A method for determining evapotranspiration response bias based on local forest cover change threshold

Abstract

The application discloses a kind of evapotranspiration response deviation determination method based on local forest coverage change threshold, it is related to the technical field of ecological hydrology remote sensing data processing.Data of evapotranspiration in multiple change periods, forest coverage data and background data are obtained, and original evapotranspiration change and forest coverage change are determined.Reference base spatial units with similar background conditions are screened, and background evapotranspiration change is estimated and deducted.A linear response model and a continuous segmented response model are established within a local analysis window, and a stable forest coverage change threshold is determined through threshold identification and resampling stability testing.Evapotranspiration response deviation is determined according to the difference between the estimated values of the two models for the same forest coverage change to be evaluated.The application can solve the problem of linear method being difficult to depict local threshold response, and realize quantitative determination of evapotranspiration response deviation.

Description

Technical Field

[0001] This invention relates to the field of eco-hydrological remote sensing data processing technology, and in particular to a method for determining evapotranspiration response deviation based on a local forest cover change threshold. Background Technology

[0002] Forest cover change is a significant land surface change affecting terrestrial water cycles, energy cycles, and land-atmosphere exchange processes. Forests influence evapotranspiration through processes such as vegetation transpiration, canopy interception, root water uptake, changes in surface resistance, and near-surface turbulent exchange. As a crucial flux for transporting moisture and latent heat energy from the land to the atmosphere, evapotranspiration not only impacts regional water resource balance but also near-surface humidity, boundary layer development, precipitation recycling, and land-atmosphere feedback processes. Therefore, in assessments of the impacts of forest restoration, afforestation, forest degradation, logging, and land use change, it is typically necessary to estimate evapotranspiration changes or evapotranspiration responses based on changes in forest cover.

[0003] In existing technologies, linear regression, average sensitivity coefficients, or empirical conversion coefficients are commonly used to directly convert forest cover change into evapotranspiration response. These methods typically implicitly assume that the marginal response of evapotranspiration to forest cover change remains constant across different levels of change. However, the impact of forest cover change on evapotranspiration is constrained by factors such as precipitation, temperature, radiation, drought severity, topography, forest type, and regional hydrothermal background. When forest cover change is within a certain range, the evapotranspiration response may change approximately linearly with forest cover change; however, when forest cover change reaches a certain threshold, the slope of the evapotranspiration response may change due to limiting factors such as water supply, canopy structure, soil moisture, canopy conductance, or turbulent exchange capacity. Therefore, at a local spatial scale, there may be a forest cover change threshold between forest cover change and evapotranspiration response.

[0004] Furthermore, the original evapotranspiration change is influenced not only by forest cover change but also by background factors such as precipitation, temperature, radiation, drought severity, topographic conditions, and remote sensing inversion errors. Directly using the original evapotranspiration change to identify forest cover change thresholds can easily misjudge regional common background changes as evapotranspiration response changes caused by forest cover change. Although background evapotranspiration change can be estimated using neighboring reference spatial units with stable land cover types and minimal forest cover change, the evapotranspiration change may still fail to represent the common background change of the target reference spatial unit when there are significant differences between the reference and target reference spatial units in terms of climate, topography, or ecological zoning. Therefore, it is necessary to strengthen the background similarity constraint between the reference and target reference spatial units while subtracting background evapotranspiration change.

[0005] Different regions exhibit variations in climate, forest type, hydrothermal conditions, and topography, resulting in spatial heterogeneity in forest cover change thresholds. Using globally uniform or regionally averaged linear relationships to estimate evapotranspiration response may extrapolate the response slope before the threshold to regions after the threshold, or describe the response process that should involve slope changes using averaged linear relationships, leading to overestimation or underestimation of the evapotranspiration response. Therefore, a method is urgently needed to identify the local threshold between forest cover change and the evapotranspiration response after background evapotranspiration reduction, and to further determine the evapotranspiration response bias caused by ignoring this local threshold. Summary of the Invention

[0006] The purpose of this invention is to provide a method for determining the evapotranspiration response deviation based on the local forest cover change threshold, which can solve the problems that linear methods are difficult to characterize local threshold responses and difficult to quantitatively determine model structure deviations, and realize the quantitative determination of evapotranspiration response deviation.

[0007] To achieve the above objectives, the present invention provides the following solution: A method for determining evapotranspiration response bias based on a local forest cover change threshold includes: Acquire evapotranspiration data, forest cover data, land cover type data, climate background data, topographic data, ecological zoning data, and spatial location data for multiple time periods within the target area; Based on evapotranspiration and forest cover data, the original changes in evapotranspiration and forest cover are determined using basic spatial units as objects. Based on land cover type data, forest cover change, climate background data, topographic data, ecological zoning data, and spatial location data, reference base spatial units with similar background conditions are selected within the neighborhood of the target base spatial unit, and the background evapotranspiration change of the target base spatial unit is estimated from the original evapotranspiration change of the reference base spatial units. Subtract the background evapotranspiration change from the original evapotranspiration change of the target basic spatial unit to obtain the evapotranspiration response after background subtraction; The target area is divided into multiple local analysis windows, and within each local analysis window, a local response sample set is constructed based on the forest cover change and the evapotranspiration response after background subtraction. Linear response models and continuous piecewise response models were established based on local response sample sets. After candidate forest cover change threshold identification and resampling stability test, the stable forest cover change threshold was determined. A threshold response model was established using a stable forest cover change threshold. The same forest cover change to be evaluated was input into the linear response model and the threshold response model respectively to obtain the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate. The evapotranspiration response bias is determined based on the difference between the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate.

[0008] Preferably, based on evapotranspiration data and forest cover data, the original changes in evapotranspiration and forest cover are determined using basic spatial units as objects, including: Using the start and end times of the change period as time references, the evapotranspiration of the basic spatial unit at the start and end times is extracted from the evapotranspiration data, and the difference between the evapotranspiration at the end time and the evapotranspiration at the start time is determined as the original evapotranspiration change. The forest cover rate of the basic spatial unit at the start and end times is extracted from the forest cover data, and the difference between the forest cover rate at the end time and the forest cover rate at the start time is determined as the forest cover change.

