Patch Aggregation Research Articles

Graph Neural Network-Based Molecular Property Prediction with Patch Aggregation.

Graph neural networks (GNNs) have emerged as powerful tools for quantum chemical property prediction, leveraging the inherent graph structure of molecular systems. GNNs depend on an edge-to-node aggregation mechanism for combining edge representations into node representations. Unfortunately, existing learnable edge-to-node aggregation methods substantially increase the number of parameters and, thus, the computational cost relative to simple sum aggregation. Worse, as we report here, they often fail to improve predictive accuracy. We therefore propose a novel learnable edge-to-node aggregation mechanism that aims to improve the accuracy and parameter efficiency of GNNs in predicting molecular properties. The new mechanism, called "patch aggregation", is inspired by the Multi-Head Attention and Mixture of Experts machine learning techniques. We have incorporated the patch aggregation method into the specialized, state-of-the-art GNN models SchNet, DimeNet++, SphereNet, TensorNet, and VisNet and show that patch aggregation consistently outperforms existing learnable and nonlearnable aggregation techniques (sum, multilayer perceptron, softmax, and set transformer aggregation) in the prediction of molecular properties such as QM9 thermodynamic properties and MD17 molecular dynamics trajectory energies and forces. We also find that patch aggregation not only improves prediction accuracy but also is parameter-efficient, making it an attractive option for practical applications for which computational resources are limited. Further, we show that Patch aggregation can be applied across different GNN models. Overall, Patch aggregation is a powerful edge-to-node aggregation mechanism that improves the accuracy of molecular property predictions by GNNs.

Response of Surface Runoff Evolution to Landscape Patterns in Karst Areas: A Case Study of Yun–Gui Plateau

To control and improve the phenomena of rocky desertification and soil erosion in karst landform areas, which are caused by a series of human factors that include social and economic development and human activities, China has successively introduced many policies, resulting in spatial and temporal changes in the landscape pattern of the southern karst area. In this study, land use transfer intensity maps, the grid method, the sample line method, the semivariogram method, and the Spearman analysis method are used to explore the spatial and temporal evolutions in surface runoff as responses to landscape pattern and policy factors in karst landform area. Therefore, this study provides theoretical and policy support for improving the regional landscape structure, optimizing the landscape layout, introducing regional policies, reducing surface runoff, and alleviating soil erosion. The results show that the best scale for the study of landscape patterns in the southern karst area is 3000 m. Forests are the land type that make up the highest proportion in the southern karst area, and they have the strongest interception capacity for surface runoff. The spatial and temporal distributions of the surface runoff are significantly different, and urban expansion has led to an increase in impervious runoff year over year. Runoff is positively correlated with the Shannon diversity index (SHDI), patch density (PD), and landscape shape index (LSI). The stronger the landscape heterogeneity, the more runoff. DIVISION is positively correlated with forest runoff and negatively correlated with other land types. The higher is the degree of aggregation of impervious patches, the higher the regional runoff rate. The more dispersed the forest patches are, the smaller the area proportion, and the greater the runoff. In addition, policy factors have a significant impact on surface runoff.

Source Apportionment and Response to Landscape Pattern of Health Risk of Cultivated Soil Heavy Metals

Heavy metal pollution mainly caused by human activities is becoming increasingly prominent and threatening human health and ecosystem safety in soil, which is a non-renewable natural resource that humans rely on for survival and development. Assessment and analysis of soil heavy metal health risk is significant for protecting human health, preventing soil pollution, and maintaining ecosystem security. Based on the investigation of heavy metals, including Cr, Pb, Cd, As, and Hg, in cultivated soil in Liuhe District, the health risk assessment model was used to identify the health risk characteristics of heavy metals, and the spatial distribution, main sources, and responses to landscape patterns were explored by using inverse distance weight interpolation, positive definite matrix factorization, landscape pattern index, and redundancy analysis. The results showed that the coefficients of variation corresponding to Cr, Pb, Cd, As, and Hg in the study area were 0.19, 0.36, 0.23, 0.52, and 1.16, respectively, all of which belonged to moderate or high variability, indicating that they had high spatial heterogeneity and were susceptible to human factors. Cr, Pb, and As were the main health risk characteristic factors in the study area, with the carcinogenic risks to children ranging from 13.307×10-6 to 38.400×10-6, 0.839×10-6 to 3.250×10-6, and 4.548×10-6 to 16.680×10-6, respectively, which were higher than those in adults. Agricultural production activities, industrial production, and transportation activities were the main sources of heavy metals, with carcinogenic risks to children of 17.946×10-6 and 12.941×10-6, respectively. Furthermore, high-risk areas caused by agricultural production activities were mainly concentrated in the northern area of Liuhe District and showed an increasing trend from south to north and from the center to the periphery. The surrounding areas caused by industrial production activities and transportation were mainly concentrated in the chemical industry park and economic development zone of Liuhe District and showed a spatial agglomeration feature of decreasing from south to north and from the core to the periphery. The cumulative explanatory value of the landscape pattern index for the comprehensive carcinogenic risk to children was 0.463, and patch density, patch proportion in landscape area, patch aggregation degree, and maximum patch index had significant effects on the comprehensive carcinogenic risk in children, and the corresponding explanatory values were 0.422, 0.274, 0.351, and 0.232, respectively. This study had important theoretical and practical significance for expanding the perspective of environmental health research, promoting the transformation of soil heavy metal management methods and safeguarding regional population health.

Hierarchical Patch Aggregation Transformer for Motion Deblurring

The encoder-decoder framework based on Transformer components has become a paradigm in the field of image deblurring architecture design. In this paper, we critically revisit this approach and find that many current architectures severely focus on limited local regions during the feature extraction stage. These designs compromise the feature richness and diversity of the encoder-decoder framework, leading to bottlenecks in performance improvement. To address these deficiencies, a novel Hierarchical Patch Aggregation Transformer architecture (HPAT) is proposed. In the initial feature extraction stage, HPAT combines Axis-Selective Transformer Blocks with linear complexity and is supplemented by an adaptive hierarchical attention fusion mechanism. These mechanisms enable the model to effectively capture the spatial relationships between features and integrate features from different hierarchical levels. Then, we redesign the feedforward network of the Transformer block in the encoder-decoder structure and propose the Fused Feedforward Network. This effective aggregation enhances the ability to capture and retain local detailed features. We evaluate HPAT through extensive experiments and compare its performance with baseline methods on public datasets. Experimental results show that the proposed HPAT model achieves state-of-the-art performance in image deblurring tasks.

A critical role of the capital green belt in constraining urban sprawl and its fragmentation measurement

Urban green belts (UGBs), an important green infrastructure of cities, have the functions of constraining urban sprawl, which are of great significance to urban green, resilient and sustainable development. However, it remains unclear how to quantify the constraining effect of UGBs on urban sprawl and the degree of fragmentation after UGBs constrain urban sprawl. Here, we proposed a UGBs extraction method to explore the spatial evolution of UGBs in the rapidly expanding megacity Beijing from 2000 to 2020. We also quantitatively investigated the role of UGBs in constraining urban sprawl and measured the degree of fragmentation of the UGBs after constraining urban sprawl with the help of the constraint assessment index, the Urban Annual Ring model, and the landscape pattern index. Results showed that: (1) Beijing’s UGBs area showed a “V”-shaped trend of first decreasing and then increasing from 2000 to 2020, decreasing from 332.61 km2 in 2000 to 303.41 km2 in 2015, and then increasing to 369.42 km2 in 2020, which is equivalent to about 54 Beijing Olympic Forest Parks. (2) The UGBs’ role in constraining urban sprawl was characterized by gradual enhancement in three periods from 2005 to 2020. The boundary containment rate (BCR) of urban sprawl was 0.48, 0.22, and 0.14 respectively. After the construction of the newly planned UGBs in 2005, the speed of urban sprawl from 2010 to 2015 decreased by 10.90%. (3) The fragmentation of UGBs showed an upward trend and connectivity showed a downward trend. The patch density (PD) characterizing fragmentation increased from 17.22 in 2000 to 31.60 in 2020, and the patch aggregation index (AI) characterizing connectivity decreased from 86.92 to 79.73. The research can provide a reference for other similar sized cities in UGB planning and government decision-making.

Greenness and its composition and configuration in association with allergic rhinitis in preschool children

BackgroundFew studies focus on the associations of green space composition and configuration with children's allergic rhinitis (AR). MethodsA multi-center population-based cross-sectional study was performed in 7 cities in mainland of China between 2019 and 2020, recruiting 36,867 preschool children. Information on the current AR symptoms and demographics were collected by questionnaire. Exposure to residential greenness was estimated by Normalized Difference Vegetation Index (NDVI, 1000 m buffer) around the residences. Greenness composition was estimated in 3 main categories: forest, grassland, shrubland. Configuration of each category and total greenness (a spatial resolution of 10 m × 10 m) was estimated by 6 landscape pattern metrics to quantify their area, shape complexity, aggregation, connectivity, and patch density. Exposure to daily ambient particulate matter (PM1, PM2.5 and PM10, a spatial resolution of 1 km × 1 km) was estimated. Multilevel logistic regression models were applied to analyze the associations of greenness and its composition and configuration with AR, and mediation effects by PMs were examined by mediation analysis models. ResultsThe prevalence of self-reported current AR in preschool children was 33.1%. Two indicators of forest, Aggregation Index of forest patches (AIforest) (odds ratio (OR):0.92, 95% Confidential Interval (CI): 0.88–0.97), and Patch Cohesion of forest (COHESIONforest) (OR: 0.93, 95% CI:0.89–0.98) showed significantly negative associations with AR symptoms. Mediation analyses found the associations were partially mediated by PMs. Age, exclusive breastfeed duration and season were the potential effect modifiers. The associations varied across seven cities. ConclusionOur findings suggest the inverse associations of the aggregation and connectivity of forest patches surrounding residence addresses with AR symptoms. Since the cross-sectional study only provides associations rather than causation, further studies are needed to confirm our results as well as the underlying mechanisms.

How Do the Dynamics of Urbanization Affect the Thermal Environment? A Case from an Urban Agglomeration in Lower Gangetic Plain (India)

Urban growth and development has significantly affected urban heat island (UHI) due to urbanization. Particularly in the cities in developing countries, the assessment of UHI has emerged as one of the core research themes as it significantly affects the ecological environment and livability in cities. Thus, the assessment of UHI is crucial for climate mitigation and sustainable urban landscape planning. This study identifies the dynamics of landscape patterns and the impact of composition and configuration on the thermal environment in English Bazar Urban Agglomeration (EBUA), Eastern India, along the urban–rural gradient (URG) approach. Geospatial approaches and spatial metrics were employed to assess the impact of the landscape pattern on the thermal environment. Descriptive and inferential statistics have also been used to find the effects of landscape patterns on the thermal environment. The result has also been validated based on the location and correlation analysis. The built-up area increased by about 63.54%; vegetation covers and water bodies declined by 56.72% and 67.99% from 2001 to 2021. Land surface temperature (LST) decreased with increasing distance from the core of the city. LST declined by about 0.45 °C per kilometer from the core of the city towards the outside. LST had a positive correlation with IS and a negative correlation with green space (GS) and blue space (BS). The mean aggregation of the impervious patches was larger (73.21%) than the GS (43.18%) and BS (49.02%). The aggregation of impervious surface (IS) was positively correlated, and aggregations of GS and BS had a negative correlation with LST. Findings suggest that the spatial composition and configuration of the impervious surface, GS, and BS must be considered in landscape planning and design framework to make the city more livable.

Linking soil erosion and sediment yield with landscape heterogeneity: Exploration using the lower Jinsha River Basin, China

Landscape heterogeneity, including compositional and spatial configuration, changes soil erosion and sediment yield (SESY) in watersheds by regulating hydrological processes. However, existing research has paid little attention to landscape composition. The Lower Jinsha River Basin (LJRB) has significant landscape heterogeneity and the highest sediment yields in the Yangtze River Basin, China. The spatially distributed soil erosion model WaTEM/SEDEM was used to evaluate the spatio-temporal changes in SESY in the LJRB from 1990 to 2020. Area information conservation method and semi-variogram were utilized to determine the optimal grain and extent size of landscape analysis. The landscape heterogeneity characteristics of the LJRB were analyzed for landscape composition and spatial configuration. Finally, statistical analysis was performed to determine the differences in SESY among different landscape composition types (LCTs) and to explore the relationship between landscape configuration and SESY. The optimal grain and extent size for analysis of the landscape pattern in the LJRB were 150 and 2100 m, respectively. Based on this, three types and ten subtypes of landscape composition were identified. Landscapes composed of forest, shrubs, grassland, and farmland were predominant in the study area, with a relatively even distribution of patches and minimum dominance of specific patch types. Additionally, the landscapes composed of forest, shrubs, grassland, farmland, and impervious surfaces were diverse, with the highest fragmentation. Landscapes comprising farmland and impervious surfaces occupied a minor proportion of the study area, but specific patches exhibited higher landscape connectivity. The magnitude of SESY in the LJRB gradually decreased during 1990–2020. Farmland and grassland were the land types with the highest magntidue of SESY in this region. Substantial sediment has been deposited in Xiangjiaba and Xiluodu reservoirs since they commenced operations. Increased proportions of farmland patches in some LCTs have significantly enhanced SESY. For LCTs dominated by farmland and impervious surfaces, reducing the dominance of the largest patch, decreasing patch aggregation, and increasing landscape-type diversity and evenness have helped mitigate SESY. The opposite is true for LCTs dominated by forests, shrubs, and grasslands. Collectively, the results of this study provide valuable progress in the capacity to support landscape restoration and comprehensive management of soil and water degradation in mountainous watersheds with similar characteristics to the LJRB.

Urban ecological environment evaluation and influencing factors analysis at the residential quarter-level in Xi'an, China

Residential quarters are a critical component of urban areas, which significantly affects people's health and quality of life. In recent decades, the high-density and crowded residential quarters have become the mainstream organizational model of construction in China, because of the rapid and massive urbanization in many cities. This results in the contradiction between the eco-environment needs and the accommodation needs. Therefore, a systematic evaluation of the eco-environment in residential quarters is urgently needed to address the drawbacks of current residential quarters' planning and design. This paper aims to explore ways to improve the eco-environment with residential quarters from the perspective of China. We first used the Remote Sensing Ecological Index (RSEI) to evaluate the eco-environmental quality of 3, 756 residential quarters in the main urban areas of Xi 'an, China. Then, we quantitatively analyzed the landscape patterns of the impervious surfaces in these residential quarters. Finally, we characterized the impacts of residential quarters layout on the ecological environment by using the vegetation coverage (FVC), the normalized difference bare soil index (NDBSI), building density (BD), the residential density (RD), and the population density (PD) as main factors. The results are as follows: (1) Only 20.5 % of residential quarters with good eco-environments were distributed between the Second and Third Ring Road, indicating the need for improvement in creating livable eco-environments. (2) The shape of the impervious surface of excellent ecological quarters is complex and broken, the average patch area is smaller, and the degree of aggregation and combination of patches is lower compared to residential quarters with poor ecological quality. Therefore, it is recommended to reduce the connectivity of the impervious surface to improve the eco-environment of residential quarters. (3) Among the factors affecting RSEI of residential quarters, FVC, and NDBSI contribute more than other factors. For the construction of excellent eco-environments in residential quarters, we suggest that vegetation coverage should be no less than 0.4, the impervious surface coverage should be no more than 0.53, the building density should be between 0.28 ∼ 0.35, residential density should be no more than 0.22, and the population density of residential quarter should be under 0.29. Overall, for open and low-density residential quarters, the living environment can be improved by enriching vegetation and reducing impervious surfaces, while compact, high-density residential quarters can create a livable eco-environment by optimizing the impervious surface landscape patterns.

Linking Anthropogenic Landscape Perturbation to Herbivory and Pathogen Leaf Damage in Tropical Tree Communities.

Anthropogenic disturbance of tropical humid forests leads to habitat loss, biodiversity decline, landscape fragmentation, altered nutrient cycling and carbon sequestration, soil erosion, pest/pathogen outbreaks, among others. Nevertheless, the impact of these alterations in multitrophic interactions, including host-pathogen and vector-pathogen dynamics, is still not well understood in wild plants. This study aimed to provide insights into the main drivers for the incidence of herbivory and plant pathogen damage, specifically, into how vegetation traits at the local and landscape scale modulate such interactions. For this purpose, in the tropical forest of Calakmul (Campeche, Mexico), we characterised the foliar damage caused by herbivores and pathogens in woody vegetation of 13 sampling sites representing a gradient of forest disturbance and fragmentation in an anthropogenic landscape from well preserved to highly disturbed and fragmented areas. We also evaluated how the incidence of such damage was modulated by the vegetation and landscape attributes. We found that the incidence of damage caused by larger, mobile, generalist herbivores, was more sensitive to changes in landscape configuration, while the incidence of damage caused by small and specialised herbivores with low dispersal capacity was more influenced by vegetation and landscape composition. In relation to pathogen symptoms, the herbivore-induced foliar damage seems to be the main factor related to their incidence, indicating the enormous importance of herbivorous insects in the modulation of disease dynamics across tropical vegetation, as they could be acting as vectors and/or facilitating the entry of pathogens by breaking the foliar tissue and the plant defensive barriers. The incidence of pathogen damage also responded to vegetation structure and landscape configuration; the incidence of anthracnose, black spot, and chlorosis, for example, were favoured in sites surrounded by smaller patches and a higher edge density, as well as those with a greater aggregation of semi-evergreen forest patches. Fungal pathogens were shown to be an important cause of foliar damage for many woody species. Our results indicate that an increasing transformation and fragmentation of the tropical forest of southern Mexico could reduce the degree of specialisation in plant-herbivore interactions and enhance the proliferation of generalist herbivores (chewers and scrapers) and of mobile leaf suckers, and consequently, the proliferation of some symptoms associated with fungal pathogens such as fungus black spots and anthracnose. The symptoms associated with viral and bacterial diseases and to nutrient deficiency, such as chlorosis, could also increase in the vegetation in fragmented landscapes with important consequences in the health and productivity of wild and cultivated plant species. This is a pioneering study evaluating the effect of disturbances on multitrophic interactions, offering key insights on the main drivers of the changes in herbivory interactions and incidence of plant pathogens in tropical forests.

Patch-based stochastic attention for image editing

Attention mechanisms have become of crucial importance in deep learning in recent years. These non-local operations, which are similar to traditional patch-based methods in image processing, complement local convolutions. However, computing the full attention matrix is an expensive step with heavy memory and computational loads. These limitations curb network architectures and performances, in particular for the case of high resolution images. We propose an efficient attention layer based on the stochastic algorithm PatchMatch, which is used for determining approximate nearest neighbors. We refer to our proposed layer as a “Patch-based Stochastic Attention Layer” (PSAL). Furthermore, we propose different approaches, based on patch aggregation, to ensure the differentiability of PSAL, thus allowing end-to-end training of any network containing our layer. PSAL has a small memory footprint and can therefore scale to high resolution images. It maintains this footprint without sacrificing spatial precision and globality of the nearest neighbors, which means that it can be easily inserted in any level of a deep architecture, even in shallower levels. We demonstrate the usefulness of PSAL on several image editing tasks, such as image inpainting, guided image colorization, and single-image super-resolution. Our code is available at: https://github.com/ncherel/psal.

Transformer-based biomarker prediction from colorectal cancer histology: A large-scale multicentric study

Deep learning (DL) can accelerate the prediction of prognostic biomarkers from routine pathology slides in colorectal cancer (CRC). However, current approaches rely on convolutional neural networks (CNNs) and have mostly been validated on small patient cohorts. Here, we develop a new transformer-based pipeline for end-to-end biomarker prediction from pathology slides by combining a pre-trained transformer encoder with a transformer network for patch aggregation. Our transformer-based approach substantially improves the performance, generalizability, data efficiency, and interpretability as compared with current state-of-the-art algorithms. After training and evaluating on a large multicenter cohort of over 13,000 patients from 16 colorectal cancer cohorts, we achieve a sensitivity of 0.99 with a negative predictive value of over 0.99 for prediction of microsatellite instability (MSI) on surgical resection specimens. We demonstrate that resection specimen-only training reaches clinical-grade performance on endoscopic biopsy tissue, solving a long-standing diagnostic problem.

A Critical Assessment of Master Development Plan and Landscape Pattern of Jaipur City

There are various factors that influence the land use/land cover changes and spatial pattern of the landscape. The land use/land cover categories of existing land use map-2009 and master development plan-2025 are reorganized into 8 land use/land cover categories to measure the changes if everything goes according to the Master Development Plan-2025. The ArcGIS 10.3 and Erdas 14 software have been used to generate statistics and maps. The Fragstats 4.2 software has been used to analyse the indices of the landscape pattern. The urban area (i.e. Jaipur city region, satellite towns, and transport corridors) was 19.94% in MDP-2011 that is expected to increase to 54.3% in MDP-2025. During 2009-2025 the JDA aims to increase built-up from 48.79% to 72.95% in JMC. The ecological area is also expected to increase from 24.55% (MDP-2011) to 30.4% (MDP-2025). The rural area is expected to decrease significantly from 55.51% (MDP-2011) to 15.31% (MDP-2025). Thematic change dynamics reveal that in the JMC most of the agricultural land and vacant land will be converted into built-up, urban greenery, and institutional land. The landscape metrics analysis reveals that land use conversion is reducing the diversity of land use/land cover patches but patches are being more spatially aggregated. The diversity and heterogeneity of patches will reduce but the aggregation of patches will cause an increase in the patch size.

Exploring the impact of changing landscape patterns on ecological quality in different cities: A comparative study among three megacities in eastern and western China

The landscape pattern emerges from the intricate interplay among natural, biological, and social elements, profoundly influencing the quality of the regional ecological environment. By comparing the spatial and temporal driving mechanisms of landscape pattern changes on ecological environment quality across various cities, we can gain valuable insights into charting sustainable development paths for cities in the future. The results show that: (1) From 1990 to 2020, the mean RSEI values for Kunming and Nanjing exhibited a certain degree of increase, suggesting an improvement in ecological quality over time. Conversely, Guangzhou displayed relatively high and stable ecological quality throughout the studied period. (2) The CA-Markov results reveal minimal variations in the ecological quality of Kunming by 2035. Conversely, Nanjing experiences a decline in ecological quality, while Guangzhou demonstrates commendable achievements in ecological development, with an area accounting for moderate or higher ecological grades reaching 71.13%. (3) During the 30 years, the polarization of the three cities intensified and gradually formed a polycentric and clustered urban form. Among them, Nanjing and Guangzhou showed a fragmented landscape pattern, while the landscape diversity of Kunming is decreasing. (4) In terms of landscape pattern changes, Nanjing and Guangzhou showed a fragmented landscape pattern and increased landscape heterogeneity, but the degree of fragmentation and the rate of urban expansion in Guangzhou have been mitigated. In contrast, the spatial distribution of landscape patches in Kunming is becoming more clustered and the landscape diversity is decreasing. (5) The results of MGWR show that improving the landscape diversity, landscape patch aggregation and connectivity is beneficial to improving ecological quality. Overall, this study provides valuable insights and serves as a relevant reference for optimizing urban spatial structure in China and other developing countries' cities, aiming to foster sustainable and ecologically conscious urban development.

Role of Landscape and Land-Use Transformation on Nonpoint Source Pollution and Runoff Distribution in the Dongsheng Basin, China

Non-point source pollution (NSP) and runoff intensities and distribution are primarily affected by landscape structure and composition. Multiple causalities hinder our ability to determine significant variables that influence NSP. Therefore, we developed an approach that integrates the Soil and Water Assessment Tool (SWAT), random forest regression model, redundancy analysis, and correlation coefficient to assess the role of landscape structure on runoff and NSP in the Dongsheng basin. We used R to calculate landscape metrics and the SWAT to simulate NSP loads from 1990 to 2019. redundancy analysis (RDA), random forest, and Pearson correlation were used to analyze the relationships among landscape metrics and NSP variables. The largest patch index (LPI) shows a significant negative correlation with NSP, with an R2 of −0.58 for TP and TN and −0.62 for sediment load. The findings indicate that landscapes with larger patch sizes, a high number of patches, and aggregation of patches largely influence pollution distribution. Overall, the results suggest that the role of landscape patterns on NSP outweighs that of runoff. Moreover, the findings infer that the aggregation and connectivity of forest patches contribute to the decline in NSP load and vice versa for cropland cover. Thus, for sustainable watershed management, it is crucial to encourage unfragmented landscapes, especially pollutant-intercepting landcovers such as forests.

Assessing the Contributions of Urban Green Space Indices and Spatial Structure in Mitigating Urban Thermal Environment

Urban green space takes a dominant role in alleviating the urban heat island (UHI) effect. Most investigations into the effects of cooling factors from urban green spaces on the UHI have evaluated the correlation between each factor and land surface temperature (LST) separately, and the contribution weights of various typical cooling factors in mitigating the thermal environment have rarely been analyzed. For this research, three periods of Landsat 8 data captured between 2014 and 2018 of Xuzhou during the summer and autumn seasons were selected along with corresponding meteorological and flux measurements. The mono-window method was employed to retrieve LST. Based on the characteristics of the vegetation and spatial features of the green space, eight factors related to green space were selected and computed, consisting of three indices that measure vegetation and five metrics that evaluate landscape patterns: vegetation density (VD), evapotranspiration (ET), green space shading degree (GSSD), patch area ratio (PLAND), largest patch index (LPI), patch natural connectivity (COHESION), patch aggregation (AI), and patch mean shape index distribution (SHPAE_MN). Linear regression and bivariate spatial autocorrelation analyses between each green space factor and LST showed that there were significant negative linear and spatial correlations between all factors and LST, which proved that the eight factors were all cooling factors. In addition, LST was strongly correlated with all factors (|r| > 0.5) except for SHPAE_MN, which was moderately correlated (0.3 < |r| < 0.5). Based on this, two principal components were extracted by applying principal component analysis with all standardized green space factors as the original variables. To determine the contribution weight of each green space factor in mitigating the urban heat island (UHI) effect, we multiplied the influence coefficient matrix of the initial variables with the standardized multiple linear regression coefficients between the two principal component variables and LST. The final results indicated that the vegetation indices of green space contribute more to the alleviation of the UHI than its landscape pattern metrics, and the contribution weights are ranked as VD ≥ ET > GSSD > PLAND ≈ LPI > COHESION > AI > SHAPE_MN. Our study suggests that increasing vegetation density is preferred in urban planning to mitigate urban thermal environment, and increasing broadleaf forests with high evapotranspiration and shade levels in urban greening is also an effective way to reduce ambient temperature. For urban green space planning, a priority is to multiply the regional green space proportion or the area of largest patches. Second, improving the connectivity or aggregation among patches of green space can enhance their ability to cool the surrounding environment. Altering the green space spatial shape is likely the least significant factor to consider.

Impact of landscape configuration and composition on pollinator communities across different European biogeographic regions

IntroductionHeterogeneity in composition and spatial configuration of landscape elements support diversity and abundance of flower-visiting insects, but this is likely dependent on taxonomic group, spatial scale, weather and climatic conditions, and is particularly impacted by agricultural intensification. Here, we analyzed the impacts of both aspects of landscape heterogeneity and the role of climatic and weather conditions on pollinating insect communities in two economically important mass-flowering crops across Europe.MethodsUsing a standardized approach, we collected data on the abundance of five insect groups (honey bees, bumble bees, other bees, hover flies and butterflies) in eight oilseed rape and eight apple orchard sites (in crops and adjacent crop margins), across eight European countries (128 sites in total) encompassing four biogeographic regions, and quantified habitat heterogeneity by calculating relevant landscape metrics for composition (proportion and diversity of land-use types) and configuration (the aggregation and isolation of land-use patches).ResultsWe found that flower-visiting insects responded to landscape and climate parameters in taxon- and crop-specific ways. For example, landscape diversity was positively correlated with honey bee and solitary bee abundance in oilseed rape fields, and hover fly abundance in apple orchards. In apple sites, the total abundance of all pollinators, and particularly bumble bees and solitary bees, decreased with an increasing proportion of orchards in the surrounding landscape. In oilseed rape sites, less-intensively managed habitats (i.e., woodland, grassland, meadows, and hedgerows) positively influenced all pollinators, particularly bumble bees and butterflies. Additionally, our data showed that daily and annual temperature, as well as annual precipitation and precipitation seasonality, affects the abundance of flower-visiting insects, although, again, these impacts appeared to be taxon- or crop-specific.DiscussionThus, in the context of global change, our findings emphasize the importance of understanding the role of taxon-specific responses to both changes in land use and climate, to ensure continued delivery of pollination services to pollinator-dependent crops.

Defect transformer: An efficient hybrid transformer architecture for surface defect detection

Surface defect detection is an extremely crucial step to ensure the quality of industrial products. Nowadays, convolutional neural networks (CNNs) based on encoder–decoder architecture have achieved tremendous success in various defect detection tasks. However, the intrinsic locality of convolution prevents them from modeling long-range interactions explicitly, making it difficult to distinguish pseudo-defects in cluttered backgrounds. Recent transformers are especially skilled at learning global image dependencies, but with limited local structural information for the refined defect location. To overcome the above limitations, we incorporate CNN and transformer into an efficient hybrid transformer architecture for defect detection, termed Defect Transformer (DefT), to capture local and non-local relationships collaboratively. Specifically, in the encoder module, a convolutional stem block is firstly adopted to retain more spatial details. Then, the patch aggregation blocks are used to generate multi-scale representation with four hierarchies, each of them is followed by a series of DefT blocks, which respectively include a locally position-aware block for local position encoding, a lightweight multi-pooling self-attention to model multi-scale global contextual relationships with good computational efficiency, and a convolutional feed-forward network for feature transformation and further local information learning. Finally, a simple but effective decoder module is constructed to gradually recover spatial details from the skip connections in the encoder. Extensive experiments on three datasets demonstrate the superiority and efficiency of our method compared with other deeper and complex CNN- and transformer-based networks.

Designing an optimized landscape restoration with spatially interdependent non-linear models

Brazilian Atlantic Forest is a biodiversity hotspot drastically fragmented due to different land use practices. Our understanding on the impacts of fragmentation and restoration practices on ecosystem functionality significantly increased during the last decades. However, it is unknown to our knowledge how a precision restoration approach, integrated with landscape metrics, will affect the decision-making process of forest restoration. Here, we applied Landscape Shape Index and Contagion metrics in a genetic algorithm for planning forest restoration in watersheds at the pixel level. We evaluated how such integration may configure the precision of restoration with scenarios related to landscape ecology metrics. The genetic algorithm worked toward optimizing the site, shape, and size of forest patches across the landscape according to the results obtained in applying the metrics. Our results, obtained by simulations of scenarios, support aggregation of forest restoration zones as expected, with priority restoration areas indicated where most of the aggregation of forest patches occurs. Our optimized solutions for the study area (Santa Maria do Rio Doce Watershed) predicted an important improvement of landscape metrics (LSI = 44 %; Contagion/LSI = 73 %). Largest shifts are suggested based on LSI (i.e., three larger fragments) and Contagion/LSI (i.e., only one well-connected fragment) optimizations. Our findings indicate that restoration in an extremely fragmented landscape will promote a shift toward more connected patches and with reduction of the surface:volume ratio. Our work explores the use of genetic algorithms to propose forest restoration based on landscape ecology metrics in a spatially explicit innovative approach. Our results indicate that LSI and Contagion:LSI ratio may affect the choice concerning precise location of restoration sites based on forest fragments scattered in the landscape and reinforce the usefulness of genetic algorithms to yield an optimized-driven solution for restoration initiatives.

Effects of Landscape, Climate and Hunting on the Occurrence of White-Browed Guan Penelope jacucaca in Central-North Caatinga, Brazil

The White-browed Guan Penelope jacucaca is an endangered member of the family Cracidae endemic to the Caatinga region of northeast Brazil. Hunting and habitat loss are the presumed major threats to the species, however, their impacts on its occurrence have been insufficiently tested. We used field observations and information on hunting activity to assess whether hunting, landscape and/or climate affect the occurrence of the White-browed Guan in the Caatinga. We sampled 23 localities within the Caatinga and recorded the species' occurrence and hunting encounters. For each locality, we extracted three landscape metrics (vegetation cover, patch aggregation, and patch richness); two climate variables (temperature seasonality and mean precipitation of the wettest month); and one topographic variable (mean slope). We then used generalized linear models based on proportion to determine if the species' occurrence was related to these environmental factors. We found a greater likelihood of detecting the species in localities with greater slope, greater heterogeneity of habitats and more stable temperatures, whereas hunting activity presented a trend of negative impact on occurrence. Our results emphasize the relevance of using different environmental metrics to implement conservation programs for this threatened species.