• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Wednesday, November 18, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Biology

Integrated Decision Modeling and Machine Learning Map Western Jilin Ecological Security Network

by
November 18, 2026
in Biology
Reading Time: 7 mins read
0
Integrated Decision Modeling and Machine Learning Map Western Jilin Ecological Security Network

Integrated Decision Modeling and Machine Learning Map Western Jilin Ecological Security Network

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

Landscape researchers have established an optimized regional ecological security pattern across the ecologically fragile agricultural-wetland ecotone of western Jilin, China, by integrating multi-criteria decision modeling, explainable machine-learning resistance surfaces, and circuit theory. The investigative framework resolved long-standing uncertainties surrounding ecological source identification and landscape resistance estimation, demonstrating how data-driven modeling can reconcile the acute spatial conflicts between national agricultural expansion mandates and vital wetland habitat preservation. By evaluating multiple source delineation methodologies and competing algorithm architectures, the research team identified twenty-three primary ecological source patches spanning 3,719.62 square kilometers, thirty least-cost wildlife corridors covering 215.98 kilometers, 243 high-current pinch points occupying 21.99 square kilometers, and 31 high-resistance ecological barriers totaling 33.33 square kilometers. These structural components were consolidated into a coherent territorial spatial architecture termed the two corridors, two belts, and three zones framework, offering an empirical foundation for ecosystem restoration and corridor maintenance across semi-arid to semi-humid transition zones.

The study focused on western Jilin, a 47,000-square-kilometer territory situated in the southwestern Songnen Plain that represents 25.5 percent of Jilin Province. Encompassing the prefecture-level administrative divisions of Songyuan and Baicheng, the terrain features elevations ranging from 22 to 672 meters across flat northern expanses and higher peripheral topography. The regional climate is characterized by a temperate continental monsoon regime, with mean annual temperatures of 4.4 degrees Celsius and annual precipitation between 384 and 521 millimeters that declines along a southeast-to-northwest gradient. In 2023, the region supported 3.59 million residents and generated a gross domestic product of 148.5 billion yuan. Historically, the area has carried the conflicting burdens of commercial grain development under provincial agricultural capacity initiatives and critical biodiversity conservation as a major stopover along the East Asian-Australasian Flyway. Between 2000 and 2020, regional cropland expanded by 3.6 percent and built-up land increased by 15.0 percent, whereas wetlands, open water bodies, and grasslands contracted by 7.2 percent, 12.6 percent, and 11.8 percent, respectively, driving severe habitat fragmentation.

To capture regional ecological multifunctionality before delineating conservation priorities, the researchers quantified five fundamental ecosystem services at a standardized 30-meter raster resolution: water conservation, habitat quality, soil conservation, carbon sequestration, and food production. Spatial modeling revealed pronounced geographical heterogeneity across these services, with four exhibiting a general downward trajectory from east to west. Water conservation mirrored precipitation patterns, concentrating primarily within high-rainfall zones and natural forests in eastern and southern Songyuan, while remaining low across Baicheng. Soil conservation reached its peak in eastern Songyuan and within managed shelterbelts in northwestern Baicheng. High carbon sequestration was documented throughout forest, grassland, and intensively cultivated agricultural tracts across Songyuan and northwestern Baicheng, whereas extensive saline-alkali land in southern Baicheng exhibited minimal carbon retention capacity despite recent land reclamation efforts. Food production was concentrated in extensive agricultural fields in eastern, central, and southern Songyuan, while habitat quality displayed no unified east-west gradient, clustering instead around river networks and wetland complexes in northern Songyuan and northeastern Baicheng.

