Method for evaluating contribution of potential roof greening to urban ecological network

Abstract

The invention provides a method for evaluating contribution of potential roof greening to an urban ecological network. The method comprises the following steps: acquiring construction suitability characteristics of urban roof greening; training the multi-task network based on the construction suitability features, and then extracting potential roof greening; constructing an urban ecological network of the current green space; and fusing the potential roof greening with the current green space, simulating the reconstruction process of the potential roof greening participating in the urban ecological network, comparing the urban ecological network before and after the participation of the potential roof greening, and evaluating the contribution degree of the potential roof greening to the improvement of the urban ecological network. According to the method, a multi-task network is trained based on the building suitability characteristics of roof greening, and potential roof greening is extracted by using the trained network. And then, by simulating a process of reconstructing the urban ecological network by potential roof greening, considering potential influences on the urban ecological network from three aspects of source expandability, corridor connectivity and source corridor convergence degree.

Description

Technical Field

[0001] The present invention relates to the technical field of urban ecological network optimization, and specifically provides a method for evaluating the contribution of potential roof greening to an urban ecological network. Background Art

[0002] The built environment of urban centers urgently needs to explore new technologies and methods to meet diverse value pursuits within limited spatial resources. Green roofs are of great significance for improving urban ecological environments, but rapid and accurate identification of potential green roofs and assessment of their ecological impacts remain challenging. Current research suffers from two significant limitations. First, traditional methods for identifying potential green roofs require mapping all sample roof types, extracting their outlines, and then gradually removing roofs unsuitable for greening by learning texture features. This multi-step, overlaying process not only reduces efficiency but also offers limited accuracy when dealing with complex or contiguous building types. Second, while some research has begun to address the role of potential green roofs in urban ecological networks, these often treat them as independent elements, analyzing their distance relationship to existing green space networks. Few consider potential green roofs as integral components of network reconstruction and their integration with the ground-level green space system. In particular, when constructing urban ecological networks, relevant research often focuses on large-scale green spaces as the core of the network, overlooking the ecological connectivity potential of small-scale, highly dispersed green roofs.

[0003] Therefore, this field needs a new method to evaluate the potential contribution of rooftop greening to urban ecological networks to solve the above problems. Summary of the Invention

[0004] In order to overcome the above-mentioned defects, the present invention is proposed to provide a solution or at least partially solve the problem of being unable to effectively obtain the contribution evaluation of potential roof greening to the optimization of urban ecological network.

[0005] The present invention provides a method for evaluating the contribution of potential roof greening to an urban ecological network, comprising the following steps: obtaining buildability characteristics of urban roof greening; training a multi-task network based on the buildability characteristics, and extracting potential roof greening based on the trained multi-task network; constructing an urban ecological network of existing green spaces; integrating the potential roof greening with the existing green spaces, simulating the process of potential roof greening participating in the construction of the urban ecological network, and obtaining a reconstructed urban ecological network; and comparing the urban ecological networks before and after the participation of the potential roof greening, and evaluating the degree of contribution of the potential roof greening to the improvement of the urban ecological network.

[0006] In one technical solution of the above-mentioned method for evaluating the contribution of potential green roofs to urban ecological networks, the process of obtaining the suitability characteristics of urban green roofs includes: obtaining preset conditions based on at least the building slope, roof color, additional structure, and number of building floors; and determining whether the urban roof meets the preset conditions. If so, the urban roof is a potential green roof suitable for construction.

[0007] In one technical solution of the above-mentioned method for evaluating the potential contribution of roof greening to the urban ecological network, the multi-task network includes an edge detection branch module and a texture feature extraction module. The method includes at least the following steps to obtain the multi-task network: obtaining key contour information of urban roofs based on the edge detection branch module; obtaining texture characteristics of the surface of urban roofs based on the texture feature extraction module; adopting a feature fusion strategy to integrate the key contour information and texture characteristics to obtain preliminary recognition results; during the training process, selecting areas with low recognition confidence in the preliminary recognition results, and adopting an incremental supplementation strategy to expand the training samples, and then obtaining the final multi-task network through repeated training iterations.

[0008] In one technical solution of the above-mentioned method for evaluating the contribution of potential rooftop greening to the urban ecological network, the process of constructing the urban ecological network of the existing green space includes: obtaining suitable first potential source patches in the city, wherein the first potential source patches are screened source patches whose area reaches a threshold and has spatial independence; obtaining first potential ecological corridors based on the first potential source patches using the minimum cumulative path method; and the first potential ecological corridors connecting the various first potential source patches to form an interconnected network structure, namely, the urban ecological network of the existing green space.

[0009] In one technical solution of the above-mentioned method for evaluating the contribution of potential roof greening to the urban ecological network, the process of obtaining a potential ecological corridor based on the first potential source patch using the minimum cumulative path method includes: constructing an ecological resistance surface based on the land cover characteristics of the existing elements; calculating the minimum cumulative resistance path from each first potential source patch to the target patch on the ecological resistance surface using the minimum cumulative path method based on the first potential source patch, and extracting the calculated minimum cumulative resistance path as a potential ecological corridor.

