) is introduced into the model to represent the time autocorrelation and the geographic position information, i.e. each space position X and the observation time t are simultaneously assigned with a weight value, and after the space-time weight matrix is constructed, v (X, t) is combined with each weighted argument omega (X) n ) The multiplication is performed and,obtaining final weight independent variable:
z(X n ,t)＝ν(x,t)×ω(X n ) (10)
fifthly, constructing a drought monitoring model taking space-time constraint into consideration:
finally, the response variable VDI is to be used c And a weight argument z (X n Substituting t) into the model to obtain a drought monitoring model taking space-time constraints into consideration as follows:
the regression coefficient of each independent variable at each space position and observation time can be obtained by using a least square method solution model according to the formula (11), so that drought monitoring and prediction can be performed, and the final high-space resolution (1 km) drought index value VDI can be obtained c 。
3) Construction of vegetation key character prediction model
Under drought conditions, the assimilation and fixation carbon in the vegetation photosynthetic process is distributed to the leaves to participate in subsequent photosynthesis according to a certain proportion on the basis of meeting the respiratory consumption, and the Leaf biomass can be described by the product of two structural property indexes, namely Leaf Area Index (LAI) and Leaf Area Index (Leaf Mass per Area, LMA), so that the Leaf Area Index LAI of the vegetation structural state Index can be expressed as:
in the sigma (GPP-R e ) Accumulating net carbon benefits for the current growing season, wherein GPP is total primary productivity, and the net carbon benefits can be estimated by a general photosynthesis estimation model Pmodel, namely, calculated by a formula (13); f (f) 1 For the ratio distributed to the leaves, a reference value of 0.3 can be set for woody plants and a reference value of 0.4 can be set for herbaceous plants; the variation trend of the specific blade weight LMA and the blade age is positively correlated, and the median 100 can be used as a reference value for arid and semiarid regions or according to the bladeAnd (5) estimating the age of the tablet.
GPP is total primary productivity, estimated by the generalized photosynthesis estimation model Pmodel as follows:
in the method, in the process of the invention,is intrinsic optical quantum efficiency, and can be expressed as temperature T a A function of(. Degree.C.) calculated from equation (14); i obs The photosynthetic effective radiation intercepted by the canopy can be calculated by a formula (15); c * A fixed constant value of 0.41; the m term represents CO 2 Limitation to photosynthesis can be calculated from equation (16); beta (theta) is a water stress term driven by the soil water content theta and can be calculated by the formula (17).
Intrinsic optical quantum efficiencyCan be calculated by the following formula:
wherein T is a For air temperature data, a 2 meter surface temperature product provided by a re-analytical weather product may be used;
photosynthetic active radiation I intercepted by canopy obs Can be calculated by the following formula:
I obs ＝fAPAR×PAR (15)
wherein fAPAR represents canopy interception capability, and a FAPAR dataset of MODIS or a higher-resolution optical remote sensing data source can be used; for photosynthetically active radiation, the SDSR data in the analysis product ERA5 can be used here to estimate PAR;
the m term can be calculated by the following formula:
wherein, c i Set to a constant value of 0.41, τ * The carbon dioxide compensation point can be obtained by the following formula:
in the method, in the process of the invention,e has a value of 4.22Pa, < >>Take the value of 27056J mol –1 R is 8.314J mol –1 K –1 ，T a For air temperature data, a 2 meter surface temperature product provided by the analysis product ERA5 may be used;
beta (θ) can be calculated by the following formula:
wherein θ represents the water content of the soil, and the soil water content product provided by the analysis product ERA5 can be used * The value is 0.6.
In summary, the low-resolution (10 km) vegetation structural property index LAI predicted value which can be estimated by the formula (12) is taken as a parameter of drought coupling prediction in the next step.
4) Construction of vegetation drought prediction coupling model
With Random Forest (RF) fitter as core, fusing the high-resolution remote sensing drought monitoring index VDI c (1 km) and predicted low resolution (10 km) vegetation leaf area index LAI, introducing a high-resolution (1 km) remote sensing drought monitoring result VDI into a vegetation state prediction trend c And obtaining a high-resolution (1 km) vegetation drought prediction result.
Firstly, preparing a long-time series historical drought remote sensing index VDI by using a drought monitoring model which is constructed in the second step and takes space-time constraints into consideration c And simultaneously acquiring a vegetation key character index LAI of the same time sequence according to the vegetation key character prediction model constructed in the third step. Due to VDI c The response to drought lags the LAI, requiring a prior lag time correction, i.e., for LAI and VDI c The historical data sequence of (2) is subjected to smoothing and filtering to obtain the phase of the periodic variation of the two, wherein the phase difference is the lag time, and the VDI corrected by the lag time c Recorded as NVDI c . The time correction comprises the following four steps:
firstly, smoothing and filtering; for LAI and VDI c Is smoothed and filtered to obtain smoothed LAI sequences SLAI and VDI c Sequence SVDI c ：
SLAI(q)＝Smoothing(LAI(q)) (20)
SVDIc(q)＝Smoothing(VDIc(q)) (21)
Wherein SLAI (q) is a smoothed LAI sequence; SVDIc (q) is smoothed VDI c A sequence; q represents a time phase; smoothing is a Smoothing function;
secondly, calculating a phase difference; calculating the periodic variation phase difference of SLAI and SVDIc by Fourier transform, etc., and recording asThe phase difference may be determined by calculating the phase difference of the harmonic components of the two signals:
wherein phase_difference is a Phase Difference function;
thirdly, calculating lag time; the lag time can be expressed as follows:
wherein T is the period of VDic;for SLAI and SVDI c A periodically varying phase difference of (2);
fourth, lag time correction; combining the above calculation results, the VDic sequence can be delayed by τ to obtain NVDIc:
NVDIc(q)＝VDIc(q-τ) (24)
where τ is the lag time and NVDIc (q) is VDI corrected by the lag time c A sequence;
then, the quantitative relation between NVDIc and LAI can be realized through an RF fitting device, and the vegetation drought prediction model has the following specific expression:
NVDI c,q+1 ＝F(NVDI c,q ,LAI q+1 )+Bias (25)
wherein q and q+1 are corresponding phases, F is a nonlinear model constructed based on RF, and NVDI of q phase is utilized c,q And phase q+1 LAI q+1 Co-predicting; bias is a residual term driven by relevant environmental factors (such as temperature, water vapor pressure difference value, radiation and the like) and is used for representing unknown processes of vegetation prediction models and counteracting influences brought by systematic deviation.
Finally, NVDI to be matched c,q And LAI q+1 The two historical data sets are randomly divided according to 60% and 40%, wherein 60% of the data are used for training a machine learning model, the remaining 40% of the data are used for testing whether the accuracy of a drought prediction model meets the requirement or not, and finally a vegetation drought prediction model F is obtained. The super parameters of the random forest can be optimized and controlled in a three-fold cross verification and grid traversal mode; the precision of the random forest algorithm is closely related to two super parameters: MAX-FEATURES and N-parameters, so that the model accuracy is guaranteed by performing traversal and parameter adjustment on the two super-parameters. The parameter-adjusting result is to adopt three-fold intersectionThe training set is verified, the average RMSE of three verification sets in cross verification is used as an evaluation index of super-parameter tuning, and the smaller the RMSE value is, the better the performance of the corresponding super-parameter combination model is; thereby obtaining the value of the optimal super parameter.
5) Acquisition of drought prediction results
After the trained prediction model (formula (25)) is obtained, the weather prediction data and the historical drought monitoring data set at q time can be used for drought prediction, and the prediction result is the drought prediction value at q+1 time.
In summary, the invention introduces a space-time kernel function and a space autocorrelation term to improve the model, so that the model relation accords with the change rule in time and space dimensions, and the improvement of drought monitoring precision is realized; the invention provides a general drought prediction method with mechanism constraint, and the drought is predicted by combining vegetation character index simulation results and utilizing long time sequence historical data, so that a high-resolution and high-precision prediction result is obtained.
The above embodiments are not intended to limit the present invention, and the present invention is not limited to the above examples, but is also intended to be limited to the following claims.

Claims (4)

1. A drought prediction method taking space-time constraint and vegetation condition into consideration is characterized in that: the drought prediction method comprises the following five steps:
1) Acquiring and processing multisource remote sensing data required by model construction;
2) Constructing a drought monitoring model considering space-time constraint, and solving a high-spatial resolution drought index;
3) Constructing a vegetation key character prediction model, and estimating to obtain a prediction result of a low-resolution vegetation key character index;
4) Fusing the high-spatial-resolution drought index and the low-resolution vegetation key character prediction result to construct a vegetation drought prediction coupling model;
5) Acquiring drought prediction results;
the construction process of the drought monitoring model considering space-time constraint comprises the following steps: based on a drought remote sensing monitoring index model, introducing time and space dimensions, assuming regression coefficients as functions of space positions and observation moments, considering space-time non-stationarity, and constructing a drought monitoring model taking space-time constraints into consideration by combining space autocorrelation, time autocorrelation and geographic position information, wherein the method comprises the following specific steps:
first, constructing an initial model: the effect of the spatial and temporal dimensions is considered, expressed as follows:
in the formula, VDI c Is the drought index at position (x, t); beta 0 (x, t) represents a constant coefficient, beta n (X, t) is an independent variable X n Regression coefficients at spatial position x and time t; x is X n (n=1, 2,..k) represents a set of explanatory variables including a temperature condition index TCI, a vegetation state index VCI, a vegetation water index WCI, a soil water index SWCI, and a precipitation index PCI, wherein k represents the number of explanatory variables; epsilon is the error term;
secondly, the model introduces space-time non-stationarity:
to consider spatiotemporal instability, we assume the regression coefficient β n (x, t) is a function of spatial position x and observation time t, fitted using a two-dimensional cubic B-spline function:
wherein beta is n,j (x) Is the regression coefficient, phi, at the spatial position x j (t)) is a B-spline function at time t, j representing the regression coefficient and the index in the spline function;
third, the model introduces spatial autocorrelation:
to take into account spatial autocorrelation, a spatial weight is introducedA weight matrix ω (X) is used to assign a weight value to each spatial position X, a spatial autocorrelation model is used to calculate ω (X), and ω (X) is then combined with each independent variable X n Multiplying to obtain weighted independent variables:
ω(X n )＝ω(x)×X n (9)
fourth, the model introduces temporal autocorrelation and geographic location information:
introducing a space-time weight matrix v (X, t) into the model to represent the time autocorrelation and geographical position information, namely, simultaneously distributing a weight value to each space position X and observation time t, constructing the space-time weight matrix, and combining v (X, t) with each weighted independent variable omega (X n ) Multiplying to obtain a final weight independent variable:
z(X n ,t)＝ν(x,t)×ω(X n ) (10)
fifthly, constructing a drought monitoring model taking space-time constraint into consideration:
will respond to variable VDI c And a weight argument z (X n Substituting t) into the model to obtain a drought monitoring model taking space-time constraints into consideration as follows:
based on the constructed drought monitoring model taking space-time constraint into consideration, a least square method solving model is used to obtain regression coefficients of each independent variable on each spatial position and observation time, drought monitoring and prediction are carried out, and a final high-spatial resolution drought index value VDI is obtained c ；
The construction process of the vegetation key character prediction model comprises the following steps: based on the ecological hydrologic optimality principle, the plant function and structural character adaptation mechanism to the environment under the condition of quantitative expression water stress is used for constructing a vegetation key character prediction model under the influence of drought stress, and the specific steps are as follows:
under drought conditions, the assimilation and fixation of carbon in the vegetation photosynthetic process is distributed to the leaves in proportion to participate in subsequent photosynthesis, and the biomass of the leaves can be described by the product of two structural property indexes, namely the leaf weight LMA and the leaf area index LAI, so that the leaf area index LAI of the vegetation structural property indexes can be expressed as:
in the sigma (GPP-R e ) Accumulating net carbon benefits for the current growing season, wherein GPP is total primary productivity, and estimating by a pervasive photosynthesis estimation model Pmodel; f (f) 1 For the ratio assigned to the leaves, a reference value of 0.3 is set for woody plants, and a reference value of 0.4 is set for herbaceous plants; the specific blade weight LMA is positively correlated with the variation trend of the blade age, and the median 100 is used as a reference value for arid and semiarid regions or estimated according to the blade age;
