A method and system for quantifying the temporal dynamics of surface temperature response to forest loss

Abstract

The present invention belongs to the technical field of forest loss assessment, and relates to a method and system for quantifying the temporal dynamics of the response of land surface temperature to forest loss. The method includes: extracting the trend component and seasonal variation component in the LST time series, and detecting the mutation of the LST trend component; spatiotemporally matching the mutation of the LST trend component with forest loss events; adopting a method combining space and time to select forest loss pixels and background forest pixels; using the mutation characteristic value and the gradual change characteristic value of the trend component of LST to quantify the long-term response of LST to forest loss. By matching the mutation of the LST trend component with forest loss events, the present invention sets a forest loss window and its subsequent temperature response interval for each forest loss event, ensuring that the observed LST dynamics are only related to forest loss and reducing the uncertainty of LST response quantification.

Description

Technical Field

[0001] The present invention belongs to the technical field of forest loss assessment. Specifically, it relates to a method and system for quantifying the temporal dynamics of the response of land surface temperature to forest loss. Background Art

[0002] Forest loss affects the land surface temperature (LST), i.e., the biophysical effect, by changing the radiative flux and non-radiative processes between the atmosphere and the land surface through altering the land surface physical properties (such as land surface albedo and roughness). Generally, it is considered that the loss of boreal forests reduces the land surface temperature due to the decrease in downward shortwave radiation, while the loss of tropical forests increases the land surface temperature due to the decrease in evapotranspiration. Understanding the climate effects and biophysical mechanisms of forest loss is of great significance for alleviating local climate change by guiding land radiation management and natural climate solutions.

[0003] Satellite remote sensing observation data can provide direct evidence for the response of land surface temperature to forest loss and are widely used to evaluate the biophysical effects of forest loss at the regional and even global scales. Currently, the methods for quantifying such biophysical effects based on satellite remote sensing observation data are divided into:

[0004] (1) Pairwise comparison of space for time: For example, extracting the spatial differences in land surface temperature between forests with different vegetation coverages or between forests and farmlands, forests and grasslands to approximately simulate the impact of deforestation or other land uses on local temperature. However, the biophysical effects extracted by this method are estimates under ideal conditions. Considering the patch distribution characteristics of different land cover types on the land surface and the differences and complexities of terrain, there are certain biases in the temperature responses obtained by such methods;

[0005] (2) Methods combining space and time: Such methods focus on real forest loss events, regarding the difference in land surface temperature before and after forest loss as the combined influence of the biophysical effect brought by forest loss and climate change, and extracting the response of land surface temperature to forest loss by removing the temperature changes of background stable forest pixels during the forest loss period (which can be regarded as the influence of climate change);

[0006] The method of combining space with time focuses on the static differences before and after forest cover change, confusing the sudden response of LST caused by forest loss events and the slow response of LST caused by subsequent land surface succession, resulting in the response of LST being highly dependent on the time setting after forest loss in static comparison. The method of substituting space for time only considers the differences between stable forest and non-forest cover, ignoring the long-term dynamic changes in the LST response caused by forest loss and its subsequent land surface succession. Due to inconsistent experimental settings or method differences in a single method, the states after forest loss or stable non-forest cover involved in existing studies are at different evolution stages after forest loss, so the signs and magnitudes of the LST response to forest loss quantified in different studies are inconsistent. Summary of the Invention

[0007] To solve the above technical problems, the present invention provides a method and system for quantifying the temporal dynamics of land surface temperature response to forest loss.

[0008] In the first aspect, the present invention provides a method for quantifying the temporal dynamics of land surface temperature response to forest loss, including:

[0009] Extracting the trend component and seasonal change component in the LST time series, and detecting the mutation of the trend component;

[0010] Selecting the forest loss pixels that trigger the mutation of the trend component of LST to evaluate the impact of forest loss on LST, performing spatio-temporal matching between the mutation of the trend component of LST and the forest loss event, and specifying the paired time before and after loss and the long-term dynamic monitoring time period for each forest loss event;

[0011] Adopting the method of combining space with time, selecting the forest loss pixels and the background forest pixels of the forest loss pixels, and calculating the mutation characteristic value and the gradual change characteristic value of the trend component of LST;

[0012] Quantifying the long-term response of LST to forest loss by using the mutation characteristic value and the gradual change characteristic value of the trend component of LST.

[0013] In the second aspect, the present invention provides a system for quantifying the temporal dynamics of land surface temperature response to forest loss, including an extraction and detection unit, a matching unit, a selection and analysis unit, and a processing unit;

[0014] The extraction and detection unit is used for extracting the trend component and seasonal change component in the LST time series, and detecting the mutation of the trend component;

[0015] The matching unit is used for selecting the forest loss pixels that trigger the mutation of the trend component of LST to evaluate the impact of forest loss on LST, performing spatio-temporal matching between the mutation of the trend component of LST and the forest loss event, and specifying the paired time before and after loss and the long-term dynamic monitoring time period for each forest loss event;

[0016] A selection and analysis unit is used to select forest loss pixels and background forest pixels of forest loss pixels by using a method that combines space and time, and calculate the mutation eigenvalue and the gradual change eigenvalue of the trend component of LST.

[0017] A processing unit is used to quantify the long-term response of LST to forest loss by using the mutation eigenvalue of the trend component of LST and the gradual change eigenvalue of the trend component.

[0018] Based on the above technical solutions, the present invention can also be improved as follows.

[0019] Furthermore, the long-term trend in the LST time series is adopted by using a piecewise linear function, and the seasonal variation in the LST time series is simulated by using multiple second harmonic functions. Let the LST of the time series be LST t , the trend component be T t , the seasonal component be S t , the residual part be e t , the LST of the time series is the sum of the trend component, the seasonal component and the residual part, t is time, n is the number of LST observations, t = 1,..., n, T t i is a linear function for simulating the trend component of LST, is a second harmonic function of the seasonal component, m is the number of breakpoints, and the trend component of LST is composed of m + 1 linear functions T t i simulates, i is the number of segments of the linear function, i = 1,…, m + 1, a i is the slope of the linear function of the i-th segment, b i is the intercept of the linear function of the i-th segment, and the LST seasonal component is composed of multiple second harmonic functions is fitted, q + 1 is the number of harmonic functions, is the seasonal breakpoint, the amplitude of the j-th segment harmonic function is the phase of the j-th segment harmonic function is tan -1 (σ j,h / γ j,h ), γ j,h and σ j,h are parameters that determine the amplitude and phase (tan -1 (σ j,h / γ j,h )) of the j-th segment harmonic function, h is the order of the harmonic function, λ is the period of the LST time series, the seasonal component is fitted by several second harmonic functions, and the trend component of LST is simulated by m + 1 linear functions, then:

[0020] LST t = Tt +S t +e t , where \(t = 1,\ldots,n\);

[0021]

[0022] Furthermore, the conditions for spatiotemporally matching the trend component mutation of LST with forest loss events include: the forest loss occurs within the confidence range of the trend component mutation of LST detected by the change detection method, and the time difference between the forest loss and the trend component mutation of LST does not exceed a set number of years; if the detected trend component mutation of LST matches the forest loss event, the forest loss window of the forest loss pixel is defined as the union of the year of the trend component mutation of LST and the year of the forest loss event.

[0023] Furthermore, the background forest pixels controlled by each forest loss pixel are selected, and the background forest pixels satisfy: the background forest pixels are within the search window at a set distance range from the forest loss pixel, the forest cover change rate of the background forest pixels is less than a set value, and at the same time less than the loss percentage of the forest loss pixel, the elevation difference between the background forest pixels and the forest loss pixel is less than a set value, and the number of background forest pixels exceeds a set ratio of all background pixels.

[0024] Furthermore, for each forest loss pixel, the change in LST caused by forest loss is obtained by calculating the difference between the change value of LST observed by the forest loss pixel and the change value of LST observed by the background forest pixels during the same period.

[0025] Furthermore, let the trend component of LST after forest loss be \(\Delta T\), the target forest loss pixel be \(o\), the background stable forest pixel be \(bg\), and the trend component of the forest loss pixel after loss be \(T\) o,after , the trend component of the forest loss pixel one year before loss be \(T_1\) o,before , the change in the trend component one year after forest loss be \(\Delta T_1\), and the annual scale difference between the trend component of the forest loss pixel after loss and the trend component of the forest loss pixel one year before loss be \(\delta T\) o , the average value of the differences in the trend components of all background forest pixels be \(\overline{\delta T}\) bg , then the time series change of the trend component of LST after forest loss is expressed as \(\Delta T\); the change \(\Delta T_1\) in the trend component one year after forest loss is selected to represent the immediate impact of forest loss on LST, and the calculation formula is:

[0026] \(\Delta T=\overline{\delta T}\) o -\overline{\delta T}\) bg ;

[0027] \(\overline{\delta T}\) o =T o,after -T_1o,before .