[0010] In one technical solution of the above-mentioned method for evaluating the contribution of potential green roofs to the urban ecological network, the process of potential green roofs participating in the construction of the urban ecological network is simulated, and the process of reconstructing the urban ecological network includes: defining participating factors, which include existing green space and potential green roofs; analyzing the relationship between potential green roofs and existing green space based on spatial adjacency; obtaining potential source patches based on the relationship between potential green roofs and existing green space, and screening suitable second potential source patches from the potential source patches; obtaining second potential ecological corridors based on the second potential source patches using the minimum cumulative path method; and connecting each second potential source patch through the second potential ecological corridor to form an interconnected network structure, thereby reconstructing the urban ecological network.

[0011] In one technical solution of the above-mentioned method for evaluating the contribution of potential green roofs to urban ecological networks, the process of obtaining potential source patches based on the connection between potential green roofs and existing green spaces includes: if the connection is that the potential green roofs and existing green spaces have a shared boundary or spatial overlap, there is no need to add an additional corridor, and the potential green roofs and existing green spaces are integrated into a second potential source patch.

[0012] In one technical solution of the above-mentioned method for evaluating the contribution of potential green roofs to the urban ecological network, the existing green space includes at least trees and grass, and the method also includes setting an extended buffer zone outside the potential green roof.

[0013] In one technical solution of the above-mentioned method for evaluating the potential contribution of roof greening to the urban ecological network, the process of evaluating the potential contribution of roof greening to improving the urban ecological network based on the urban ecological network of the existing green space and the reconstruction of the urban ecological network includes: analyzing the urban ecological network of the existing green space and the improvement effect of the reconstruction of the urban ecological network on the basic element characteristics, and the basic element characteristics at least include: the scalability from the source node and the connectivity of the corridor; analyzing the urban ecological network of the existing green space and the reconstruction of the corridor convergence of all source nodes in the urban ecological network and its change value.

[0014] The above one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

[0015] In implementing the technical solution of the present invention, a method for evaluating the contribution of potential green roofs to an urban ecological network is provided, comprising: obtaining buildability characteristics of urban green roofs; training a multi-task network based on the buildability characteristics, and extracting potential green roofs based on the trained multi-task network; constructing an urban ecological network of existing green spaces; integrating the potential green roofs with the existing green spaces, simulating the process of potential green roofs participating in the construction of the urban ecological network, and obtaining a reconstructed urban ecological network; and comparing the urban ecological networks before and after the participation of potential green roofs to evaluate the contribution of potential green roofs to improving the urban ecological network. Compared with the prior art, the method for evaluating the contribution of potential green roofs to an urban ecological network provided by the present invention has the following advantages: the method trains a multi-task network based on the buildability characteristics and uses the trained network to extract potential green roofs. Compared with traditional methods for obtaining potential green roofs, which first require mapping all sample roof types and then extracting their outlines, and then gradually removing roofs unsuitable for greening by learning texture features, the multi-step overlay process not only reduces work efficiency but also has limited accuracy when processing complex or contiguous building types. This method directly learns the key characteristics of potential green roofs to achieve direct and accurate extraction, breaking away from the complex "extraction followed by exclusion" process of traditional methods. By simulating the process of potential green roofs participating in the construction of urban ecological networks, the potential impact of potential green roofs on the entire urban ecological network can be evaluated from two dimensions: fundamental element improvement and overall structural optimization; and from three perspectives: source expansion capacity, corridor connectivity, and source corridor convergence. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The disclosure of the present invention will become more easily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0017] Additionally, like numerals are used to denote like parts throughout the drawings, wherein:

[0018] Figure 1 1 is a flow chart of the main steps of a method for evaluating the potential contribution of green roofs to urban ecological networks according to one embodiment of the present invention;

[0019] Figure 2 is a schematic diagram of a process for obtaining preset conditions based on at least building slope, roof color, additional structure, and number of building floors according to one embodiment of the present invention;

[0020] Figure 3 is a spatial distribution map of rooftop greening potential in the central urban area of ​​Tianjin according to one embodiment of the present invention;

[0021] Figure 4is a kernel density map of potential green roof spaces according to one embodiment of the present invention;

[0022] Figure 5 is a local spatial autocorrelation analysis diagram of potential rooftop greening area according to one embodiment of the present invention;

[0023] Figure 6 is the potential green roof coverage at a community scale according to one embodiment of the present invention;

[0024] Figure 7 is a hotspot analysis map of potential rooftop greening coverage at a community scale according to one embodiment of the present invention;

[0025] Figure 8 is a spatial distribution map of current source areas according to an embodiment of the present invention;

[0026] Figure 9 is a diagram of the current urban ecological network construction according to an embodiment of the present invention;

