PlanX Suitability Lab
Academic Reference Manual

1.2 Vector Rasterizer

GDAL Rasterize vector-to-raster attribute burn

Theory & Method

Converts vector features (points, lines, polygons) to a single-band Float32 raster via GDAL's gdal_rasterize. A specified attribute field or constant burn value is assigned to cells whose centre falls within a feature. This is the primary bridge from vector GIS data to the raster MCDA domain.

Interpretation: Output extent matches the input vector layer bounds. Cells not covered by any feature are nodata (0). For categorical land-use data, use the field containing the suitability score directly.

1.3 Proximity Builder

GDAL Proximity Euclidean distance continuous surface

Theory & Method

Computes Euclidean distance from every cell to the nearest target pixel (cells whose band-1 value matches user-specified integer targets). The algorithm uses the two-pass chamfer distance transform internal to GDAL Proximity with Euclidean metric, producing a continuous proximity surface.

Typical use: Distance to roads, transit stops, water bodies, or hazard zones — as normalisable suitability criteria.

1.5 Constraint Builder (0/1 Mask)

Boolean mask thresholding exclusion

Theory & Method

Applies a Boolean threshold \(R(x,y) \ge \tau\) to a criterion raster, producing a binary mask where 1 = suitable (passes constraint) and 0 = excluded (fails constraint). This implements the “hard constraint” layer in GIS-MCDA: areas that are absolutely infeasible regardless of other criteria scores (e.g., protected areas, water bodies, slopes exceeding regulatory limits).

Multiple constraint masks can be multiplied to create a composite exclusion layer. Constraint-failing cells in MCDA scoring are forced to 0.

1.6 Synthetic Sample Raster Pack

test data smoke test synthetic

Theory & Method

Generates five synthetic Float32 GeoTIFFs (four criteria rasters with controlled spatial patterns and one binary constraint mask) for pipeline testing without real data. Each criterion is computed from normalised coordinate grids with distinct spatial structures: eastward decay, northward increase, diagonal ridge, and radial hotspot, all with Gaussian noise (seed = 7). The constraint mask excludes the bottom-left 15% by 10% of the study area. All rasters are EPSG:3857, clipped 0–100, nodata -9999.

Output files: criterion_1.tif through criterion_4.tif and constraint_mask.tif.

1.7 Synthetic End-to-End Pipeline Runner

integration test smoke test verification

Theory & Method

Chain-runs 35+ sub-algorithms in deterministic sequence: synthetic data generation, four MCDA methods (WLC, OWA, TOPSIS, VIKOR) with equal weights, Monte Carlo uncertainty (120 iterations, sigma 0.08), AHP/PCA/Entropy/CRITIC weight derivation, ensemble synthesis, scenario consensus/disagreement/quadrant/intervention mapping, risk-opportunity matrix, regret fragility, opportunity loss, priority zoning (baseline + refined), multi-scenario benchmarking, robustness ranking, budget phasing, deployment readiness gating, and the full reporting suite (HTML report, Markdown brief, beta evidence HTML). Produces 70+ output files.

Typical use: One-click verification that every algorithm executes correctly after installation or dependency updates.

2.2 PCA-derived Weights (Variance Proxy)

eigendecomposition PC1 loading data-driven

Theory & Method

Samples up to ~8000 cells from the aligned raster stack, standardises (z-score), and computes the covariance matrix. Eigendecomposition yields principal components; the absolute loadings of PC1, scaled by its explained variance ratio, become the weights. This variance-proxy approach assigns higher weight to criteria driving the largest spatial differences. If fewer than 3 valid sample cells exist, equal weights are returned.

Interpretation: High weight = criterion strongly loads on the dominant spatial-variance axis. Low weight = criterion is spatially uniform or anti-correlated with the dominant pattern. This method rewards criteria with high spatial variability, which may not always align with decision relevance.

2.5 Weight Consolidator

normalisation activation step

Reads any weight table (from AHP, PCA, Entropy, or CRITIC) and re-normalises all weights to sum to 1. This is a pass-through activation step: relative proportions are preserved, but the output is guaranteed ready for downstream MCDA scoring. Guards against zero-sum edges by substituting a divisor of 1.0 when the total sum is zero. Output: criterion_id, weight. Reference: Malczewski (1999).

2.6 Scenario Weight Pack

multi-method merge blended average sensitivity analysis

Merges AHP, PCA, Entropy, and CRITIC weight tables into a unified scenario pack. Each method's weights are independently normalised to sum to 1. An additional BLENDED_AVG scenario computes the per-criterion arithmetic mean of normalised weights across all supplied methods, creating a simple ensemble that balances subjective and objective approaches. Output columns: method, criterion_id, weight_raw, weight_norm. Use: Prepare multi-scenario weight sets for sensitivity analysis or consensus auditing.

2.7 Weight Consensus Audit

coefficient of variation cross-method agreement robustness

Computes per-criterion descriptive statistics (mean, median, min, max, range, population std) and the coefficient of variation \(CV = \sigma / \bar{w} \times 100\%\) across all weighting methods in a Scenario Weight Pack. Agreement labels: HIGH (CV ≤ 10%), MEDIUM (CV ≤ 25%), LOW (CV > 25%). Identifies robust criteria (weight stable regardless of method) vs. contested criteria (weight depends strongly on the method), guiding stakeholder discussion.

3.6 Opportunity Loss Map (Hard Constraint Trade-off)

constraint relaxation potential forfeited trade-off quantification

Theory & Method

Quantifies suitability forfeited by hard constraints by computing, for each constraint-failing cell, the suitability value that would have been assigned if the constraint were removed (the “opportunity loss”). This information supports constraint-revision negotiations: if high-suitability land is blocked by a modifiable constraint, the loss map provides quantitative evidence for policy reconsideration.

Output: loss score (0–100) and loss class (0=no loss, 1=low, 2=medium, 3=high).

3.7 Candidate Site Extractor

connected components BFS labelling polygonisation compactness

Theory & Method

Thresholds suitability, labels contiguous components via BFS (4- or 8-connectivity), filters by minimum area (hectares), ranks by mean suitability (area as tie-break), takes top-N, polygonises via GDAL, and computes compactness (\(4\pi A/P^2\)). The BFS component labelling is implemented directly in NumPy/stdlib for testability. Outputs a vector layer with site_id, rank, cells, area_ha, mean_suit, max_suit, and compactness.

3.8 Priority Zoning from Suitability

tiered planning three-axis fusion action zoning

Theory & Method

Fuses three spatial axes (suitability \(S\), confidence \(C\), opportunity loss \(O\)) into a priority index and a 5-class zone map. The index is \(\text{PI} = 0.60S + 0.30C - 0.20O\) (clipped 0–100). Zones: 4=constraint-revisit (high loss), 3=immediate priority (high S + high C), 2=detailed feasibility (medium S + C), 1=longlist (medium S), 0=low priority. Optional confidence and opportunity-loss rasters default to 75 and 0 respectively.

3.9 Scenario Consensus Zone

agreement count cell-wise consensus planning zones

Theory & Method

Maps per-cell scenario agreement: counts how many scenarios meet medium (\(\ge\) med) and high (\(\ge\) high) suitability thresholds. Zones by agreement ratio: 0=none, 1=≥50% medium, 2=≥75% medium, 3=≥60% high, 4=≥85% near-unanimous high. Input: 2–10 scenario rasters. Threshold defaults: medium=60, high=75.

3.10 Scenario Disagreement Attention

cross-scenario conflict std-based score critical conflict

Theory & Method

Quantifies cross-scenario conflict per cell as \(\text{score} = \min(100, (\sigma_S / \sigma_{\text{ref}}) \times 100)\). Attention classes: 0=low (<20), 1=watch (≥20), 2=review (≥40), 3=high conflict (≥60), 4=critical (≥75 with high mean suitability). The “critical” class flags cells that are both highly contested and highly suitable on average — high-stakes areas needing stakeholder deliberation.

3.11 Decision Quadrant Map

2×2 matrix strategic screening suitability vs. conflict

Theory & Method

Cross-classifies cells by suitability (high/low) and disagreement (high/low): 1=confident opportunity (high S, low D), 2=contested opportunity (high S, high D), 3=confident avoidance (low S, low D), 4=contested avoidance (low S, high D). This is the spatial analogue of a 2×2 strategic management matrix (e.g., BCG growth-share matrix adapted to spatial MCDA).

3.12 Intervention Conflict Matrix

action classification priority × conflict implementation guidance

Theory & Method

Fuses Priority Zoning (0–4) with Disagreement Attention (0–4) into seven intervention classes: 0=Hold, 1=Quick Win (high priority + low conflict), 2=Deliberate Design, 3=Resolve First (high priority + high conflict), 4=Monitor, 5=Co-Design, 6=Re-screen. This directly supports the “action prioritisation” phase of spatial planning workflows.

3.13 Risk-Opportunity Matrix Map

portfolio screening robust suitability × uncertainty investment logic

Theory & Method

Classifies cells by robust suitability (high/low) and uncertainty std (high/low): 1=Invest now (high S, low σ), 2=Validate then invest (high S, high σ), 3=Low strategic value (low S, low σ), 4=Defer and monitor (low S, high σ). This applies a mean-variance decision logic to spatial portfolio selection, conceptually related to Markowitz portfolio theory adapted from finance to spatial planning.

3.14 Regret Fragility Map

spread fragility max-min range scenario sensitivity

Theory & Method

Maps per-cell spread (max − min) across 2–12 scenario rasters. Score = spread (0–100); classes: 0=stable (spread < low), 1=watch (between), 2=fragile (spread ≥ high). Identifies cells whose suitability is scenario-dependent: stable cells are reliable regardless of scenario choice; fragile cells need contingency plans.

3.15 Uncertainty-Adjusted Suitability

mean-variance risk aversion lambda penalty

Theory & Method

Produces risk-adjusted suitability via the simplified mean-variance rule: \(S_{\text{robust}} = \text{clip}(S_{\text{mean}} - \lambda \cdot \sigma_S, \; 0, 100)\). The risk aversion factor \(\lambda\) (0–5) controls the uncertainty penalty: \(\lambda=0\) returns the mean unchanged; \(\lambda=1\) (default) subtracts one standard deviation; higher values increasingly penalise cells with large Monte Carlo spread, favouring proven, low-variance locations.

3.16 Scenario Comparator (Delta)

pairwise comparison delta map B−A

Theory & Method

Computes a pairwise delta map (scenario B minus baseline A) with running summary statistics (valid cell count, min, max, mean delta). Chunk-aware block processing accumulates statistics across the entire raster. The delta can be negative (scenario B rates the cell lower). Use to map spatial patterns of gain/loss between two planning alternatives.

5.1 Global Sensitivity

variance-based criterion ranking sampled importance

Theory & Method

Samples each criterion raster at a regular stride (default every 8 cells), computes spatial standard deviation, multiplies by weight to produce a weighted sensitivity proxy, and normalises to an importance score (0–1, sum = 1). Ranks criteria by contribution to overall suitability variation. Low-importance criteria with small weighted_std may be candidates for model simplification.

5.2 Local Contribution Decomposition

per-cell breakdown contribution share dominance map

Theory & Method

Decomposes the final suitability into per-criterion contribution rasters. For each criterion \(j\), the absolute weighted contribution is \(|X_j \cdot w_j|\) and the percentage share is \((\text{abs\_contribution}_j / \sum_k \text{abs\_contribution}_k) \times 100\). Produces \(2n + 1\) rasters in an output folder: \(n\) absolute contribution rasters, \(n\) percentage rasters, and one denominator raster. A criterion contributing over 50% at a given cell warrants closer scrutiny at that location.

5.3 GeoShapley Lite

Shapley value coalitional game theory spatial attribution marginal contribution

Theoretical Background

GeoShapley applies coalitional game theory to spatial MCDA: each criterion is treated as a “player” in a cooperative game where the “payout” is the suitability score produced by a coalition (subset) of criteria. The Shapley value (Shapley, 1953) computes each criterion's fair marginal contribution to the total suitability, averaged over all possible orders in which criteria could be added. Unlike simpler contribution measures, the Shapley value accounts for interactions: a criterion that is redundant with others (their joint contribution is not the sum of individual contributions) receives a lower value.

Where \(N\) is the full set of \(n\) criteria, \(T\) is a coalition (subset), and \(v(T)\) is the suitability score produced using only criteria in \(T\). The algorithm samples a regular grid of cells and computes the Shapley value per criterion from the weighted linear combination with the subset's weights re-normalised. Results are summarised as a table with criterion-level mean and median Shapley values.

Parameters

Interpretation Guide

• High Shapley value: Criterion uniquely contributes to explaining or driving the suitability pattern that no other criterion replicates.
• Low or zero Shapley value: Criterion is redundant (other criteria cover its information) or contributes negligibly.
• Compare with Global Sensitivity: A criterion may have high spatial variation (high importance in Global Sensitivity) but low Shapley value if that variation is perfectly correlated with another criterion. Shapley values penalise redundancy.

References

6.1 Suitability HTML Report

composite report HTML output decision summary

Assembles a self-contained HTML document integrating project metadata, weight methodology description, QA summary, class area statistics, and reference paths to all intermediate diagnostic tables (similarity, dominance, benchmark, robustness, portfolio diversity, minimax regret, alignment audit, redundancy, constraint impact, transition matrix, fragmentation, consensus/attention/quadrant/intervention/risk/regret area stats, threshold curve, KPI, policy gate, scorecard, scenario selection, governance audit, wave plan, deployment gate, policy frontier, budget phasing, action playbook, and release narrative). Designed as a single-file decision briefing document.

6.2 Numerical QA Check

quality assurance value range validation outlier detection

Performs numerical quality checks on a suitability raster against Monte Carlo percentile rasters: reports min, max, mean, std, count of valid cells, count outside 0–100, negative values, non-finite values, and the percentage of cells where the base suitability falls outside the P05–P95 envelope (potential outliers or constraint-affected cells). Outputs a plain-text QA summary string.

6.3 Beta Evidence Report

audit trail artifact inventory evidence pack

Generates an HTML evidence pack cataloguing all artifacts in an output directory, with pipeline summary text and embedded references to QA results, class statistics, weight audits, and all diagnostic tables. Designed as an auditable record of the complete suitability modelling workflow for regulatory or peer-review submission.

6.4 Scenario Markdown Brief

executive summary Markdown output

Produces a Markdown-formatted executive brief comparing a baseline scenario against one or more alternative scenarios, with embedded QA results, weight audit paths, scenario similarity and dominance tables, method benchmark leaderboard, robustness rankings, portfolio diversity, minimax regret, and governance audit references. Ready for direct inclusion in version-controlled documentation.

6.5 Suitability Class Area Stats

area tabulation class breaks zonal statistics

Classifies a suitability raster by user-specified break values (comma-separated, e.g. "20,40,60,80") and tabulates cell count, area (in raster units and hectares), and percentage of valid area per class. Block-processed for memory efficiency. Outputs a GeoPackage table plus a summary text string.

6.6 Scenario Similarity Matrix

cross-scenario correlation Pearson r cosine similarity

Computes the pairwise Pearson correlation coefficient and cosine similarity between 2–12 scenario rasters over all valid (non-nodata, mask-passing) cells. Outputs a GeoPackage matrix table with scenario pairs, Pearson r, cosine similarity, and the effective sample size. Block-processed for large rasters. High similarity (ρ > 0.95) indicates the scenarios are functionally identical; low similarity flags genuine differences in spatial prioritisation.

6.7 Scenario Dominance Matrix

pairwise dominance ε-dominance strict pareto

Computes a pairwise dominance matrix between scenarios: scenario A dominates B if A's suitability values are greater than or equal to B's for all valid cells, with strict inequality for at least an ε-fraction of cells (default ε = 0.5% of valid cells). Reports the dominance relation, the fraction of cells where A > B, A = B, A < B, and the mean difference. A scenario that dominates all others is the unambiguously preferred choice.

6.8 Scenario Robustness Leaderboard

composite ranking multi-criterion synthesis

Synthesises the benchmark, dominance, minimax regret, and portfolio diversity tables into a weighted robustness score per scenario. Ranks scenarios from most robust to least. The composite ranking provides a single defensible ordering for decision-makers overwhelmed by multiple diagnostic metrics. Output: GeoPackage with scenario, benchmark_score, dominance_wins, minimax_regret, diversity, robustness_score, and rank.

6.9 Scenario Governance Audit

decision documentation governance trail multi-perspective audit

Audits the decision governance process by combining robustness ranking, benchmark performance, minimax regret, scenario selection recommendation, and deployment readiness status into a unified governance table. Documents which scenarios were evaluated, how they ranked, which was selected, and whether deployment conditions are satisfied.

6.10 Method Benchmark Leaderboard

multi-metric ranking high-suitability coverage mean/std ratio

Benchmarks 2–12 scenario rasters on five metrics: mean suitability, std, high-suitability area fraction (cells ≥ threshold, default 70), coefficient of variation (CV%), and a composite benchmark score. The composite score rewards high mean, high coverage, low CV, and moderate discrimination. Output: GeoPackage leaderboard with rank.

6.11 Scenario Portfolio Diversity

Shannon diversity scenario distinctiveness

Computes per-cell normalised suitability shares across scenarios, then calculates the Shannon entropy (diversity) of these shares. A cell where all scenarios agree has low diversity; a cell where scenarios strongly disagree has high diversity. Reports mean, median, and max diversity across the study area, plus the scenario-level marginal contribution to total diversity. Output: GeoPackage with scenario-level diversity statistics.

6.12 Minimax Regret Selector

Savage criterion robust decision rule worst-case optimisation

Applies the minimax regret decision criterion (Savage, 1951): for each cell, compute the regret of choosing scenario \(k\) as the difference between the best achievable suitability (across all scenarios) and scenario \(k\)'s value. The scenario that minimises the maximum regret across the study area is the minimax-optimal choice. Reports per-scenario max regret, mean regret, and a minimax rank. This is the most conservative scenario selection rule: it protects against the worst outcome.

6.13 Policy Sensitivity Frontier

lambda parameter sweep risk-return trade-off efficient frontier

Sweeps the risk aversion parameter \(\lambda\) across a user-specified list (e.g., "0,0.5,1,1.5,2") and computes for each \(\lambda\) value the mean and standard deviation of the resulting risk-adjusted suitability raster. Produces a GeoPackage table tracing the mean-variance trade-off curve (the “policy frontier”). This enables decision-makers to visualise the cost (in expected suitability) of increasing risk aversion, analogous to an efficient frontier in portfolio theory.

6.14 Action Playbook Generator

Markdown playbook zone descriptions implementation guidance

Generates a structured Markdown action playbook from priority zoning, similarity matrix, and policy frontier data. Includes zone-by-zone descriptions, recommended actions per priority tier, scenario comparison insights, and implementation notes. Designed as a ready-to-share decision document for stakeholder review.

6.15 Intervention Budget Phasing

cost-weighted phasing cumulative coverage budget planning

Computes cumulative cost and area coverage as priority zones are implemented in descending priority-index order, with user-specified per-zone unit costs (comma-separated). Produces a stepwise budget phasing table (GeoPackage) showing at each percentile step the cumulative area, cumulative cost, and remaining budget. Steps are defined by the user (default every 10% of area). Supports phased implementation planning with budget constraints.

6.16 Release Narrative Builder

versioned changelog release documentation

Generates a Markdown release narrative for a software version, embedding project name, version, key highlights, policy frontier data, budget phasing, and scenario similarity insights. Described as a changelog/documentation hybrid for plugin release notes.

6.17 Criterion Redundancy Diagnostics

inter-criterion correlation VIF redundancy detection

Samples criterion rasters (up to 120,000 cells, block-processed), computes the Pearson correlation matrix between all pairs, and produces two output tables: pairs (criterion A, criterion B, Pearson r, rsquared, n_samples) and criteria (criterion, mean_abs_correlation, max_abs_correlation, VIF). Criteria with VIF > 10 or pairwise |r| > 0.90 are flagged as potentially redundant. Reference: Dormann, C.F. et al. (2013). Collinearity: a review of methods. Ecography, 36(1), 27–46. DOI: 10.1111/j.1600-0587.2012.07348.x

6.18 Constraint Impact Audit

exclusion quantification constraint tightening impact analysis

Computes the fraction of valid area excluded by one or more binary constraint masks, the mean suitability of excluded vs. retained cells, and the percentage of high-suitability cells (above threshold) lost to constraints. Outputs a GeoPackage table with per-constraint and combined impact metrics. Answers: “How much high-potential land are we excluding?”

6.19 Alignment Audit

grid alignment check CRS/extent/resolution verification

Checks that all criterion rasters share identical CRS, extent, and cell size relative to a snap raster. Reports per-raster alignment status, pixel offsets, and resolution mismatches. A pre-flight check ensuring the Data Harmonizer step was successful.

6.20 Priority Transition Matrix

zone-to-zone transitions cross-tabulation refinement impact

Cross-tabulates two priority zone rasters (typically baseline vs. refined) to quantify how many cells transition between priority classes. Produces a transition matrix (FROM zone, TO zone, cell count, area) showing which zones were promoted, demoted, or remained stable after refinement.

6.21 Spatial Fragmentation Audit

patch metrics landscape ecology contiguity analysis

Thresholds a suitability raster (default ≥70), labels connected components (4- or 8-connectivity), and computes patch-level landscape metrics: number of patches, mean patch area, largest patch index, edge density, and a fragmentation index. High fragmentation (many small patches) may indicate that suitable land is scattered rather than clustered, affecting implementation feasibility.

6.22 Consensus Zone Area Stats

zone tabulation consensus area

Tabulates the area (cell count, map units, hectares) and percentage of each consensus zone class (0–4) from a Scenario Consensus Zone raster. Complements the map with numerical diagnostics.

6.23 Threshold Coverage Curve

cumulative coverage sensitivity to threshold trade-off curve

Computes the cumulative fraction of valid area exceeding each suitability threshold from 0 to 100 at a user-specified step (default 5). Produces a GeoPackage table tracing the coverage curve: threshold, cells_above, area_above, pct_above. This is the spatial equivalent of a receiver operating characteristic (ROC)-like threshold sensitivity curve.

6.24 Attention Class Area Stats

class area tabulation

Tabulates area per attention class (0–4) from a Scenario Disagreement Attention raster. Provides numerical backing for the attention map: how much area falls in each conflict category.

6.25 Executive KPI Snapshot

dashboard metrics summary statistics decision KPIs

Computes a one-page executive KPI table from suitability, robust suitability, constraint mask, consensus zone, and attention class rasters. Metrics include: mean suitability, high-suitability area %, robust area %, constrained area %, consensus level, conflict area %, and a composite decision confidence score. Designed for dashboard integration or slide-ready reporting.

6.26 Policy Gate Recommender

gate criteria policy compliance threshold recommendation

Recommends a suitability threshold (policy gate) based on the threshold coverage curve and policy sensitivity frontier, subject to user-specified constraints: target coverage percentage (e.g., 25% of area) and maximum acceptable critical conflict percentage (e.g., 10%). Returns the recommended gate value, the coverage it achieves, and whether it passes/fails the conflict constraint.

6.27 Decision Quadrant Area Stats

quadrant tabulation

Tabulates area per decision quadrant class (1–4) from a Decision Quadrant Map raster.

6.28 Intervention Conflict Area Stats

intervention class tabulation

Tabulates area per intervention class (0–6) from an Intervention Conflict Matrix raster.

6.29 Risk Opportunity Area Stats

risk-opportunity tabulation

Tabulates area per risk-opportunity class (1–4) from a Risk-Opportunity Matrix Map raster.

6.30 Regret Fragility Area Stats

fragility class tabulation

Tabulates area per fragility class (0–2) from a Regret Fragility Map raster.

6.31 Scenario Selection Recommender

multi-criterion recommendation consensus rule

Synthesises the benchmark leaderboard, portfolio diversity, minimax regret, policy gate, actionability scorecard, and robustness leaderboard into a weighted recommendation table. Each scenario receives a composite selection score and rank. The top-ranked scenario is the one that best balances performance, robustness, minimax protection, policy compliance, and actionability.

6.32 Implementation Wave Planner

phased rollout risk-sequenced implementation

Plans phased implementation by sequencing priority zones within each risk-opportunity class. High-priority + low-risk zones are scheduled in early waves; low-priority + high-risk zones are deferred. Produces a GeoPackage table with wave number, priority zone, risk class, cell count, area, and cumulative coverage. Supports staged project delivery.

6.33 Deployment Readiness Gate

go/no-go decision readiness checklist

Evaluates whether conditions are met for deployment: synthesises the actionability scorecard, scenario selection recommendation, minimax regret, wave plan, and regret fragility statistics into a binary readiness gate with supporting evidence. Returns a GeoPackage with gate criteria, thresholds, actual values, and pass/fail status per criterion.

6.34 Actionability Scorecard

implementation feasibility scorecard metrics

Computes an actionability scorecard from the executive KPI snapshot, budget phasing table, and policy gate recommendation. Metrics include: spatial concentration (are top zones clustered?), fiscal feasibility (budget coverage), policy alignment (gate status), and an overall actionability index (0–100). Designed as the bridge between analytical suitability and implementation planning.