Construction of an evaluation model and prediction method based on subtropical forest fire intensity
By acquiring survey data on subtropical forest fire intensity and processing remote sensing images, combined with a random forest model, the problem of assessing and predicting fire intensity in subtropical regions has been solved, enabling accurate assessment and prediction of the spatial distribution of forest fire intensity and improving the scientific nature of fire management.
Patent Information
- Application Number
- CN202211656160.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-12-22
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2042-12-22
AI Technical Summary
In subtropical regions, there is limited research on the spatial patterns of fire intensity, and existing technologies are insufficient to effectively assess and predict the influencing factors of forest fire intensity, especially the differences under different geographical environments and vegetation conditions.
By acquiring survey sample data on wildfire intensity in subtropical forests, radiometric calibration and atmospheric correction were performed using remote sensing images, remote sensing spectral indices were calculated, a comprehensive combustion index regression model was established, and the impact of environmental factors on the spatial distribution pattern of forest fire intensity was assessed and predicted using a random forest model.
It has enabled accurate assessment of subtropical forest fire intensity and prediction of spatial distribution patterns, provided a theoretical basis for post-fire vegetation restoration and management, filled a research gap in subtropical regions, and improved the scientific nature of fire management.
Smart Images

Figure CN115994618B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of fire intensity evaluation, and in particular relates to an evaluation model construction and prediction method based on subtropical forest fire intensity. Background Art
[0002] Forest fire intensity refers to the impact or damage of fire on forest ecosystems, which directly affects various ecological processes, such as vegetation succession and recovery, C and N cycle, and decomposition of dead branches and leaves. Quantitative evaluation of forest fire intensity helps to reveal the development and changes of various ecological processes in forest ecosystems after forest fire disturbance and the formation mechanism of forest landscape pattern.
[0003] Forest fire intensity usually shows spatial heterogeneity in the landscape, which profoundly affects various ecological processes in the forest landscape. It is generally believed that forest fire intensity at the regional scale is usually affected by meteorological factors, while forest fire intensity at the local scale is mainly affected by terrain and vegetation type. Quantitative analysis of the spatial pattern of forest fire intensity helps to provide certain theoretical basis and data support for forest fire prevention and management. At present, the spatial pattern of fire intensity is carried out in the cold temperate forest area, where the fire area is large and the intensity is high. In contrast, the fire area in the subtropical evergreen broad-leaved forest area is relatively small and the fire intensity is lower. Studies have shown that the spatial pattern of fire intensity is often different under different geographical environments and vegetation conditions. However, there are few reports on the spatial pattern of fire intensity in subtropical forest areas and the relative influence of vegetation and terrain on it. Summary of the invention
[0004] In order to evaluate the fire intensity of subtropical forests and obtain the influencing factors of the spatial distribution pattern of the fire intensity of subtropical forests, the present invention provides the following scheme: a method for constructing an evaluation model and predicting the fire intensity of subtropical forests, comprising:
[0005] Obtain survey sample data and initial remote sensing images of wild fire intensity in subtropical forests; randomly group the survey sample data in a ratio of 7:3 to obtain a first group of survey sample data and a second group of survey sample data; perform radiation calibration and atmospheric correction processing on the initial remote sensing images to obtain target remote sensing images;
[0006] Calculate and obtain a remote sensing spectral index based on the target remote sensing image; establish a relationship model between a comprehensive combustion index regression model and the remote sensing spectral index based on the first group of survey sample data;
[0007] Evaluate the fitting effect of the relationship module on the comprehensive combustion index regression model and the remote sensing spectral index according to preset evaluation indicators, and determine the optimal relationship model and the optimal remote sensing spectral index;
[0008] Draw a distribution map of the fire intensity level according to the optimal relationship model and the optimal remote sensing spectral index; and evaluate the accuracy of the fire intensity level using a confusion matrix through the second group of survey sample data;
[0009] The random forest model was used to evaluate the relative impact of environmental factors on the spatial distribution pattern of forest fire intensity, and the spatial distribution pattern of forest fire intensity was predicted based on the environmental factors to obtain the prediction results.
[0010] Preferably, the survey sample data includes surface combustibles and soil layer, herbs, low shrubs and tree seedlings, tall shrubs and tree layers, and forest canopy layer.
[0011] Preferably, after the initial remote sensing image is subjected to radiation calibration and atmospheric correction processing, obtaining the target remote sensing image also includes calculating the reflection index, thermal index, and mixing index of the target remote sensing image;
[0012] The reflectance index includes normalized burn ratio, differential normalized burn ratio, relative differential normalized burn ratio, relativized burn ratio, normalized vegetation index, differential normalized vegetation index, enhanced vegetation index, and differential enhanced vegetation index;
[0013] The thermal index includes surface temperature and differential surface temperature.
[0014] Preferably, the evaluation index includes a coefficient of determination and a residual standard deviation.