[0027] Figure 10 is a spatial distribution map of potential sources according to one embodiment of the present invention;

[0028] Figure 11 is a potential urban ecological network construction diagram according to one embodiment of the present invention;

[0029] Figure 12 is a distribution map of potential sources of different contribution types according to an embodiment of the present invention;

[0030] Figure 13 is a distribution diagram of potential corridors of different contribution types according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] Some embodiments of the present invention are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0032] Example 1

[0033] See attached Figure 1 , Figure 1 FIG. 1 is a flow chart showing the main steps of a method for evaluating the potential contribution of green roofs to urban ecological networks according to an embodiment of the present invention. Figure 1 As shown, the method for evaluating the potential contribution of rooftop greening to the urban ecological network in the embodiment of the present invention mainly includes the following steps S1 to S5.

[0034] Step S1: Obtaining the suitability characteristics of urban rooftop greening. This step aims to collect and summarize various relevant characteristics that determine whether urban rooftops are suitable for greening construction.

[0035] Step S2: Train a multi-task network based on the suitability features, and extract potential green roofs based on the trained multi-task network. In this step, the multi-task network is a neural network architecture capable of simultaneously processing multiple related tasks. Here, the collected green roof suitability feature data is used as input, and the network learns the inherent correlation patterns between these features and the suitability of a roof for greening. Once the multi-task network is trained, the relevant feature data of numerous roofs in the city can be input into this trained network. Based on the learned patterns, the network will determine whether each roof is potentially suitable for greening, thereby screening and extracting those roofs with a high feasibility for greening.

[0036] Step S3: construct an urban ecological network of existing green spaces.

[0037] Step S4: Simulate the process of potential green roofs participating in the construction of the urban ecological network to obtain a reconstructed urban ecological network. Specifically, the urban ecological network is a complex system composed of various green spaces in the city and the connections between them. In this step, the previously extracted potential green roofs are treated as new ecological elements and added to the existing urban ecological network. Their connections and interactions with existing green spaces and ecological corridors are analyzed.

[0038] Step S5: Evaluate the potential contribution of potential green roofs to improving the urban ecological network based on the existing green space urban ecological network and the reconstructed urban ecological network. This step compares and analyzes the existing green space urban ecological network and the reconstructed urban ecological network including potential green roofs. This allows for an assessment of the potential contribution of potential green roofs from two dimensions and three aspects.

[0039] In one embodiment, the process of step S1, obtaining the suitability characteristics of urban roof greening, includes: Figure 2 As shown, the preset conditions are obtained based on at least the building slope, roof color, additional structure, and number of building floors; it is determined whether the urban roof meets the preset conditions. If so, the urban roof is a potential roof greening suitable for construction.

[0040] In this example, roofs that meet all of the following characteristics are labeled as positive samples: flat roof, black-gray, gray-green, or red-orange roof color, no more than 1 / 2 of the roof is covered with additional structure, and are not (super) high-rise buildings. Roofs that do not meet any of these conditions are classified as negative samples. Specifically, the pre-set condition for building slope: flat roofs are considered suitable for rooftop greening and are therefore considered positive samples; pitched roofs and irregularly shaped roofs are considered unsuitable and are therefore considered negative samples. The pre-set condition for roof color: roof color can indirectly reflect the roof material. For example, in Tianjin, it has been observed that most flat roofs appear black-gray or gray-green, with a small number appearing red-orange. These colors correspond to roof materials suitable for rooftop greening, so these roofs are classified as positive samples. Roofs that appear blue, red-orange, or are made of color-coated steel plates or other metal or clay tiles are not suitable for greening due to their material characteristics and are therefore classified as negative samples. The pre-set condition for additional structure: the proportion of the roof area covered by additional structure is determined. If the additional structure accounts for no more than half of the roof area, such a roof is relatively more suitable for greening in terms of space utilization and load-bearing capacity, and is therefore designated as a positive example. If the additional structure accounts for more than half, it will adversely affect greening construction, such as insufficient space and inconvenient layout, and is therefore classified as a negative example. This forms the precondition for roof suitability under the additional structure dimension. Precondition for the number of floors: In Google's high-resolution remote sensing imagery, high-rise buildings have distinct side elevations, shadows, and relatively complex forms, and their rooftop equipment and structures are also prominent. These characteristics make implementing greening on high-rise roofs difficult, such as construction difficulties and complex assessments of the impact on the overall building structure. Low-rise, multi-story, and low-rise buildings, on the other hand, have fewer or no side elevations, relatively simple roofs, and fewer structures, making them more suitable for rooftop greening. Therefore, the roofs of low-rise, multi-story, and low-rise buildings are classified as positive examples, and the roofs of high-rise buildings are classified as negative examples.

