Method and medium for monitoring mine site ecosystems based on net primary productivity of vegetation

Abstract

The present application provides a kind of based on vegetation net primary productivity mining area ecosystem monitoring method and storage medium, comprising: by solving and processing the continuous spatiotemporal resolution spectral image set of spatial resolution not less than 60 meters per pixel, obtain continuous vegetation net primary productivity pixel data set, according to the continuous vegetation net primary productivity pixel data set and pixel driving factor data, pixel by pixel calculation and processing obtain continuous pixel moran index set and further obtain continuous cluster pixel region data set, by continuous cluster pixel region data set training, obtain vegetation net primary productivity prediction nonlinear regression model, utilize the vegetation net primary productivity prediction nonlinear regression model and obtain prediction instance set, to obtain driving factor contribution map and driving factor target threshold according to interpreter model again. Visible, the present application provides a kind of based on vegetation net primary productivity mining area ecosystem monitoring method, can realize the monitoring of mining area ecosystem.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of remote sensing monitoring and ecological restoration technology for mining areas, and particularly to a method and medium for monitoring mining area ecosystems based on net primary productivity of vegetation. Background Technology

[0002] Ecosystem restoration and reconstruction have become an important task in the construction of national ecological civilization. Net primary productivity (NPP) is a core indicator characterizing the carbon sequestration capacity and health status of terrestrial ecosystems and is widely used in the quantitative assessment of the effectiveness of ecological restoration and reconstruction.

[0003] However, mining area ecosystems are characterized by high spatial heterogeneity, fragmented surfaces, and high levels of human intervention. Their surface landscapes are fragmented, with reclaimed plots, spoil heaps, and natural landforms often interspersed. The net primary productivity of vegetation in mining areas is not only subject to natural factors such as temperature and precipitation, but also faces unique human influences such as surface deformation and groundwater level changes. Therefore, existing technologies are insufficient for detecting mining area ecosystems.

[0004] Therefore, how to monitor the ecosystem of mining areas has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] This invention provides a method for monitoring mining area ecosystems based on net primary productivity of vegetation, comprising: acquiring a continuous spatiotemporal resolution spectral image set, pixel driving factor data, and total solar radiation data of the target mining area; the continuous spatiotemporal resolution spectral image set includes pixel red band reflectance data and pixel near-infrared band reflectance data, the spatial resolution of the continuous spatiotemporal resolution spectral image set is not less than 60 meters, the pixel driving factor data includes pixel driving factor values ​​corresponding one-to-one with pixels in the continuous spatiotemporal resolution spectral image set, and the pixel driving factor data includes pixel environmental driving factor data and pixel human activity driving factor data; A continuous high spatiotemporal resolution spectral image set is solved pixel by pixel to obtain a continuous normalized vegetation index pixel dataset. Based on the continuous normalized vegetation index pixel dataset, pixel driving factor data and total solar radiation data, a continuous net primary productivity pixel dataset is calculated pixel by pixel. Based on the continuous vegetation net primary productivity pixel dataset and pixel driving factor data, the bivariate local Moran index is calculated pixel by pixel to obtain the continuous pixel Moran index set; A continuous clustered pixel region dataset is obtained from the continuous pixel Moran index set. A nonlinear regression model for predicting net primary productivity of vegetation is trained based on the continuous clustered pixel region dataset. The independent variable of the nonlinear regression model for predicting net primary productivity of vegetation is the pixel driving factor data, and the dependent variable is the net primary productivity of vegetation. A set of prediction instances is obtained through a nonlinear regression model for predicting net primary productivity of vegetation. Based on the set of prediction instances and the nonlinear regression model for predicting net primary productivity of vegetation, a driving factor contribution map is obtained through an interpreter model. The horizontal axis of the driving factor contribution map represents the driving factor value, and the vertical axis represents the driving factor contribution value. Based on the driving factor contribution map, the target threshold of the driving factor for monitoring the mining area ecosystem is obtained.

[0006] As can be seen, the mining area ecosystem monitoring method based on net primary productivity of vegetation provided by this invention obtains a continuous net primary productivity pixel dataset by calculating and processing a continuous spatiotemporal resolution spectral image set with a spatial resolution of not less than 60 meters for the target mining area pixel by pixel. Based on the continuous net primary productivity pixel dataset and pixel driving factor data, a continuous pixel Moran index set is calculated and processed pixel by pixel, and a continuous clustered pixel region dataset is further obtained. The continuous clustered pixel region dataset is used as training data to train a nonlinear regression model for predicting net primary productivity of vegetation. The prediction instance set is obtained using the nonlinear regression model for predicting net primary productivity of vegetation. Then, the driving factor contribution map is obtained according to the interpreter model, and the target threshold of the driving factor for monitoring the mining area ecosystem is obtained according to the driving factor contribution map. Thus, the mining area ecosystem monitoring method based on net primary productivity (NPMP) provided in this embodiment of the invention can obtain target thresholds for driving factors used in mining area ecosystem monitoring. By using these target thresholds, the overall level of the mining area ecosystem can be monitored. Furthermore, based on the target thresholds and the continuous pixel Moran index set, pixel-level monitoring of the mining area ecosystem can be achieved. This allows for monitoring of the mining area ecosystem tailored to its specific characteristics. The calculation of the continuous pixel Moran index set is a step in the process of obtaining the target thresholds for driving factors and does not need to be calculated separately, thereby improving the efficiency of mining area ecosystem monitoring. In addition, obtaining the target thresholds for driving factors through a nonlinear regression model and interpreter model for NPMP prediction not only improves the prediction accuracy of NPMP under different pixel driving factor data in the mining area ecosystem and obtains the contribution value of each pixel driving factor data to the prediction, but also allows for further monitoring at the overall level of the mining area ecosystem, achieving a technical effect greater than the sum of its parts. Furthermore, the nonlinear regression model for predicting net primary productivity of vegetation was trained using pixel-driven factor data and a continuously clustered pixel region dataset. The pixel-driven factor data included both environmental and human activity-driven factor data. This approach, combined with monitoring the natural environment, addresses the high level of human intervention in the mining area ecosystem, further improving the prediction accuracy of the nonlinear regression model. Finally, the continuously clustered pixel region dataset was obtained by processing a set of continuous spatiotemporal resolution spectral images with a spatial resolution of at least 60 meters. This approach, considering the fragmented surface and human intervention characteristics of the mining area, uses a set of continuous spatiotemporal resolution spectral images with a temporally continuous resolution of at least 60 meters as the foundation and initial step for data processing, improving the accuracy of the continuously clustered pixel region dataset and the entire subsequent monitoring of the mining area ecosystem. Attached Figure Description

[0007] Figure 1This is a schematic flowchart of a mining area ecosystem monitoring method based on vegetation net primary productivity provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the driving factor contribution of the monitoring method for mining area ecosystems based on net primary productivity of vegetation provided in the embodiments of the present invention. Detailed Implementation

[0008] 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.

[0009] Please refer to Figure 1 and Figure 2 , Figure 1 This is a schematic flowchart of a mining area ecosystem monitoring method based on vegetation net primary productivity provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the driving factor contribution of the monitoring method for mining area ecosystems based on net primary productivity of vegetation provided in the embodiments of the present invention.

[0010] As shown in the figure, the method provided in this embodiment of the invention includes the following steps: In step S1, a continuous spatiotemporal resolution spectral image set, pixel driving factor data, and total solar radiation data of the target mining area are acquired.

[0011] The target mining area is the mining area used as an ecological monitoring target. A continuous spatiotemporal resolution spectral image set is an image set containing multiple spectral images with the same spatial resolution. Same spatial resolution means that the mining area corresponding to each pixel in each spectral image is of the same size. Continuous spatiotemporal resolution refers to the consistent and continuous time intervals between the spectral images in this continuous spatiotemporal resolution spectral image set. For example, a continuous spatiotemporal resolution spectral image set includes 365 spectral images with the same spatial resolution, and the time interval between each spectral image, i.e., the temporal resolution, is 24 hours. Pixel-driven factor data refers to data accurate to the pixel level that affects net primary productivity (NPP) of vegetation. For example, taking precipitation as an example, each pixel in all spectral images in the continuous spatiotemporal resolution spectral image set must have a corresponding precipitation amount. Total solar radiation data is the total solar radiation to the mining area, usually calculated based on solar radiation and the area of ​​the mining area. It is worth noting that in this invention, terms containing "pixel" refer to those accurate to the pixel level, and will not be elaborated further below.

[0012] It is worth noting that, in this invention, the continuous spatiotemporal resolution spectral image set includes pixel red band reflectance data and pixel near-infrared band reflectance data. The spatial resolution of the continuous spatiotemporal resolution spectral image set is not less than 60 meters. The pixel driving factor data includes pixel driving factor values ​​that correspond one-to-one with the pixels in the continuous spatiotemporal resolution spectral image set. The pixel driving factor data includes pixel environmental driving factor data and pixel human activity driving factor data.

[0013] A continuous spatiotemporal resolution spectral image set includes pixel red band reflectance data and pixel near-infrared band reflectance data. Specifically, it refers to the red band and infrared band reflectance data of each pixel within all spectral images in the set. These can be red band and infrared band images obtained separately by a satellite's red band detector and near-infrared band detector, respectively. It's important to note that red band and infrared band images captured at the same time are considered the same spectral image within the continuous spatiotemporal resolution spectral image set, only located in different spectral bands.

[0014] The spatial resolution of the continuous spatiotemporal resolution spectral image set is no less than 60 meters, that is, each pixel corresponds to a mining area of ​​60×60 or smaller.

[0015] Pixel driving factor data includes pixel driving factor values ​​that correspond one-to-one with pixels in the continuous spatiotemporal resolution spectral image set. It refers to driving factor data at the pixel scale, which is consistent not only with the size of the pixels in the continuous spatiotemporal resolution spectral image set, but also with the mining area corresponding to the pixel.