[0015] Preferably, the relationship module includes a linear model, a quadratic polynomial model, a cubic polynomial model, a logarithmic model and a saturated growth model.
[0016] Preferably, the fire intensity level is divided into unburned area, low intensity, medium intensity and high intensity based on the comprehensive combustion index.
[0017] Preferably, the process of using the random forest model to assess the relative impact of environmental factors on the spatial distribution pattern of forest fire intensity includes:
[0018] The measured values of altitude, slope and aspect of the burned area were collected, and the pre-fire normalized difference vegetation index value was calculated. The random forest model was used with the relativized burn rate as the dependent variable and altitude, aspect, slope and pre-fire normalized difference vegetation index as independent variables to model the factors controlling the spatial pattern of fire intensity and evaluate the relative influence of each variable on the spatial pattern of fire intensity.
[0019] Preferably, the slope includes a flat slope, a gentle slope, an inclined slope, and a steep slope;
[0020] The slope directions include flat land, shady slopes, and sunny slopes.
[0021] The present invention discloses the following technical effects:
[0022] The present invention provides an evaluation model construction and prediction method based on subtropical forest fire intensity. The remote sensing spectral index is calculated through remote sensing images to evaluate the forest fire intensity in subtropical areas and the influencing factors of its spatial distribution, which fills the research gap on different forest fire intensities in subtropical areas. By accurately evaluating the level of forest fire intensity, it has certain guiding significance for the restoration and management of vegetation after the fire. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0024] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION
[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0026] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0027] like Figure 1 As shown, the present invention provides an evaluation model construction and prediction method based on subtropical forest fire intensity, including:
[0028] Obtain survey sample data and initial remote sensing images of subtropical forest wildfire intensity; among them, the survey sample data include surface combustibles and soil layers, herbs, low shrubs and tree seedlings, tall shrubs and tree layers, and forest canopy layers.
[0029] Furthermore, the surface combustibles and soil layer includes the dead combustible consumption of dead branches and leaves, semi-humus, coarse combustibles, cover of new exposed rocks, and soil color; the herbs, low shrubs and tree seedlings include the percentage of leaf change, survival rate, plant colonization, and species composition; the tall shrubs and tree layer includes the percentage of leaf change, survival rate, percentage of cover change, and species composition; the canopy layer includes the green canopy percentage, black canopy percentage, brown canopy percentage, canopy mortality rate, and blackening height.
[0030] The survey sample data were randomly grouped in a ratio of 7:3 to obtain the first group of survey sample data and the second group of survey sample data; 70% of the first group of survey sample data were used to build the regression model, and the remaining 30% of the second group of survey sample data were used for accuracy verification. Different comprehensive combustion index thresholds were distributed with sample points to meet the needs of model building and accuracy verification. The initial remote sensing image was subjected to radiation calibration and atmospheric correction to obtain the target remote sensing image; the reflectance index, thermal index, and hybrid index of the target remote sensing image were calculated;
[0031] Among them, the reflectance index includes normalized burn ratio, difference normalized burn ratio, relative difference normalized burn ratio, relativized burn ratio, normalized vegetation index, difference normalized vegetation index, enhanced vegetation index, and difference enhanced vegetation index; the thermal index includes surface temperature and difference surface temperature.
[0032] The remote sensing spectral index is calculated based on the target remote sensing image; the relationship model between the comprehensive combustion index regression model and the remote sensing spectral index is established based on the first group of survey sample data; the relationship modules include linear model, quadratic polynomial model, cubic polynomial model, logarithmic model and saturated growth model.
[0033] Linear model: y = a(x) + b
[0034] Quadratic polynomial model: y = a(x) 2 +b(x)+c
[0035] Cubic polynomial model: y = a(x) 3 +b(x) 2 +c(x)+d
[0036] Logarithmic model: y = aln(x) + b
[0037] Saturated growth model: y = x (a | x | + b) -1
[0038] In the above formula, X represents the comprehensive combustion index CBI, and y represents the fire intensity evaluation index.
[0039] The fitting effect of the comprehensive combustion index regression model and the remote sensing spectral index was evaluated by the relationship module according to the determination coefficient and residual standard deviation, and the optimal relationship model and the optimal remote sensing spectral index were determined;
[0040] The distribution map of the fire intensity level was drawn according to the optimal relationship model and the optimal remote sensing spectral index; the fire intensity level was divided into unburned area (CBI≤0.25), low intensity (0.25<CBI≤1.25), medium intensity (1.25<CBI≤2.25), and high intensity (CBI>2.25) based on the comprehensive combustion index. The second group of survey sample data was used to evaluate the accuracy of the fire intensity level using the confusion matrix; the confusion matrix was used to evaluate the accuracy of automatic identification of fire intensity. The evaluation indicators include overall accuracy (OA) and kappa. The overall accuracy is the ratio of the total number of pixels correctly classified by the classifier to the total number of pixels. Kappa is an indicator that combines production accuracy and user accuracy. The larger the OA and kappa values, the smaller the difference between the classification result and the actual value, and the higher the classification accuracy of the classifier.