[0028] Furthermore, let the linear slope of the trend component of the LST of the forest loss pixels before the loss be T_slope o,before , and the linear slope of the trend component of the LST of the forest loss pixels after the loss be T_slope o,after , and the change in the linear slope of the trend component of the LST of the forest loss pixels after and before the loss be δT_slope o , and the linear slope from after the loss to the next mutation or the end of the time series be T_slope o,after , and the linear slope of the trend component of the LST of the background forest pixels before the loss be T_slope bg,before , and the linear slope of the trend component of the LST of the background forest pixels after the loss be T_slope bg,after , and the difference in the linear slope of the trend component of the LST of the background forest pixels after and before the loss be δT_slope bg , the forest loss pixels have the same climate signal as the background forest pixels before the loss, i.e., T_slope o,before = T_slope bg,before , the change in the linear slope of the trend component of the LST after the loss is ΔT_slope, and the gradual change of the trend component of the LST is quantified by the change in the linear slope of the trend component of the LST after the loss, i.e.:

[0029] ΔT_slope = δT_slope o - δT_slope bg ;

[0030] δT_slope o = T_slope o,after - T_slope o,before .

[0031] The beneficial effects of the present invention are as follows:

[0032] (1) By applying the change detection method to the mechanism of space combined with time, the present invention can distinguish the sudden impact of forest loss on LST and the response of LST to subsequent terrestrial vegetation evolution, depict the long-term dynamic pattern of LST response, and help to reconcile the inconsistent results in the existing research on the impact of forest loss;

[0033] (2) By matching the mutation of the trend component of LST with the forest loss event, the present invention sets a forest loss window and its subsequent temperature response interval for each forest loss event, ensuring that the observed LST dynamics are only related to forest loss, which increases the comparability of LST responses among different pixels and reduces the uncertainty of LST response quantification;

[0034] (3) The method of the present invention for jointly detecting changes and quantifying the temperature response by combining spatial and temporal aspects has generalizability and is applicable to any land use and land cover change, and can effectively evaluate the biophysical effects of these disturbances on the local climate;

[0035] (4) Since the climate-forest loss feedback of the present invention is climate-specific, loss type-dependent, and time-varying, it helps to improve the land surface model to more accurately predict the climate, which is of great significance for making correct decisions on intervention measures against the intensification of climate change. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic diagram of a method for quantifying the temporal dynamics of the surface temperature response to forest loss provided in Embodiment 1 of the present invention;

[0037] Figure 2 It is a schematic diagram of the result of LST change detection;

[0038] Figure 3 It is a simulation diagram of the long-term dynamic response of LST to multiple forest loss types in different climate regions of the world;

[0039] Figure 4 It is a schematic diagram of a system for quantifying the temporal dynamics of the surface temperature response to forest loss provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention described and illustrated herein can generally be arranged and designed in a variety of different configurations.

[0041] Embodiment 1

[0042] As an embodiment, as shown in the attached... Figure 1 To solve the above technical problems, this embodiment provides a method for quantifying the temporal dynamics of the surface temperature response to forest loss, including:

[0043] Extract the trend component and seasonal change component in the LST time series, and detect the mutation of the trend component;

[0044] Select the forest loss pixels that cause the mutation of the trend component of LST to evaluate the impact of forest loss on LST, spatiotemporally match the mutation of the trend component of LST with the forest loss events, and assign paired times before and after the loss and a long-term dynamic monitoring time period to each forest loss event;

[0045] Adopt a method that combines space and time, select forest loss pixels and the background forest pixels of forest loss pixels, and calculate the mutation eigenvalue and the gradual change eigenvalue of the trend component of LST;

[0046] Use the mutation eigenvalue and the gradual change eigenvalue of the trend component of LST to quantify the long-term response of LST to forest loss.

[0047] The present invention proposes a method for jointly detecting changes and combining space and time to quantify the response of LST to forest loss. First, use the change detection method to extract the trend component in the LST time series retrieved by satellite remote sensing and detect the mutation of the trend component of LST. By spatiotemporally matching the mutation of the trend component of LST with forest loss events, assign paired times before and after the loss and the long-term dynamic monitoring time period for each forest loss event. Secondly, adopt a method that combines space and time, and use the mutation and gradual change of the trend component of LST to quantify the long-term response of LST to forest loss.