[0016] Pixel-driven factor data includes pixel environmental-driven factor data and pixel human activity-driven factor data, which means that the factors influencing vegetation net primary productivity include not only environmental factors but also human activity factors.

[0017] It is readily understood that the continuous spatiotemporal resolution spectral image set acquired by this invention is an image set with many requirements. In some embodiments where it can be directly detected, it can be directly acquired and used; however, in other embodiments where it cannot be directly detected, it can be obtained through data processing. Therefore, in some embodiments, acquiring the continuous spatiotemporal resolution spectral image set of the target mining area includes: In step S11, a first spectral image set and a second spectral image set of the target mining area are acquired. The spatial resolution of the first spectral image set is higher than that of the second spectral image set, and the temporal resolution of the second spectral image set is higher than that of the first spectral image set. Both the first spectral image set and the second spectral image set include red band reflectance and near-infrared band reflectance images.

[0018] It's easy to understand that while the first spectral image set is clear, the long time intervals between its captures limit the information it can provide in the continuous temporal dimension. The second spectral image set, with its high temporal resolution, provides more images, but its low spatial resolution means that each image is less clear than those in the first spectral image set. Specifically, the first spectral image set could be the Landsat satellite image set, and the second spectral image set could be the MODIS satellite image set. Both Landsat and MODIS satellite image sets are satellite imagery products that are practically available.

[0019] In step S12, the second spectral image set is resampled until the spatial resolution is the same as that of the first spectral image set, thus obtaining a second spectral resampled image set corresponding to the pixels of the first spectral image set.

[0020] Resampling can be achieved using bilinear interpolation.

[0021] In step S13, the first spectral image set and the second spectral resampled image set are fused using a high-resolution spatiotemporal fusion algorithm to obtain the continuous spatiotemporal resolution spectral image set.

[0022] High-resolution spatiotemporal fusion algorithms can be any algorithms capable of fusing different temporal and spatial resolutions. In some embodiments, they can be super-resolution reconstruction models of high-resolution spatiotemporal maps, treating each pixel as a graph node and constructing a target energy function containing a data term and a smoothness term. The data term constrains the predicted image to approximate the real Landsat image set at a known time, while the smoothness term utilizes the rich temporal information of the MODIS image set to constrain the trend of change of the predicted image in the temporal dimension to be consistent with the trend of change observed by MODIS, thereby fusing and reconstructing a continuous spatiotemporal resolution spectral image set.

[0023] Furthermore, the target mining area may be located in a geographically affected area during the rainy season. The heavy cloud cover during this period could cause the already low temporal resolution of the first spectral imagery to become invalid or lost over a long period. Therefore, in some embodiments, step S13 includes: In step S131, the first spectral image set and the second spectral resampled image set are fused using a high-resolution spatiotemporal fusion algorithm to obtain a fused image set.

[0024] The fused image atlas is a dataset obtained by fusing the first spectral image set and the second spectral resampled image set when the first spectral image is lost due to the rainy season. The fusion step itself is no different from the fusion step of Rongguo's high-resolution spatiotemporal fusion algorithm in step S13. The only difference is the description because the data loss during the rainy season makes it discontinuous.

[0025] In step S132, the continuous spatiotemporal resolution spectral image set is obtained by filling the fused image set with the Bayesian maximum a posteriori probability algorithm.

[0026] By combining the Bayesian prediction framework and using the second spectral image as prior information, the maximum a posteriori probability of the first spectral image at the missing time is estimated to complete the fused image atlas and obtain a continuous spatiotemporal resolution spectral image set.

[0027] In this way, by completing the Bayesian prediction framework, the monitoring of the mining area ecosystem during the rainy season can be reduced.

[0028] In this way, when a continuous spatiotemporal resolution spectral image set cannot be directly detected, the first spectral image set and the second spectral image set can be fused to obtain a continuous spatiotemporal resolution spectral image set with higher temporal resolution than the first spectral image set and higher spatial resolution than the second spectral image set, providing richer spectral information support for subsequent monitoring of the mining area ecosystem.

[0029] In step S2, the continuous high spatiotemporal resolution spectral image set is solved pixel by pixel to obtain the continuous normalized vegetation index pixel dataset. Based on the continuous normalized vegetation index pixel dataset, pixel driving factor data and total solar radiation data, the continuous net primary productivity pixel dataset is calculated pixel by pixel.

[0030] The pixel-by-pixel calculation of a continuous high spatiotemporal resolution spectral image set involves solving the normalized vegetation index (NVI) for each pixel in the set according to the following formula. The NVI data of all pixels constitute a continuous normalized vegetation index pixel dataset: ; in, It is the normalized vegetation index. It is a near-infrared reflectance image. It is an infrared reflectance image.