[0041] In one embodiment, the multi-task network includes an edge detection branch module and a texture feature extraction module, and the method includes at least the following steps to obtain the multi-task network: obtaining key contour information of urban roofs based on the edge detection branch module; obtaining texture characteristics of the urban roof surface based on the texture feature extraction module; using a feature fusion strategy to integrate the key contour information and texture characteristics to obtain preliminary recognition results; during the training process, selecting areas with low recognition confidence in the preliminary recognition results, and using an incremental supplement strategy to expand the training samples, and then obtaining the final multi-task network after repeated training iterations.

[0042] In this embodiment, the encoder uses a pre-trained ResNet50 as the backbone network and introduces the SE attention mechanism after the convolutional layers to dynamically adjust the feature responses of different channels and enhance feature representation. Furthermore, the improved residual block incorporates the attention mechanism, and residual connections further enhance the network's ability to extract features of potential green roof areas. The ASPP module is integrated after the backbone network. By combining convolutions with multiple dilation rates (including one 1x1 convolution and three 3x3 dilated convolutions with different dilation rates) with global average pooling, it effectively captures multi-scale contextual information. This effectively improves the network's ability to perceive potential green roof areas at different scales. An attention gating mechanism is employed to guide the decoder to focus on key details in low-level features. During the decoding process, stepwise upsampling and convolution operations restore the spatial resolution of the feature maps, and skip connections are used to fuse low-level features from the encoder. Furthermore, a context enhancement module and a hierarchical attention fusion module are introduced to further enhance the decoder's context modeling. At the same time, in order to enhance the refinement of the potential green roof outline, a parallel edge detection branch is designed in the decoder. The edge features are extracted through the convolution layer and the BatchNorm layer, and the edge feature map is finally upsampled to the same size as the input image. In view of the texture features of the potential green roof, a multi-scale texture module and a texture attention module are designed. The former extracts texture features at different scales through convolution kernels of different sizes. The latter uses the attention mechanism to adaptively enhance the expression ability of texture features. This branch enhances the model's ability to understand the details of the roof texture. The main segmentation branch, the edge detection branch, and the texture enhancement branch are fused together in a cascade manner. The feature maps from different sources are fused using 1x1 convolution. The fused multi-branch features enable the model to comprehensively utilize roof feature information of different granularities and types to obtain the potential green roof extraction results. An adaptive weight-based multi-task learning framework is used for the extraction of potential green roofs. The framework designs a region segmentation loss L region , edge detection loss L edge and texture enhancement loss L texture In order to automatically balance the relative importance of each task, learnable weight parameters w1, w2, and w3 are introduced. The total loss function is defined as:

[0043]

[0044] The region segmentation loss combines the binary cross entropy (BCE) loss, the Dice loss, and the Focal loss to address pixel-level classification, overall segmentation accuracy, and class imbalance, respectively. The edge detection loss uses BCE to optimize boundary extraction, while the texture enhancement loss maintains texture consistency by calculating the mean and standard deviation of the feature map. The exponential form of the weight parameters ensures positive weights and prevents them from approaching zero. This adaptive weighting mechanism eliminates the need for manual weight adjustment and dynamically adjusts the importance of each task during training. Regions with low recognition confidence are selected. These regions often indicate that the network's judgment accuracy needs improvement and are prone to misclassification. For these regions, an incremental supplementation strategy is used to expand the training samples. This involves collecting more rooftop data related to these low-confidence regions and adding it to the training sample set. The multi-task network then undergoes repeated training iterations based on this expanded training sample set. During each iteration, the network continuously adjusts its internal parameters (such as weights and biases in the neural network) to gradually enhance its ability to learn and judge roof features. After multiple such training iterations, the network's recognition accuracy reaches a relatively reliable and stable state, ultimately obtaining a multi-task network that meets the requirements and can accurately extract potential roof greening areas.

[0045] In one embodiment, step S3, the process of constructing the urban ecological network of the existing green space, includes: obtaining suitable first potential source patches in the city, wherein the first potential source patches are screened source patches whose area reaches a threshold and has spatial independence; obtaining a first potential ecological corridor based on the first potential source patches using the minimum cumulative path method; the first potential ecological corridor connects each first potential source patch to form an interconnected network structure, namely, the urban ecological network of the existing green space.

[0046] In this embodiment, in landscape ecology, the urban landscape can be disassembled into three important parts: source, corridor and matrix. The source and corridor can be abstracted into nodes and edges, and together they build the basic skeleton structure of the ecological network of the urban center. The source patch plays an extremely critical role in the entire ecosystem. It is not only a place for many animals and plants to survive and live, but also an important place for various species to move, reproduce and other ecological processes to occur. According to the "area-species" theory, it can be known that the size of the source patch has a decisive influence on urban biodiversity, and the two show an obvious positive correlation, that is, the larger the area of ​​the source patch, the more species richness and population size will increase. And the study found that in order to ensure the normal survival of the biological community and maintain a certain level of biodiversity, the area of ​​the source patch cannot be less than 50,000 m 2Otherwise, biodiversity will decline significantly. Therefore, when constructing an urban ecological network of existing green spaces, it is particularly important to screen source patches.

[0047] Specifically, the first potential source patch screening criterion is that its area must reach a specified threshold, which is set at 50,000 m 2 At the same time, these patches also need to have spatial independence, meaning they are relatively independent in spatial distribution and have clear boundaries with other surrounding areas. The first potential ecological corridor plays the role of connecting various first potential source patches. Numerous first potential source patches are interconnected through the first potential ecological corridor, ultimately forming a complete and interconnected network structure. This network structure is the urban ecological network of the existing green space.

[0048] In one embodiment, the process of obtaining a potential ecological corridor based on the first potential source patch using the minimum cumulative path method includes: constructing an ecological resistance surface based on the land cover characteristics of the current elements; calculating the minimum cumulative resistance path from each first potential source patch to the target patch on the ecological resistance surface based on the first potential source patch using the minimum cumulative path method, and extracting the calculated minimum cumulative resistance path as a potential ecological corridor.

[0049] In this embodiment, according to the different land cover types in various areas of the city, corresponding resistance values ​​are assigned to them. For each first potential source patch, it is used as the starting point. Then, on the constructed ecological resistance surface, a specific model (such as the commonly used LCP model or its improved model, etc., these algorithms can find the shortest path in a weighted network diagram, where the weight is the resistance value corresponding to the land cover) is used to calculate the minimum cumulative resistance path between it and each target patch. During the calculation, the first potential source patch is considered to start from, and how these resistances accumulate when passing through areas of different land cover types (that is, areas with different resistance values), and ultimately the path that minimizes the cumulative resistance is found. After the above calculation process, the minimum cumulative resistance path between the target patches can be obtained for each first potential source patch. These paths are the most ecologically conducive channels for species to move between different source patches. They avoid areas with high resistance and connect different source patches along areas with low resistance or relatively easy passage as much as possible. These calculated minimum cumulative resistance paths are extracted and can be regarded as potential ecological corridors.

[0050] In one embodiment, the process of simulating the participation of potential green roofs in the construction of an urban ecological network to obtain a process of reconstructing the urban ecological network includes: defining participating factors, which include existing green space and potential green roofs; analyzing the relationship between potential green roofs and existing green space based on spatial adjacency; obtaining potential source patches based on the relationship between potential green roofs and existing green space, and screening suitable second potential source patches from the potential source patches; obtaining second potential ecological corridors based on the second potential source patches using the minimum cumulative path method; and connecting each second potential source patch through the second potential ecological corridor to form an interconnected network structure, thereby reconstructing the urban ecological network.

[0051] In this embodiment, an area of ​​50,000 m 2 A threshold is used to screen the potential source patches. After identifying the second potential source patches, an ecological corridor connecting them, known as the second potential ecological corridor, is constructed. Here, the minimum cumulative path method is also used, based on the land cover characteristics after incorporating potential green roofs. The integration of potential green roofs changes the urban land cover, and different areas (such as newly added green roof areas, integrated source patches, and other unchanged land types) present new resistances to ecological processes such as species migration. By analyzing these new land cover characteristics, each area is assigned a corresponding resistance value. Then, starting from the second potential source patch, an algorithm is used to calculate the minimum cumulative resistance paths between them. This method identifies the paths between source patches that are most conducive to species migration and ecological connectivity under the new urban ecological landscape. These paths are known as the second potential ecological corridor. Finally, the second potential ecological corridor plays a key role in connecting various second potential source patches. Numerous second potential source patches are interconnected with the help of these second potential ecological corridors, thus forming a complete and interconnected network structure. This network structure is what we call the reconstructed urban ecological network.

[0052] In one embodiment, the process of obtaining potential source patches based on the connection between potential green roofs and existing green spaces includes: if the connection is that the potential green roofs and existing green spaces have a shared boundary or a spatial overlap, there is no need to add an additional corridor, and the potential green roofs and existing green spaces are integrated into a second potential source patch.

[0053] In this embodiment, when the potential green roof and the existing green space share a boundary or have spatial overlap, it means that they are closely connected in space and can be regarded as being able to spontaneously exchange information in terms of ecological functions. Therefore, there is no need to build new corridors to connect them, and the two can be integrated.

[0054] In one embodiment, the existing green space includes at least trees and grass, and the method further includes providing an extended buffer zone outside the potential green roof.

[0055] In this example, potential green roofs are small and scattered, so they need to be integrated with the surrounding area in a certain way. The green roofs themselves and the 3m buffer zone (set to take into account the geometric errors caused by facade interference due to the oblique angle in Google high-resolution images. This buffer zone can ensure the accuracy and completeness of the analysis to a certain extent) are participating factors. The purpose is to enable them to better integrate into the urban ecological network and play an ecological role together with other green spaces.