[0009] Preferably, within the neighborhood of the target basic spatial unit, reference basic spatial units with similar background conditions are selected, including: Based on spatial location data, the basic spatial unit is determined within the neighborhood of the target basic spatial unit; Based on land cover type data, select basic spatial units from the neighborhood of the target basic spatial unit that have the same land cover type at the start and end of the change period, and have the same land cover type as the target basic spatial unit at the end of the change period. Based on the forest cover change, basic spatial units whose absolute value of forest cover change is not greater than a preset reference change threshold are selected from the basic spatial units after land cover type screening. Based on climate background data, topographic data, and ecological zoning data, basic spatial units with similar background conditions to the target basic spatial unit are selected from the basic spatial units after screening by forest cover change, and these are used as reference basic spatial units.

[0010] Preferably, based on climate background data, topographic data, and ecological zoning data, reference spatial units are selected from the basic spatial units screened by forest cover change that have background conditions similar to the target basic spatial unit, including: Based on the climate background data, at least one of the differences in precipitation, temperature, radiation, drought index and potential evapotranspiration between the basic spatial unit screened by forest cover change and the target basic spatial unit is determined. When each of the determined differences meets the corresponding preset climate similarity threshold, the basic spatial unit screened by forest cover change is determined to meet the climate similarity condition. Based on topographic data, at least one of the differences in altitude, slope, and aspect between the basic spatial unit after forest cover change screening and the target basic spatial unit is determined. When each of the determined differences meets the corresponding preset topographic difference threshold, the basic spatial unit after forest cover change screening is determined to meet the topographic similarity condition. Based on ecological zoning data, if the basic spatial unit after screening by forest cover change has the same zoning result as the target basic spatial unit in at least one of the zoning types of climate zone, ecological zone, hydroclimate zone and forest type zone, the basic spatial unit after screening by forest cover change is determined to meet the ecological zoning consistency condition. The basic spatial unit that simultaneously meets the climate similarity condition, the terrain similarity condition, and the ecological zoning consistency condition is determined as the reference basic spatial unit.

[0011] Preferably, estimating the background evapotranspiration change of the target base space unit from the original evapotranspiration change of the reference base space unit includes: Determine the spatial distance between the target basic spatial unit and each reference basic spatial unit based on spatial location data; Determine distance weights that are negatively correlated with spatial distance based on spatial distance; Based on distance weights, the original evapotranspiration changes of each reference base space unit are calculated by weighted average to obtain the background evapotranspiration changes of the target base space unit.

[0012] Preferably, the background evapotranspiration change is subtracted from the original evapotranspiration change of the target basic space unit to obtain the evapotranspiration response after background subtraction, including: The original evapotranspiration change of the target basic spatial unit is spatially correlated with the background evapotranspiration change of the target basic spatial unit; The background evapotranspiration response is obtained by subtracting the background evapotranspiration change of the target basic space unit from the original evapotranspiration change of the target basic space unit.

[0013] Preferably, the target area is divided into multiple local analysis windows, and within each local analysis window, a local response sample set is constructed based on forest cover change and evapotranspiration response after background subtraction, including: Within each local analysis window, the changes in forest cover and the evapotranspiration response after background subtraction are spatially and temporally matched for multiple time periods. The matched forest cover change and the background-subtracted evapotranspiration response were used as local response samples. A set of local response samples is constructed based on multiple local response samples.

[0014] Preferably, a linear response model and a continuous piecewise response model are established based on a local response sample set. After candidate forest cover change threshold identification and resampling stability testing, a stable forest cover change threshold is determined, including: Based on a local response sample set, a linear response model is established to characterize the linear relationship between forest cover change and evapotranspiration response after background subtraction. Based on the local response sample set, a continuous segmented response model is established with candidate forest cover change thresholds as segmentation points and maintaining continuity at the segmentation points. The candidate threshold search range is determined based on the empirical distribution of forest cover change in the local response sample set. Multiple candidate thresholds are generated within the candidate threshold search range, and a continuous piecewise response model corresponding to each candidate threshold is fitted. When the continuous piecewise response model satisfies the preset model improvement conditions relative to the linear response model, and the difference between the response slope before the threshold and the response slope after the threshold satisfies the preset slope change conditions, the corresponding candidate threshold is determined as the candidate forest cover change threshold. The candidate forest cover change thresholds are resampled for stability testing to determine the stable forest cover change threshold.

[0015] Preferably, a resampling stability test is performed on the candidate forest cover change thresholds to determine the stable forest cover change threshold, including: Multiple resampling with replacement is performed on the local response sample set within the local analysis window with candidate forest cover change thresholds to obtain multiple resampled sample sets; For each resampled sample set, candidate forest cover change threshold identification and continuous piecewise response model fitting are performed again to obtain the resampled forest cover change threshold, the response slope before the threshold, and the response slope after the threshold. The effective threshold fitting ratio is determined by the ratio between the number of effective resampling forest cover change thresholds and the total number of resamplings. The directional consistency ratio is determined by the proportion of times the directional consistency ratio is consistent with the directional consistency ratio between the directional consistency ratio of the response slope before and after the effective resampled forest cover change threshold and the directional consistency ratio of the response slope before and after the candidate forest cover change threshold. When the effective threshold fitting ratio is not lower than the preset effective ratio and the directional consistency ratio is not lower than the preset directional consistency ratio, the stable forest cover change threshold is determined based on multiple effective resampled forest cover change thresholds.

[0016] Preferably, the evapotranspiration response deviation is determined based on the difference between the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate, including: Subtracting the threshold-constrained evapotranspiration response estimate from the linear evapotranspiration response estimate yields the evapotranspiration response bias corresponding to the forest cover change to be evaluated. When the evapotranspiration response deviation is greater than zero, it is determined that the linear response model overestimates the evapotranspiration response compared to the threshold response model. When the evapotranspiration response deviation is less than zero, it is determined that the linear response model underestimates the evapotranspiration response compared to the threshold response model. When the evapotranspiration response deviation is equal to zero, it is determined that the linear response model and the threshold response model have no difference in their estimation of the evapotranspiration response.

[0017] The present invention discloses the following beneficial effects: This invention acquires evapotranspiration data, forest cover data, land cover type data, climate background data, topographic data, ecological zoning data, and spatial location data corresponding to multiple time periods within a target area. It then determines the original evapotranspiration and forest cover changes using basic spatial units as objects, ensuring that forest cover changes and evapotranspiration changes correspond under the same spatial objects and time periods, avoiding the need to calculate evapotranspiration response solely based on regional averages. By using land cover type data, forest cover changes, climate background data, topographic data, ecological zoning data, and spatial location data, it selects reference basic spatial units with similar background conditions within the neighborhood of the target basic spatial unit. The background evapotranspiration change is then estimated from the original evapotranspiration change of the reference basic spatial units and subtracted from the original evapotranspiration change of the target basic spatial unit. This reduces the influence of non-forest cover change factors such as climate background, topographic background, and ecological zoning background on the evapotranspiration response. The method incorporates quantitative data; by constructing a local response sample set based on forest cover change and background subtraction evapotranspiration response within the local analysis window, and establishing linear response models and continuous piecewise response models, a stable forest cover change threshold is determined through candidate forest cover change threshold identification and resampling stability test. This allows for modeling situations where the evapotranspiration response slope changes after forest cover change reaches a certain level at the local scale, rather than directly extrapolating the linear response relationship to the entire range of changes. By inputting the same forest cover change to be evaluated into the linear response model and the threshold response model respectively, and determining the evapotranspiration response bias by the difference between the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate, the overestimation or underestimation caused by ignoring the local forest cover change threshold can be transformed into a bias value under the same forest cover change, thereby solving the problem that existing linear methods are difficult to characterize local threshold responses and difficult to quantitatively represent model structure bias. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of the method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the stable forest cover change threshold output format provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the output format of the response slope change before and after the threshold provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the evaporation response deviation output format provided in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] This embodiment addresses the assessment of the hydrological effects of forest cover change, applicable to regions on a global or regional scale where forest cover increases, decreases, or ecological restoration projects have significant impacts. Based on evapotranspiration data, forest cover data, land cover type data, climate background data, topographic data, ecological zoning data, and spatial location data, this method first obtains the evapotranspiration response after background subtraction by filtering neighboring reference base spatial units and subtracting background evapotranspiration changes. Then, within a local analysis window, a local response sample set is constructed between the forest cover change and the background-subtracted evapotranspiration response. Stable forest cover change thresholds are identified through candidate threshold scanning, continuous piecewise response model fitting, and resampling stability testing. Finally, based on the estimation differences of the same forest cover change to be assessed under linear response models and threshold response models, the evapotranspiration response bias caused by ignoring the threshold response is determined.

[0023] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1As shown, this embodiment provides a method for determining evapotranspiration response deviation based on a local forest cover change threshold, including: Step 100: Obtain evapotranspiration data, forest cover data, land cover type data, climate background data, topographic data, ecological zoning data, and spatial location data for multiple time periods within the target area; Step 200: Based on evapotranspiration data and forest cover data, determine the original evapotranspiration change and forest cover change using basic spatial units as objects; Step 300: Based on land cover type data, forest cover change, climate background data, topographic data, ecological zoning data, and spatial location data, select reference base spatial units with similar background conditions within the neighborhood of the target base spatial unit, and estimate the background evapotranspiration change of the target base spatial unit from the original evapotranspiration change of the reference base spatial units. Step 400: Subtract the background evapotranspiration change from the original evapotranspiration change of the target basic space unit to obtain the evapotranspiration response after background subtraction; Step 500: Divide the target area into multiple local analysis windows, and within each local analysis window, construct a local response sample set based on forest cover change and evapotranspiration response after background subtraction; Step 600: Establish a linear response model and a continuous piecewise response model based on the local response sample set. After candidate forest cover change threshold identification and resampling stability test, determine the stable forest cover change threshold. Step 700: Establish a threshold response model based on a stable forest cover change threshold. Input the same forest cover change to be evaluated into the linear response model and the threshold response model respectively to obtain the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate. Step 800: Determine the evapotranspiration response deviation based on the difference between the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate.

[0024] Specifically, in step 100, evapotranspiration data, forest cover data, land cover type data, climate background data, topographic data, ecological zoning data, and spatial location data corresponding to multiple time periods within the target area are obtained.

[0025] In this embodiment, the target area is the global land area, the effective data time range is from 2001 to 2020, and the variation period is the variation period between adjacent years, including 2001-2002, 2002-2003, and up to 2019-2020. The basic spatial unit is a 0.05°×0.05° raster pixel, and the unified coordinate system is the WGS84 geographic coordinate system. The local analysis window is a 2°×2° regular spatial window.

[0026] In this embodiment, the evapotranspiration data uses PML annual-scale evapotranspiration products, with a raster data format, a time resolution of one year, and a unit of millimeters per year, to characterize the evapotranspiration of each basic spatial unit in the corresponding year.

[0027] In this embodiment, the forest cover data uses the Tree Cover Percentage data layer in the MODIS Vegetation Continuous Fields product. The data format is raster data, the time resolution is year, and the unit is percentage, which is used to characterize the forest cover rate of each basic spatial unit in the corresponding year.

[0028] In this embodiment, the land cover type data adopts the MODIS Land Cover Type annual land cover classification product. The data format is raster data, the time resolution is year, and the value is the land cover type code, which is used to determine the land cover type and its stability of the basic spatial unit in the beginning and end years of the change period.

[0029] In this embodiment, the climate background data uses TerraClimate gridded climate data, specifically including precipitation data, temperature data, shortwave radiation data, potential evapotranspiration data, and drought index data. Precipitation data is used to calculate the multi-year average precipitation difference between the target and reference spatial units, in millimeters per year; temperature data is used to calculate the multi-year average temperature difference between the target and reference spatial units, in degrees Celsius; shortwave radiation data is used to calculate the multi-year average shortwave radiation difference between the target and reference spatial units, in watts per square meter; potential evapotranspiration data is used to calculate the multi-year average potential evapotranspiration difference between the target and reference spatial units, in millimeters per year; drought index data is calculated based on precipitation and potential evapotranspiration, or obtained using drought-related variables in TerraClimate, and is a dimensionless index. The time range of the climate background data is from 2001 to 2020, and the climate background conditions of the base spatial units are represented by the multi-year average values ​​from 2001 to 2020.

[0030] In this embodiment, the terrain data is obtained using GMTED2010 digital elevation model data, including elevation, slope, and aspect. Elevation is directly obtained from the GMTED2010 digital elevation model and is in meters; slope and aspect are calculated by the digital elevation model using terrain factors, with slope and aspect in degrees. The terrain data is spatially static data, used to characterize the terrain background conditions of the basic spatial unit throughout the entire analysis period.

[0031] In this embodiment, the ecological zoning data uses Köppen-Geiger climate zone classification data. The ecological zoning data is a discrete categorical variable used to determine whether the reference and target base spatial units are located in the same climate zone.

