CN-122020393-A - Parameter calibration method for evapotranspiration model of paddy field and reed wetland

CN122020393ACN 122020393 ACN122020393 ACN 122020393ACN-122020393-A

Abstract

The invention discloses a parameter calibration method for an evapotranspiration model of paddy fields and reed wetlands, and belongs to the technical field of ecological hydrologic simulation. The method comprises the steps of S1, collecting multi-source data of a paddy field and reed wetland observation stations and processing the collected data, S2, carrying out sensitivity analysis on parameters of a PML model and a PT-JPL model, S3, dividing a growth stage according to different growth periods of the paddy field and the reed, setting a dynamic parameter optimization range, grading and calibrating key parameters of the PML model and the PT-JPL model based on measured evapotranspiration data and a Bayesian optimization algorithm for years, S4, substituting the calibrated optimal parameter combination into the corresponding PML model or the PT-JPL model to simulate wetland evapotranspiration, S5, verifying a simulation effect of the model through multiple reference indexes, and screening the optimal model and the parameter combination. By adopting the method, parameters can be accurately calibrated, the simulation accuracy and suitability of the wetland evapotranspiration are improved, and the ecological hydrologic research is supported.

Inventors

YU WENYING
ZHAO KEXIN
YAN GUOFENG
CAI FU
ZHANG YONGSHENG
ZHOU LI
JIA QINGYU
ZHAO XIANLI
WANG XIAOYING
TIAN YUNPENG

Assignees

中国气象局沈阳大气环境研究所

Dates

Publication Date: 20260512
Application Date: 20260209

Claims (9)

1. The parameter calibration method of the evapotranspiration model for the paddy field and the reed wetland is characterized by comprising the following steps of: s1, collecting multi-source data of paddy fields and reed wetland observation stations, including meteorological data, flux data and leaf area index data, processing the collected data, and dividing the collected data into two groups of calibration data and verification data according to time periods; S2, combining the growth characteristics of rice and reed, performing sensitivity analysis on each parameter of a PML model and a PT-JPL model, determining sensitive parameters, and setting the empirical value of non-sensitive parameters; s3, dividing the growth stages according to different growth periods of rice and reed, setting a dynamic parameter optimization range by combining with wetland characteristics, and grading key parameters of a PML model and a PT-JPL model based on measured evapotranspiration data of years and a Bayesian optimization algorithm; S4, selecting other years or time periods which do not participate in the calibration, substituting the calibrated optimal parameter combination into a corresponding PML model or PT-JPL model respectively, and simulating wetland evaporation and emission by combining processed data and model inherent physical calculation logic; S5, based on the actual measurement evaporation data which are not involved in the calibration, the simulation effect of the model is verified through multiple reference indexes, and the optimal model and parameter combination of the adaptive target wetland are screened.
2. The method for calibrating parameters of an evapotranspiration model for paddy fields and reed wetlands according to claim 1, wherein S1 specifically comprises: S11, acquiring leaf area index remote sensing data based on MOD15A2H products, comparing the leaf area index remote sensing data of the paddy field and the reed wetland with the field measured leaf area index data after extracting the leaf area index remote sensing data, and optimizing the leaf area index data by a linear interpolation method to ensure consistency of the leaf area index remote sensing data and the field measured leaf area index data; S12, eliminating extreme abnormal values in meteorological elements and actually measured evaporative data based on a 3 sigma rule, manually deleting error data in a super physical reasonable range caused by instrument faults and invalid data with negative values of available energy absorbed by the earth surface; S13, uniformly converting all the processed data into units required by calculation of a PML model and a PT-JPL model, aligning the data in a time dimension by taking the date as a key word, and ensuring that meteorological data, actually measured evapotranspiration data and leaf area index data of the same date are in one-to-one correspondence; s14, dividing the data processed in the S13 into two groups according to the observation period or year, wherein one group is the rated data of participation rating, and the other group is the verification data of participation verification.
3. The method for calibrating parameters of an evapotranspiration model for paddy fields and reed wetlands according to claim 2, wherein S2 specifically comprises: The PML model parameters comprise an extinction coefficient k A of available energy, an extinction coefficient k Q of short-wave radiation, a water vapor pressure difference D 50 when the air pore conductivity is half of the maximum value, a visible radiation flux Q 50 when the air pore conductivity is half of the maximum value, a soil water stress factor f and a maximum air pore conductivity g sx of a blade, wherein g sx is a sensitive parameter, k A 、k Q 、D 50 、Q 50 is a non-sensitive parameter, the non-sensitive parameter is set as an empirical value, f is set as1 based on the long-term supersaturation characteristic of soil moisture of a target wetland, the PT-JPL model parameters comprise a light energy utilization coefficient k 1 , a water stress coefficient k 2 , an evaporation regulating coefficient beta and a vegetation optimal growth temperature T opt , wherein k 1 、k 2 、β、T opt is the sensitive parameter, and the empirical value of a bush is selected as an initial value of the PT-JPL model parameter based on the similarity of the wetland and the bush environment.
4. A method for calibrating parameters of an evapotranspiration model for paddy fields and reed wetlands according to claim 3, wherein S3 specifically comprises: S31, dividing the growth stage of the natural wetland vegetation reed into a germination stage, a vegetative growth stage, a reproductive growth stage and a mature stage-a yellow initial stage; S32, setting a dynamic parameter optimization range based on rice and reed growth characteristics, wherein g sx is respectively set to be 0.55-0.85, k 2 is set to be 0.75-0.95, beta is set to be 1.43-2.2, and T opt is set to be 20-28 ℃ according to growth stages divided in S31; s33, performing Bayesian optimization grading calibration based on the data processed in the S1 and the dynamic parameter optimization range of vegetation in the S32, wherein the method specifically comprises the following steps: s331, constructing a parameter search space based on the parameter optimization range set in the S32 by taking g sx 、k 1 、k 2 、β、T opt as an optimization variable; S332, taking the error of the model simulation value and the site actual measurement value as an evaluation standard, and selecting a Root Mean Square Error (RMSE) or a negative Nash coefficient (NSE) to construct an objective function f (x), wherein the optimization target is minimization f (x); S333, selecting 5-10 groups of initial parameter combinations x i in a parameter search space by using a Latin hypercube sampling method, respectively operating a PML model and a PT-JPL model, and calculating an objective function value y i corresponding to the model to form an initial data set D= { x i ,y i }; S334, fitting a Gaussian process agent model based on the initial data set D, outputting a prediction mean and a prediction variance of any parameter point by the model, and determining a next parameter combination point x next to be evaluated by optimizing an acquisition function; S335, respectively running a PML model and a PT-JPL model at x next , acquiring a real objective function value y next , adding (x next ,y next ) into a data set D and updating a Gaussian process proxy model; S336, repeating S334-S335 until convergence conditions are met or preset maximum evaluation times are reached, and selecting a parameter combination which enables the objective function f (x) to be optimal as a calibration result of a corresponding PML model or PT-JPL model.
5. The method for calibrating parameters of an evapotranspiration model for paddy fields and reed wetlands according to claim 4, wherein S4 specifically comprises: Selecting a non-rated year or period, substituting the optimal parameter combination obtained in the step S336 into a PML model and a PT-JPL model respectively, and simultaneously combining the daily-scale meteorological data and the leaf area index data processed in the step S1 to calculate the daily-scale total evapotranspiration, wherein the PML model is based on Penman-Monteth principle, combines the canopy conductivity and the soil water stress factor, calculates the total evapotranspiration through an energy balance relation, and the PT-JPL model is based on Priestley-Taylor theory, and respectively simulates vegetation canopy transpiration, soil evaporation and interception evaporation and superposition to obtain a total evapotranspiration simulation value.
6. The method for calibrating parameters of an evapotranspiration model for paddy fields and reed wetlands according to claim 5, wherein S5 specifically comprises: S51, verifying a model simulation effect by calculating reference indexes based on the actually measured evaporative data which are not involved in the calibration, wherein the reference indexes comprise a determinable coefficient R 2 , a Nash coefficient NSE and a root mean square error RMSE, R 2 represents the correlation between a simulation value and an actually measured value, and the formula is as follows: ; Wherein, the For the number of samples, For the index of the sample, , Is the first The measured amount of evapotranspiration of each sample, For the average value of all the measured amounts of evapotranspiration, Is the first The simulated evapotranspiration of the individual samples, Average value of all simulated evapotranspiration; NSE represents the efficiency of model simulation, and its formula is as follows: ; RMSE represents the deviation between the analog and measured values, and its formula is as follows: ; S52, screening out a model with R 2 being more than or equal to A, NSE being more than or equal to B and RMSE being less than or equal to C and a corresponding parameter combination as a candidate set, judging the trend consistency of the time sequence through calculating the linear regression slope deviation rate of the simulation value and the actual measurement value, and selecting a model with the deviation rate being less than or equal to M and the trend fitting degree exceeding any fitting degree and the parameter combination as an optimal result, wherein A is a critical value of R 2 , B is a critical value of NSE, C is a critical value of RMSE, the unit is mm.d -1 , and M is a deviation rate critical value.
7. The method for calibrating parameters of the evapotranspiration model for paddy fields and reed wetlands according to claim 2, wherein the daily-scale meteorological driving data comprise near-ground air temperature, relative humidity, saturated water vapor pressure difference VPD, net radiation R n , soil heat flux G and air pressure P.
8. The method for calibrating parameters of an evapotranspiration model for paddy fields and reed wetlands according to claim 4, wherein the dynamic adjustment of the parameter optimization range is suitable for physiological activity differences of paddy fields and reed in different growth periods, and the evapotranspiration characteristics of the wetland vegetation in each growth stage are matched in response to photosynthesis demand enhancement and evapotranspiration cooling mechanisms after temperature rise.
9. The method for calibrating parameters of an evapotranspiration model for paddy fields and reed wetlands according to claim 4, wherein the converging condition of Bayesian optimization is that the objective function value is continuously iterated R times with the variation amplitude < q or the preset maximum evaluation times not less than S times, and the iteration is stopped when any condition is met, wherein q is the variation amplitude critical value of the objective function value.

Description

Parameter calibration method for evapotranspiration model of paddy field and reed wetland Technical Field The invention relates to the technical field of ecological hydrologic simulation, in particular to a parameter calibration method for an evapotranspiration model of paddy fields and reed wetlands. Background The evaporation is a key link of hydrologic cycle and energy balance, the evaporation of accurately simulating wetland vegetation has important significance for water resource management and ecological protection of a river basin, a Penman-Montetith-Leuning model (PML model for short), a Priestley-Taylor Jet Propulsion Laboratory model (PT-JPL model for short) and the like are common tools for regional evaporation estimation, but the method still has obvious limitations in application of the wetland vegetation, the prior art mainly uses common experience values or calibration results of other regions, the parameter ranges are not optimized according to specific characteristics of ' moisture saturation and vegetation uniformity ' of paddy fields and reed wetland, a parameter system is constructed based on non-wetland scenes such as a semiarid river basin, and the like, the parameter system is not matched with the actual environment of the wetland, high-moisture pressure difference, vegetation growth period stability and other characteristics, so that the adaptation of the model is insufficient, meanwhile, most researches adopt single fixed parameter calibration modes, physiological activity differences of different wetland vegetation types and different growth periods are not considered, parameter ' average ' errors ' are easy to generate, the starting processes of each stage are difficult to attach, and the traditional algorithm is not fully combined with the model for the single optimization algorithm, the overall optimization algorithm is not accurately limited by the application of the traditional algorithm, and the evaporation algorithm is difficult to be further limited. Disclosure of Invention The invention aims to provide a parameter calibration method of an evapotranspiration model for paddy fields and reed wetlands, which solves the technical problems. In order to achieve the above purpose, the invention provides a parameter calibration method of an evapotranspiration model for paddy fields and reed wetlands, which comprises the following steps: S1, collecting multi-source data of paddy fields and reed wetland observation stations, including meteorological data, flux data and leaf area index data, processing the collected data, and dividing the data into two groups of calibration data and verification data according to time periods; S2, combining the growth characteristics of rice and reed, performing sensitivity analysis on each parameter of a PML model and a PT-JPL model, determining sensitive parameters, and setting the empirical value of non-sensitive parameters; s3, dividing the growth stages according to different growth periods of rice and reed, setting a dynamic parameter optimization range by combining with wetland characteristics, and grading key parameters of a PML model and a PT-JPL model based on measured evapotranspiration data of years and a Bayesian optimization algorithm; S4, selecting other years or time periods which do not participate in the calibration, substituting the calibrated optimal parameter combination into a corresponding PML model or PT-JPL model respectively, and simulating wetland evaporation and emission by combining processed data and model inherent physical calculation logic; S5, based on the actual measurement evaporation data which are not involved in the calibration, the simulation effect of the model is verified through multiple reference indexes, and the optimal model and parameter combination of the adaptive target wetland are screened. Preferably, S1 specifically includes: S11, acquiring leaf area index remote sensing data based on MOD15A2H products, comparing the leaf area index remote sensing data of the paddy field and the reed wetland with the field measured leaf area index data after extracting the leaf area index remote sensing data, and optimizing the leaf area index data by a linear interpolation method to ensure consistency of the leaf area index remote sensing data and the field measured leaf area index data; S12, eliminating extreme abnormal values in meteorological elements and actually measured evaporative data based on a 3 sigma rule, manually deleting error data in a super physical reasonable range caused by instrument faults and invalid data with negative values of available energy absorbed by the earth surface; S13, uniformly converting all the processed data into units required by calculation of a PML model and a PT-JPL model, aligning the data in a time dimension by taking the date as a key word, and ensuring that meteorological data, actually measured evapotranspiration data and leaf area index data of the same date are in one-t