estimating and obtaining a predicted value of a leaf area index LAI of the low-resolution vegetation structural property index through a formula (12) as a parameter of drought coupling prediction in the next step;
the vegetation drought prediction coupling model is constructed by the following steps:
preparation of high spatial resolution drought index VDI of long time series using drought monitoring model taking into account space-time constraints c Meanwhile, acquiring a vegetation key character leaf area index LAI of the same time sequence according to a vegetation key character prediction model; the vegetation drought prediction model constructed by the quantitative expression of the RF fitting device has the following specific expression:
NVDI c,q+1 ＝F(NVDI c,q ,LAI q+1 )+Bias (25)
wherein q and q+1 are corresponding phases, F is a nonlinear model constructed based on RF, and NVDI of q phase is utilized c,q And phase q+1 LAI q+1 Co-predicting; bias is a residual term driven by related environmental elements and is used for representing an unknown process of a vegetation prediction model and counteracting the influence caused by systematic deviation.
2. The drought prediction method considering space-time constraints and vegetation conditions as claimed in claim 1, whichIs characterized in that: considering that the construction of a space-time constraint drought monitoring model requires long-time sequence of medium-low resolution remote sensing data, including a vegetation index NDVI of a MODIS daily product 0 And surface temperature LST 0 Near infrared ρ NIR0 And short wave infrared reflectance data ρ SWIR0 ；
Calculating a near infrared band and a short wave infrared band to obtain a vegetation water index NDWI reflecting the water content of a vegetation canopy:
wherein ρ is NIR For near infrared reflectance products ρ SWIR Is a short wave infrared reflectivity product;
calculating a temperature condition index TCI, a vegetation state index VCI and a vegetation water index WCI respectively by using a surface temperature data set LST, a vegetation index data set NDVI and a vegetation water index NDWI, wherein the temperature condition index TCI, the vegetation state index VCI and the vegetation water index WCI are expressed as follows:
in LST max And LST min Respectively, historical maximum and minimum values of LST, NDVI max And NDVI min Respectively, historical maximum and minimum values of NDVI, NDWI max And NDWI min Historical maximum and minimum values for NDWI, respectively; LST (least squares) i 、NDVI i 、NDWI i LST, NDVI and NDWI corresponding to the ith pixel respectively;
the soil moisture index SWCI and precipitation index PCI were calculated as follows:
in SM max And SM min Respectively historical maximum value and minimum value of soil moisture in vegetation root zone, rain max And Rain min Respectively historical maximum and minimum values of precipitation; SM (SM) i 、Rain i SM and Rain corresponding to the ith pixel; using a long-time series vegetation root zone soil moisture and precipitation data set in analyzing the product ERA 5;
thus, the construction of the interpretation variable set required by the construction of the drought monitoring model is completed.
3. The drought prediction method of claim 2 wherein the space-time constraints and vegetation conditions are considered: the data required for constructing the vegetation key character prediction model are as follows: and selecting proper monitoring stations according to the requirements, collecting environmental element data such as temperature, water vapor pressure difference value, radiation and the like of the monitoring stations, collecting FAPAR data sets of MODIS, and analyzing 2 m air temperature, soil water content and short wave radiation SDSR data in the ERA5 product.
4. The drought prediction method considering space-time constraints and vegetation conditions as claimed in claim 1, wherein: VDI (VDI) c The response to drought lags the LAI, requiring a prior lag time correction, i.e., for LAI and VDI c The historical data sequence of (2) is smoothed and filtered to obtain the phase of the periodic variation of the two, the phase difference is the delay time, and the VDI after the delay time correction c Recorded as NVDI c ；
The time correction comprises the following four steps:
firstly, smoothing and filtering; for LAI and VDI c Is smoothed and filtered to obtain smoothed LAI sequences SLAI and VDI c Sequence SVDI c ：
SLAI(q)＝Smoothing(LAI(q)) (20)
SVDIc(q)＝Smoothing(VDIc(q)) (21)
Wherein SLAI (q) is a smoothed LAI sequence; SVDIc (q) is smoothed VDI c A sequence; q represents a time phase; smoothing is a Smoothing function;
secondly, calculating a phase difference; calculating the periodic variation phase difference of SLAI and SVDIc by Fourier transform, etc., and recording asThe phase difference may be determined by calculating the phase difference of the harmonic components of the two signals:
wherein phase_difference is a Phase Difference function;
thirdly, calculating lag time; the lag time can be expressed as follows:
wherein T is VDI c Is a period of (2);for SLAI and SVDI c A periodically varying phase difference of (2);
fourth, lag time correction; with the calculation result, VDI can be obtained c Sequence lag τ time to NVDI c ：
NVDIc(q)＝VDIc(q-τ) (24)
Where τ is the lag time and NVDIc (q) is VDI corrected by the lag time c Sequence.
CN202310526893.2A 2023-05-11 2023-05-11 Drought prediction method considering space-time constraint and vegetation condition Active CN116522145B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310526893.2A CN116522145B (en) 2023-05-11 2023-05-11 Drought prediction method considering space-time constraint and vegetation condition