[0041] The random forest model was used to evaluate the relative impact of environmental factors on the spatial distribution pattern of forest fire intensity. Specifically, the altitude (divided into 3 categories, <300m, 300-600m, >600m), slope (5°, 15°, 25°) and slope aspect of the burned area were collected, and the pre-fire normalized vegetation index value was calculated; the slope included flat slope (<5°), gentle slope (5-15°), slope (15-25°), and steep slope (25-35°); the slope aspect included flat land (-1°), shady slope (0-135°, 315-360°), and sunny slope (135-315°). The random forest model was used with the relativized burn rate as the dependent variable, and the altitude, slope aspect, slope and pre-fire normalized vegetation index as the independent variables. The factors controlling the spatial pattern of fire intensity were modeled to evaluate the relative impact of each variable on the spatial pattern of fire intensity. The spatial distribution pattern of forest fire intensity is predicted by comprehensive environmental factors to obtain prediction results.
[0042] The calculation of the random forest model is shown as follows:
[0043]
[0044] In the formula, GI m is the Gini index of node m, K means there are K categories, p mk Indicates the proportion of category k in node m.
[0045] The embodiments described above are only descriptions of the preferred modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.
Claims
1. A method for constructing an evaluation model and predicting the intensity of subtropical forest fires, characterized in that: Including: Obtaining the survey sample data of the wildfire intensity in subtropical forests and the initial remote sensing images; Randomly grouping the said survey sample data according to the ratio of 7:3 to obtain the first group of survey sample data and the second group of survey sample data; performing radiometric calibration and atmospheric correction processing on the said initial remote sensing images to obtain the target remote sensing images; Calculating the remote sensing spectral index based on the said target remote sensing images; establishing the relationship model between the comprehensive combustion index regression model and the said remote sensing spectral index according to the first group of survey sample data; Evaluating the fitting effect of the relationship module in fitting the comprehensive combustion index regression model and the remote sensing spectral index according to the preset evaluation indicators, and determining the optimal relationship model and the optimal remote sensing spectral index; Drawing the wildfire intensity grade distribution map according to the said optimal relationship model and the optimal remote sensing spectral index; evaluating the accuracy of the wildfire intensity grade by using the confusion matrix through the second group of survey sample data; Evaluating the relative influence of environmental factors on the spatial distribution pattern of forest fire intensity by using the random forest model, and comprehensively predicting the spatial distribution pattern of forest fire intensity by the said environmental factors to obtain the prediction result; The said survey sample data includes surface combustibles and soil layers, herbs, low shrubs and tree seedlings, tall shrubs and tree layers, and forest canopies; After performing radiometric calibration and atmospheric correction processing on the said initial remote sensing images to obtain the target remote sensing images, it further includes calculating the reflection index, thermal index, and mixture index for the said target remote sensing images; The said reflection index includes the normalized burn ratio, differential normalized burn ratio, relative differential normalized burn ratio, relative burn ratio, normalized difference vegetation index, differential normalized difference vegetation index, enhanced vegetation index, and difference enhanced vegetation index; The said thermal index includes surface temperature and difference surface temperature; The process of evaluating the relative influence of environmental factors on the spatial distribution pattern of forest fire intensity by using the random forest model includes collecting the altitude of the burned area, divided into 3 categories, <300m, 300 - 600m, >600m, the measured values of slope 5°, 15°, 25° and aspect, and calculating the pre-fire normalized difference vegetation index value; using the random forest model, with the relative burn ratio as the dependent variable, altitude, aspect, slope, and pre-fire normalized difference vegetation index as independent variables, to model the factors controlling the spatial pattern of forest fire intensity and evaluate the relative influence of each variable on the spatial pattern of forest fire intensity; The said evaluation indicators include the coefficient of determination and the residual standard deviation; the said relationship module includes linear model, quadratic polynomial model, cubic polynomial model, logarithmic model, and saturation growth model; Linear model: y = a(x) + b; Quadratic polynomial model: y = a(x) 2 +b(x)+c; Cubic polynomial model: y = a(x) 3 +b(x) 2 +c(x)+d; Logarithmic model: y = aln(x) + b; Saturated growth model: y = x (a | x | + b) -1 ; In the above formula, X represents the comprehensive combustion index CBI, and y represents the wildfire intensity evaluation index; The said wildfire intensity grade is divided into unburned area CBI ≤ 0.25, low intensity 0.25 < CBI ≤ 1.25, medium intensity 1.25 < CBI ≤ 2.25, and high intensity CBI > 2.25 based on the comprehensive combustion index; The said slope includes flat slope <5°, gentle slope 5 - 15°, sloping slope 15 - 25°, and steep slope 25 - 35°; The slope directions include -1° for flat ground, 0-135°, 315-360° for shady slopes, and 135-315° for sunny slopes.