[0056] In one embodiment, the process of evaluating the potential contribution of potential rooftop greening to improving the urban ecological network based on the existing green space urban ecological network and the reconstruction of the urban ecological network includes: analyzing the improvement effect of the existing green space urban ecological network and the reconstruction of the urban ecological network on basic element characteristics, and the basic element characteristics include at least: the scalability from the source node and the connectivity of the corridor; analyzing the optimization effect of the existing green space urban ecological network and the reconstruction of the urban ecological network on the overall structure, and the overall structure optimization refers to the corridor convergence degree of all source nodes and its change value.

[0057] In this example, when potential green roofs participate in the construction of the urban ecological network, they are integrated with existing surrounding green spaces based on shared boundaries or spatial overlap. This integration of previously dispersed potential green roof areas with existing green spaces creates new, larger-scale potential source patches, increasing the number of source nodes and broadening their distribution. When evaluating potential contributions, it's important to compare and analyze the effects of these fundamental improvements with overall structural optimization.

[0058] Specifically, the assessment of source elements is based on a classification and quantification of their evolutionary characteristics. Initially, the number and area of ​​existing source areas M0 are defined as N0 and A0, respectively. After potential green roofs are integrated into the ecological network, the potential source areas are restructured and transformed into three types: original source areas M1, expanded source areas M2, and newly added source areas M3. M1 maintains its spatial and positional stability compared to its initial values, providing a stable support for the ecological network. M2, through potential green roofs, has expanded in scale, enhancing the ecological function of the source area, and the increase in area represents the degree of strengthening of the network node. M3, through potential green roofs, transforms areas that previously did not meet the source threshold into new ecological nodes, demonstrating its potential for expansion. By comparing the change in number ΔN, area increase ΔA, and spatial distribution of these three types of source areas, the improvement effect of potential green roofs on existing source areas is quantitatively assessed.

[0059] The improvement of corridor connectivity is achieved by subdividing and evaluating the corridor through the properties of the source nodes. Considering that the expanded source M2 essentially represents the evolution of the initial ecological map, and has the same spatial position stability as the original source M1, M1 and M2 are uniformly defined as existing sources. The newly added source M3 is an incremental source. Therefore, corridors can be divided into three categories: EE corridors, EI corridors, and II corridors. Among them: EE corridors connect two existing sources, and the changes in their cumulative resistance value, corridor length, and spatial distribution reflect the improvement effect of potential roof greening on the existing ecological network; EI corridors connect existing sources and incremental sources, representing the connection between the newly added source and the existing ecological network; Class II corridors connect corridors between two incremental sources and reflect the network structure formed between the newly added sources.

[0060] The Corridor Convergence Index (CCI) characterizes the potential frequency of source nodes as pathways for biological migration. A higher CCI value indicates that the source node has a stronger ability to control the overall ecological network and is a key central node in the network. For source node i, its Corridor Convergence Index is the number of corridors that spatially intersect with source node i among all potential ecological corridors obtained through minimum cost path analysis. Its calculation formula is:

[0061]

[0062] Among them, CCI i represents the corridor convergence degree of source node i, δ ijk Indicates whether the minimum cost path connecting node j and node k intersects with node i. If it intersects, the value is 1; if it does not intersect, the value is 0. n is the total number of source patches. When j = k or i∈{j,k}, O ijk =0.

[0063] The following is a specific example: fine recognition based on deep learning multi-task network, and setting 50m 2 The above area threshold screening conditions, such as Figure 3 A total of 21,244 building roofs with potential for green roofs were identified, with a total area of ​​13,452,243 m 2 , with an average area of ​​633m 2 Among them, the largest potential roof area is 24,927m 2 , is a commercial complex located in Hexi District.

[0064] Kernel density analysis results of potential green roofs (such as Figure 4 The results (shown in Figure 2) reveal that potential green roofs exhibit a "multi-center, decentralized" spatial distribution pattern. Several highly concentrated areas have formed in the six districts of Tianjin, and clear agglomeration centers have also emerged in some peripheral areas.

[0065] In order to explore the spatial correlation of potential roof greening area, based on spatial autocorrelation analysis, the results showed that the global Moran index of potential roof greening area in the study area was 0.031624, the Z value was 32.600684, and the P value was less than 0.000001, indicating that there was a statistically significant and very weak positive spatial autocorrelation. Further LISA local spatial autocorrelation analysis found (such as Figure 5 (as shown in the figure), most of the potential green roofs in the study area do not have a significant spatial correlation pattern, but a small number of high-value clusters (High-High) and low-value clusters (Low-Low) and some spatial anomaly areas (High-Low and Low-High) appear in local areas of the six districts in the city. This shows that although the potential green roof area only shows weak spatial dependence as a whole, it still forms a significant local spatial agglomeration phenomenon in specific areas.

