Measuring Climate Exposure in Property Due Diligence: A Building-Level Framework for Flood Assessment and Valuation-Relevant Evidence

1. Introduction

Climate exposure in property due diligence is often reduced to a binary screen: a property is either inside or outside a mapped flood zone. For transaction decisions, that simplification is frequently inadequate. Buyers, lenders, insurers, and valuers may need to know how exposure reaches the building, which spaces are affected, whether access is cut off, and how credible the underlying evidence is. Two assets in the same polygon may differ materially because one has a basement, low thresholds, or vulnerable road access while the other does not.

This narrow problem matters because hydrological hazards dominate the climate-risk valuation literature. The systematic map of climate risk and real-estate valuation reports that physical risk studies, especially flood studies, are most common, and that exposure is usually measured with official hazard maps or proxies such as elevation and shoreline distance. The same source identifies a need for granular exposure information and notes that compound and cascading risks are rarely analyzed explicitly. The practical gap is therefore clear: valuation research shows that exposure matters, but it does not yet provide a transferable building-level protocol for pre-transaction measurement. [1]

This article proposes a building-level flood exposure framework for due diligence rather than a citywide hazard-management model. The design synthesizes valuation evidence, urban flood digital-twin methods, fine-grained property data, occupant-centered exposure logic used here only as a methodological analogy, and remote-inspection principles. The objective is to specify a falsifiable measurement architecture and prospective testing protocol; no calibration or empirical performance test is reported. [2] [3] [4] [5]

2. Research question and evidence scope

The main research question is: how can climate exposure in property due diligence be measured at building level, with emphasis on flood hazards, in a form that is interpretable for valuation-linked decisions and transferable across urban contexts? Two subsidiary questions follow. First, which variables should enter a due-diligence exposure measure when the purpose is transaction screening rather than explanation of observed prices? Second, how should static geospatial evidence, dynamic flood modeling, and inspection-derived building information be combined without overstating precision?

The evidence scope is targeted rather than exhaustive. The valuation map is used to identify how climate exposure has been operationalized in real-estate studies. The Galveston geospatial digital twin is used as direct evidence that flood depth, extent, duration, and velocity can be modeled at hyperlocal scale with building relevance. The Nanning housing dataset is used as evidence that structural, environmental, and accessibility variables can be integrated at property scale through GIS. Methodological support comes from the occupant-centered air-quality metric, used here only for the transferable idea of linking conditions to a subject-specific trajectory rather than a static average; from the remote-inspection review, which distinguishes monitoring, inspection, and digital-twin architectures; and from the drainage digital twin, which contributes a framework for infrastructure sensing, state monitoring, and scenario analysis. [1] [2] [3] [4] [5] [6]

These sources are methodologically heterogeneous, and that heterogeneity matters. The Nanning dataset is cross-sectional and contains a cross-year mismatch in one contextual variable. The drainage digital twin reports validation in a single city and notes limited evidence of universality across climates and urban sizes. The occupant-centered metric was developed for indoor air quality rather than flooding. Accordingly, the present article treats these studies as transferable methodological evidence, not as direct validation of the proposed framework. [3] [6] [4]

3. Prior methods and their relevance to due diligence

The dominant valuation-style approach is parsimonious exposure measurement. The systematic map reports that hydrological hazards, especially coastal and fluvial floods, are most frequently studied, and that exposure is commonly represented by official hazard maps or broad proxies. Some studies also use proximity measures to hazard hotspots or adaptation interventions when direct risk measures are unavailable. These approaches are useful for large-sample econometric work, but they are too coarse for due diligence when entry points, below-grade spaces, and access conditions drive decision relevance. [1]

Digital-twin flood studies show a different measurement logic. In Galveston, 1D-2D coupled hydrodynamic simulations integrate LiDAR, land cover, storm-sewer geometries, and precipitation scenarios to model flood depth, extent, duration, and velocity. The study argues that static maps may underestimate future damage under changing climate and urbanization conditions and emphasizes not only whether an area floods, but how deep, how fast, and with what timing. For due diligence, the transferable lesson is that building exposure should be represented as a scenario-conditioned vector of intensities rather than as zone membership alone. [2]

Fine-grained property datasets contribute a second lesson: exposure-relevant and market-context variables can be assembled together at parcel or community scale. The Nanning dataset includes structural variables, distances to amenities, water area within a 1 km buffer, interpolated air quality, and population density extracted through GIS overlay. This supports a due-diligence design in which flood exposure is analyzed alongside locational context. At the same time, the source warns that cross-sectional timing and variable mismatch can limit interpretation, which implies that due-diligence measurement should record temporal uncertainty instead of concealing it. [3]

Two further sources provide methodological transfer. The occupant-centered exposure metric shifts analysis from room conditions to an individual trajectory across rooms, combining instantaneous conditions with cumulative exposure over time. In this manuscript, that logic is transferred only conceptually: flood due diligence can move from area averages to building-specific, horizon-specific exposure without implying that an indoor air-quality validation also validates flood measurement. The remote-inspection review distinguishes monitoring from inspection and describes digital twins as combining data acquisition, simulation, prediction, and visualization. That distinction is useful because due diligence usually relies on discrete inspections supplemented by external data rather than on continuous sensor coverage. [4] [5]

4. Conceptual model of building-level flood exposure

Building-level flood exposure for due diligence is defined here as the degree to which a specific property improvement, together with its essential access and service interfaces, is subject to inundation or flood-induced operational disruption under specified scenarios over a stated time horizon. This is narrower than full risk because it does not require complete monetary loss modeling, and broader than hazard presence because it includes pathways, receptors, and selected off-site dependencies.

Three distinctions organize the concept. First, exposure is scenario-conditioned: a property can be low exposure under one rainfall event and high exposure under another. Second, exposure is horizon-specific: valuation-oriented discussion of flood probability shows that low annual probability can become material across a mortgage or holding period. Third, exposure is evidence-weighted: a coarse polygon screen and a validated building-scale model should not be treated as equally reliable measurements. [1] [2]

The proposed framework separates five analytical layers. Hazard intensity captures external water depth, velocity, duration, and event probability. Pathway captures how water reaches the building, such as overland flow, sewer surcharge, or basement ingress. Receptor captures building features that condition exposure, including below-grade space and vulnerable occupied or service areas. Dependency captures off-site elements required for functional use, especially roads and drainage. Evidence quality captures the granularity, currency, and validation status of the assembled information. The output is therefore a structured exposure profile, not a single unlabeled hazard tag. [6] [5]

5. Proposed methodology: B-FEAT

The proposed method is Building Flood Exposure Assessment for Transactions, or B-FEAT. The unit of analysis is one building or one transacting property complex. The minimum data package contains parcel geometry, building footprint, use class, any known below-grade space, surrounding elevation information if available, at least one flood-hazard source, and basic road context. Optional inputs include remote-inspection imagery, LiDAR-derived elevations, drainage-network information, critical-equipment location, and historical flood observations.

Operational variables are grouped into six domains. Domain A contains receptor variables: use type, first occupied floor, basement indicator, estimated finished-floor relation to grade, age, and observable envelope condition. Domain B contains site morphology variables: local slope, depression potential, impervious cover, distance to surface water, and water area within a fixed buffer. Domain C contains hydraulic hazard variables: external depth, velocity, inundation duration, annual exceedance probability, and hazard type. Domain D contains infrastructure dependency variables: access-road depth, proximity to road disruption, drainage bottleneck status where known, and identifiable utility exposure. Domain E contains market-context variables such as population density and accessibility, used for comparison rather than direct exposure scoring. Domain F contains evidence-quality variables: source type, spatial resolution, temporal currency, validation status, and degree of direct building observation. [3] [2] [5]

The core index for building b under scenario s and horizon H is FEI(b,s,H)=wD·D+wV·V+wT·T+wP·P+wR·R, where each component lies in [0,1] and the active weights sum to 1. D is normalized flood depth, V normalized velocity, T normalized duration, P pathway susceptibility, and R receptor-dependency susceptibility. Evidence quality is excluded from FEI because reliability and severity are analytically separate. The default full-information screening vector is preregistered as w=(0.30, 0.15, 0.15, 0.15, 0.25) for (D,V,T,P,R). When reduced-form application lacks V, T, or both, the omitted components are removed and the remaining weights are renormalized proportionally over the observed set. Thus, if V is missing, the active weights become (0.3529, 0.1765, 0.1765, 0.2941) over (D,T,P,R); if both V and T are missing, they become (0.4286, 0.2143, 0.3571) over (D,P,R). No empirical calibration is claimed. A future study may compare this default with a constrained sensitivity set in which each full-information weight varies by ±0.05, subject to non-negativity and unit sum, and reduced-form weights are generated only by the same proportional renormalization rule.

Normalization is explicit. Let d be scenario-specific maximum water depth at the building interface, v maximum local velocity, and t inundation duration above a trace-water threshold. Then D=min(d/1.0,1), V=min(v/1.0,1), and T=min(t/24,1), with units meters, meters per second, and hours. These caps are screening anchors rather than universal damage thresholds. Raw values d, v, and t are always retained and reported even when the normalized component is set to 0 or 1. Missing-data handling is rule-based. Depth is critical for all assets; if D cannot be assigned, no FEI is computed. Pathway information is critical when below-grade occupied or service space is present; if P cannot be assigned, the case escalates. Receptor information is critical when occupied space, critical equipment, or loading access is on or below grade; if R cannot be assigned, the case escalates. Velocity and duration are non-critical for reduced-form use; when unavailable, FEI is computed with the renormalized reduced-form weights defined above, but the resulting score must be labeled reduced-form. Evidence quality is always scored separately; if fewer than three Q sub-scores can be assigned, FEI may be reported only as provisional and cannot receive acceptable status under the decision rule.

One exact operational rule is as follows: if modeled or observed water depth at every building access point and around the full footprint is less than 0.05 m for a scenario, then D is set to 0 for that scenario’s normalized score, while the raw depth observations remain in the scenario table. This trace-water rule prevents minor wetting from being treated as inundation in FEI. Pathway susceptibility P is scored by rule: P=1.00 if the property has a basement and depth exceeds 0.15 m at the perimeter or an access point, or if sewer surcharge evidence exists; P=0.75 if no basement is known but multiple low-threshold entries are exposed above 0.15 m; P=0.50 if ponding plausibly reaches service-critical space; P=0.25 for shallow external ponding with no identified route to critical or occupied space; and P=0 if no pathway is identified. These are proposed screening defaults for prospective testing, not validated universal thresholds. [2] [6]

Receptor-dependency susceptibility is defined as R=0.6·BR+0.4·IR, where BR is the building receptor score and IR the infrastructure dependency score, each on [0,1]. BR is assigned by the highest applicable band, with ties resolved upward when two bands are equally supported: BR=1.00 if any occupied room is below grade, or if critical building equipment necessary for habitability or operation is located below grade or at grade in a space directly exposed to modeled or observed water entry; BR=0.75 if the lowest occupied floor is at grade with at least one low-threshold entry or opening likely to admit water, or if near-grade equipment appears exposed; BR=0.50 if occupied use begins above grade but common areas, storage, parking, or noncritical services at or below grade appear exposed; BR=0.25 if no occupied or service-critical space is identified at or below grade but minor exterior contact with the envelope is plausible; BR=0 if no receptor vulnerability is identified from available evidence. IR is assigned by the highest applicable band: IR=1.00 if all practical vehicular and pedestrian access routes are predicted or observed to be disrupted for the scenario, or if drainage malfunction or surcharge affecting the site is directly evidenced; IR=0.75 if the main access route is disrupted but at least one inferior or uncertain route may remain, or if a probable drainage bottleneck is identified from topography or network information; IR=0.50 if shallow road flooding or localized drainage stress is likely to impede access intermittently; IR=0.25 if access remains passable with only minor nuisance ponding; IR=0 if no access or drainage dependency issue is identified. For IR, disagreement between road and drainage evidence is resolved conservatively by the higher supported band because either mechanism can independently disable use of the asset. This conflict rule is confined to IR; other domain disagreements are not automatically resolved upward and instead require explicit note in the evidence statement.

Evidence quality is formalized separately as Q = 0.25*S1 + 0.20*S2 + 0.20*S3 + 0.20*S4 + 0.15*S5. S1 is source directness, from building-specific modeled or observed data down to proxy-only inference. S2 is spatial resolution adequacy. S3 is temporal currency. S4 is validation status, including whether the model or source has been checked against historical events or equivalent procedures. S5 is direct observation of building configuration through inspection, remote reconstruction, or reliable records. Q is reported alongside FEI as a confidence score, not as a severity component. If fewer than three sub-scores can be assigned, Q is indeterminate and the numeric FEI cannot support an acceptable classification. [5] [2]

A second exact rule links annual event probability to holding-period relevance under a stated assumption set. If p is a constant annual exceedance probability and annual event occurrences are treated as independent across years, then the probability of at least one occurrence during a holding period of H years is PH=1-(1-p)^H. B-FEAT reports both annual scenario labels and this horizon-conditioned probability only when those stationarity and independence assumptions are adopted for screening. If the analyst expects non-stationary hazard conditions or serial dependence, PH is not computed from this formula and the report must state that only annual scenario information is used. Scenario outputs are handled in two stages. First, each scenario is classified into one of three analyst-defined bins before any cross-scenario summary: routine scenarios, defined as the ordinary due-diligence screening set for the asset and locality; stress scenarios, defined as severe but decision-relevant contingencies; and exploratory scenarios, which are reported descriptively only. Second, FEI*=max_s FEI(b,s,H) is computed only within the routine scenario bin. Stress-scenario maxima are reported separately as FEI†, and exploratory scenarios never determine status. This rule prevents a remote worst case from dominating routine triage while preserving transparency about more severe contingencies. [1]

6. Decision sequence and comparator

The decision sequence has eight steps. Step 1: define transaction purpose, asset class, holding period, and the scenario register before reviewing results. The register must list each scenario, its hazard type, source, annual probability if available, and its bin as routine, stress, or exploratory. Step 2: assemble baseline spatial data and one or more flood sources. Step 3: derive receptor variables from records, imagery, or remote inspection. Step 4: extract depth, velocity, and duration from a dynamic model where available, or derive a reduced-form profile from maps and topographic proxies. Step 5: score pathway and receptor-dependency susceptibility. Step 6: score evidence quality separately. Step 7: compute FEI by scenario, identify FEI*=max within the routine scenario bin, and report FEI† separately for the stress-scenario bin if such scenarios were prespecified. Step 8: assign a decision status by the following default rule. Status=acceptable if FEI*<0.30 and Q≥0.60, provided no critical variable is missing, Q is determinate, and no access-cutoff band of IR=1.00 is triggered in the routine scenario that defines FEI*. Status=review required if 0.30≤FEI*<0.60, or if Q is between 0.40 and 0.59, or if FEI* is reduced-form rather than full-form, or if any single component among D, P, or R is at least 0.75 even when FEI* is lower. Status=escalation to engineering assessment if FEI*≥0.60, or Q<0.40, or Q is indeterminate, or any critical variable is missing, or D=1 with P≥0.75, or IR=1.00 for the routine scenario used for FEI*. These cutoffs are proposed by domain judgment for conservative screening, not by optimization or pilot-data calibration; they are therefore defaults for preregistered testing rather than validated operating thresholds.

The principal comparator is the simplified approach that assigns exposure through static hazard-zone membership or a single proxy such as elevation. That comparator is justified because the valuation map identifies hazard maps and broad proxies as common methods. B-FEAT does not claim demonstrated predictive superiority. Its proposed advantage is more explicit measurement structure: it distinguishes buildings within the same zone, incorporates road and drainage dependencies, and states evidentiary confidence instead of leaving these factors implicit. [1]

7. Valuation linkage and due-diligence outputs

B-FEAT is an exposure measure, not a valuation model, but it is intended for valuation-relevant downstream use. The climate-risk valuation map identifies transaction prices, asking prices, appraisals, expected damages, and cash-flow variables as outcomes used in prior studies. In that context, a due-diligence exposure measure may inform appraisal review, insurance inquiry, adaptation planning, and debt-structure assessment without directly inferring a price effect. That restraint is necessary because capitalization depends on disclosure, salience, and information frictions. [1]

The due-diligence output should contain four linked items. First, a scenario table reports raw intensities d, v, and t, normalized components D, V, and T, the scores P and R, the separate confidence score Q, FEI for each prespecified scenario, and FEI* as the maximum across scenarios for the stated horizon. Second, a short valuation-interface note records the intended downstream uses of the output, such as appraisal review, insurance inquiry, adaptation planning, or financing review, without asserting any price effect. Third, an evidence statement specifies whether the result is map-based, model-based, inspection-supported, or engineering-grade and identifies any critical missing fields. Fourth, the decision status and the rule that triggered it are recorded verbatim so that another reviewer can reproduce the classification.

The framework also improves comparability across assets because context variables such as accessibility, water area, and population density can be derived consistently in GIS. This helps separate flood exposure from neighborhood desirability, which is important where amenity value and hydrological proximity coincide. [3] [1]

8. Prospective validation protocol and limits

No empirical validation is claimed here, so a prospective protocol is essential. Validation should address three questions: measurement validity, decision validity, and transfer validity. Measurement validity asks whether FEI corresponds to observed or engineer-judged building impacts in historical events. Decision validity asks whether B-FEAT improves triage relative to static hazard-zone screening. Transfer validity asks which components of the framework remain usable across different topographies, drainage systems, and asset types.

A future measurement-validity study should assemble cases with post-event observations, claims records where accessible, or engineering reports, and classify outcomes such as no material impact, exterior-only impact, below-grade impact, or occupied-space impact. Historical flood observations have two distinct roles. If they predate the study protocol and are used to check whether an already specified B-FEAT score matches observed impact, they serve validation. If they are used to alter weights, thresholds, or scoring bands, they serve calibration and must be kept separate from the validation set. The main comparator should be hazard-zone classification alone. Relevant outcomes include false-negative rate for materially affected buildings, ordinal agreement, and stability when reduced-form and full-form B-FEAT are applied to the same cases. Because Q is not part of FEI, validation should test severity discrimination and confidence screening separately: FEI against observed impact bands, and Q against the frequency of later score revision after additional evidence is collected. For the preregistered threshold-sweep analysis, the estimands should be stated explicitly: change in continuous FEI, change in decision status, and change in agreement with observed impact bands relative to the prespecified unswept default. [2] [1]

Transfer limits are substantial. The drainage digital twin notes limited universality across climates and city sizes, and the Galveston twin indicates that larger or hillier cities may require different model structures or computing resources. Remote inspection can document geometry and visible conditions but may miss concealed vulnerabilities such as interior floor levels or backflow protection. The valuation map also shows that existing evidence is heavily concentrated in single-family housing, which restricts direct transfer to more complex multifamily and commercial properties. Compound and cascading risks remain only partly represented through dependency variables and multi-scenario profiling. The proposed defaults, including the 0.05 m trace-water rule, 0.15 m pathway trigger, and decision cutoffs, are therefore screening values for prospective testing rather than validated thresholds. Any empirical study should preregister a threshold-sweep sensitivity analysis at minimum over trace-water thresholds of 0.02, 0.05, and 0.10 m, pathway triggers of 0.10, 0.15, and 0.20 m, and acceptable/review and review/escalation FEI cutoffs shifted by ±0.05 around the default values, with resulting changes in FEI, status, and agreement with observed impact all reported. [6] [2] [5] [1]

9. Conclusion

This article proposes a bounded conceptual framework for measuring climate exposure in property due diligence with a narrow emphasis on building-level flood assessment. The synthesis supports three limited propositions. Prior real-estate valuation studies usually measure flood exposure through hazard maps or broad proxies. Digital-twin studies show that flood exposure can be represented through depth, duration, velocity, and scenario analysis at hyperlocal scale. Building- and neighborhood-level variables can be assembled in a due-diligence workflow, while ideas from occupant-centered monitoring and digital inspection can be transferred as methodological analogies rather than as direct performance evidence for flood assessment. [1] [2] [3] [4] [5]

On that basis, the article proposes B-FEAT, a sequential method with defined variables, formulas, scoring rules, a separate evidence-quality layer, missing-data rules, scenario reporting, and decision thresholds for transaction screening. The manuscript's contribution is limited to method design for a specified problem: building-level flood exposure measurement for due diligence in acquisition, lending, underwriting, and appraisal review. It does not equate exposure with monetary loss or market price effects.

Important elements remain untested. The proposed weights, normalization anchors, thresholds, cutoffs, and rubrics are provisional. The method has not been calibrated against observed building impacts, compared empirically with standard hazard-zone screening in live transactions, or tested across asset classes and cities. FEI* is intentionally limited to the maximum within a prespecified routine scenario set, while stress scenarios are reported separately, to avoid allowing an exploratory worst case to determine routine triage. The holding-period probability transformation is intended only for screening cases that explicitly assume time-invariant annual exceedance probabilities and independence across years. The next research step is a preregistered evaluation that separates calibration from validation, reports sensitivity in both scores and classifications, and identifies when the added information of B-FEAT justifies its data requirements.

References

1. Piseddu, Tommaso; Vanhuyse, Fedra (2026). Existing evidence on the impact of climate risk on real estate valuations: a systematic map. Environmental Evidence. 15(1), 8. doi:10.1186/s13750-026-00389-6.
2. Mandal, Debayan; Zou, Lei; Wadhwa, Abhinav; Wilkho, Rohan Singh; Cai, Zhenhang; Zhou, Bing; Ye, Xinyue; Newman, Galen; et al. (2026). FlowsDT: A geospatial digital twin for navigating urban flood dynamics. Computers, Environment and Urban Systems. 126, 102414. doi:10.1016/j.compenvurbsys.2026.102414.
3. Lin, Wenbo; Wu, Kaichuang; Zhang, Xiaolu; Gu, Xiaoling; Jiang, Ming; Wang, Jun; Han, Feng (2026). A comprehensive dataset for uncovering housing price drivers in emerging cities: Nanning as a case study. Data in Brief. 66, 112884. doi:10.1016/j.dib.2026.112884.
4. Ruffa, Filippo; Iadarola, Grazia; De Capua, Alberto; De Capua, Claudio (2026). A Method for Optimized Monitoring of Indoor Air Quality in Public Buildings. Sensors. 26(14), 4559. doi:10.3390/s26144559.
5. Ribeiro, Diogo; Rakoczy, Anna M.; Cabral, Rafael; Hoskere, Vedhus; Narazaki, Yasutaka; Santos, Ricardo; Tondo, Gledson; Gonzalez, Luis; et al. (2025). Methodologies for Remote Bridge Inspection—Review. Sensors. 25(18), 5708. doi:10.3390/s25185708.
6. Zhao, Yuanjing; Yang, Min (2026). Modelling and simulation of smart city drainage system based on digital twin five-dimensional models. PLOS One. 21(7), e0352787. doi:10.1371/journal.pone.0352787.