[0048] Optionally, adopt the long-term trend in the LST time series with a piecewise linear function, use multiple second harmonic functions to simulate the seasonal changes in the LST time series. Let the LST of the time series be LST t , the trend component be T t , the seasonal component be S t , the residual part be e t , the LST of the time series is the sum of the trend component, the seasonal component and the residual part. t is the time, n is the number of LST observations, t = 1,..., n, T t i is a linear function for simulating the trend component of LST, is a second harmonic function of the seasonal component, m is the number of breakpoints, the trend component of LST is simulated by m + 1 linear functions T t i , i is the number of segments of the linear function, i = 1,…, m + 1, a i is the slope of the linear function of the i-th segment, b i is the intercept of the linear function of the i-th segment, the LST seasonal component is fitted by multiple second harmonic functions , q + 1 is the number of harmonic functions, is the seasonal breakpoint, the amplitude of the j-th segment harmonic function is the phase of the j-th segment harmonic function is tan -1 (σ j,h / γ j,h ), γ j,h and σ j,h are to determine the amplitude and phase (tan -1 (σ j,h / γj,h The parameters of ( )), h is the order of the harmonic function, λ is the period of the LST time series, the seasonal component is fitted by several second-order harmonic functions, and the trend component of LST is simulated by m + 1 linear functions. Then:

[0049] LST t = T t + S t + e t , where t = 1, …, n;

[0050]

[0051] Select the method of decomposing the trend and seasonal components and detecting mutations based on intensive time series (The improved Breaks For Additive Seasonal and Trend method, improved BEAST) to extract the long-term trend of the LST time series, extract the trend component and seasonal change component in the LST time series, and detect the mutations of the trend component and seasonal change component. Apply the improved BEAST to the global LST time series observed by MODIS, decompose the LST into trend and seasonal components, and detect the first three most significant mutations (i.e., m = q = 3). As shown in the attached Figure 2 Figure showing the result of LST change detection for an example pixel. The horizontal axis is t, unit: year, and the vertical axis is the LST observation value Y of the time series t is decomposed into the trend component T t , the seasonal component S t and the remainder e t . The detected mutations are shown as vertical lines, and the confidence intervals of these mutations are represented by the time periods between the two vertical lines.

[0052] Optionally, the conditions for spatio-temporally matching the trend component mutation of LST with the forest loss event include: the forest loss occurs within the confidence range of the trend component mutation of LST detected by the change detection method, and the time difference between the forest loss and the trend component mutation of LST does not exceed the set number of years; if the detected trend component mutation of LST matches the forest loss event, the forest loss window of the forest loss pixel is defined as the union of the year of the trend component mutation of LST and the year of the forest loss event.

[0053] Select the forest loss pixels that trigger the LST trend mutation to evaluate the impact of forest loss on LST. The forest loss data is sourced from the global high-resolution forest map. This map extracts the global forest extent and the forest loss and forest growth from 2000 to 2021 through time series analysis of Landsat satellite data. Exemplarily, the matching criteria for LST trend mutation and forest loss event are as follows:

[0054] (a) Forest loss occurred within the confidence range of the LST trend mutation detected by the change detection method;

[0055] (b) The time difference between forest loss and the LST trend mutation does not exceed 2 years;

[0056] (c) For a certain pixel, if the detected LST mutation matches the forest loss event, the forest loss window is defined as the union of the LST trend mutation year and the forest loss event year. Therefore, the research period for this pixel starts from the year before the loss window and ends until the next mutation or the end of the time series, so as to explore the evolution of the LST response for up to 14 years.

[0057] As shown in the appendix Figure 3 If the sample pixel experienced deforestation in 2005 and the LST trend component mutated in 2005, the forest loss window for this pixel spans from the beginning of 2005 to the end of 2006.

[0058] Optionally, select the background forest pixels controlled by each forest loss pixel. The background forest pixels meet the following conditions: the background forest pixels are located within the search window with a set distance range from the forest loss pixel, the forest cover change rate of the background forest pixels is less than the set value and less than the loss percentage of the forest loss pixel, the elevation difference between the background forest pixels and the forest loss pixel is less than the set value, and the number of background forest pixels exceeds the set ratio of all background pixels.

[0059] Under the assumption that the climate change signal is consistent within the regional scope, estimate the climate signal of the forest loss pixel through the LST change of adjacent areas with stable forest cover (i.e., background forest pixels). The background forest pixels used to control each forest loss pixel should meet the following conditions:

[0060] (1) Located within the search window 25 km - 50 km away from the forest loss pixel;

[0061] (2) Having a stable forest land coverage rate, with the forest change percentage less than the minimum value of 2% and the loss percentage of the forest loss pixel;

[0062] (3) The elevation difference from the forest loss pixel is less than 100 m. In addition, the number of background forest pixels should exceed 5% of all background pixels to ensure reliable statistical analysis. Otherwise, exclude the paired samples from the evaluation.

[0063] The land cover map of the forest land coverage rate is derived from the global land cover map with a 300 m spatial resolution provided by the European Space Agency's Global Land Cover data, and the elevation data is from the Shuttle Radar Topography Mission with a resolution of 90 m.

[0064] Optionally, for each forest loss pixel, the LST change caused by forest loss is obtained by calculating the difference between the LST change value observed for the forest loss pixel and the LST change value observed for the background forest pixels during the same period.

[0065] Optionally, let the trend component of LST after forest loss be ΔT, the target forest loss pixel be o, the background stable forest pixel be bg, and the trend component of the forest loss pixel after loss be T o,after , and the trend component of the forest loss pixel one year before the loss be T_1 o,before , the change in the trend component one year after the forest loss be ΔT_1, and the annual-scale difference between the trend component of the forest loss pixel after the loss and the trend component of the forest loss pixel one year before the loss be δT o , and the average value of the differences in the trend components of all background forest pixels be δT bg , then the time series change in the trend component of LST after forest loss is expressed as ΔT; the change in the trend component one year after the forest loss, ΔT_1, is selected to represent the immediate impact of forest loss on LST, and the calculation formula is:

[0066] ΔT = δT o - δT bg ;

[0067] δT o = T o,after - T_1 o,before .

[0068] Optionally, let the linear slope of the trend component of LST of the forest loss pixel before the loss be T_slope o,before , and the linear slope of the trend component of LST of the forest loss pixel after the loss be T_slope o,after , and the change in the linear slope of the trend component of LST of the forest loss pixel after and before the loss be δT_slope o , and the linear slope from after the loss to the next mutation or the end of the time series be T_slope o,after , and the linear slope of the trend component of LST of the background forest pixel before the loss be T_slope bg,before , and the linear slope of the trend component of LST of the background forest pixel after the loss be T_slope bg,after , and the difference in the linear slope of the trend component of LST of the background forest pixel after and before the loss be δT_slope bg , and the forest loss pixel has the same climate signal as the background forest pixel before the loss, that is, T_slope o,before = T_slope bg,before, the linear slope change of the trend component of LST after loss is ΔT_slope, and the gradual change of the trend component of LST is quantified by the linear slope change of the trend component of LST after loss, that is:

[0069] ΔT_slope = δT_slope o -δT_slope bg ;

[0070] δT_slope o = T_slope o,after -T_slope o,before .

[0071] The method for quantifying the response of LST to forest loss based on joint change detection and space - combined - time evaluated the long - term dynamic response of LST to multiple forest loss types in different climate regions of the world. The climate region type division includes the mid - latitudes of the Southern Hemisphere (SM, 60°S - 20°S), low latitudes (LL, 20°S - 20°N), mid - latitudes of the Northern Hemisphere (NM, non - boreal climate regions above 20°N), and the boreal region (boreal, boreal climate regions above 20°N); forest loss types include commodity - driven forest - to - farmland loss (sCRO), urbanization (URB), temporary transfer from forest to farmland (shiftAG), forestry management (forestry), and fire. As shown in the Figure 3 simulation diagram of the long - term dynamic response of LST to multiple forest loss types in different climate regions of the world. The horizontal axis is the time after loss, unit: year, and the vertical axis is the trend component ΔT of LST after forest loss in the mid - latitudes of the Southern Hemisphere (SM), low latitudes (LL), mid - latitudes of the Northern Hemisphere (NM), and the boreal region (boreal), unit: K. The permanent forest loss driven by forest - to - farmland loss is the main factor causing warming, with the largest warming amplitude. The trend component mutation ΔT_1 values of the trend components of LST in the LL and SM latitudes are +0.39K and +0.41K respectively; in the long run, ΔT shows an increasing trend. URB usually leads to enhanced warming, with the highest ΔT_slope: for example, the ΔT_slope of URB in the northern latitude region is +0.03K / 10y; the linear slope change ΔT_slope of the trend component of LST after loss of URB in the boreal region is +0.08K / 10y.

[0072] Temporary forest loss results in multiple dynamic patterns of ΔT. Temporary forest loss in the NM region leads to a weakening of warming. Specifically, the temporary transfer from forest to farmland, forestry management, and fires cause a sudden increase in ΔT, followed by a recovery trend (i.e., ΔT_1>0, ΔT_slope<0), indicating that this warming weakens as the forest recovers. Among these disturbances, fires exhibit the most significant abrupt warming (ΔT_1 = +0.30K) and the most obvious downward trend (ΔT_slope = -0.14K / 10y). In the northern region, due to forest recovery, the temporary transfer from forest to farmland results in a recovery trend of negative ΔT (i.e., weakening cooling). However, forestry management and fires do not exhibit a recovery dynamic. Forestry has an enhanced cooling effect with a ΔT_slope of -0.16K / 10y; compared with forestry management, fires initially cause a weakening of warming and then enhance cooling, and fires have a greater downward trend (ΔT_slope = -0.34k / 10y). Forestry and fires in the northern region cause ΔT to continuously deviate from zero, indicating lower resilience in this region compared to low and mid latitudes.

[0073] By applying the change detection method to the spatial-temporal mechanism, the present invention can distinguish the sudden impact of forest loss on LST and the response of LST to subsequent terrestrial vegetation evolution, characterize the long-term dynamic pattern of LST response, and help reconcile inconsistent results in existing studies on the impact of forest loss;

[0074] By matching the trend component mutation of LST with forest loss events, the present invention sets a forest loss window and its subsequent temperature response interval (i.e., from the year before the loss to the end of the time series or the next sudden change) for each forest loss event, ensuring that the observed LST dynamics are only related to forest loss, which increases the comparability of LST responses among different pixels and reduces the uncertainty in LST response quantification;

[0075] The method of the present invention for jointly detecting changes and quantifying temperature responses in a spatial-temporal manner is generalizable and applicable to any land use and land cover change, and can effectively evaluate the biophysical effects of these disturbances on local climate;

[0076] Based on the fact that the climate-forest loss feedback is climate-specific, loss type-dependent, and time-varying, the present invention helps to improve land surface models to more accurately predict climate, which is of great significance for making correct decisions on intervention measures against climate change intensification.

[0077] Example 2

[0078] Based on the same principle as the method shown in Example 1 of the present invention, as shown in the appendix Figure 4As shown, in an embodiment of the present invention, a system for quantifying the temporal dynamics of the surface temperature response to forest loss is further provided, including an extraction and detection unit, a matching unit, a selection and analysis unit, and a processing unit;

[0079] The extraction and detection unit is used to extract the trend component and seasonal change component in the LST time series, and detect the trend component mutation;

[0080] The matching unit is used to select the forest loss pixels that trigger the trend component mutation of LST to evaluate the impact of forest loss on LST, perform spatio-temporal matching between the trend component mutation of LST and the forest loss event, and assign the paired times before and after the loss and the long-term dynamic monitoring time period for each forest loss event;

[0081] The selection and analysis unit is used to adopt a method combining space and time to select the forest loss pixels and the background forest pixels of the forest loss pixels, and calculate the mutation eigenvalue and the gradual change eigenvalue of the trend component of LST;

[0082] The processing unit is used to quantify the long-term response of LST to forest loss by using the mutation eigenvalue and the gradual change eigenvalue of the trend component of LST.

[0083] Optionally, the long-term trend in the LST time series is adopted by a piecewise linear function, and multiple second harmonic functions are used to simulate the seasonal changes in the LST time series. Let the LST of the time series be LST t , the trend component be T t , the seasonal component be S t , the residual part be e t , the LST of the time series is the sum of the trend component, the seasonal component and the residual part. t is the time, n is the number of LST observations, t = 1,..., n, T t i is a linear function for simulating the trend component of LST, is a second harmonic function of the seasonal component, m is the number of breakpoints, and the trend component of LST is simulated by m + 1 linear functions T t i , i is the number of segments of the linear function, i = 1,…, m + 1, a i is the slope of the i-th segment of the linear function, b i is the intercept of the i-th segment of the linear function, and the LST seasonal component is fitted by multiple second harmonic functions , q + 1 is the number of harmonic functions, is the seasonal breakpoint, the amplitude of the j-th segment of the harmonic function is The phase of the j-th segment of the harmonic function is tan -1 (σ j,h / γ j,h ), γ j,hand σ j,h are parameters that determine the amplitude of the j-th harmonic function and the phase (tan -1 (σ j,h / γ j,h )). h is the order of the harmonic function, λ is the period of the LST time series, the seasonal component is fitted by several second-order harmonic functions, and the trend component of LST is simulated by m + 1 linear functions. Then:

[0084] LST t = T t + S t + e t , t = 1, …, n;

[0085]

[0086] Optionally, the conditions for spatiotemporally matching the trend component mutation of LST with the forest loss event include: the forest loss occurs within the confidence range of the trend component mutation of LST detected by the change detection method, and the time difference between the forest loss and the trend component mutation of LST does not exceed the set number of years; if the detected trend component mutation of LST matches the forest loss event, the forest loss window of the forest loss pixel is defined as the union of the year of the trend component mutation of LST and the year of the forest loss event.

[0087] Optionally, the background forest pixels controlled by each forest loss pixel are selected. The background forest pixels satisfy: the background forest pixels are within the search window within the set distance range from the forest loss pixel, the forest cover change rate of the background forest pixels is less than the set value, and at the same time less than the loss percentage of the forest loss pixel, the elevation difference between the background forest pixels and the forest loss pixel is less than the set value, and the number of background forest pixels exceeds the set ratio of all background pixels.