[0066] In order to gain a deeper understanding of the spatial characteristics of potential green roofs at the community scale, the potential green roof coverage rate at the community scale, i.e., the ratio of potential green roof area to community area, was quantified based on 686 community units in the central urban area of ​​Tianjin. Figure 6 The results show that the average community-level potential rooftop greening coverage rate is 0.10, with the highest coverage rate of 0.32 in Rongcaili Community, Wangchuanchang Street, Hebei District. However, 13 communities have a potential rooftop greening coverage rate of 0, showing significant spatial heterogeneity. Hot spot analysis based on Getis-OrdGi* statistics ( Figure 7 The study further found that, at a 99% confidence level, areas with high potential green roof coverage showed clear spatial clustering, with a stable core of high values ​​forming at the junction of Nankai, Hexi, and Heping districts. This identification of hot and cold areas of potential green roof coverage at the community level provides the most fundamental scientific basis for tailoring differentiated green roof renewal strategies.

[0067] Results of urban ecological network construction before and after the participation of potential green roofs: A total of 100 source patches were identified in the study area ( Figure 8 ). The area ranges from 50,192 to 549,703 m 2 The average area is 109,881m 2 The total area of ​​the current source patch is 11,206,172m 2 , accounting for 5% of the total area of ​​the study area and 24% of the total area of ​​existing green space. These source patches show obvious "edge-dense-center-sparse" spatial distribution characteristics. Based on the minimum cumulative resistance path model, 4950 potential ecological corridors were identified, which is a highly complex ecological network ( Figure 9The corridor length distribution ranges from 0.23 to 25.13 km, with an average length of 10.63 km. The corridor network density reaches 271 km / km. 2 In areas with sparse water sources, the corridors are mainly connected by roads and green belts beside rivers, forming a "grid" structure; while in areas with dense water sources, a complex interwoven network structure is formed.

[0068] By integrating the potential green roofs with the existing green spaces, the ecological network structure of the study area has been significantly restructured. The number of source nodes has increased from 100 to 143. In terms of spatial distribution, urban built-up areas (such as Heping District) that originally had scarce sources have formed new potential source nodes through the strategic participation of potential green roofs, effectively supplementing the existing sources ( Figure 10 At the same time, new source node clusters were formed in source-intensive areas (such as the Yingbin Hotel area in Hexi District and the North Sports Institute area), further strengthening the ecological function of the region and improving the homogeneity of network nodes ( Figure 11 ). The number of corridors increased from 4950 to 10153. The corridor length distribution range is 0.23-25.58km, and the maximum value has increased. This is because the newly added source nodes have generated longer corridors. However, the average length of the corridors has decreased overall, and the vast majority of corridor paths have been shortened. In addition, the average minimum resistance cost of the corridors has increased. This is because 1) the increase in the total number of corridors has covered some areas with relatively large resistance, which has increased the average minimum resistance of the entire ecological network; 2) the addition of source nodes in source-dense areas has made the local network structure more complex, which may have led to the resistance value of the corridor paths in the area being further amplified. When high-resistance paths are included in the calculation of the average minimum resistance, the overall average resistance value is raised.

[0069] Therefore, integrating potential green rooftops into the urban ecological network supplements existing source nodes, increasing their number and area, and optimizing their spatial configuration, particularly in urban core areas where source nodes are relatively scarce. The increased number of corridors and the reduced average distance between them mean that organisms can spread between urban source sites through more diverse and shorter pathways. This dual increase in nodes and corridors significantly enhances the network's resilience to interference.

[0070] The improvement effect of potential green roof participation on the urban ecological network: The participation of potential green roofs significantly expanded the source nodes of the urban ecological network. This includes 63 potential expanded source nodes after potential green roof participation, and 43 new source nodes formed due to the participation of green roofs ( Figure 12 ). Changes in corridor connectivity ( Figure 13), the average cumulative minimum resistance value of EE corridors (connections between existing sources) increased from 13,634 to 14,468, and the average length increased slightly from 10.63 km to 10.68 km. The network density also increased accordingly from 271.08 to 272.18. This change is actually due to the spatial structural reorganization caused by source expansion: when the original source area expands, the shape and scope of the source change, which in turn affects the calculation of the minimum cumulative resistance path between sources. Although source expansion increases the patch area, it also changes the spatial form and center of gravity of the source, which may cause the original optimal path to no longer be the optimal choice, requiring a more circuitous route, resulting in a slight increase in the average cumulative resistance value. However, the average minimum cumulative resistance value and average length of EI and II corridors have decreased, indicating that they have shortened the connection between nodes to a certain extent, providing more possible paths for species migration and improving the redundancy of the network.

[0071] Results show that the inclusion of potential green roofs significantly optimizes the functional structure of the urban ecological network. Looking at changes in the corridor convergence index (CCI), the average CCI of existing green space nodes was 260.3. This increased to 443.8 after the inclusion of potential green roofs, a 70.5% increase. Spatially, the vast majority of the original 102 nodes saw significant increases in CCI values, with some nodes (such as those near number 20) experiencing particularly significant increases, reaching a maximum of approximately 1600. The 41 newly added green roof nodes also exhibited high CCI values, reaching a maximum of nearly 1500. While the CCI values ​​of most nodes in the existing network ranged from 200 to 400, indicating relatively balanced connectivity, the inclusion of green roofs resulted in the formation of several key central nodes with CCI values ​​exceeding 1000, demonstrating strong control over the overall ecological network. Furthermore, the inclusion of potential green roofs not only enhances the connectivity of existing nodes but also fills gaps in the original network by adding new nodes. The overall increase in the CCI value indicates increased network redundancy and the formation of more alternative pathways, which is beneficial for improving the stability and anti-interference ability of urban ecological networks and promoting the migration and diffusion of organisms. These results indicate that potential green roofs, as a supplementary strategy for urban green space systems, can effectively optimize the functional structure of urban ecological networks and enhance their overall connectivity and stability.

[0072] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the original technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A method for evaluating the potential contribution of green roofs to urban ecological networks, characterized by: include: Obtain the suitability characteristics of urban green roofs; Training a multi-task network based on the adaptability features, extracting potential green roofs based on the trained multi-task network, and achieving end-to-end extraction of potential green roofs; Constructing an urban ecological network of existing green spaces; By integrating potential green roofs with existing green spaces, the process of potential green roofs participating in the construction of urban ecological networks was simulated, and a reconstructed urban ecological network was obtained. The urban ecological networks before and after the participation of potential green roofs are compared to evaluate the contribution of potential green roofs to improving the urban ecological network.

2. The method according to claim 1, characterized in that The process of obtaining the suitability characteristics of urban green roofs includes: Obtain preset conditions based on at least building slope, roof color, additional structures, and number of building floors; It is determined whether the urban roof meets the preset conditions. If so, the urban roof is a potential green roof suitable for construction.

3. The method according to claim 1, characterized in that The multi-task network includes an edge detection branch module and a texture feature extraction module. The method at least includes the following steps to obtain the multi-task network: Acquire key contour information of urban roofs based on the edge detection branch module; Acquire the texture characteristics of urban roof surfaces based on texture feature extraction module; A feature fusion strategy is used to integrate key contour information and texture characteristics to obtain preliminary recognition results; During the training process, areas with low recognition confidence in the preliminary recognition results are selected, and the training samples are expanded using an incremental supplement strategy. After repeated training iterations, the final multi-task network is obtained.

4. The method according to claim 1, wherein The process of constructing an urban ecological network of existing green spaces includes: Obtaining a suitable first potential source patch in the city, wherein the first potential source patch is a screened source patch whose area reaches a threshold and has spatial independence; Based on the first potential source patch, the minimum cumulative path method is used to obtain the first potential ecological corridor; The first potential ecological corridor connects each of the first potential source patches to form an interconnected network structure, namely, the urban ecological network of the existing green space.

5. The method according to claim 4, characterized in that The process of obtaining potential ecological corridors based on the first potential source patch using the minimum cumulative path method includes: Constructing ecological resistance surface based on land cover characteristics of current elements; Based on the first potential source patch, the minimum cumulative path method is used to calculate the minimum cumulative resistance path from each first potential source patch to the target patch on the ecological resistance surface, and the calculated minimum cumulative resistance path is extracted as a potential ecological corridor.

6. The method according to claim 3, characterized in that The process of simulating the potential green roof participation in the construction of the urban ecological network is obtained, and the process of reconstructing the urban ecological network includes: Defining participation elements, including existing green space and potential green roofs; Analyze the connection between potential green roofs and existing green spaces based on spatial adjacency; Obtaining potential source patches based on the connection between the potential green roof and the existing green space, and screening out suitable second potential source patches from the potential source patches; Based on the second potential source patch, the minimum cumulative path method is used to obtain the second potential ecological corridor; The second potential ecological corridor connects each of the second potential source patches to form an interconnected network structure, that is, reconstructing the urban ecological network.

7. The method according to claim 6, characterized in that The process of obtaining potential source patches based on the connection between the green roof and existing green space includes: If the connection is that the potential green roof and the existing green space share a boundary or have spatial overlap, there is no need to add an additional corridor, and the potential green roof and the existing green space can be integrated into a second potential source patch.

8. The method according to claim 6, characterized in that The existing green space includes at least trees and grass, and the method further includes setting an extended buffer zone outside the potential roof greening.

9. The method according to any one of claims 1 to 8, characterized in that The process of evaluating the potential contribution of potential rooftop greening to improving the urban ecological network based on the urban ecological network of the existing green space and the reconstructed urban ecological network includes: Analyze the urban ecological network of the existing green space and the improvement effect of the reconstructed urban ecological network on basic element characteristics, wherein the basic element characteristics include at least: scalability from the source node and corridor connectivity; The urban ecological network of the existing green space and the corridor convergence degrees and their change values ​​of all source nodes in the reconstructed urban ecological network are analyzed.

Images