To overcome the limitations of conventional source extraction, the researchers conducted a comparative assessment among three distinct methodological approaches: ordered weighted averaging, morphological spatial pattern analysis, and ecosystem service significance assessment. Ordered weighted averaging applied fuzzy multi-criteria evaluation across nine decision scenarios ranging from risk-averse to risk-prone assumptions, using order weights and trade-off calculations to balance competing services. Morphological spatial pattern analysis applied mathematical morphological operators to land-cover rasters to classify image pixels into seven structural categories, extracting core areas based on topological geometry and evaluating patch importance via the probability of connectivity and delta probability of connectivity metrics. The ecosystem service significance approach applied min-max normalization and equal-weight spatial overlays across two variations: a five-service model that integrated agricultural food production and a four-service model that excluded food production, defining potential ecological sources from the upper twenty percent quantile of comprehensive values.

Comparative evaluations demonstrated that ordered weighted averaging Scenario 4 achieved superior conservation efficiency and balanced representation of multiple ecosystem services. Ordered weighted averaging Scenario 4 achieved the highest conservation efficiency for biodiversity at 368.22 percent, Ramsar-listed wetlands at 269.97 percent, and nature reserves at 202.13 percent, alongside the strongest performance across habitat-quality efficiency indicators. In contrast, morphological spatial pattern analysis delineated the largest total source territory at 11,884.68 square kilometers, consisting of 8,161 patches with a large mean patch size of 1.46 square kilometers and capturing the highest nominal coverage of biodiversity priority areas at 77.71 percent, Ramsar wetlands at 58.52 percent, and nature reserves at 44.60 percent. However, it neglected small, ecologically vital wetland patches and provisioning services. The five-service significance model encompassed 8,744.45 square kilometers across 201,698 highly fragmented patches with an average size of only 0.04 square kilometers, wherein cropland represented 85.98 percent of the area, skewing conservation priorities toward agricultural production rather than natural preservation.

Following the selection of ordered weighted averaging Scenario 4 as the optimal candidate scheme, the initial 9,056.70 square kilometers of potential sources—representing 18.95 percent of the regional landscape—underwent topological filtering in Conefor 2.6 software. Sensitivity analyses spanning patch area thresholds from 0.1 to 100 square kilometers and delta probability of connectivity thresholds from 0.1 to 5 determined that retaining patches larger than 10 square kilometers with connectivity values greater than 1 established the optimal trade-off between connectivity maintenance and computational parsimony. This filtering yielded twenty-three final ecological sources encompassing 3,719.62 square kilometers, or 7.78 percent of western Jilin, with the largest contiguous patch measuring 986.24 square kilometers. Ecological land types accounted for 93.46 percent of the total source extent, dominated by wetlands at 52.34 percent and open water bodies at 23.02 percent, with forests and grasslands providing supplementary coverage and cropland restricted to 6.54 percent. These sources formed continuous structural corridors in northeastern Baicheng and northern Songyuan while functioning as stepping-stone habitats in western Baicheng.

To model regional resistance to species dispersal without relying on arbitrary expert-assigned friction scores, the authors evaluated five machine-learning algorithms: Logistic Regression, Random Forest, Extreme Gradient Boosting, Support Vector Machine, and Decision Tree. Models were trained using 10,000 random points within potential ecological sources as positive presence samples and 10,000 points outside sources as pseudo-absence samples, maintaining minimum spatial separation distances above 100 meters. To account for spatial autocorrelation across predictors, the team implemented a nested spatial block cross-validation framework with an outer five-fold evaluation loop and an inner four-fold hyperparameter tuning loop based on an empirical block size of 50 kilometers. Across cross-validation trials, Random Forest achieved the highest area under the receiver operating characteristic curve at 0.743 plus or minus 0.043, alongside precision of 0.689 and specificity of 0.732. Extreme Gradient Boosting exhibited closely comparable discriminative accuracy, attaining an area under the curve of 0.742 plus or minus 0.047 with higher recall of 0.627, negative predictive value of 0.655, and F1-score of 0.649.

Because Random Forest achieved superior overall accuracy while Extreme Gradient Boosting provided a more balanced factor weighting structure, both models were retained to construct an integrated resistance surface. Shapley Additive Explanations analysis of both algorithms revealed that population density, distance to water bodies, and gross domestic product represented the primary drivers governing landscape permeability, with normalized integrated relative importance values of 0.2475, 0.2342, and 0.1616, respectively. Digital elevation model values, distance to roads, vegetation greenness, and slope contributed moderate effects, whereas nighttime light intensity exerted negligible influence. Predicted presence probabilities were transformed into relative friction values between 1 and 100, generating high resistance across urbanized cores in Taobei and Ningjiang districts and low resistance within wetland sanctuaries. Independent empirical validation utilizing bird monitoring records gathered between 2018 and 2022 confirmed the ecological fidelity of the machine-learning surface: mean resistance across documented bird observation localities was 16.88, substantially lower than the study-area mean of 44.07 to 44.45 and random-point benchmarks of 44.20 to 44.64.

Using Linkage Mapper software to integrate minimum cumulative resistance modeling and circuit theory, the researchers extracted thirty least-cost ecological corridors with a combined length of 215.98 kilometers. Individual corridor lengths exhibited considerable dispersion, ranging from a maximum path of 24.3 kilometers connecting major northern wetlands to an abbreviated link of just 0.085 kilometers between contiguous stepping stones. Northern corridors in Zhenlai and Da’an counties traversed extensive distances across flat terrain characterized by low human disturbance, whereas western corridors in Taonan City connected smaller, isolated habitat fragments. Because least-cost vectors lack intrinsic physical widths, buffer analysis evaluated land-cover composition across widths ranging from 30 to 3,000 meters. At 30 meters, natural land covers represented 40.78 percent of corridor area, but agricultural and saline-alkali lands rapidly dominated wider swaths, causing total required implementation land to swell from 6.14 square kilometers to 678.07 square kilometers and leading the authors to recommend realistic operational corridor widths between 30 and 300 meters.

Circuit theory modeling through the Pinchpoint Mapper and Barrier Mapper modules delineated critical functional nodes across the landscape matrix. Current flow simulations across cost-weighted distance thresholds selected a parameter of 50,000 as optimal, where cumulative current density delineated 243 high-priority pinch points covering 21.99 square kilometers. These bottlenecks, where species movement is heavily constricted by surrounding land uses, were situated primarily across saline-alkali land at 25.14 percent, agricultural cropland at 24.02 percent, wetlands at 23.70 percent, and grasslands at 11.02 percent. Barrier detection using moving windows with a 600-meter search radius identified 31 primary impediment zones covering 33.33 square kilometers, which were heavily dominated by cultivated farmland at 71.08 percent and saline-alkali soil at 18.12 percent. These barriers exert disproportionate obstruction on regional ecological flows, indicating that targeted farmland-to-wetland conversion, wildlife underpass installations, and soil desalination represent essential interventions to restore landscape connectivity.

Synthesizing these spatial elements, the authors formulated the two corridors, two belts, and three zones optimization model. The two corridors comprise the Songhua River Ecological Corridor, which links the two largest regional sources measuring 986.24 and 590.11 square kilometers across a 19.47-kilometer path containing the single largest pinch point of 3.5 square kilometers, and the Tao’er River Ecological Corridor, anchored by an elongated 371.08-square-kilometer source with a perimeter of 1,695.6 kilometers. The two belts include the Central Lake-Wetland Ecological Restoration Belt, containing eleven sources, ten corridors, seventy pinch points, and eleven barriers linking Xianghai to Chagan Lake, and the Western Wetland-River Ecological Security Barrier Belt, which spans five sources and thirteen pinch points to buffer against eastward encroachment by the Horqin Sandy Land. The three zones encompass the Momoge Wetland Ecological Restoration Zone with 152 pinch points, the Chagan Lake Ecological Protection Zone with 148 pinch points, and the Xianghai Wetland Ecological Restoration Zone with fifteen pinch points, jointly encompassing 2,044.89 square kilometers and securing 54.98 percent of western Jilin’s core ecological source territory.

The authors emphasized several methodological and empirical caveats inherent in their modeling workflow. The assessment was constrained to five ecosystem services due to spatial data availability, omitting cultural and recreational values that also shape regional land management. Additionally, resampling multiple regional data layers to a standardized 30-meter grid inevitably smoothed fine-scale micro-topographical variations and rapid hydrologic transitions among saline flats, marshes, and agricultural ditches, potentially obscuring narrow habitat patches utilized by staging waterbirds. Some computational endogeneity persisted because predictor variables utilized in resistance modeling overlapped partially with indicators informing source identification. Furthermore, biodiversity validation was restricted primarily to ornithological monitoring datasets, lacking corroborating tracking records for aquatic fauna, amphibians, and terrestrial plant distributions. Future investigations are encouraged to deploy high-resolution remote sensing, multi-taxa telemetry, and empirical movement genetics to validate these regional corridors and support coordinated spatial governance, including the ongoing development of the regional Crane Hometown National Park.

Subject of Research: Ecological security pattern optimization and machine-learning resistance modeling in western Jilin, China

Article Title: Optimizing ecological security patterns in western Jilin, China: comparative ecological source identification and explainable machine-learning resistance modeling

Article References: Zhu, M., Yu, H., Ni, B., Zhang, P., & Zou, Y. (2026). Optimizing ecological security patterns in western Jilin, China: comparative ecological source identification and explainable machine-learning resistance modeling. Landscape Ecology, 41(11), Article 215. https://doi.org/10.1007/s10980-026-02470-4

Image Credits: AI Generated

DOI: 10.1007/s10980-026-02470-4

Keywords: ecological security patterns, western Jilin, Songnen Plain, ordered weighted averaging, morphological spatial pattern analysis, circuit theory, random forest, XGBoost, ecological corridors, pinch points, ecological barriers, wetland conservation, Ramsar wetlands, Momoge National Nature Reserve, Xianghai National Nature Reserve, landscape connectivity, ecosystem services, SHAP analysis

News Source: Teresa Odom. (November 18, 2026). Integrated Decision Modeling and Machine Learning Map Western Jilin Ecological Security Network. Scienmag.

Tags: circuit theoryecological barriersEcological corridorsecological security patternsEcosystem Serviceslandscape connectivityMomoge National Nature ReserveMorphological Spatial Pattern Analysisordered weighted averagingpinch pointsRamsar wetlandsRandom ForestSHAP analysisSongnen Plainwestern Jilinwetland conservationXGBoostXianghai National Nature Reserve
Share12Tweet7Share2ShareShareShare1

Related Posts

Exercise Improves Rheumatoid Arthritis Symptoms but Guideline Dose Effects Remain Uncertain

Exercise Improves Rheumatoid Arthritis Symptoms but Guideline Dose Effects Remain Uncertain

November 18, 2026
Slaughterhouse Veterinarians Face Public Invisibility and Stigma in Austrian Qualitative Study

Slaughterhouse Veterinarians Face Public Invisibility and Stigma in Austrian Qualitative Study

November 18, 2026

Neonatal Brain Injury Reshapes Gut Microbiome and Immune System Communication

November 18, 2026

Ethanol and Sublethal Heat Boost GABA Production in Lactiplantibacillus plantarum

November 18, 2026

POPULAR NEWS

  • Coal Gangue Aggregates Can Replace Natural Stone in Concrete Applications

    Coal Gangue Aggregates Can Replace Natural Stone in Concrete Applications

    29 shares
    Share 12 Tweet 7
  • Cluster-Based Training Selection Outperforms Active Learning for Material Property Prediction

    29 shares
    Share 12 Tweet 7
  • Cesarean Section Rates Reached Forty Percent Among Surveyed Mothers in Njombe Region

    29 shares
    Share 12 Tweet 7
  • Testosterone Injections Exacerbate Pre-Neoplastic Prostate Lesions in NKX3.1-Deficient Mice

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Exercise Improves Rheumatoid Arthritis Symptoms but Guideline Dose Effects Remain Uncertain

US Registry Shows Only Twelve Percent of Glaucoma Patients Receive Procedures

Zanubrutinib Associated With Improved Real-World Outcomes in Older CLL Patients

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 88 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.