[0032] In this embodiment, the spatial location data consists of the registered raster row and column numbers, the longitude of the pixel center, the latitude of the pixel center, and spatial reference information, which are used to determine the neighborhood relationship, spatial distance, local analysis window affiliation relationship, and spatialization result output location between basic spatial units.

[0033] Specifically, spatial registration, spatial resampling, temporal matching, spatial pruning, and invalid value processing are performed on different data, including: unifying evapotranspiration data, forest cover data, land cover type data, climate background data, topographic data, ecological zoning data, and spatial location data to the same target area, the same spatial reference system, and the same basic spatial unit scale; resampling data with different spatial resolutions to the basic spatial units; aggregating or matching data from different time scales to the same period of change; and removing basic spatial units that are missing evapotranspiration data, missing forest cover data, invalid land cover types, missing climate background data, missing topographic data, missing ecological zoning data, or have abnormal spatial locations or exceed the target area.

[0034] In this embodiment, outliers and invalid values ​​are removed according to the following rules: when evapotranspiration data is NoData, the quality control identifier is invalid, or the evapotranspiration is less than 0, the corresponding basic spatial unit is marked as invalid for that year; when forest cover data is NoData, the forest cover rate is less than 0% or greater than 100%, the corresponding basic spatial unit is marked as invalid for that year; when land cover type data has an invalid product code, an unclassified code, or a missing code, the corresponding basic spatial unit is marked as invalid for that year; when climate background data, topographic data, or ecological zoning data is NoData or exceeds the valid value range of the corresponding data product, the corresponding basic spatial unit is marked as invalid. For any period of change, if a basic spatial unit lacks evapotranspiration data, forest cover data, land cover type data, or necessary background similarity screening data at the start or end time, then that basic spatial unit will not be included in the subsequent calculations for that period of change.

[0035] For continuous variable data, including evapotranspiration, forest cover, precipitation, temperature, shortwave radiation, potential evapotranspiration, drought index, elevation, and slope, spatial resampling uses bilinear interpolation. For discrete categorical variable data, including land cover type, Köppen-Geiger climate zone classification, and aspect grouping, spatial resampling uses the mode method.

[0036] For data with different time scales, matching is performed according to the year corresponding to the period of change. For monthly climate background data, it is first aggregated into annual data, and then multi-year average climate background data is calculated based on the corresponding years from 2001 to 2020, which is used to determine the climate similarity of the reference basic spatial unit.

[0037] Specifically, in step 200, adjacent years are used as the time periods of change. For any basic spatial unit... Let the starting year of a certain period of change be... The end year is Calculate the original change in evapotranspiration based on the evapotranspiration data:

[0038] Calculate forest cover change based on forest cover data:

[0039] in, Represents the basic spatial unit. Indicates the start time of the period of change. Indicates the end time of the period of change. Represents the basic spatial unit The original change in evapotranspiration, Represents the basic spatial unit Changes in forest cover and Representing the basic spatial units Evaporation at the start and end times and Representing the basic spatial units Forest cover at the start and end times.

[0040] In this embodiment, evapotranspiration is expressed in millimeters per year, forest cover is expressed as a percentage, and the unit for forest cover change is percentage points. Basic spatial units with an absolute forest cover change greater than 2 percentage points are designated as forest cover change units to be evaluated; basic spatial units with an absolute forest cover change of no more than 2 percentage points and meeting the subsequent screening criteria for reference basic spatial units are designated as candidate reference basic spatial units.

[0041] Specifically, in step 300, reference base spatial units are selected within the neighborhood of the target base spatial unit. The reference base spatial units are used to estimate the background evapotranspiration changes of the target base spatial unit during the corresponding change period caused by common climate fluctuations, topographic background, ecological zoning background, and common data errors.

[0042] In this embodiment, the reference basic spatial unit satisfies the basic screening conditions and the background similarity screening conditions.

[0043] The basic screening criteria include: the reference basic spatial unit is located within the preset neighborhood search radius of the target basic spatial unit; the reference basic spatial unit has the same land cover type in the start and end years of the corresponding change period; the reference basic spatial unit has the same land cover type as the target basic spatial unit in the end year of the corresponding change period; the absolute value of the forest cover change of the reference basic spatial unit is not greater than the preset reference change threshold; and the number of reference basic spatial units corresponding to the target basic spatial unit is not less than the preset minimum number of reference units.

[0044] In this embodiment, the preset neighborhood search radius is used. Take 50km as an example, and set a reference change threshold. We set a 2% threshold and a minimum preset number of reference units of 10. The preset neighborhood search radius ensures that the reference base spatial unit and the target base spatial unit are within their local neighborhood. The preset reference change threshold ensures that the reference base spatial unit itself does not exhibit significant forest cover changes. The preset minimum number of reference units ensures that the estimation of background evapotranspiration changes has sufficient reference samples.

[0045] In this embodiment, the background similarity screening conditions include climate similarity conditions, terrain similarity conditions, and ecological zoning consistency conditions.

[0046] Specifically, the climate similarity conditions are determined based on TerraClimate gridded climate data, including differences in precipitation, temperature, shortwave radiation, potential evapotranspiration, and drought index, and each difference indicator is required to simultaneously meet the corresponding preset climate similarity threshold.

[0047] In a preferred embodiment, the difference in multi-year average precipitation between the reference base space unit and the target base space unit from 2001 to 2020 is no greater than 20% of the multi-year average precipitation of the target base space unit; the difference in multi-year average temperature between 2001 and 2020 is no greater than 2°C; the difference in multi-year average shortwave radiation between 2001 and 2020 is no greater than 15% of the multi-year average shortwave radiation of the target base space unit; the difference in multi-year average potential evapotranspiration between 2001 and 2020 is no greater than 15% of the multi-year average potential evapotranspiration of the target base space unit; and the difference in multi-year average drought index between 2001 and 2020 is no greater than 0.2.

[0048] Specifically, the terrain similarity conditions are determined based on GMTED2010 digital elevation model data, including differences in altitude, slope, and aspect, and each difference index is required to simultaneously meet the corresponding preset terrain difference threshold.

[0049] In a preferred embodiment, the altitude difference between the reference base space unit and the target base space unit is no greater than 300m, the slope difference is no greater than 10°, and the aspect difference is no greater than 45°.

[0050] Specifically, the ecological zoning consistency conditions are determined based on Köppen-Geiger climate zone classification data, including that the reference base spatial unit and the target base spatial unit belong to the same climate zone.

[0051] For the target basic space unit The reference basic spatial unit set is determined according to the following formula:

[0052] in, Represents the basic spatial unit of the target The corresponding set of reference basic spatial units, Indicates the reference base space unit. Represents the basic spatial unit of the target With reference base space unit Spatial distance between them Indicates the preset neighborhood search radius. and Representing the reference base space unit Land cover type at the start and end times, Represents the basic spatial unit of the target Land cover type at the end of the time, Indicates reference base space unit Changes in forest cover This indicates a preset reference change threshold. Units and The units are consistent. Represents the basic spatial unit of the target With reference base space unit Background similarity discriminant function between them.

[0053] When the target basic space unit With reference base space unit When the background similarity filtering conditions are met, ;otherwise, .

[0054] When the number of reference base space units corresponding to the target base space unit is less than 10, the target base space unit is marked as having no effective background evapotranspiration change during the corresponding change period and is not included in the background subtraction processing for that change period.

[0055] Specifically, in step 400, the distance weight is determined based on the spatial distance between the reference base space unit and the target base space unit, and the background evapotranspiration change of the target base space unit is estimated based on the original evapotranspiration change of the reference base space unit.

[0056] Distance weights are determined according to the following formula:

[0057] The change in background evapotranspiration is determined by the following formula:

[0058] in, Represents the basic spatial unit of the target Background evapotranspiration change Indicates reference base space unit The original change in evapotranspiration, Represents the basic spatial unit of the target With reference base space unit Distance weights between them Adopting target basic space unit With reference base space unit The spherical distance between the center points is determined. This is used to avoid positive numbers with a denominator of zero.

[0059] Subtracting the background evapotranspiration change from the original evapotranspiration change of the target basic spatial unit yields the evapotranspiration response after background subtraction:

[0060] in, Represents the basic spatial unit of the target The evapotranspiration response after background subtraction. The evapotranspiration response after background subtraction is used to characterize the remaining evapotranspiration change after subtracting the changes in the common background evapotranspiration in the neighborhood.

[0061] Specifically, in step 500, the target area is divided into multiple local analysis windows, and within each local analysis window, the forest cover change corresponding to multiple change periods is spatially and temporally matched with the evapotranspiration response after background subtraction to form a local response sample set.

[0062] No. The set of local response samples within a local analysis window is represented as follows:

[0063] in, Indicates the first A set of local response samples within a local analysis window Indicates the first Forest cover change corresponding to each sample Indicates the first Evapotranspiration response after background subtraction for each sample Indicates the first The number of valid samples within a local analysis window.

[0064] In this embodiment, within each 2°×2° local analysis window, the forest cover change and evapotranspiration response after background subtraction are collected for multiple adjacent years to construct a local response sample set. Number of valid samples within each local analysis window If the value is less than 300, the local analysis window will not participate in the identification of candidate forest cover change thresholds; if the difference between the 95th quartile and the 5th quartile of the forest cover change is less than 4 percentage points, the local analysis window will not participate in the identification of candidate forest cover change thresholds.

[0065] Specifically, in step 600, a linear response model and a continuous piecewise response model are established based on the local response sample set, respectively.

[0066] The linear response model is established according to the following formula:

[0067] The continuous piecewise response model is established according to the following formula:

[0068] in, This indicates the change in forest cover. This represents the evapotranspiration response after background subtraction. Indicates the threshold for candidate forest cover change. , , , and Indicates model parameters, This represents the slope of the response before the threshold. Indicates the slope of the response after the threshold; when When the continuous piecewise response model has equal expressions on both sides, the model remains continuous at the candidate forest cover change threshold.

[0069] According to the The empirical distribution of forest cover change within a local analysis window determines the candidate threshold search range, which is as follows:

[0070] in, and They represent the first The first local analysis window of forest cover change quantiles and the quantiles, and .

[0071] In this embodiment, Take 5%, Taking 95%, the candidate threshold search range is the [number]th [th element]. The 5th to 95th quantiles of forest cover change within a local analysis window were used. Within the candidate threshold search range, a sample quantile scanning method was employed to form a candidate threshold set. Specifically, multiple quantile probability points were generated at quantile intervals of 0.5 percentage points within the 5% to 95% quantile probability range, and the corresponding sample quantile values ​​of forest cover change were used as candidate thresholds. When duplicate candidate thresholds existed, they were deduplicated. For each candidate threshold... If satisfied The number of samples is less than or satisfy The number of samples is less than If the candidate threshold is not included in the comparison of continuous piecewise response model fitting and model fitting evaluation index, then the candidate threshold will not be included in the comparison of continuous piecewise response model fitting and model fitting evaluation index.

[0072] For each candidate threshold, a continuous piecewise response model is fitted using ordinary least squares, and the sum of squared residuals is calculated. The sum of squared residuals is determined according to the following formula:

[0073] in, Indicates candidate threshold The corresponding sum of squared residuals, Indicates candidate threshold The established continuous piecewise response model for the first The estimated value of a sample.

[0074] The candidate threshold with the smallest sum of squared residuals is determined as the initial forest cover change threshold:

[0075] in, Indicates the first The initial forest cover change threshold for each local analysis window.

[0076] To avoid the continuous piecewise response model being selected solely due to an increase in the number of parameters, this embodiment sets model improvement conditions. The model improvement conditions are determined according to the following formula:

[0077] in, This represents the Akaike Information Criterion value for the linear response model. This represents the Akaike Information Criterion value for the continuous piecewise response model.

[0078] This embodiment also includes a slope variation condition. The slope variation condition is determined according to the following formula:

[0079] in, This is a positive number used to avoid conditions where the denominator is zero, which is used to prevent changes in slope.

[0080] When the initial forest cover change threshold satisfies both the model improvement condition and the slope change condition, the initial forest cover change threshold is determined as the candidate forest cover change threshold; otherwise, the corresponding local analysis window does not output the candidate forest cover change threshold.

[0081] In this embodiment, forest cover change is expressed as a percentage point, and evapotranspiration response after background subtraction is expressed in millimeters per year. Therefore, the slope of the response before the threshold is... Response slope after threshold The unit is millimeters per year per percentage point.

[0082] Specifically, in step 600, a resampling stability test is performed on the local analysis window with candidate forest cover change thresholds.

[0083] In this embodiment, the number of resampling times Take 500. For the first... Each local analysis window is used for resampling, and samples are taken from within that local analysis window. Sampling with replacement is performed from the valid sample to draw... Each sample forms a resampled sample set. For each resampled sample set, the candidate threshold scan, continuous piecewise response model fitting, model improvement condition determination, and slope change condition determination are re-executed to obtain the resampled forest cover change threshold, the response slope before the threshold, and the response slope after the threshold.

[0084] A resampling result is considered valid when the following conditions are met: the resampling sample set can fit a continuous piecewise response model; the difference between the 95th quantile and the 5th quantile of forest cover change within the resampling sample set is not less than 4 percentage points; the resampling forest cover change threshold is located between the 2nd and 98th quantiles of forest cover change; and the number of samples before the threshold is not less than [number missing]. The number of samples after the threshold is not less than The corresponding continuous piecewise response model satisfies the model improvement condition and the slope change condition.

[0085] The effective threshold fitting ratio is determined according to the following formula:

[0086] in, Indicates the first The effective threshold fitting ratio of each local analysis window Indicates in The number of times the effective resampled forest cover change threshold is obtained in the resampling.

[0087] The directional consistency ratio is determined by the following formula:

[0088] in, Indicates the first The proportion of consistency in the direction of change of the slope of the response before and after the threshold within a local analysis window. This represents the set of resampling times used to obtain an effective resampling threshold for forest cover change. This indicates the direction of the slope change in the response before and after the candidate forest cover change threshold. Indicates the first The direction of the slope change of the response before and after the threshold corresponding to each effective resampling. This is an indicator function that takes the value 1 when the condition inside the parentheses is true, and 0 otherwise.

[0089] In this embodiment, the preset effective ratio is 80%, and the preset directional consistency ratio is 90%. and At that time, the first The local analysis window is determined as the stable threshold window; otherwise, the first local analysis window is determined as the stable threshold window. Each local analysis window is marked as a threshold window for unstable forest cover change.

[0090] For the stable threshold window, the stable forest cover change threshold is determined based on the median of multiple valid resampled forest cover change thresholds:

[0091] in, Indicates the first The stable forest cover change threshold for each local analysis window (i.e., the stable forest cover change threshold within the k-th local analysis window). Values ), , , , Indicates the first Multiple effective resampling forest cover change thresholds within a local analysis window.

[0092] like Figure 2 As shown, Figure 2 The output format is used to display the threshold for stable forest cover change. Figure 2 The values ​​within each effective grid represent the stable forest cover change threshold for the corresponding local analysis window, in percentage points; the color scale ranges from 0 to 32 percentage points, with the color from light to dark indicating the stable forest cover change threshold from low to high. Figure 2 In row R1, columns C3 and C4 have values ​​of 20.4 and 5.3 respectively; in row R2, columns C1, C2, C4, C5, C6, and C7 have values ​​of 10.8, 11.5, 11.3, 11.3, 9.8, and 10.7 respectively; and in row R3, columns C1, C2, C3, C4, C5, C6, and C10 have values ​​of 17.6, 11.2, 13.7, 14.8, 21.4, and 10.0 respectively. The values ​​for R4 rows and columns C2, C3, C4, C5, C6, C7, and C9 are 7.0, 13.0, 8.1, 18.7, 4.8, 7.8, and 9.8, respectively; the values ​​for R5 rows and columns C6 and C7 are 4.6 and 20.7, respectively; the values ​​for R6 rows and columns C4, C5, C6, C7, C8, and C10 are 2.9, 2.8, 15.4, 5.7, 2.3, and 5.1, respectively. Figure 2 The gray grid and short horizontal lines in the text indicate that the corresponding local analysis window did not obtain a stable forest cover change threshold, or that the threshold result was not output due to insufficient number of effective samples, insufficient distribution range of forest cover change, unmet model improvement conditions, unmet slope change conditions, or failure to pass the resampling stability test.

[0093] like Figure 3 As shown, Figure 3 This output format is used to display the change in the slope of the response before and after a threshold. Figure 3 The values ​​within each effective grid represent the percentage change in the response slope after the threshold relative to the response slope before the threshold within the corresponding local analysis window; the color scale ranges from 20% to 120%, with the color from light to dark indicating the magnitude of the change in the response slope before and after the threshold from low to high. Figure 3 In the table, columns C3 and C4 of row R1 contain 120 and 93 respectively; columns C1, C2, C4, C5, C6, and C7 of row R2 contain 120, 120, 115, 98, 89, and 115 respectively; and columns C1, C2, C3, C4, C5, C6, and C10 of row R3 contain 111, 79, 118, 82, 120, 115, and 56 respectively. The values ​​of columns C2, C3, C4, C5, C6, C7, and C9 in row R4 are 104, 109, 104, 103, 92, 120, and 113, respectively; the values ​​of columns C6 and C7 in row R5 are 102 and 120, respectively; and the values ​​of columns C4, C5, C6, C7, C8, and C10 in row R6 are 98, 99, 120, 94, 100, and 103, respectively. Figure 3 The gray grid and short horizontal lines in the text indicate that the corresponding local analysis window did not obtain a stable forest cover change threshold, or did not output effective results of the slope change before and after the threshold.

[0094] Specifically, in step 700, a stable forest cover change threshold is used. Using fixed segmentation points, the continuous segmented response model is refitted based on the local response sample set within the corresponding local analysis window to obtain the threshold response model.

[0095] In this embodiment, the amount of forest cover change to be assessed To correspond to the forest cover change of the target basic spatial unit within the local analysis window during the period to be assessed, the unit and Consistent, both are percentage points. In scenario assessment, Data can also be derived from land use scenario data, ecological engineering planning data, vegetation dynamic model output data, or remote sensing prediction data. When the source data is expressed as a scale of 0 to 1, it should be converted to percentage points before being input into the linear response model and threshold response model.

[0096] For the forest cover change to be assessed The estimated linear evapotranspiration response value corresponding to the linear response model is:

[0097] The threshold-constrained evapotranspiration response estimate for the threshold response model is:

[0098] Specifically, in step 800, the evapotranspiration response deviation is determined based on the difference between the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate. The evapotranspiration response deviation is determined according to the following formula:

[0099] in, Indicates the amount of forest cover change to be assessed. The corresponding evapotranspiration response deviation, This represents the threshold for stable forest cover change. and Represents the parameters of the linear response model. , and This represents the threshold response model parameters.

[0100] when When, it indicates that the linear response model overestimates the evapotranspiration response compared to the threshold response model; when When the linear response model underestimates the evapotranspiration response compared to the threshold response model, the evapotranspiration response bias is the model structure deviation produced by the linear response model compared to the threshold response model; when When the value is 0, it indicates that the linear response model and the threshold response model have no difference in estimating the evapotranspiration response.

[0101] This embodiment outputs at least one of the following results: stable forest cover change threshold, response slope before threshold, response slope after threshold, change in response slope before and after threshold, linear evapotranspiration response estimate, threshold-constrained evapotranspiration response estimate, and evapotranspiration response deviation. These results can be output as raster data, tabular data, vector data, or a virtual grid schematic.

[0102] like Figure 4 As shown, Figure 4 The output format is used to demonstrate the deviation in evapotranspiration response. Figure 4 The values ​​within each effective grid represent the difference in evapotranspiration response estimates obtained after inputting the same forest cover change to the linear response model and the threshold response model, in millimeters per year; the color scale ranges from -1.5 to 1.5 millimeters per year, with the color changing from blue to red indicating that the evapotranspiration response deviation changes from negative to positive. Figure 4 In row R1, columns C3 and C4 are 0.05 and -0.56 respectively; in row R2, columns C1, C2, C4, C5, C6, and C7 are -0.01, 0.40, -0.99, -0.55, -0.35, and -0.40 respectively; in row R3, columns C1, C2, C3, C4, C5, C6, and C10 are 2.07, 0.57, 0.56, 1.58, 0.86, -0.33, and 0.78 respectively; R... The values ​​for columns C2, C3, C4, C5, C6, C7, and C9 in row 4 are -0.35, 0.45, -0.66, 0.47, -0.58, -1.28, and 0.03, respectively; the values ​​for columns C6 and C7 in row R5 are -0.17 and -0.22, respectively; and the values ​​for columns C4, C5, C6, C7, C8, and C10 in row R6 are -0.56, -0.06, 0.40, -0.37, -0.16, and -1.09, respectively. Figure 4 When values ​​of 2.07 in row R3, column C1 and 1.58 in row R3, column C4 exceed the color scale range, they are displayed according to the color of the positive endpoint. A positive evaporation response deviation indicates that the linear response model overestimates the evaporation response compared to the threshold response model; a negative evaporation response deviation indicates that the linear response model underestimates the evaporation response compared to the threshold response model. Figure 4 The gray grid and short horizontal lines in the diagram indicate that the corresponding local analysis window did not obtain effective evapotranspiration response deviation results.

[0103] In this embodiment, Figures 2 to 4This is used to demonstrate the format of the output results of the method of the present invention and is not intended to be used as a real geospatial distribution map. Figures 2 to 4 The spatialized output of the method of the present invention is represented by a virtual grid region, which does not include real geographical boundaries, national boundaries, provincial boundaries, administrative boundaries, coastlines or identifiable geographical base maps. Figures 2 to 4 In the figure, the horizontal axis C1 to C10 represent the column numbers of the virtual grid region, and the vertical axis R1 to R6 represent the row numbers of the virtual grid region; each grid represents a local analysis window; gray grids and short horizontal lines indicate that the corresponding position does not meet the stable threshold output condition, or there is no valid output result. The values ​​in the figure are used to show the data structure, units, color scale representation, and invalid window display method of the output results of the method of this invention.

[0104] In other embodiments, the target area can be a regional-scale forest area, an ecological engineering implementation area, a watershed forest change area, or other areas with significant forest cover changes; the basic spatial unit can be other raster scales, vector plots, ecological plots, or sub-watershed units; the climate background data, topographic data, and ecological zoning data can be other data products with the same or similar functions as those in this embodiment; the neighborhood search radius, reference change threshold, local analysis window scale, candidate threshold search range, model improvement conditions, slope change conditions, resampling times, preset effective ratio, and preset directional consistency ratio can be adjusted according to the data resolution, target area scale, distribution of forest cover change, and number of local response samples.

[0105] This embodiment first estimates the background evapotranspiration change of the target base spatial unit using a reference base spatial unit with stable land cover type, minimal forest cover change, and similar background conditions. Then, it subtracts the background evapotranspiration change from the original evapotranspiration change of the target base spatial unit to obtain the evapotranspiration response after background subtraction. Subsequently, a linear response model and a continuous piecewise response model are established within the local analysis window between the forest cover change and the background-subtracted evapotranspiration response, and a stable forest cover change threshold is determined through a resampling stability test. Finally, the same forest cover change to be evaluated is input into both the linear response model and the threshold response model, and the difference between the estimation results of the two models is used to determine the evapotranspiration response bias. This processing method can reduce the impact of common background changes on the evapotranspiration response and quantitatively represent the estimation bias of the linear response model relative to the threshold response model when the local forest cover change threshold is ignored.

[0106] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0107] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for determining evapotranspiration response deviation based on a local forest cover change threshold, characterized in that, include: Acquire evapotranspiration data, forest cover data, land cover type data, climate background data, topographic data, ecological zoning data, and spatial location data for multiple time periods within the target area; Based on the evapotranspiration data and the forest cover data, the original evapotranspiration change and forest cover change are determined using basic spatial units as objects. Based on the land cover type data, the forest cover change, the climate background data, the topography data, the ecological zoning data, and the spatial location data, reference basic spatial units with similar background conditions are selected within the neighborhood of the target basic spatial unit, and the background evapotranspiration change of the target basic spatial unit is estimated from the original evapotranspiration change of the reference basic spatial units. Subtract the background evapotranspiration change from the original evapotranspiration change of the target basic space unit to obtain the evapotranspiration response after background subtraction; The target area is divided into multiple local analysis windows, and within each local analysis window, a local response sample set is constructed based on the forest cover change and the evapotranspiration response after background subtraction. Based on the local response sample set, a linear response model and a continuous piecewise response model are established. After candidate forest cover change threshold identification and resampling stability test, the stable forest cover change threshold is determined. A threshold response model is established based on the stable forest cover change threshold. The same forest cover change to be evaluated is input into the linear response model and the threshold response model respectively to obtain the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate. The evapotranspiration response bias is determined based on the difference between the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate.

2. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 1, characterized in that, Based on the evapotranspiration data and the forest cover data, the original evapotranspiration change and forest cover change are determined using basic spatial units as objects, including: Using the start and end times of the change period as time references, the evapotranspiration of the basic spatial unit at the start and end times is extracted from the evapotranspiration data, and the difference between the evapotranspiration at the end time and the evapotranspiration at the start time is determined as the original evapotranspiration change. The forest cover rate of the basic spatial unit is extracted from the forest cover data at the start time and the end time, and the difference between the forest cover rate at the end time and the forest cover rate at the start time is determined as the forest cover change.

3. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 1, characterized in that, Within the neighborhood of the target basic spatial unit, select reference basic spatial units with similar background conditions, including: Based on the spatial location data, a basic spatial unit is determined within the neighborhood of the target basic spatial unit; Based on the land cover type data, select basic spatial units from the neighborhood of the target basic spatial unit that have the same land cover type at the start and end of the change period, and have the same land cover type as the target basic spatial unit at the end of the change period. Based on the forest cover change, select basic spatial units from the basic spatial units after land cover type screening where the absolute value of the forest cover change is not greater than a preset reference change threshold. Based on the climate background data, the topographic data, and the ecological zoning data, basic spatial units with similar background conditions to the target basic spatial unit are selected from the basic spatial units after screening by forest cover change, and these are used as the reference basic spatial units.

4. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 3, characterized in that, Based on the climate background data, the topographic data, and the ecological zoning data, basic spatial units with similar background conditions to the target basic spatial unit are selected from the basic spatial units filtered by forest cover change, and these are used as the reference basic spatial units, including: Based on the climate background data, at least one of the differences in precipitation, temperature, radiation, drought index, and potential evapotranspiration between the basic spatial unit screened by forest cover change and the target basic spatial unit is determined. When each of the determined differences meets the corresponding preset climate similarity threshold, the basic spatial unit screened by forest cover change is determined to meet the climate similarity condition. Based on the terrain data, at least one of the differences in altitude, slope, and aspect between the basic spatial unit after forest cover change screening and the target basic spatial unit is determined. When each of the determined differences meets the corresponding preset terrain difference threshold, the basic spatial unit after forest cover change screening is determined to meet the terrain similarity condition. Based on the ecological zoning data, if the basic spatial unit after forest cover change screening has the same zoning result as the target basic spatial unit in at least one of the zoning types of climate zone, ecological zone, hydroclimate zone and forest type zone, the basic spatial unit after forest cover change screening is determined to meet the ecological zoning consistency condition. The basic spatial unit that simultaneously meets the climate similarity condition, the terrain similarity condition, and the ecological zoning consistency condition is determined as the reference basic spatial unit.

5. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 1, characterized in that, Estimating the background evapotranspiration change of the target base space unit from the original evapotranspiration change of the reference base space unit includes: The spatial distance between the target basic spatial unit and each of the reference basic spatial units is determined based on the spatial location data; Determine the distance weight that is negatively correlated with the spatial distance based on the spatial distance; Based on the distance weights, the original evapotranspiration changes of each of the reference base space units are calculated by weighted average to obtain the background evapotranspiration changes of the target base space unit.

6. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 1, characterized in that, Subtracting the background evapotranspiration change from the original evapotranspiration change of the target basic space unit yields the background-subtracted evapotranspiration response, including: The original evapotranspiration change of the target basic spatial unit is spatially correlated with the background evapotranspiration change of the target basic spatial unit; The background evapotranspiration response is obtained by subtracting the background evapotranspiration change of the target basic space unit from the original evapotranspiration change of the target basic space unit.

7. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 1, characterized in that, The target area is divided into multiple local analysis windows, and within each local analysis window, a local response sample set is constructed based on the forest cover change and the evapotranspiration response after background subtraction, including: Within each of the local analysis windows, the changes in forest cover and the evapotranspiration response after background subtraction are spatially and temporally matched for multiple time periods. The matched forest cover change and the background-subtracted evapotranspiration response were used as local response samples. The local response sample set is constructed based on multiple local response samples.

8. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 1, characterized in that, Based on the local response sample set, a linear response model and a continuous piecewise response model are established. After candidate forest cover change threshold identification and resampling stability testing, a stable forest cover change threshold is determined, including: Based on the aforementioned local response sample set, a linear response model is established to characterize the linear relationship between forest cover change and evapotranspiration response after background subtraction. Based on the local response sample set, a continuous segmented response model is established with candidate forest cover change thresholds as segmentation points and maintaining continuity at the segmentation points. The candidate threshold search range is determined based on the empirical distribution of forest cover change in the local response sample set. Multiple candidate thresholds are generated within the candidate threshold search range, and a continuous piecewise response model corresponding to each candidate threshold is fitted. When the continuous segmented response model satisfies the preset model improvement condition relative to the linear response model, and the difference between the response slope before the threshold and the response slope after the threshold satisfies the preset slope change condition, the corresponding candidate threshold is determined as the candidate forest cover change threshold. The candidate forest cover change thresholds are subjected to a resampling stability test to determine the stable forest cover change threshold.

9. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 8, characterized in that, To determine the stable forest cover change threshold, a resampling stability test is performed on the candidate forest cover change thresholds, including: Multiple resampling with replacement is performed on the local response sample set within the local analysis window that has the candidate forest cover change threshold to obtain multiple resampled sample sets. For each set of resampled samples, candidate forest cover change threshold identification and continuous segmented response model fitting are performed again to obtain the resampled forest cover change threshold, the response slope before the threshold, and the response slope after the threshold. The effective threshold fitting ratio is determined by the ratio between the number of effective resampling forest cover change thresholds and the total number of resamplings. The directional consistency ratio is determined based on the proportion of times the directional change in the slope of the response before and after the effective resampled forest cover change threshold is consistent with the directional change in the slope of the response before and after the threshold of the candidate forest cover change threshold. When the effective threshold fitting ratio is not lower than the preset effective ratio and the directional consistency ratio is not lower than the preset directional consistency ratio, the stable forest cover change threshold is determined based on multiple effective resampled forest cover change thresholds.

10. The method for determining evapotranspiration response deviation based on a local forest cover change threshold according to claim 1, characterized in that, The evapotranspiration response bias is determined based on the difference between the linear evapotranspiration response estimate and the threshold-constrained evapotranspiration response estimate, including: Subtracting the threshold-constrained evapotranspiration response estimate from the linear evapotranspiration response estimate yields the evapotranspiration response bias corresponding to the forest cover change to be evaluated. When the evapotranspiration response deviation is greater than zero, it is determined that the linear response model overestimates the evapotranspiration response relative to the threshold response model. When the evapotranspiration response deviation is less than zero, it is determined that the linear response model underestimates the evapotranspiration response relative to the threshold response model. When the evapotranspiration response deviation is equal to zero, it is determined that the linear response model and the threshold response model have no difference in their estimation of the evapotranspiration response.

Images