Extended Formula Derivation

The Conflict Risk Index is a normalized concentration metric derived from descriptive statistics. Beginning from a vector of conflict counts X = (x₁, x₂, …, xₙ), the index is:

CRI = x̄ / x_max = (1/n)·Σx_i / max(x)

Equivalent forms include:

• Severity-weighted CRI: CRI_w = Σ(w_i·x_i) / (n · max(w·x)), where w_i is a severity weight (e.g., 1 = crop, 3 = livestock, 5 = human injury).
• Effort-standardised CRI: CRI_e = (Σx_i/E_i) / (n · max(x/E)), where E_i is the sampling effort for unit i.
• Pareto-aware concentration: The 80/20 metric P₈₀ = (k/n) where k is the smallest count of top units whose cumulative incidents reach 80% of Σx.

Assumptions

• Sampling effort is standardised across units (or has been corrected via the effort-standardised form).
• Detection probability is comparable across units — under-reporting bias is not a confounder.
• The maximum value x_max is not a measurement artifact (e.g., one unit double-counted incidents).
• Non-negative integer counts (incidents cannot be fractional).

Limitations

• Sample-size sensitivity — CRI from n < 5 is unstable; report bootstrap CIs for n < 15.
• Outlier sensitivity — a single anomalously high x_max drives CRI down without changing the underlying conflict landscape; always inspect the ranked count distribution before interpretation.
• Temporal aggregation — annual CRI hides seasonal pulses; report seasonally where data permit.
• Spatial-scale dependence — CRI computed at the village scale and at the grid-cell scale will differ; always state the unit definition.
• Causality — CRI describes the burden but not the cause; pair with covariate models (logistic regression, MaxEnt, GLM) to identify drivers.

Decision Checklist

• ✓ You have conflict-incident counts from a defined set of sampling units (villages, grid cells, ranches, months).
• ✓ You want to identify hotspots for prioritised management intervention.
• ✓ Your sampling effort is standardised, or you can correct for effort variation.
• ✓ You need a publication-ready, dimensionless metric for an ecology, conservation, or wildlife management journal.
• ✗ Do NOT use if your sampling effort differs greatly between units and you cannot adjust — first standardise to incidents-per-unit-effort.
• ✗ Do NOT use if you have only presence/absence data — use occupancy models instead.
• ✗ Do NOT use as the sole basis for management — pair with wildlife abundance, livelihood-cost, and stakeholder-perception data.

Real-World Examples

1. Wolf–Livestock Conflict, Greater Yellowstone (USA) — Compensation records from 12 ranches in Montana's Paradise Valley, used to identify the 2–3 ranches driving most of the cattle losses for targeted livestock-guarding-dog deployment.
2. Black Bear Property Damage, Florida (USA) — Bear damage incidents across 12 counties, used to prioritise bear-resistant trash-can grants in the highest-CRI counties.
3. Mountain Lion–Pet Conflict, California (USA) — Predation incidents across 12 foothill zones, informing the design of public-awareness campaigns and livestock-shelter requirements.
4. Elk–Crop Conflict, Colorado Rocky Mountains (USA) — Hay and alfalfa damage across 10 farms near Estes Park, used to prioritise game-fencing assistance.
5. Coyote–Sheep Conflict, West Texas (USA) — Sheep depredation across 10 ranches, identifying the small subset of ranches accounting for 80% of total losses.

Sampling Design Guidance

• Minimum recommended n: 5 sampling units (CRI is interpretable but unstable; report bootstrap CIs).
• Reliable hotspot identification: 15–30 units.
• Robust spatial pattern testing: 50+ units (allows Moran's I, Getis–Ord G*).
• Time period: minimum 12 months to capture seasonal pulses; 2–3 years recommended for trend testing.
• Effort standardisation: if effort varies, report incidents-per-effort (e.g., per livestock-night, per ha-month).

Related Metrics — Decision Tree

Need a single number for conflict burden?  → Conflict Risk Index (CRI)
  → Want to weight severity differently? → CRI weighted form
  → Sampling effort varies among units?   → CRI effort-standardised form
Need to identify hotspots only?           → Getis–Ord G* / Moran's I
Need to predict where conflict will occur?→ Logistic regression / MaxEnt
Need to compare conflict between species? → Encounter Rate (per km) / RAI
Need to link conflict to wildlife density?→ Camera Trap RAI + CRI overlay