[0088] Optionally, for each forest loss pixel, the change in LST caused by the forest loss is obtained by calculating the difference between the change value of LST observed in the forest loss pixel and the change value of LST observed in the background forest pixels during the same period.

[0089] Optionally, let the trend component of LST after the forest loss be ΔT, the target forest loss pixel be o, the background stable forest pixel be bg, the trend component of the forest loss pixel after the loss be T o,after , the trend component of the forest loss pixel one year before the loss be T_1 o,before , the change in the trend component one year after the forest loss be ΔT_1, the annual-scale difference between the trend component of the forest loss pixel after the loss and the trend component of the forest loss pixel one year before the loss be δT[[ID=4l]] o , and the average value of the differences in the trend components of all background forest pixels be δTbg Then, the time series change of the trend component of LST after forest loss is denoted as ΔT; the change in the trend component ΔT_1 in the year after forest loss is selected to represent the immediate impact of forest loss on LST, and the calculation formula is:

[0090] ΔT = δT o -δT bg ;

[0091] δT o = T o,after -T_1 o,before .

[0092] Optionally, let the linear slope of the trend component of LST of the forest loss pixel before loss be T_slope o,before , and the linear slope of the trend component of LST of the forest loss pixel after loss be T_slope o,after , the change in the linear slope of the trend component of LST of the forest loss pixel after and before loss be δT_slope o , the linear slope from after loss to the next mutation or the end of the time series be T_slope o,after , the linear slope of the trend component of LST of the background forest pixel before loss be T_slope bg,before , and the linear slope of the trend component of LST of the background forest pixel after loss be T_slope bg,after , the difference in the linear slope of the trend component of LST of the background forest pixel after and before loss be δT_slope bg , the forest loss pixel has the same climate signal as the background forest pixel before loss, that is, T_slope o,before = T_slope bg,before , the change in the linear slope of the trend component of LST after loss be ΔT_slope, and the gradual change of the trend component of LST is quantified by the change in the linear slope of the trend component of LST after loss, that is:

[0093] ΔT_slope = δT_slope o -δT_slope bg ;

[0094] δT_slope o = T_slope o,after -T_slope o,before .

[0095] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method to quantify the temporal dynamics of surface temperature response to forest loss, characterized in that include: Extract trend components and seasonal change components from LST time series, and detect sudden changes in trend components; The trend and seasonal component decomposition and mutation detection method based on dense time series is applied to the LST time series observed by satellite. A piecewise linear function is used to simulate the long-term trend in the LST time series, and multiple second-order harmonic functions are used to simulate the seasonal variation in the LST time series. The impact of forest loss on LST was assessed by selecting forest loss pixels that triggered a sudden change in the trend component of LST. The sudden change in the trend component of LST was spatially and temporally matched with forest loss events. For each forest loss event, a pairing time before and after the loss and a long-term dynamic monitoring period were assigned. Using the method of combining space and time, forest loss pixels and background forest pixels of forest loss pixels were selected to calculate the sudden change characteristic value and gradual change characteristic value of the trend component of LST; The abrupt change characteristic value and gradual change characteristic value of the trend component of LST were used to quantify the long-term response of LST to forest loss.

2. A method for quantifying the temporal dynamics of surface temperature response to forest loss according to claim 1, characterized in that: The long-term trend in the LST time series is simulated using piecewise linear functions, and the seasonal variation in the LST time series is simulated using multiple second-order harmonic functions. Let the LST of the time series be LST t , the trend group is divided into T t , seasonal groups are divided into S t , the residual is e t The LST of the time series is the sum of the trend component, seasonal component and residual part, t is the time, n is the number of LST observations, t=1,...,n, T t i is a linear function simulating the trend component of LST, S t j is the second-order harmonic function of the seasonal component, m is the number of breakpoints, and the trend component of LST is composed of m+1 linear functions T t i Simulation, i is the number of segments of the linear function, i=1,…,m+1, a i is the slope of the linear function of the i-th segment, b i is the intercept of the linear function of the i-th segment. The LST seasonal component is composed of multiple second-order harmonic functions S t j fitting, q+1 is the number of harmonic functions, is the seasonal breakpoint, and the amplitude of the j-th harmonic function is The phase of the jth harmonic function is tan -1 (σ j,h / γ j,h ), γ j,h and σ j,h It determines the amplitude of the jth harmonic function and phase (tan -1 (σ j,h / γ j,h )), h is the order of the harmonic function, λ is the period of the LST time series, the seasonal component is fitted by several second-order harmonic functions, and the trend component of LST is simulated by m+1 linear functions, then: LST t =T t +S t +e t ,t=1,…,n; 3. A method for quantifying the temporal dynamics of surface temperature response to forest loss according to claim 1, characterized in that: The conditions for temporally and spatially matching LST trend component mutations with forest loss events include: forest loss occurs within the confidence range of LST trend component mutations detected by the change detection method, and the time difference between forest loss and LST trend component mutations does not exceed the set number of years; if the detected LST trend component mutation matches the forest loss event, the forest loss window of the forest loss pixel is defined as the union of the LST trend component mutation year and the forest loss event year.

4. A method for quantifying the temporal dynamics of surface temperature response to forest loss according to claim 1, characterized in that: Background forest pixels controlled by each forest loss pixel are selected, and the background forest pixels meet the following requirements: the background forest pixel is located within the search window of the set distance range from the forest loss pixel, the forest cover change rate of the background forest pixel is less than the set value and is also less than the loss percentage of the forest loss pixel, the altitude difference between the background forest pixel and the forest loss pixel is less than the set value, and the number of background forest pixels exceeds the set ratio of all background pixels.

5. A method for quantifying the temporal dynamics of surface temperature response to forest loss according to claim 1, characterized in that: For each forest loss pixel, the LST change caused by forest loss was obtained by calculating the difference between the LST change value observed in the forest loss pixel and the LST change value observed in the background forest pixel during the same period.

6. A method for quantifying the temporal dynamics of surface temperature response to forest loss according to claim 5, characterized in that: Assume that the trend component of LST after forest loss is ΔT, the target forest loss pixel is o, the background stable forest pixel is bg, and the trend component of the forest loss pixel after loss is T o,after The trend group of forest loss pixels one year before the loss is T_1 o ,before The change in the trend component one year after forest loss is ΔT_1, and the annual scale difference between the trend component of the forest loss pixel after the loss and the trend component of the forest loss pixel one year before the loss is δT o The average difference in trend components of all background forest pixels is δT bg , then the time series change of the trend component of LST after forest loss is expressed as ΔT; the change of the trend component one year after forest loss ΔT_1 is selected to represent the immediate impact of forest loss on LST, and the calculation formula is: ΔT=δT o -δT bg ; δT o =T o,after -T_1 o,before 。 7. A method for quantifying the temporal dynamics of surface temperature response to forest loss according to claim 5, characterized in that: Assume that the linear slope of the trend component of the forest loss pixel LST before loss is T_slope o,before The linear slope of the trend component of the forest loss pixel LST after loss is T_slope o,after The linear slope of the trend component of the forest loss pixel LST before and after the loss is δT_slope o , the linear slope from the loss to the next mutation or the end of the time series is T_slope o,after The linear slope of the trend component of the background forest pixel LST before loss is T_slope bg,before , the linear slope of the trend component of the background forest pixel LST after loss is T_slope bg,after The difference between the linear slope of the trend component of the background forest pixel LST before and after the loss is δT_slope bg , the forest loss pixel has the same climate signal as the background forest pixel before the loss, that is, T_slope o,before =T_slope bg,before , the linear slope change of the trend component of LST after the loss is ΔT_slope, and the trend component gradient of LST is quantified by the linear slope change of the trend component of LST after the loss, that is: ΔT_slope=δT_slope o -δT_slope bg ; δT_slope o =T_slope o,after -T_slope o,before 。 8. A system for quantifying the temporal dynamics of land surface temperature responses to forest loss, characterized in that It includes extraction and detection unit, matching unit, selection and analysis unit and processing unit; Extraction and detection unit, used to extract trend components and seasonal change components in LST time series, and detect sudden changes in trend components; The trend and seasonal component decomposition and mutation detection method based on dense time series is applied to the LST time series observed by satellite. A piecewise linear function is used to simulate the long-term trend in the LST time series, and multiple second-order harmonic functions are used to simulate the seasonal variation in the LST time series. The matching unit is used to select forest loss pixels that trigger a sudden change in the trend component of LST to evaluate the impact of forest loss on LST, match the sudden change in the trend component of LST with the forest loss event in time and space, and specify the pairing time before and after the loss and the long-term dynamic monitoring period for each forest loss event; Selection and analysis units are used to select forest loss pixels and background forest pixels of forest loss pixels using a spatial-temporal method to calculate the sudden change characteristic value and gradual change characteristic value of the trend component of LST; The processing unit is used to quantify the long-term response of LST to forest loss using the abrupt change characteristic value and the gradual change characteristic value of the trend component of LST.

Images