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310526893.2A CN116522145B (en) 2023-05-11 2023-05-11 Drought prediction method considering space-time constraint and vegetation condition

Publications (2)

Publication Number Publication Date
CN116522145A CN116522145A (en) 2023-08-01
CN116522145B true CN116522145B (en) 2024-02-13

Family

ID=87402794

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310526893.2A Active CN116522145B (en) 2023-05-11 2023-05-11 Drought prediction method considering space-time constraint and vegetation condition

Country Status (1)

Country Link
CN (1) CN116522145B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117350082B (en) * 2023-12-04 2024-03-22 南京大学 Calculation method for net ecological system productivity

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107782701A (en) * 2017-09-20 2018-03-09 北京师范大学 A kind of agricultural arid monitoring method of multi- source Remote Sensing Data data
CN112858632A (en) * 2021-01-14 2021-05-28 中国科学院空天信息创新研究院 Grassland drought monitoring method comprehensively considering temperature and water stress
CN115062997A (en) * 2022-06-27 2022-09-16 西南石油大学 Agricultural drought monitoring method based on remote sensing technology and machine learning and suitable for Tibet region

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4050334A1 (en) * 2021-02-26 2022-08-31 Tata Consultancy Services Limited System and method for root zone soil moisture estimation for vegetation cover using remote sensing

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107782701A (en) * 2017-09-20 2018-03-09 北京师范大学 A kind of agricultural arid monitoring method of multi- source Remote Sensing Data data
CN112858632A (en) * 2021-01-14 2021-05-28 中国科学院空天信息创新研究院 Grassland drought monitoring method comprehensively considering temperature and water stress
CN115062997A (en) * 2022-06-27 2022-09-16 西南石油大学 Agricultural drought monitoring method based on remote sensing technology and machine learning and suitable for Tibet region

Also Published As

Publication number Publication date
CN116522145A (en) 2023-08-01

Similar Documents

Publication Publication Date Title
CN111727443B (en) Soil available nutrient inversion method based on deep neural network
Huang et al. Assimilation of remote sensing into crop growth models: Current status and perspectives
CN110751094B (en) Crop yield estimation method based on GEE comprehensive remote sensing image and deep learning method
Wei et al. Comparative evaluation of drought indices for monitoring drought based on remote sensing data
Manivasagam et al. Practices for upscaling crop simulation models from field scale to large regions
CN109711102B (en) Method for rapidly evaluating crop disaster loss
CN105912836B (en) A kind of watershed hydrologic cycle analogy method of pure remotely-sensed data driving
Choudhary et al. Agriculture phenology monitoring using NDVI time series based on remote sensing satellites: a case study of Guangdong, China
CN103345707A (en) Crop maturation stage remote sensing prediction method based on multi-source remote sensing data
CN110633841B (en) Provincial range plot scale data assimilation yield prediction method based on set sampling
CN116522145B (en) Drought prediction method considering space-time constraint and vegetation condition
CN107423850A (en) Region corn maturity period Forecasting Methodology based on time series LAI curve integral areas
Hu et al. Detecting regional GPP variations with statistically downscaled solar-induced chlorophyll fluorescence (SIF) based on GOME-2 and MODIS data
CN115375036A (en) Crop maturity prediction method based on fusion of remote sensing and light energy utilization rate model and weather
Liu et al. Winter wheat yield estimation based on assimilated Sentinel-2 images with the CERES-Wheat model
Cui et al. Estimating maize yield in the black soil region of Northeast China using land surface data assimilation: integrating a crop model and remote sensing
CN114120137A (en) Wetland element space-time evolution monitoring method based on time sequence main remote sensing image
Ji et al. Integrating the temperature vegetation dryness index and meteorology parameters to dynamically predict crop yield with fixed date intervals using an integral regression model
Zhang et al. Enhanced soil moisture improves vegetation growth in an arid grassland of Inner Mongolia Autonomous Region, China
CN116151454A (en) Method and system for predicting yield of short-forest linalool essential oil by multispectral unmanned aerial vehicle
CN109359862A (en) A kind of real-time yield estimation method of cereal crops and system
CN112949182B (en) Method for calculating regional scale maximum pore conductance of coupling surface soil water content
CN115661601A (en) Agricultural soil water stress discrimination and monitoring method based on CEEMDAN
Huang et al. Markov chain monte carlo and four-dimensional variational approach based winter wheat yield estimation
Chen et al. Cotton growth monitoring and yield estimation based on assimilation of remote sensing data and crop growth model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant