Gimpo Habitat Profile: Does the Incheon AVIX-AI Model Transfer?

Abstract

Korea's K-UAM Roadmap 2030 designates Gimpo Airport as a primary inner-metropolitan vertiport anchor — the western gateway to the Han River corridor and a relief valve for Incheon's capacity constraints. The natural assumption for operators entering this space is that wildlife-strike mitigation models validated at Incheon can be ported to Gimpo with minimal rework. That assumption is wrong, and the cost of getting it wrong at scale is measurable: a species-class miss in the 0–150 m final approach band is not an academic error, it is a certification liability and a demonstrated density-cascade risk. This article presents a habitat-profile fitness study — not a product endorsement — examining whether AVIX-AI BirdThreat's Incheon training corpus is transferable to Gimpo's Han River basin ecology. The verdict is: partial transfer, with a mandatory recalibration pass. The article details the specific ecological deltas between sites, identifies the regulatory constraints that make a copy-paste deployment non-compliant, and outlines the structured assessment process that bridges the gap before the 2027 commercial window closes.

---

1. Operational Anchor — Gimpo Airport and the Inner-Metropolitan Vertiport Role

The Site

Gimpo Airport (IATA: GMP) occupies 16.4 km² in Gangseo-gu, Seoul, operating domestically under Korea Airports Corporation (KAC) management. As of 2026, it handles approximately 21 million annual passengers on the Gimpo–Jeju trunk route, with helicopter and business-aviation aprons already active on the northern perimeter. MOLIT's K-UAM infrastructure planning positions Gimpo as the primary western Seoul eVTOL node, feeding into the Yeouido financial district and the Mapo–Sangam demand corridor. Unlike Incheon International, Gimpo is embedded within a fully urbanised municipality — its eastern fence line abuts residential Gangseo-gu, and its northern boundary runs within 800 m of the Han River channel.

Environmental Read

The Han River basin supplies three persistent environmental variables that distinguish Gimpo from Incheon. First, the river's riparian corridor functions as a directional flight highway for waterbirds transiting between estuary staging grounds and inland reservoir roost sites — creating a predictable lateral threat vector across runway 14L/32R that has no Incheon equivalent. Second, the Han River Estuary Wetland Reserve (Ramsar-listed) sits 12 km west-northwest, generating a high-magnitude pulse of protected waterfowl during October–November Baikal Teal passage. Third, the reed-bed communities along the river's Gangseo marsh inserts generate thermal uplift signatures and acoustic clutter profiles that differ fundamentally from Incheon's tidal mudflat baseline.

Differential Factor

The critical ecological difference is guild structure, not just species composition. Incheon's dominant threat guild is intertidal shorebirds: compact, fast-moving flocks of Dunlin and Red-knot with predictable flock geometry and consistent wingbeat acoustic profiles. Gimpo's dominant threat guild is freshwater waterbirds: large-bodied Grey Herons and Black-crowned Night Herons operating solitarily at low altitude, supplemented by dense Eurasian Coot rafts and seasonally massive Baikal Teal flocks. These species present fundamentally different radar cross-sections, thermal signatures, and behavioural flight envelopes. An AI entity classifier calibrated on the Incheon guild will exhibit systematic false-negative rates on Gimpo's heron and coot classes — which is precisely the failure mode that matters most for eVTOL operators flying sub-150 m final approaches.

Modern Bridge

For vertiport operators submitting environmental management plans to MOLIT and KAC ahead of 2027 commercial licensing, this distinction is not theoretical. KAC's Wildlife Strike Prevention Programme requires site-specific strike risk assessments grounded in local species inventories — a generic Incheon-derived submission will not satisfy the Gimpo permit envelope. The operational question is therefore: what recalibration investment closes the gap between the validated Incheon model and a Gimpo-compliant deployment? That question frames every section that follows.

---

2. Problem Definition — The Habitat Generalisation Gap, Quantified

The assumption that AI wildlife-detection models transfer freely between aerodrome sites is not unique to Korea; it is a recurring failure mode in airport wildlife management globally. ICAO Doc 9332 explicitly cautions that site-specific ecological surveys are a prerequisite for any probabilistic strike-risk model, precisely because avian community composition varies across habitat types at scales as small as 5–10 km.

At Gimpo, the quantitative gap is significant. KAC's published wildlife strike data for Gimpo (2018–2023) show that waterbirds — primarily herons, egrets, and ducks — account for 41% of recorded strikes, compared with 18% at Incheon, where shorebirds and gulls dominate. This inversion means that a model architecture optimised for Incheon's threat profile will systematically under-weight the dominant strike category at Gimpo.

The seasonal concentration risk is compounding. The EAAF Baikal Teal staging event — historically concentrated in the Han River estuary during late October and early November — has recorded single-day counts exceeding 300,000 individuals at the broader estuary complex (EAAFP monitoring data). Even a small fraction of this mass redirected over Gimpo's approach corridors during weather displacement events represents a density-cascade scenario. An AI detection system carrying miscalibrated species priors for large waterfowl will fail to trigger the deterrence pipeline at the correct density threshold, exposing the vertiport to the simultaneous multi-aircraft grounding scenario that makes wildlife strike a systemic, not individual, operational risk.

The regulatory cost is compounding further. Under KAS Part 25 compatibility requirements for vertiport structural and operational certification, operators must demonstrate that safety-critical sensing systems are validated for the specific operating environment. A model validation certificate from Incheon Technopark (commit fbcb327, 2026-04-20) constitutes evidence of methodological soundness, but not site-specific environmental coverage. The 2027 commercial window creates a hard deadline: operators who cannot submit site-validated wildlife-risk mitigation plans during the MOLIT licensing sprint will face delayed or conditional approvals.

---

3. UAM KoreaTech Solution — Recalibration Architecture for Gimpo

AVIX-AI BirdThreat's 4-stage habitat treatment pipeline was designed with site portability as a first-order constraint, not an afterthought. The distinction matters here: the Incheon validation (19/19 HTTP 200 at Incheon Technopark) demonstrates that the pipeline architecture is sound; the recalibration exercise for Gimpo updates the environmental parameters without rebuilding the inference stack.

The recalibration sequence for a Gimpo deployment centres on three technical interventions. First, species-prior updating: the Animal-class entity taxonomy published natively into Anduril Lattice is extended to include the Gimpo-specific threat guild — Grey Heron, Black-crowned Night Heron, Eurasian Coot, and Baikal Teal — with flight-vector priors drawn from Han River basin ornithological survey data (NIER monitoring transects, eBird Korea records, and KAC strike logs). Second, acoustic-thermal fusion recalibration: the reed-bed and riparian thermal clutter signatures at Gimpo require updated background suppression filters to prevent false-positive suppression of large-bodied, slow-moving waterbirds that present a thermal profile similar to background vegetation during dawn and dusk roost-departure events. Third, density-threshold adjustment: the flock-density alerting thresholds must be reconfigured to account for Baikal Teal passage volumes, which can produce near-instantaneous density spikes an order of magnitude above Incheon's gull-flock baseline.

The Acoustic Vibration Mat (KAS Part 25 compatible, 90% absorption at 8–40 Hz) serves a complementary function at Gimpo that is more critical than at Incheon: rooftop vertipad noise signatures in the sub-40 Hz band are precisely the frequency range that attracts large waterbirds roosting in adjacent reed beds. Suppressing the low-frequency ground-borne vibration signature during pre-departure cycles reduces passive attraction before AVIX-AI BirdThreat's active deterrence pipeline needs to engage — a prevention-first posture that reduces deterrence event frequency and associated wildlife disturbance logging obligations under protected-species regulations.

Critically, the provenance discipline established at Incheon — structured commit logging, HTTP 200 validation at each pipeline stage — must be replicated at Gimpo to produce a certification-grade audit trail. MOLIT and KAC will require this documentation for the Gimpo vertiport licensing submission.

---

4. Strategic Context — Why the Gimpo Generalisation Question Defines the 2027 Window

The K-UAM Roadmap 2030 projects 200+ vertiports across the Korean peninsula, the majority of which will be sited along or adjacent to the EAAF flyway pinch corridor — a geographic reality driven by the concentration of Korean population and infrastructure in coastal and riverine lowlands. This means that the Incheon-to-Gimpo generalisation question is not a one-off technical exercise: it is the prototype for every subsequent vertiport deployment along the Seoul–Incheon–Gyeonggi corridor.

Operators and VCs scoping the 2027 commercial window need to recognise that the wildlife-mitigation certification backlog will be a rate-limiting factor. MOLIT's K-UAM working group has signalled that environmental management plan (EMP) submissions will be evaluated on a site-specific basis, and KAC's Wildlife Strike Prevention Programme does not accept cross-site extrapolation without empirical validation. The operators who enter 2027 with site-validated wildlife mitigation systems — including Lattice-published entity taxonomies, certified vibration-mat installations, and structured audit trails — will have a material lead over competitors attempting to compress this process during the licensing sprint.

The Han River basin corridor also has political salience that Incheon's industrial perimeter does not. The Ramsar-listed Han River Estuary is a high-visibility conservation asset; any vertiport operator perceived to be deploying inadequate or inappropriate wildlife-mitigation tooling adjacent to this site will face regulatory and public-affairs headwinds that go beyond technical compliance. This is a dual-use constraint: the same site conditions that make Gimpo ecologically challenging make a well-documented, peer-reviewable mitigation approach a competitive differentiator rather than a cost centre.

---

5. Forward Outlook

The 12-month path to a Gimpo-validated AVIX-AI BirdThreat deployment follows three sequential milestones. By Q3 2026, a baseline ornithological survey aligned with ICAO Doc 9332 survey protocols should be completed, establishing the site-specific species inventory against which model priors will be updated. By Q4 2026, acoustic-thermal sensor installation and 60-day baseline data collection should produce the calibration dataset needed to update the freshwater clutter suppression filters. By Q1 2027, a recalibrated model validation run — structured to mirror the Incheon Technopark protocol — should generate the HTTP 200 audit record and Lattice entity-publication log required for MOLIT EMP submission.

The Acoustic Vibration Mat accelerometer audit at install can be sequenced in parallel with sensor baseline collection, allowing the full vertipad acoustic profile to be documented within the same Q4 2026 window. This sequencing is achievable within the 2027 commercial licensing timeline if site access agreements with KAC are initiated no later than August 2026.

---

Conclusion

The Incheon AVIX-AI model does not transfer to Gimpo by assumption — it transfers by structured recalibration, and that distinction is the entire content of this fitness study. For operators building their 2027 licensing submissions, the Han River basin's freshwater guild structure and EAAF staging significance are not obstacles to be minimised; they are site-specific inputs that, properly addressed, produce a certified wildlife-mitigation posture that no late-entry competitor can replicate without investing the same field-validation time. The 2027 commercial window is narrow, and the documentation lead — built through disciplined habitat profiling and provenance-grade audit trails — compounds into a durable regulatory moat.