[0031] In a continuous normalized vegetation index (NVI) pixel dataset, "continuous" means that the NVI pixel data within the dataset have the same temporal resolution. In this invention, datasets with "continuous" in their names all indicate that the data within the dataset has a consistent temporal resolution, which will not be elaborated further. It is worth noting that in some embodiments, the temporal resolution of the continuous NVI pixel dataset can be the same as that of the continuous spatiotemporal resolution spectral image set, while in other embodiments it can be greater than the temporal resolution of the continuous spatiotemporal resolution spectral image set. For example, the temporal resolution of the continuous spatiotemporal resolution spectral image set may be 24 hours, while the temporal resolution of the continuous NVI pixel dataset can be monthly. This is achieved by obtaining the NVI pixel data for each month based on the spectral images contained within each month of the continuous spatiotemporal resolution spectral image set, and then assembling the continuous NVI pixel dataset.

[0032] In some embodiments, to obtain a more accurate continuous net primary productivity pixel dataset for a target mining area, the step of calculating the continuous net primary productivity pixel dataset for vegetation pixel based on the continuous normalized vegetation index pixel dataset, the pixel driving factor data, and the total solar radiation data includes: The absorption ratio of photosynthetically active radiation by vegetation is obtained from the continuous normalized vegetation index (NVI) pixel dataset. The absorption ratio of photosynthetically active radiation by vegetation is input pixel by pixel. The pixel driving factor data and the total solar radiation data are then fed into a pre-set pixel-based net primary productivity (NPMP) calculation model to obtain a continuous NMP pixel dataset. The NMP calculation model includes: ; ; ; in: It is the net primary productivity of vegetation. It is the photosynthetically active radiation absorbed by the vegetation. It is the actual light energy utilization rate. It is the proportion of photosynthetically active radiation absorbed by vegetation. This is the total solar radiation data. It is the pixel driving factor stress coefficient. It is the maximum light energy utilization rate of vegetation. It is the pixel position. It's time.

[0033] It is worth noting that, since the pixel driving factors include both pixel environmental driving factor data and pixel human activity driving factor data, the stress coefficient of the pixel driving factors at this location... .

[0034] Thus, by using the pixel-based net primary productivity calculation model for vegetation provided in this embodiment of the invention, which takes into account both pixel-based human activity driving factor data and rainy season correction, a more accurate continuous net primary productivity pixel dataset for vegetation can be obtained.

[0035] In addition, the continuous vegetation net primary productivity pixel dataset, where continuous means temporal continuity, means that the time intervals between the net primary productivity pixels of each vegetation in the continuous vegetation net primary productivity pixel dataset are the same.

[0036] It is worth noting that, in this embodiment of the invention, the continuous vegetation net primary productivity pixel dataset of the target mining area has been obtained through steps S1 and S2. However, this embodiment of the invention does not directly monitor the mining area ecosystem based on the continuous vegetation net primary productivity data at the pixel level, but rather performs subsequent processing based on the continuous vegetation net primary productivity pixel dataset.

[0037] In step S3, based on the continuous vegetation net primary productivity pixel dataset and pixel driving factor data, the bivariate local Moran index is calculated pixel by pixel to obtain the continuous pixel Moran index set.

[0038] In some embodiments, the formula for calculating the bivariate local Moran index per pixel is: .in, It is a pixel. yes The neighboring pixels, It is a pixel Moran's index, It is a pixel Net primary productivity of vegetation It is a neighboring pixel Pixel driving factor data, It is a pixel With neighboring pixels The spatial weight value. Wherein Based on the continuous vegetation net primary productivity pixel dataset. This is obtained from pixel-driven factor data. Substituting different pixel-driven factors, the sign of the Moran index I indicates whether the pixel-driven factor has a unidirectional or antidirectional spatial clustering relationship with the pixel's net primary productivity of vegetation. The absolute value represents the strength of the relationship. For example, if the pixel-driven factor data includes pixel precipitation, pixel radiation, and pixel road distance, substituting these factors respectively yields the pixel Moran index for net primary productivity of vegetation with the pixel precipitation driver, the pixel Moran index for net primary productivity of vegetation with the pixel radiation driver, and the pixel Moran index for net primary productivity of vegetation with the pixel road distance driver.

[0039] In step S4, a continuous clustered pixel region dataset is obtained based on the continuous pixel Moran index set, and a nonlinear regression model for predicting net primary productivity of vegetation is trained based on the continuous clustered pixel region dataset to obtain the nonlinear regression model for predicting net primary productivity of vegetation.

[0040] The continuous clustered pixel region dataset is a dataset containing pixel clustering information obtained by clustering based on the continuous pixel Moran index set. Taking pixel driving factor data, including pixel precipitation driving factors, pixel radiation driving factors, and pixel road distance driving factors, as an example: For the pixel Moran index, which characterizes pixel net primary productivity (NPMP) and pixel precipitation driving factors, pixels with large absolute values ​​of the Moran index, consistent positive and negative signs, and adjacent pixels are identified to form cluster regions. Cluster regions with positive Moran indices indicate strong precipitation and high NPMP for pixels within the region, while cluster regions with negative Moran indices indicate low precipitation but high NPMP. Similarly, clustered pixel regions for radiation and road distance driving factors are identified, collectively forming the continuous clustered pixel region dataset.

[0041] The independent variable of the nonlinear regression model for predicting net primary productivity of vegetation is pixel-driven factor data, and the dependent variable is net primary productivity of vegetation. In some embodiments, the nonlinear regression model for predicting net primary productivity of vegetation includes an extreme gradient boosting model. In some embodiments, the interpreter model includes a tree interpreter model.

[0042] To improve the quality of continuously clustered pixel region datasets, in some embodiments, the continuously clustered pixel region dataset is obtained based on the continuously clustered pixel Moran index set, including: In step S41, a Monte Carlo permutation test is performed pixel by pixel based on the continuous pixel Moran index set to obtain a significant continuous pixel Moran index set that passes the Monte Carlo permutation test.

[0043] In step S42, based on the Moran index set of significant continuous pixels, high-high cooperative gain pixel clusters, low-low cooperative stress pixel clusters, high-low restricted gain pixel clusters, and low-high increasing stress pixel clusters are identified to obtain a continuous clustered pixel region dataset.

[0044] In this way, by setting the Monte Carlo permutation test, non-significant noise is removed, thereby improving the quality of the continuous clustered cell region dataset.

[0045] Thus, in steps S3 and S4, based on the continuous net primary productivity pixel dataset of vegetation obtained in step S2, a continuous clustered pixel region dataset is processed and identified, and a nonlinear regression model for predicting net primary productivity of vegetation is trained using the continuous clustered pixel region dataset.

[0046] In step S5, a set of prediction instances is obtained through the nonlinear regression model for predicting net primary productivity of vegetation. Based on the set of prediction instances and the nonlinear regression model for predicting net primary productivity of vegetation, a driving factor contribution map is obtained through the interpreter model. The horizontal axis of the driving factor contribution map represents the driving factor value, and the vertical axis represents the driving factor contribution value. Based on the driving factor contribution map, the target threshold of the driving factor for monitoring the mining area ecosystem is obtained.

[0047] The target threshold for driving factors is a threshold that is significant for monitoring the ecosystem of the mining area.

[0048] In some embodiments, obtaining the target thresholds for driving factors used for monitoring the mining area ecosystem based on the driving factor contribution map includes: In step S51, a fitting trend curve of the driving factor contribution value changing with the driving factor value is fitted in the driving factor contribution graph.

[0049] In step S52, the target threshold of the driving factor is identified on the fitted trend curve. The target threshold of the driving factor includes the driving factor limit threshold corresponding to the derivative of the fitted trend curve being zero, and the driving factor critical threshold corresponding to the driving factor contribution value being zero.

[0050] The critical threshold of driving factors can indicate the tolerance threshold of the mining area ecosystem, while the limit threshold of driving factors can indicate the optimal adaptation point of the mining area ecosystem. Combining the two can provide monitoring information of the overall level of the mining area ecosystem, such as the climate adaptability limit or the environmental carrying capacity boundary.

[0051] Thus, the vegetation net primary productivity prediction nonlinear regression model obtained in step S4 can be used to input different pixel driving factor data to obtain a rich set of prediction instances. Looking at steps S1 to S5, the role of the vegetation net primary productivity prediction nonlinear regression model is primarily to act as a precise amplifier. Based on the limited and specific measured instance of the continuous spatiotemporal resolution spectral image set, it obtains a rich and comprehensive set of prediction instances, reducing the cost of obtaining the massive number of comprehensive instances required to acquire the target threshold of the driving factors, or in other words, improving the accuracy of the target threshold of the driving factors.

[0052] As can be seen, the mining area ecosystem monitoring method based on net primary productivity of vegetation provided by this invention obtains a continuous net primary productivity pixel dataset by calculating and processing a continuous spatiotemporal resolution spectral image set with a spatial resolution of not less than 60 meters for the target mining area pixel by pixel. Based on the continuous net primary productivity pixel dataset and pixel driving factor data, a continuous pixel Moran index set is calculated and processed pixel by pixel, and a continuous clustered pixel region dataset is further obtained. The continuous clustered pixel region dataset is used as training data to train a nonlinear regression model for predicting net primary productivity of vegetation. The prediction instance set is obtained using the nonlinear regression model for predicting net primary productivity of vegetation. Then, the driving factor contribution map is obtained according to the interpreter model, and the target threshold of the driving factor for monitoring the mining area ecosystem is obtained according to the driving factor contribution map.

[0053] Thus, the mining area ecosystem monitoring method based on net primary productivity (NPMP) provided in this embodiment of the invention can obtain target thresholds for driving factors used in mining area ecosystem monitoring. By using these target thresholds, the overall level of the mining area ecosystem can be monitored. Furthermore, based on the target thresholds and the continuous pixel Moran index set, pixel-level monitoring of the mining area ecosystem can be achieved. This allows for monitoring of the mining area ecosystem tailored to its specific characteristics. The calculation of the continuous pixel Moran index set is a step in the process of obtaining the target thresholds for driving factors and does not need to be calculated separately, thereby improving the efficiency of mining area ecosystem monitoring. In addition, obtaining the target thresholds for driving factors through a nonlinear regression model and interpreter model for NPMP prediction not only improves the prediction accuracy of NPMP under different pixel driving factor data in the mining area ecosystem and obtains the contribution value of each pixel driving factor data to the prediction, but also allows for further monitoring at the overall level of the mining area ecosystem, achieving a technical effect greater than the sum of its parts. Furthermore, the nonlinear regression model for predicting net primary productivity of vegetation was trained using pixel-driven factor data and a continuously clustered pixel region dataset. The pixel-driven factor data included both environmental and human activity-driven factor data. This approach, combined with monitoring the natural environment, addresses the high level of human intervention in the mining area ecosystem, further improving the prediction accuracy of the nonlinear regression model. Finally, the continuously clustered pixel region dataset was obtained by processing a set of continuous spatiotemporal resolution spectral images with a spatial resolution of at least 60 meters. This approach, considering the fragmented surface and human intervention characteristics of the mining area, uses a set of continuous spatiotemporal resolution spectral images with a temporally continuous resolution of at least 60 meters as the foundation and initial step for data processing, improving the accuracy of the continuously clustered pixel region dataset and the entire subsequent monitoring of the mining area ecosystem.

[0054] In some embodiments, the cell artificial activity driving factor includes the cell nighttime light driving factor and the cell road distance driving factor, and therefore the driving factor target threshold includes the nighttime light driving factor target threshold and the road distance driving factor target threshold.

[0055] Thus, the factors that significantly influence the net primary productivity of vegetation are dust and nighttime light, but dust-driven factors are difficult to obtain. By replacing dust-driven factors with pixel-road distance-driven factors, the distance from pixels to roads can be obtained through a continuous spatiotemporal resolution spectral image set with a spatial resolution of not less than 60 meters. This transforms the pixel dust radius-driven factor into a pixel-road distance-driven factor, thereby improving the accuracy of monitoring the mining area ecosystem.

[0056] In some embodiments, the pixel-level artificial activity driving factors include the pixel-level nighttime light driving factor and the pixel-level road distance driving factor; therefore, the target thresholds for the driving factors include the target threshold for the nighttime light driving factor and the target threshold for the road distance driving factor. The mining area ecosystem monitoring method provided by this invention further includes: In step S6, the core remediation area is divided according to the target threshold of the road distance driving factor and the pixel driving factor data.

[0057] The pixel driving factor data includes the pixel road distance driving factor, which is the distance of each pixel from the mining area road.

[0058] In step S7, based on the target threshold of the nighttime light driving factor and the pixel driving factor data, the area outside the core remediation area is divided into an edge buffer zone and a natural recovery zone.

[0059] Thus, based on the target threshold of the road distance driving factor, the limit distance of road impact, i.e., the dust radius, is obtained. Areas smaller than this distance are designated as core remediation zones, where subsequent measures such as engineering sand fixation, water spraying for dust suppression, and artificial irrigation can be implemented. Similarly, based on the target threshold of the nighttime light driving factor, the limit distance of nighttime light impact is obtained. This distance is usually greater than the dust radius. Therefore, the area between this limit distance of nighttime light impact and the dust radius is designated as an edge buffer zone, where subsequent protection measures such as enclosure can be implemented to prevent degradation. Pixels larger than the limit distance of nighttime light impact are designated as natural recovery zones.

[0060] Please continue to refer to this. Figure 2 In some embodiments, the method for monitoring mining area ecosystems based on net primary productivity of vegetation provided by the present invention, wherein training a nonlinear regression model for predicting net primary productivity of vegetation based on the continuously clustered pixel region dataset to obtain the nonlinear regression model for predicting net primary productivity of vegetation includes: The continuous clustered pixel region dataset is divided into at least two sets of continuous time period clustered pixel region datasets according to the time span. The nonlinear regression model for predicting net primary productivity of vegetation is trained on each set of continuous time period clustered pixel region datasets to obtain the corresponding nonlinear regression model for predicting net primary productivity of vegetation.

[0061] The process involves obtaining a prediction instance set through the nonlinear regression model for predicting net primary productivity of vegetation, and based on this prediction instance set, combined with the nonlinear regression model for predicting net primary productivity of vegetation, obtaining a driving factor contribution map through an interpreter model. The horizontal axis of the driving factor contribution map represents the driving factor value, and the vertical axis represents the driving factor contribution value. Furthermore, the driving factor target threshold is obtained based on the driving factor contribution map, including: By using the nonlinear regression models for predicting net primary productivity of vegetation corresponding to the at least two sets of clustered pixel region datasets for consecutive time periods, corresponding prediction examples are obtained. By combining the corresponding nonlinear regression models for predicting net primary productivity of vegetation with the interpreter model, the corresponding driving factor contribution map is obtained. The horizontal axis of the driving factor contribution map is the driving factor value, and the vertical axis is the driving factor contribution value. The corresponding driving factor target threshold is obtained based on the corresponding driving factor contribution map.

[0062] Figure 2 The target threshold for the driving factor is abbreviated as target threshold.

[0063] In some embodiments, the mining area ecosystem monitoring method based on vegetation net primary productivity further includes: Based on the driving factor contribution maps and driving factor target thresholds corresponding to the at least two sets of clustered pixel region datasets over consecutive time periods, the displacement of the driving factor target threshold and the evolution variables of the environmental carrying capacity boundary are obtained. In this way, at least two sets of clustered pixel region datasets over consecutive time periods, along with the driving factor contribution maps and driving factor target thresholds, can be obtained, thereby enabling the monitoring of the evolution of the climate adaptability limit or the environmental carrying capacity boundary.

[0064] The present invention also provides a storage medium storing a program for implementing the above-described method for monitoring mining ecosystems based on net primary productivity of vegetation.

[0065] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is accorded the widest scope consistent with the principles and novel features disclosed herein.

[0066] While the embodiments of the present invention have been disclosed above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A method for monitoring mining area ecosystems based on net primary productivity of vegetation, characterized in that, include: Acquire a continuous spatiotemporal resolution spectral image set, pixel driving factor data, and total solar radiation data for the target mining area. The continuous spatiotemporal resolution spectral image set includes pixel red band reflectance data and pixel near-infrared band reflectance data. The spatial resolution of the continuous spatiotemporal resolution spectral image set is not less than 60 meters. The pixel driving factor data includes pixel driving factor values ​​that correspond one-to-one with the pixels in the continuous spatiotemporal resolution spectral image set. The pixel driving factor data includes pixel environmental driving factor data and pixel human activity driving factor data. The continuous high spatiotemporal resolution spectral image set is solved pixel by pixel to obtain a continuous normalized vegetation index pixel dataset. Based on the continuous normalized vegetation index pixel dataset, the pixel driving factor data and the total solar radiation data, the continuous net primary productivity pixel dataset is calculated pixel by pixel. Based on the continuous vegetation net primary productivity pixel dataset and the pixel driving factor data, the bivariate local Moran index is calculated pixel by pixel to obtain the continuous pixel Moran index set; A continuous clustered pixel region dataset is obtained based on the continuous pixel Moran index set. A nonlinear regression model for predicting net primary productivity of vegetation is trained based on the continuous clustered pixel region dataset. The independent variable of the nonlinear regression model for predicting net primary productivity of vegetation is the pixel driving factor data, and the dependent variable is the net primary productivity of vegetation. A prediction instance set is obtained through the nonlinear regression model for predicting net primary productivity of vegetation. Based on the prediction instance set and the nonlinear regression model for predicting net primary productivity of vegetation, a driving factor contribution map is obtained through the interpreter model. The horizontal axis of the driving factor contribution map represents the driving factor value, and the vertical axis represents the driving factor contribution value. Based on the driving factor contribution map, the target threshold of the driving factor for monitoring the mining area ecosystem is obtained.

2. The method for monitoring mining area ecosystems based on net primary productivity of vegetation as described in claim 1, characterized in that, The acquisition of the continuous spatiotemporal resolution spectral image set of the target mining area includes: Acquire a first spectral image set and a second spectral image set of the target mining area. The spatial resolution of the first spectral image set is higher than that of the second spectral image set, and the temporal resolution of the second spectral image set is higher than that of the first spectral image set. Both the first spectral image set and the second spectral image set include red band reflectance and near-infrared band reflectance images. The second spectral image set is resampled to have the same spatial resolution as the first spectral image set to obtain a second spectral resampled image set corresponding to the pixels of the first spectral image set. The continuous spatiotemporal resolution spectral image set is obtained by fusing the first spectral image set and the second spectral resampled image set using a high-resolution spatiotemporal fusion algorithm.

3. The method for monitoring mining area ecosystems based on net primary productivity of vegetation as described in claim 2, characterized in that, The continuous spatiotemporal resolution spectral image set is obtained by fusing the first spectral image set and the second spectral resampled image set using a high-resolution spatiotemporal fusion algorithm, including: The first spectral image set and the second spectral resampled image set are fused together using a high-resolution spatiotemporal fusion algorithm to obtain a fused image set. The fused image set is then filled with a Bayesian maximum a posteriori probability algorithm to obtain the continuous spatiotemporal resolution spectral image set. The step of calculating the continuous net primary productivity pixel dataset for vegetation pixel based on the continuous normalized vegetation index pixel dataset, the pixel driving factor data, and the total solar radiation data pixel data pixel by pixel includes: The absorption ratio of photosynthetically active radiation by vegetation is obtained from the continuous normalized vegetation index (NVI) pixel dataset. The absorption ratio of photosynthetically active radiation by vegetation is input pixel by pixel. The pixel driving factor data and the total solar radiation data are then fed into a pre-set pixel-based net primary productivity (NPMP) calculation model to obtain a continuous NMP pixel dataset. The NMP calculation model includes: ； ； ； in: It is the net primary productivity of vegetation. It is the photosynthetically active radiation absorbed by the vegetation. It is the actual light energy utilization rate. It is the proportion of photosynthetically active radiation absorbed by vegetation. This is the total solar radiation data. It is the pixel driving factor stress coefficient. It is the maximum light energy utilization rate of vegetation. It is the pixel position. It's time.

4. The method for monitoring mining area ecosystems based on net primary productivity of vegetation as described in claim 3, characterized in that, The step of obtaining the continuous clustered pixel region dataset based on the continuous pixel Moran index set includes: Based on the continuous pixel Moran index set, a Monte Carlo permutation test is performed pixel by pixel to obtain a significant continuous pixel Moran index set that passes the Monte Carlo permutation test. Based on the Moran index set of significant continuous pixels, we identify high-high cooperative gain pixel clusters, low-low cooperative stress pixel clusters, high-low restricted gain pixel clusters, and low-high increasing stress pixel clusters, thus obtaining a continuous clustered pixel region dataset.

5. The method for monitoring mining area ecosystems based on net primary productivity of vegetation as described in claim 3, characterized in that, The step of obtaining the target thresholds for driving factors used in monitoring the mining area ecosystem based on the driving factor contribution map includes: Fit the trend curve of the driving factor contribution value as the driving factor value changes in the driving factor contribution graph; The target thresholds of driving factors are identified on the fitted trend curve. The target thresholds of driving factors include the driving factor limit threshold corresponding to the derivative of the fitted trend curve being zero, and the driving factor critical threshold corresponding to the driving factor contribution value being zero.

6. The method for monitoring mining area ecosystems based on net primary productivity of vegetation as described in any one of claims 1-5, characterized in that, The step of training a nonlinear regression model for predicting net primary productivity of vegetation based on the continuously clustered pixel region dataset to obtain the nonlinear regression model for predicting net primary productivity of vegetation includes: The continuous clustered pixel region dataset is divided into at least two groups of continuous time period clustered pixel region datasets according to the time span. The nonlinear regression model for predicting net primary productivity of vegetation is trained on each group of continuous time period clustered pixel region datasets to obtain the corresponding nonlinear regression model for predicting net primary productivity of vegetation. The process involves obtaining a prediction instance set through the nonlinear regression model for predicting net primary productivity of vegetation, and based on this prediction instance set, combined with the nonlinear regression model for predicting net primary productivity of vegetation, obtaining a driving factor contribution map through an interpreter model. The horizontal axis of the driving factor contribution map represents the driving factor value, and the vertical axis represents the driving factor contribution value. Furthermore, the driving factor target threshold is obtained based on the driving factor contribution map, including: By using the nonlinear regression model for predicting net primary productivity of vegetation corresponding to the at least two sets of clustered pixel region datasets for consecutive time periods, the corresponding prediction implementation examples are obtained. By combining the corresponding nonlinear regression model for predicting net primary productivity of vegetation, the corresponding driving factor contribution map is obtained through the interpreter model. The horizontal axis of the driving factor contribution map is the driving factor value, and the vertical axis is the driving factor contribution value. The corresponding driving factor target threshold is obtained based on the corresponding driving factor contribution map. The method for monitoring mining area ecosystems based on net primary productivity of vegetation also includes: Based on the driving factor contribution maps and driving factor target thresholds corresponding to the at least two sets of clustered pixel region datasets for consecutive time periods, the driving factor target threshold displacement and environmental carrying capacity boundary evolution variables are obtained.

7. The method for monitoring mining area ecosystems based on net primary productivity of vegetation as described in any one of claims 1-5, characterized in that, The pixel artificial activity driving factors include the pixel nighttime light driving factor and the pixel road distance driving factor; The target thresholds for driving factors include the target thresholds for nighttime headlight driving factors and the target thresholds for road distance driving factors.

8. The method for monitoring mining area ecosystems based on net primary productivity of vegetation as described in claim 7, characterized in that, Also includes: Based on the target threshold of the road distance driving factor and the pixel driving factor data, the core remediation area is divided; Based on the target threshold of the nighttime light driving factor and the pixel driving factor data, the area outside the core remediation area is divided into an edge buffer zone and a natural recovery zone.

9. The method for monitoring mining area ecosystems based on net primary productivity of vegetation as described in any one of claims 1-5, characterized in that, The nonlinear regression model for predicting net primary productivity of vegetation to be trained includes an extreme gradient boosting model to be trained, and the interpreter model includes a tree interpreter model.

10. A storage medium, characterized in that, The storage medium stores a program to implement the mining area ecosystem monitoring method based on vegetation net primary productivity as described in any one of claims 1-9.

Images