Enterprise AI Analysis
Harmonizing Community Science Datasets to Model Highly Pathogenic Avian Influenza (HPAI) in Birds in the Subantarctic
Community science observational datasets are useful in epidemiology and ecology for modeling species distributions, but the heterogeneous nature of the data presents significant challenges for standardization, data quality assurance and control, and workflow management. In this paper, we present a data workflow for cleaning and harmonizing multiple community science datasets, which we implement in a case study using eBird, iNaturalist, GBIF, and other datasets to model the impact of highly pathogenic avian influenza in populations of birds in the subantarctic. We predict population sizes for several species where the demographics are not known, and we present novel estimates for potential mortality rates from HPAI for those species, based on a novel aggregated dataset of mortality rates in the subantarctic.
Executive Impact: Key Metrics & Insights
Our analysis reveals the power of harmonized community science data in critical ecological modeling.
0
Bird Observations on eBird
0
Biodiversity Observations on iNaturalist
0
Key Species Modeled for HPAI
0
Est. Brown Skua Population (SINZ)
0
Predicted Skua HPAI Mortality Rate
0
Est. Wandering Albatross Pop (SINZ)
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Introduction & Data Challenges
Methodology & Harmonization
Results & Implications
Biases & Future Work
Community science platforms like eBird and iNaturalist have amassed enormous datasets, offering unprecedented scale for ecological research. However, their crowd-sourced nature introduces specific biases that must be addressed for reliable scientific use.
1 Billion+
Verified Bird Observations on eBird
250 Million+
Crowd-sourced Biodiversity Observations on iNaturalist
| Bias Type | Description | Impact on Data |
|---|---|---|
| Sampling Bias | Observations concentrated where people are (e.g., weekends, populated areas); charismatic species are over-reported. | Skews geographic and species distribution, potentially misrepresenting actual populations. |
| Skill Bias | Observers vary widely in expertise, with no confidence levels for reporting. | Introduces variability in species identification and counting accuracy. |
| Detection Bias | Some birds are easier to find (near humans, larger, diurnal) than others (pelagic, nocturnal). | Undercounts cryptic or remote species, overcounts common or visible ones. |
| Duplicated Effort | Multiple observers can report the same individual without shared metadata, leading to inflated counts. | Creates skewed distributions and overestimates of population sizes. |
| Over-reliance on AI | Automated identification tools can be inaccurate, leading naïve users to report wrong species. | Introduces misidentification errors into the dataset. |
| Gamification | Platform features encouraging competitive logging can prioritize quantity over data quality. | May lead to rushed or less careful observations, degrading overall data reliability. |
Our methodology involved a rigorous multi-stage pipeline to clean, harmonize, and integrate data from various community science platforms for robust ecological modeling of HPAI in subantarctic bird populations.
Enterprise Process Flow: Data Harmonization Pipeline
Download Data
→
Clean & Standardize Datasets
→
Dataset Filtering
→
Geographic Filtering
→
Taxonomic Reconciliation
→
Taxonomic Filtering
Case Study: HPAI H5Nx Modeling in the Subantarctic
Highly Pathogenic Avian Influenza (HPAI) H5Nx strains are decimating bird populations globally, recently spreading to the subantarctic regions like Falkland Islands, SGSSI, Crozets, Kerguelen, and PEI. This study focuses on the Subantarctic Islands of New Zealand (SINZ), an area not yet affected by HPAI but vulnerable due to fragile ecosystems and existing species declines.
Key species like Brown Skua, Kelp Gull, Southern Giant Petrels, and Wandering Albatrosses are identified as potential vectors, with Skuas being migratory and kleptoparasitic, making them a primary concern. The model predicts population sizes and potential mortality rates for these species in SINZ based on data from affected regions and multiple community science datasets.
Our novel estimates show a predicted 17.74% mortality rate for Brown Skuas and 1.46% for Wandering Albatrosses in SINZ, emphasizing the severe potential impact if HPAI reaches these vulnerable populations. These projections are critical for informing conservation efforts and public policy.
Our modeling efforts yielded novel population estimates and critical HPAI mortality rate predictions for key subantarctic bird species, informing potential impacts if the virus spreads further.
3,565
Estimated Brown Skua Population (SINZ, Model 2)
7,507
Estimated Wandering Albatross Population (SINZ, Model 3)
17.74%
Predicted Brown Skua HPAI Mortality Rate (Model 2)
0.03%
Predicted King Penguin HPAI Mortality Rate (Macquarie, Model 1)
While our models provide valuable insights, it's crucial to acknowledge the inherent biases and limitations in community science data and the specific model assumptions. Addressing these will be key for future research.
| Limitation | Description | Impact on Model |
|---|---|---|
| Sparse Data Robustness | Data for subantarctic islands is limited, leading to less robust inferences. | May reduce confidence in population estimates for under-surveyed areas. |
| Platform Differences | iNaturalist observations skew towards charismatic, easy-to-photograph species; eBird protocols vary. | Introduces observational biases affecting species representation. |
| Island Specificity | Each island has unique geophysical and ecological properties not fully accounted for. | Generalizations across islands may overlook critical local factors. |
| Species Differences | Mortality impacts vary by species (e.g., albatross chicks vs. skuas). | Assumed similar impacts may not hold true across diverse species. |
| Seabird Mortality at Sea | Many seabirds die at sea, but shore counts are typically unsystematic. | Underestimation of total mortality if not accounting for at-sea deaths. |
| HPAI Impact Variation | HPAI may affect different bird populations in distinct ways due to varying immunities or exposure. | Extrapolations from limited HPAI data may not universally apply. |
First-ever
Published HPAI Mortality Projections for SINZ
Calculate Your Potential AI Impact
Estimate the transformative power of AI-driven insights on your operational efficiency and cost savings.
Your Enterprise Profile
Projected Annual Impact
Est. Annual Cost Savings
$0
Est. Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A structured approach to integrating AI, from strategy to sustainable impact.
Phase 1: Discovery & Strategy
Duration: 2-4 Weeks. Comprehensive assessment of existing infrastructure, data landscape, and business objectives. Identification of high-impact AI opportunities and development of a tailored AI strategy roadmap.
Phase 2: Pilot & Validation
Duration: 6-12 Weeks. Design and implementation of a targeted AI pilot project. Focus on proving concept, validating ROI, and collecting user feedback for iterative refinement.
Phase 3: Scaled Deployment
Duration: 3-6 Months. Full-scale integration of validated AI solutions across the enterprise. Includes robust data pipeline development, security hardening, and comprehensive user training programs.
Phase 4: Optimization & Future-Proofing
Duration: Ongoing. Continuous monitoring, performance optimization, and adaptation of AI models. Exploration of advanced AI capabilities and proactive planning for evolving business needs and technological advancements.
Ready to Transform Your Enterprise with AI?
Unlock the full potential of your data and drive significant business impact with a customized AI strategy.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat