ecological-agent-skills 3.1.0

package/skills/ecological-data-foundation/scripts/download_from_obis.py

@@ -0,0 +1,318 @@
+# ecological-agent-skills / Copyright (C) 2026 Francisco Diego Barros Barata
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Download marine occurrence records from OBIS (Ocean Biodiversity Information System).
+
+Usage: python download_from_obis.py <species_name_or_list_csv> <output_dir> [year_from] [year_to] [wkt_geometry]
+
+Arguments:
+    species_name_or_list_csv : Species name (e.g., "Chelonia mydas") or path to CSV
+                               with column 'scientificName'
+    output_dir               : Directory for outputs (created if absent)
+    year_from                : Minimum year of observation (default: 1950)
+    year_to                  : Maximum year of observation (default: current year)
+    wkt_geometry             : WKT polygon to restrict query (optional)
+
+Outputs (per species):
+    occurrences_raw_OBIS_{species}_{date}.csv  — standardised occurrence records
+    download_metadata_OBIS_{species}.txt        — download provenance and citation
+
+Standard output schema:
+    species, decimalLatitude, decimalLongitude, eventDate, countryCode,
+    basisOfRecord, coordinateUncertaintyInMeters, datasetName, occurrenceID,
+    source, download_doi
+Extra OBIS columns:
+    depth, marine
+"""
+
+import logging
+import sys
+from datetime import datetime
+from pathlib import Path
+
+SKILL_NAME = "ecological-data-foundation"
+_LOG_DIR   = Path("logs")
+_LOG_DIR.mkdir(parents=True, exist_ok=True)
+_log_file  = _LOG_DIR / f"skill_{SKILL_NAME}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.log"
+logging.basicConfig(
+    level=logging.INFO,
+    format="[%(asctime)s] [%(levelname)s] [" + SKILL_NAME + "] %(message)s",
+    datefmt="%Y-%m-%d %H:%M:%S",
+    handlers=[
+        logging.StreamHandler(sys.stdout),
+        logging.FileHandler(_log_file, encoding="utf-8"),
+    ],
+)
+logger = logging.getLogger(SKILL_NAME)
+
+def log_step(n: int, desc: str) -> None:
+    logger.info("-- STEP %d: %s", n, desc)
+
+def log_decision(var: str, val, why: str) -> None:
+    logger.info("DECISION | %s = %s | %s", var, val, why)
+
+
+from datetime import date
+
+try:
+    import requests
+    import pandas as pd
+except ImportError as e:
+    logger.error(
+        "Dependencia ausente: %s\n  Instale com: pip install requests pandas\n  Skill anterior: ecological-data-foundation",
+        e,
+    )
+    sys.exit(1)
+
+
+# ── Constants ─────────────────────────────────────────────────────────────────
+OBIS_API_BASE = "https://api.obis.org/v3"
+PAGE_SIZE     = 5000  # OBIS max per request
+BAD_FLAGS     = {"NO_COORD", "ZERO_COORD", "ON_LAND", "DEPTH_EXCEEDS_BATH"}
+
+
+# ── Helper functions ──────────────────────────────────────────────────────────
+
+def fetch_obis(scientificname: str, year_from: int, year_to: int,
+               wkt_geometry: str | None) -> list[dict]:
+    """Fetch all OBIS records for a species, handling pagination."""
+    url    = f"{OBIS_API_BASE}/occurrence"
+    offset = 0
+    all_results: list[dict] = []
+
+    while True:
+        params: dict = {
+            "scientificname": scientificname,
+            "absence":        "exclude",
+            "startdate":      f"{year_from}-01-01",
+            "enddate":        f"{year_to}-12-31",
+            "size":           PAGE_SIZE,
+            "offset":         offset,
+        }
+        if wkt_geometry:
+            params["geometry"] = wkt_geometry
+
+        try:
+            resp = requests.get(url, params=params, timeout=60)
+            resp.raise_for_status()
+            data = resp.json()
+        except requests.RequestException as e:
+            logger.error(
+                "Falha na requisicao OBIS (offset=%d, especie='%s'): %s\n  Causa provavel: sem conexao com a internet ou API OBIS indisponivel.\n  Verifique: https://api.obis.org/\n  Skill anterior: ecological-data-foundation",
+                offset, scientificname, e,
+            )
+            raise
+
+        results = data.get("results", [])
+        if not results:
+            break
+
+        all_results.extend(results)
+        logger.info("Pagina offset=%d: %d registros recuperados (total: %d)",
+                    offset, len(results), len(all_results))
+
+        total = data.get("total", 0)
+        if len(all_results) >= total:
+            break
+        if len(results) < PAGE_SIZE:
+            break
+        offset += PAGE_SIZE
+
+    return all_results
+
+
+def apply_quality_flags(records: list[dict]) -> list[dict]:
+    """Remove records with known OBIS quality issues."""
+    filtered = []
+    n_flagged = 0
+    for rec in records:
+        flags_raw = rec.get("flags", "") or ""
+        flags = set(str(flags_raw).upper().split(","))
+        if flags & BAD_FLAGS:
+            n_flagged += 1
+            continue
+        filtered.append(rec)
+    if n_flagged > 0:
+        logger.warning("%d registros removidos por flags OBIS de qualidade.", n_flagged)
+    return filtered
+
+
+def standardise_records(records: list[dict], species_name: str) -> "pd.DataFrame":
+    """Convert OBIS record dicts to the standard occurrence schema."""
+    rows = []
+    for rec in records:
+        rows.append({
+            "species":                       species_name,
+            "decimalLatitude":               rec.get("decimalLatitude"),
+            "decimalLongitude":              rec.get("decimalLongitude"),
+            "eventDate":                     rec.get("eventDate") or rec.get("date_start", ""),
+            "countryCode":                   rec.get("countryCode", ""),
+            "basisOfRecord":                 rec.get("basisOfRecord", "OCCURRENCE"),
+            "coordinateUncertaintyInMeters": rec.get("coordinateUncertaintyInMeters"),
+            "datasetName":                   rec.get("datasetName", ""),
+            "occurrenceID":                  str(rec.get("occurrenceID", rec.get("id", ""))),
+            "source":                        "OBIS",
+            "download_doi":                  None,
+            "depth":                         rec.get("depth"),
+            "marine":                        True,
+        })
+
+    df = pd.DataFrame(rows)
+    df["decimalLatitude"]  = pd.to_numeric(df["decimalLatitude"],  errors="coerce")
+    df["decimalLongitude"] = pd.to_numeric(df["decimalLongitude"], errors="coerce")
+    df = df.dropna(subset=["decimalLatitude", "decimalLongitude"])
+    return df.reset_index(drop=True)
+
+
+def save_metadata(output_dir: Path, species_name: str, n_records: int,
+                  year_from: int, year_to: int, wkt_geometry: str | None) -> None:
+    safe_name = species_name.replace(" ", "_")
+    today = date.today().isoformat()
+    year  = date.today().year
+    lines = [
+        f"Species: {species_name}",
+        f"Source: Ocean Biodiversity Information System (OBIS) — https://obis.org",
+        f"API endpoint: {OBIS_API_BASE}/occurrence",
+        f"Absence records excluded: True",
+        f"OBIS quality flags applied: True (NO_COORD, ZERO_COORD, ON_LAND, DEPTH_EXCEEDS_BATH removed)",
+        f"Year range: {year_from} - {year_to}",
+        f"WKT geometry: {wkt_geometry or 'none (global)'}",
+        f"n_records: {n_records}",
+        f"Download date: {today}",
+        (f"Citation: OBIS ({year}) Ocean Biodiversity Information System. "
+         "Intergovernmental Oceanographic Commission of UNESCO. "
+         f"www.obis.org. Accessed on {today}."),
+        "License: CC0 1.0 (https://creativecommons.org/publicdomain/zero/1.0/)",
+    ]
+    meta_path = output_dir / f"download_metadata_OBIS_{safe_name}.txt"
+    try:
+        meta_path.write_text("\n".join(lines), encoding="utf-8")
+        logger.info("Metadados gravados: %s", meta_path)
+    except OSError as e:
+        logger.error(
+            "Falha ao gravar metadados: %s\n  Skill anterior: ecological-data-foundation", e,
+        )
+        raise
+
+
+# ── Main download logic ────────────────────────────────────────────────────────
+
+def download_species(species_name: str, output_dir: Path,
+                     year_from: int, year_to: int,
+                     wkt_geometry: str | None) -> None:
+    logger.info("--- Iniciando download OBIS: %s ---", species_name)
+    today_str = date.today().strftime("%Y%m%d")
+    safe_name = species_name.replace(" ", "_")
+
+    log_step(1, f"Buscar registros OBIS para '{species_name}'")
+    records = fetch_obis(species_name, year_from, year_to, wkt_geometry)
+    logger.info("Registros brutos recuperados: %d", len(records))
+
+    if not records:
+        logger.warning("Nenhum registro OBIS encontrado para '%s'.", species_name)
+        return
+
+    log_step(2, "Aplicar filtros de qualidade OBIS")
+    records = apply_quality_flags(records)
+    logger.info("Registros apos filtros OBIS: %d", len(records))
+
+    log_step(3, "Padronizar registros para schema de saida")
+    df = standardise_records(records, species_name)
+    n_final = len(df)
+    logger.info("Registros com coordenadas validas: %d", n_final)
+
+    if n_final < 30:
+        logger.warning(
+            "Registros insuficientes para analise confiavel (n = %d). Considere relaxar filtros.",
+            n_final,
+        )
+
+    log_step(4, "Gravar CSV de ocorrencias")
+    csv_path = output_dir / f"occurrences_raw_OBIS_{safe_name}_{today_str}.csv"
+    try:
+        df.to_csv(csv_path, index=False)
+        logger.info("Gravado: %s (%d registros)", csv_path, n_final)
+    except OSError as e:
+        logger.error(
+            "Falha ao gravar CSV: %s\n  Skill anterior: ecological-data-foundation", e,
+        )
+        raise
+
+    log_step(5, "Gravar metadados do download")
+    save_metadata(output_dir, species_name, n_final, year_from, year_to, wkt_geometry)
+
+
+# ── Entry point ────────────────────────────────────────────────────────────────
+
+def main():
+    logger.info("Script: download_from_obis.py | Skill: %s", SKILL_NAME)
+
+    argv = sys.argv[1:]
+
+    if len(argv) < 2:
+        species_input = "Chelonia mydas"
+        output_dir    = Path("output/obis")
+        year_from     = 1950
+        year_to       = date.today().year
+        wkt_geometry  = None
+        logger.warning("Menos de 2 argumentos fornecidos. Usando valores padrao para teste.")
+    else:
+        species_input = argv[0]
+        output_dir    = Path(argv[1])
+        year_from     = int(argv[2]) if len(argv) >= 3 else 1950
+        year_to       = int(argv[3]) if len(argv) >= 4 else date.today().year
+        wkt_geometry  = argv[4] if len(argv) >= 5 and argv[4] else None
+
+    logger.info("Species input  : %s", species_input)
+    logger.info("Output dir     : %s", output_dir)
+    logger.info("Year range     : %d - %d", year_from, year_to)
+    logger.info("WKT geometry   : %s", wkt_geometry or "nenhum (global)")
+
+    log_decision("absence", "exclude",
+                 "apenas registros de presenca confirmada")
+    log_decision("quality_flags", str(BAD_FLAGS),
+                 "registros com flags de qualidade OBIS sao removidos")
+
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    # Build species list
+    log_step(0, "Construir lista de especies")
+    if species_input.endswith(".csv") and Path(species_input).exists():
+        try:
+            df_sp = pd.read_csv(species_input)
+            if "scientificName" not in df_sp.columns:
+                logger.error(
+                    "Coluna 'scientificName' nao encontrada em: %s\n  Skill anterior: ecological-data-foundation",
+                    species_input,
+                )
+                sys.exit(1)
+            species_list = df_sp["scientificName"].dropna().unique().tolist()
+            logger.info("Modo batch: %d especies carregadas", len(species_list))
+        except Exception as e:
+            logger.error(
+                "Falha ao ler lista de especies: %s\n  Skill anterior: ecological-data-foundation", e,
+            )
+            sys.exit(1)
+    else:
+        species_list = [species_input.strip()]
+        logger.info("Modo especie unica: %s", species_list[0])
+
+    for sp in species_list:
+        try:
+            download_species(sp, output_dir, year_from, year_to, wkt_geometry)
+        except FileNotFoundError as e:
+            logger.error(
+                "Arquivo de entrada nao encontrado ao processar '%s': %s\n  Skill anterior: ecological-data-foundation",
+                sp, e,
+            )
+        except Exception as e:
+            logger.error(
+                "Falha ao baixar '%s' do OBIS: %s\n  Causa provavel: problema de rede ou especie nao encontrada.\n  Skill anterior: ecological-data-foundation",
+                sp, e,
+            )
+
+    logger.info("Todos os downloads OBIS concluidos. Verifique: %s", output_dir)
+
+
+if __name__ == "__main__":
+    main()

package/skills/ecological-impact-assessment/SKILL.md

@@ -0,0 +1,123 @@
+---
+name: ecological-impact-assessment
+description: "Quantifies ecological impacts using BACI designs, landscape fragmentation metrics, and pressure indices. Use this skill when the user mentions BACI analysis, before-after-control-impact, impact assessment, disturbance effects, land use change impacts, fragmentation metrics, landscape metrics, pressure or threat indices, intervention effectiveness, management outcome evaluation, or control-impact comparisons."
+skill_version: 1.0.0
+---
+
+# Skill: ecological-impact-assessment
+
+**Domain:** BACI · Before/After · Fragmentation · Pressure · Impact synthesis  
+**Phase:** 2 — Modeling  
+**Used by:** assess-ecological-impact, build-fire-risk-map, analyze-environmental-change
+
+---
+
+## Purpose
+
+Guides the agent through the design and analysis of ecological impact assessments: BACI frameworks, before/after comparisons, landscape fragmentation analysis, anthropogenic pressure quantification, and synthesis of impact magnitude.
+
+---
+
+## When to Invoke
+
+- Quantifying the effect of a disturbance, land-use change, or management intervention
+- Designing a BACI study or interpreting BACI data
+- Computing landscape fragmentation or connectivity metrics
+- Building a composite pressure or threat index
+
+---
+
+## Inputs
+
+| Input | Format | Required |
+|-------|--------|----------|
+| Ecological indicator time series (control and impact sites) | CSV | Yes |
+| Disturbance event date(s) | Text / date | Yes |
+| Land cover maps (pre/post) | GeoTIFF | Conditional |
+| Pressure / threat layers | GeoTIFF, SHP | Optional |
+
+---
+
+## Outputs
+
+| Output | Description |
+|--------|-------------|
+| `baci_results.csv` | BACI interaction term with CI and p-value |
+| `fragmentation_metrics.csv` | Patch area, perimeter, connectivity indices |
+| `pressure_index.tif` | Composite pressure map |
+| `impact_synthesis.md` | Narrative summary of impact magnitude |
+| `impact_plots/` | Before/after and control/impact comparison plots |
+
+---
+
+## Steps
+
+### 1. Define the Impact Hypothesis
+- Identify the disturbance and its expected effect (direction, magnitude, timing)
+- Define the ecological indicator(s) to measure
+- Define control sites (unaffected by disturbance, similar baseline conditions)
+
+### 2. BACI Design Assessment
+- Verify: measurements Before AND After disturbance, at Control AND Impact sites
+- Compute the BACI interaction: (After−Before)_Impact − (After−Before)_Control
+- If multiple control or impact sites: use mixed model with site as random effect
+- Test for parallel trends in the pre-disturbance period (assumption check)
+
+### 3. Statistical Analysis
+- Linear model: `indicator ~ period * treatment + (1|site)`
+- The interaction term (`period:treatment`) is the BACI effect
+- Report coefficient, 95% CI, and p-value for the interaction
+- Effect size: Cohen's d or partial η²
+
+### 4. Landscape Fragmentation (if applicable)
+- Compute patch-level metrics: area, perimeter/area ratio, shape index
+- Compute landscape-level metrics: largest patch index, edge density, COHESION, MESH
+- Compare pre/post land cover using transition matrix
+- Report habitat loss and fragmentation separately
+
+### 5. Pressure Index (if applicable)
+- Identify pressure layers: road density, agricultural intensity, human footprint, fire frequency
+- Normalise each layer (0–1)
+- Compute weighted sum (document weights and rationale)
+- Classify pressure zones (low / moderate / high / critical)
+
+### 6. Synthesis
+- Combine BACI results, fragmentation metrics, and pressure index
+- Classify overall impact as: none / minor / moderate / major / critical
+- Describe spatial distribution of impact
+- Identify most impacted areas and most important drivers
+
+---
+
+## Key Decisions to Document
+
+- Control site selection criteria
+- BACI model specification (fixed vs. random effects)
+- Fragmentation metrics chosen and software version
+- Pressure layer sources and weighting scheme
+- Impact classification thresholds
+
+---
+
+## Tools and Libraries
+
+**R:** `lme4`, `emmeans`, `landscapemetrics`, `lsm`, `ggplot2`  
+**Python:** `pylandstats`, `statsmodels`, `rasterio`  
+**CLI:** FRAGSTATS (landscape metrics)
+
+---
+
+## Resources
+
+- `resources/baci-design-guide.md` — BACI design requirements and common pitfalls
+- `resources/fragmentation-metrics-reference.md` — FRAGSTATS metric definitions
+- `resources/pressure-index-template.md` — standard pressure layer sources
+- `examples/impact/` — worked BACI and fragmentation example
+
+---
+
+## Notes
+
+- BACI requires pre-disturbance data; post-hoc designs without pre-disturbance are not BACI
+- Multiple control sites are strongly preferred; single control site severely limits inference
+- Landscape metrics are scale-dependent; always report the spatial resolution used

package/skills/ecological-impact-assessment/examples/example-prompts.md

@@ -0,0 +1,32 @@
+# Example Invocation Prompts — ecological-impact-assessment
+
+## BACI Analysis
+
+```
+Load skill: ecological-impact-assessment
+Task: BACI analysis of deforestation impact on stream macroinvertebrate richness.
+
+Data: data/baci_invertebrates.csv
+Columns: site_id, treatment (control/impact), period (before/after), richness (count), date
+
+Model: richness ~ period * treatment + (1|site), family = negative binomial
+Check parallel trends using pre-disturbance data only.
+Report: BACI interaction estimate, 95% CI, p-value, % change on original scale.
+Generate: before/after comparison plots for control and impact sites.
+```
+
+## Fragmentation Analysis
+
+```
+Load skill: ecological-impact-assessment
+Task: Compute landscape fragmentation metrics for Atlantic Forest before (2000) and after (2022) period.
+
+Input layers:
+  - data/landcover_2000.tif (MapBiomas class 3 = Atlantic Forest)
+  - data/landcover_2022.tif (same classification)
+  - data/spatial/study_area.shp
+
+Metrics: total area (ha), number of patches, mean patch size, largest patch index,
+  edge density, COHESION, MESH (effective mesh size).
+Report change between years. Output: fragmentation_metrics.csv, fragmentation_plots.png
+```

package/skills/ecological-impact-assessment/resources/baci-design-guide.md

@@ -0,0 +1,55 @@
+# BACI Design Requirements and Common Pitfalls
+
+## Core Requirements
+
+BACI (Before-After-Control-Impact) is the gold standard for ecological impact assessment.
+
+### Minimum design:
+- **Before:** ≥ 1 sampling period before the disturbance
+- **After:** ≥ 1 sampling period after the disturbance
+- **Control:** ≥ 1 site not affected by the disturbance, similar to impact site
+- **Impact:** ≥ 1 site affected by the disturbance
+
+### Recommended design (for inference):
+- Multiple control sites (≥ 3 preferred)
+- Multiple impact sites (≥ 3 preferred)
+- ≥ 3 time points before and after
+- Sites monitored for the same duration before and after
+
+## Statistical Model
+
+```
+indicator ~ period * treatment + (1 | site)
+```
+
+- `period`: factor with levels `before` / `after`
+- `treatment`: factor with levels `control` / `impact`
+- `period:treatment`: **the BACI interaction — the effect of interest**
+- `(1 | site)`: random intercept for repeated measures at sites
+
+The BACI effect = difference in before-after change between impact and control sites.
+
+## Assumptions
+
+1. **Parallel trends:** Control and impact sites had similar trajectories *before* the disturbance. **Test this** by checking the significance of `time × treatment` interaction in the pre-disturbance period only.
+2. **No spillover:** The disturbance does not affect control sites (e.g., no upstream pollution affecting downstream controls).
+3. **Independence among sites** (partially relaxed by random effects).
+
+## Common Pitfalls
+
+| Pitfall | Consequence | Fix |
+|---------|------------|-----|
+| Only one control site | Cannot distinguish site effect from treatment effect | Use ≥ 3 control sites |
+| No pre-disturbance data | Cannot implement BACI; only Before-After | Document limitation explicitly |
+| Control site affected by disturbance | BACI estimate biased toward zero | Verify spatial separation |
+| Very short pre-disturbance baseline | Parallel trend assumption unverifiable | Report as limitation |
+| Pseudoreplication (subsamples as sites) | Inflated degrees of freedom, false significance | Use random effects or aggregate |
+| No random effects for repeated sites | Underestimated standard errors | Always include `(1|site)` |
+
+## Effect Size Reporting
+
+For the BACI interaction (β_BACI):
+
+- **Linear scale (Gaussian):** