Ecological Informatics最新文献

{"title":"A photogrammetric approach to the estimation of distance to animals in camera trap images","authors":"B. Mirka ,&nbsp;C.D. Lippitt ,&nbsp;G.M. Harris ,&nbsp;R. Converse ,&nbsp;M. Gurule ,&nbsp;S.E. Sesnie ,&nbsp;M.J. Butler ,&nbsp;D.R. Stewart ,&nbsp;Z. Rossman","doi":"10.1016/j.ecoinf.2025.103120","DOIUrl":"10.1016/j.ecoinf.2025.103120","url":null,"abstract":"<div><div>The ability to assess accurate distance measurements, meaning the distance from an animal to the observer, is critical for wildlife monitoring methods such as distance sampling. This process has typically been done through manual measurements or using reference points, making fieldwork and analysis labor intensive and error prone. Using Structure from Motion (SfM) photogrammetry and geotagged imagery captured from smartphones offers a simple and cost-effective solution to distance estimation from camera traps. This research evaluates the potential of SfM to create 3D models of the areas within a camera trap field-of-view to produce depth maps that can be used to automate the distance estimation process from bounding box-labeled imagery. These methods have the potential, when paired with deep learning algorithms or crowdsourcing, to automate the process of wildlife identification and distance estimation, reducing the need for manual fieldwork or analysis while increasing the precision and replicability of population estimates. To test the accuracy of these methods and evaluate best practices, three types of GNSS integration into the SfM model were assessed at three camera trap sites placed within the Sevilleta National Wildlife Refuge in New Mexico, USA. The evaluation focused on the geometric accuracy of the resulting model, quantified using Root Mean Square Error (RMSE) against known ground control points, the accuracy of distance estimates relative to field-measured distances, and the comparison of modeled distances of detected animals to ground control points. The first method used only the GNSS data embedded (geotagged) cell phone images (ImageGNSS), the second used ground control points (GCPGNSS), and the third used both GNSS sources (AllGNSS). Models created using ImageGNSS had the highest accuracy, with an overall RMSE of 1.25 pixels and a mean absolute error (MAE) of 1.76 m for distance estimation from GCPs. A comparison of the modeled distance for 10 animals to known distance points produced minimal error (MAE of 3.31 m), demonstrating the potential of photogrammetric approaches to make accurate distance estimations from camera trap imagery. Unlike other digital distance estimation techniques, error rates do not increase significantly (<em>p</em> = 0.36) at distances greater than 15 m to at least a distance of 35 m. The biggest driver of distance error was scene complexity (complex topography, dense vegetation, etc.)</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"89 ","pages":"Article 103120"},"PeriodicalIF":5.8,"publicationDate":"2025-03-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143892204","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detection of anthropogenic noise pollution as a possible chronic stressor in Antarctic Specially Protected Area N°150, Ardley Island","authors":"Maximiliano Anzibar Fialho ,&nbsp;Martín Rocamora ,&nbsp;Lucía Ziegler","doi":"10.1016/j.ecoinf.2025.103117","DOIUrl":"10.1016/j.ecoinf.2025.103117","url":null,"abstract":"<div><div>Anthropogenic noise pollution is emerging as an important environmental stressor with the potential effect of disrupting natural ecosystems, since many taxa rely on acoustic signals for social interaction and communication. Antarctic wildlife is increasingly experiencing the impact of growing human presence on the continent, especially near populated areas such as research stations. Until now, most studies on the sound impact in Antarctica have focused on marine ecosystems, with a clear paucity of studies at the level of terrestrial environments. In this study, we analyze the presence of a specific anthropogenic sound source, a power generator, in the soundscape of the Antarctic Specially Protected Area (ASPA) N°150, Ardley Island. We used Audiomoth recorders to hourly monitor the soundscape in Ardley Island and create a simple yet effective detection method based on spectral features of the source. We cross-validate the detection algorithm with human perception classification of the source presence in the recordings, obtaining a Pearson correlation coefficient of 0.61 between the two methods. Further, we relate the detection with wind velocity and direction, concluding that under certain meteorological conditions, the source can be clearly heard from Ardley. Our results suggest that the soundscape of Ardley Island is altered by the near presence of an anthropogenic noise source which could represent an impact on animal life in the ASPA. We consider this kind of study to be relevant in bringing awareness of noise pollution in Antarctic ecosystems and improving management plans in the ASPAs.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103117"},"PeriodicalIF":5.8,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143715225","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessment of the green development level and the identification of obstacles to grass-based livestock husbandry in the farming–pastoral ecotone of northern China","authors":"Dan Liu ,&nbsp;Jianjun Jin ,&nbsp;Xuan Zhang ,&nbsp;Xin Qiu ,&nbsp;Rui He ,&nbsp;Jie Yang","doi":"10.1016/j.ecoinf.2025.103112","DOIUrl":"10.1016/j.ecoinf.2025.103112","url":null,"abstract":"<div><div>The green development of livestock husbandry represents the balance between livestock production and environmental protection. In this study, a comprehensive framework and evaluation indices were developed for assessing the green development level of grass-based livestock husbandry (GDL-GLiH), including the green growth degree (GGD), green carrying capacity (GCC) and green guarantee capability (GGC). On the basis of the combined weight Technique Order Preference by Similarity to an Ideal Solution (TOPSIS) model and an analysis of obstacles, the GDL-GLiH in the farming–pastoral ecotone of northern China (FPEN) was evaluated, and the main obstacles were identified. The results indicated a general upward trend in the GDL-GLiH across the FPEN, increasing from 0.343 in 2010 to 0.416 in 2020, reflecting a growth rate of 21.138 %. Among the three dimensions, the GGC showed the most substantial increase of 95.937 %, whereas the GCC exhibited minimal growth of 1.006 %. Spatial variations were observed, with livestock-dominated systems exhibiting higher average levels (0.398) but lower growth rates (19.740 %) than crop-dominated systems (0.356; 22.252 %). Additionally, the production of milk (average obstacle degree: 12.970 %), the proportion of forage cultivation in crop cultivation (11.312 %), the total mechanical power per unit agricultural sown area (10.081 %) and the availability of purebred bovine and ovine breeding stock (9.034 %) were identified as the key obstacles. This study provides a holistic assessment framework for green livestock development and serves as a reference for formulating green development strategies in the FPEN, as well as in similar agricultural systems.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103112"},"PeriodicalIF":5.8,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143715227","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Toward causal artificial intelligence approach for PM2.5 interpretation: A discovery of structural causal models","authors":"Mallika Kliangkhlao ,&nbsp;Apaporn Tipsavak ,&nbsp;Thanathip Limna ,&nbsp;Racha Dejchanchaiwong ,&nbsp;Perapong Tekasakul ,&nbsp;Kirttayoth Yeranee ,&nbsp;Thanyabun Phutson ,&nbsp;Bukhoree Sahoh","doi":"10.1016/j.ecoinf.2025.103115","DOIUrl":"10.1016/j.ecoinf.2025.103115","url":null,"abstract":"<div><div>Understanding the causal mechanisms underlying PM_2.5 generation is critical for developing effective prevention strategies, necessitating an approach that goes beyond prediction and seeks deeper causal explanations to support decision-making. This study addresses these concerns through a novel causal artificial intelligence framework employing structural causal models (SCMs) to interpret PM_2.5 dynamics. The research uncovers hidden cause-and-effect relationships between meteorological factors and PM_2.5 exposure by leveraging a data-driven causal structure discovery approach, effectively representing complex data-generating processes. The proposed SCMs undergo systematic validation across two critical dimensions: demonstrating human-like intelligence understanding and achieving significant alignment with real-world observations. The PC-based SCM is particularly outstanding when compared to other algorithms like GES- and Chow-Lui-based SCMs, delivering a remarkable performance in discovering cause-and-effect relationships with an F-measure of approximately 80 % compared to the gold-standard SCM. Statistical validation provided robust evidence of the model's reliability, with fit indices—including <em>NFI</em>, <em>TLI</em>, <em>CFI</em>, <em>GFI</em>, and <em>AGFI</em>—reaching approximately 0.98 and <em>RMSEA</em> approximating 0.05. These findings demonstrate that SCM can encode human-like reasoning and naturally align with real-world meteorological systems. This method is especially effective for urban air quality monitoring, where accessible meteorological data and transparent causal relationships are essential. Its capacity to inform evidence-based policy decisions makes it a powerful tool for creating intelligent decision-support systems in PM2.5 analysis and environmental mitigation strategies.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103115"},"PeriodicalIF":5.8,"publicationDate":"2025-03-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143679825","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Analysing the impact of large mammal herbivores on vegetation structure in Eastern African savannas combining high spatial resolution multispectral remote sensing data and field observations","authors":"Helena M. Back ,&nbsp;Isabel Pérez-Postigo ,&nbsp;Clemens Geitner ,&nbsp;Almut Arneth","doi":"10.1016/j.ecoinf.2025.103113","DOIUrl":"10.1016/j.ecoinf.2025.103113","url":null,"abstract":"<div><div>It is well understood, that grazing and browsing in the African savanna ecosystem modulates tree-grass ratios. However, many of the large mammals are under pressure due to land use and climate change. It is challenging to predict how their altered abundance or range shifts will interact with savanna structure. Herbivore exclusion experiments can help to better understand the impacts of herbivores of different sizes on vegetation structure and composition, including interactions with rainfall. Here, we combine field data of the Ungulate Herbivory Under Rainfall Uncertainty (UHURU) exclusion experiment and high spatial resolution satellite images of the PlanetScope and Sentinel-2 series to investigate the impacts of herbivores on vegetation structure in a Kenyan savanna. Field data as well as NDVI values derived from Sentinel-2 and NDVI contrast values of PlanetScope show that presence of herbivores lowers vegetation cover and modify the woody vegetation structure depending on which herbivores are present. The vegetation grew tallest when mega-sized herbivores were absent but meso-sized and small herbivores were present, which resulted in high NDVI contrast values. The absence of herbivores resulted in fewer bare ground patches and increased green biomass, such as a higher mean canopy width, which led to higher NDVI values. Few studies have explored the potential of passive remote sensing data to assess herbivory impacts beyond the plot scale and over longer time-periods; however, these previous studies solely focused on the NDVI. Here we demonstrate the added value of also using GLCM texture measures to investigate effects on a savanna ecosystem in response to presence or absence of herbivores. Combining these data with plot measurements our study demonstrates the benefits of combining field and space perspectives in ecosystem studies<em>.</em></div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103113"},"PeriodicalIF":5.8,"publicationDate":"2025-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143687623","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Seasonal and inter-annual dynamics of a Macrocystis pyrifera forest in Concepcion Bay, Chile","authors":"Daniel Gonzalez-Aragon ,&nbsp;Richard Muñoz ,&nbsp;Henry Houskeeper ,&nbsp;Kyle Cavanaugh ,&nbsp;Wirmer García-Tuñon ,&nbsp;Laura Farías ,&nbsp;Carlos Lara ,&nbsp;Bernardo R. Broitman","doi":"10.1016/j.ecoinf.2025.103103","DOIUrl":"10.1016/j.ecoinf.2025.103103","url":null,"abstract":"<div><div>Kelp forest are foundation species that deliver key ecosystem services for coastal habitats. Chile is one of the largest exporters of kelp biomass, which relies on the harvesting of wild populations. The vast and rugged coastline of Chile hinders field-based studies of the seasonal and spatial dynamics of kelp biomass, yet remote sensing approaches can provide an effective tool to study temporal patterns of kelp distribution and biomass. Our study aimed to establish the basic patterns of variation in the surface area and biomass of a <em>Macrocystis pyrifera</em> forest off Concepcion Bay, Central Chile. Using archival data from the Landsat series we constructed a long-term series of annual kelp canopy cover and assessed patterns of interannual, and a seasonal variation with the more recent Sentinel 2 data using Google Earth Engine. We validated satellite observations of the kelp forest in the field by recording local temperature and nutrient concentrations and through a sample of blades and stipes, which we used to estimate whole-individual <em>in situ</em> biomass through allometric relationships. Finally, we related decadal to interannual changes in canopy cover to local and regional drivers using data from public repositories. Our 24-year annual time series revealed large year-to-year variability in kelp forest area that did not show a significant association with different El Niño-Southern Oscillation indices, but the deviance explained increased notably with a 1-year lag. The seasonal time series exhibited clear seasonal patterns with cover peaking during summer. We found a significant influence of local environmental variables such as temperature, wave height, nitrate concentration, and solar radiation on kelp forest area. Furthermore, blade counts appeared as the most reliable metric for estimating <em>M. pyrifera</em> biomass. Interestingly, we found no evidence of temperature or nutrient stress during the summer biomass peak, hence seasonal variation in <em>M. pyrifera</em> abundance appears to be primarily influenced by solar radiation and wave activity in our study population. Our results provide a basis to derive seasonal time series across Chile’s kelp forests and suggest that understanding local stressors is key to ensure harvesting practices that promote the sustainable management of these key habitats. As ongoing climate change and overexploitation threaten kelp forest habitats, remote sensing emerges as a promising tool for the monitoring and management of extensive and remote coastlines.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103103"},"PeriodicalIF":5.8,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143680200","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Coral reef detection using ICESat-2 and machine learning","authors":"Gabrielle A. Trudeau ,&nbsp;Kim Lowell ,&nbsp;Jennifer A. Dijkstra","doi":"10.1016/j.ecoinf.2025.103099","DOIUrl":"10.1016/j.ecoinf.2025.103099","url":null,"abstract":"<div><div>As anthropogenic impacts threaten natural habitats, effective monitoring strategies are crucial. Coral reefs, among the most vulnerable ecosystems, traditionally employ monitoring techniques that are labor-intensive and costly, prompting the exploration of remote sensing as a cost-effective alternative. Launched in October 2018, the Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) provides high-resolution, high-frequency data, with its green laser offering unprecedented opportunities for bathymetric and coral reef applications. This study investigates the use of ICESat-2 data for atoll coral reef detection, utilizing Heron Island in the Great Barrier Reef, AU, and employing machine learning models. A binary logistic regression (BLR) model and convolutional neural network (CNN) were tested for determining coral reef presence, with the CNN outperforming the BLR in accuracy (85.4%), F1 score (43%), and false positive rate (13.1%). A challenge of the study included the difficulty of balancing false positive rates in predictive models to avoid over- or underestimations of reef extent. These obstacles were mitigated through the integration of algorithmically derived pseudo-rugosity and slope metrics as innovative proxies for seafloor complexity, significantly improving predictive performance. Feature importance analysis identified satellite-derived bathymetry (SDB) depth as the most critical predictor of coral presence, followed by pseudo-rugosity, slope, and various other depth measurements. This research establishes a new application of ICESat-2 data combined with advanced machine learning techniques as a promising method for efficient and cost-effective coral reef monitoring. Future work should refine algorithms and incorporate additional environmental variables to improve model performance across various reef types.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103099"},"PeriodicalIF":5.8,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143680203","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Refining landsat-based annual NDVImax estimation using shape model fitting and phenological metrics","authors":"Lihao Zhang ,&nbsp;Miaogen Shen ,&nbsp;Licong Liu ,&nbsp;Xuehong Chen ,&nbsp;Ruyin Cao ,&nbsp;Qi Dong ,&nbsp;Yang Chen ,&nbsp;Jin Chen","doi":"10.1016/j.ecoinf.2025.103107","DOIUrl":"10.1016/j.ecoinf.2025.103107","url":null,"abstract":"<div><div>The annual maximum normalized difference vegetation index (NDVI_max) is widely used as a surrogate for annual aboveground net primary productivity (ANPP) of summer-green vegetation. Landsat data, with its 30-m spatial resolution and high temporal consistency, have revealed long-term changes in NDVI_max and ANPP. However, in cloudy regions with summer-green vegetation, such as the Tibetan Plateau, the scarcity of cloud-free Landsat NDVI observations complicates NDVI_max estimation, particularly due to interannual variations in phenology and NDVI_max. This study proposed a shape model fitting method that integrates interannual phenological similarity to estimate Landsat NDVI_max, using the Tibetan Plateau as an example. For a given target year, an annual NDVI shape model was constructed using all cloud-free Landsat NDVI observations from that year and phenologically similar years, identified using phenological metrics derived from MODIS and GIMMS NDVI datasets. The model was then fitted to the target year's cloud-free NDVI time series to correct seasonal biases in NDVI observations. Validations with simulated and real images indicated that the proposed method outperformed several commonly used approaches in estimating NDVI_max and detecting temporal trends across various conditions. The method more accurately captured the true annual NDVI trajectory and NDVI_max date for the target year. It enabled the retrieval of long-term high-resolution NDVI_max series for summer-green vegetation on the Tibetan Plateau and provided a reference for Landsat NDVI_max extraction in other summer-green vegetation regions. Additionally, by addressing the observational biases, the method corrected previous overestimates of greening on Tibetan Plateau, thereby improving global change studies on summer-green vegetation.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103107"},"PeriodicalIF":5.8,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143680204","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The ‘everything is everywhere’ framework: Holistic network analysis as a marine spatial management tool","authors":"Molly K. James ,&nbsp;Gennadi Lessin ,&nbsp;Muchamad Al Azhar ,&nbsp;Michael Bedington ,&nbsp;Charlotte H. Clubley ,&nbsp;Paul Somerfield ,&nbsp;Antony M. Knights","doi":"10.1016/j.ecoinf.2025.103105","DOIUrl":"10.1016/j.ecoinf.2025.103105","url":null,"abstract":"<div><div>The North Sea hosts numerous man-made structures, some recently installed and others nearing end-of-life, with decisions about their decommissioning at the centre of current debate. Further there are plans for significant expansion of structures relating in particular to offshore wind energy. Here, using a combination of hydrodynamic modelling, particle tracking, and graph network analysis, we evaluate connectivity under two scenarios: existing structures – releasing particles from cells where structures are currently present – and “everything is everywhere” – releasing particles from every cell in the domain. Additionally, we introduce a Connectivity Importance Index (CII) to assess both current and potential future connectivity within the region. The CII under the ‘everything is everywhere’ scenario revealed cells with high potential connectivity that align with, but also extend beyond, those identified under the existing structures scenario, pointing to potentially valuable regions for future structure placement. The relocatable methodology described in this paper allows for the quantification of potential networks, applicable with or without existing habitat data, offering valuable insights for ecologically coherent marine spatial management strategies.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103105"},"PeriodicalIF":5.8,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143680198","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluation of a CNN model to map vegetation classification in a subalpine coniferous forest using UAV imagery","authors":"Weibo Shi ,&nbsp;Xiaohan Liao ,&nbsp;Shaoqiang Wang ,&nbsp;Huping Ye ,&nbsp;Dongliang Wang ,&nbsp;Huanyin Yue ,&nbsp;Jianli Liu","doi":"10.1016/j.ecoinf.2025.103111","DOIUrl":"10.1016/j.ecoinf.2025.103111","url":null,"abstract":"<div><div>Unmanned aerial vehicle (UAV) remote sensing based on deep learning has increasingly been applied for forest vegetation classification. However, existing studies have focused mainly on simple woodlands, and accurately mapping the vegetation distribution in complex natural forests remains challenging. To address this, we conducted a study in a natural alpine forest in Southwest China, leveraging high-resolution UAV imagery and deep learning for vegetation classification. We systematically assessed the effects of patch size, spatial resolution, and rotation angle on the model performance, considering their interactions. Our results demonstrate that UAVs combined with deep learning techniques achieve high classification accuracy in natural forests, with a mean F1-score of 0.91. Patch size has a significant influence on accuracy, although its impact diminishes as the spatial resolution decreases. As the patch size increased from 128 × 128 to 256 × 256, the model F1-score improved by 18 % at a 5 cm resolution, whereas it improved by only 3 % at a 10 cm resolution. A higher spatial resolution does not necessarily enhance model accuracy, and the effect of patch size also needs to be considered. The Rotation angle, as a data augmentation strategy, is crucial when training data are limited and can significantly increase model performance. These findings highlight the potential of combining deep learning and UAV remote sensing for natural forests. This approach facilitates more reliable access to forest information in forest areas where access is difficult. Overall, this study provides an efficient and cost-effective method for monitoring and protecting natural forests, serving as a reference for selecting appropriate parameters in UAV-based deep learning remote sensing.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"87 ","pages":"Article 103111"},"PeriodicalIF":5.8,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143687622","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}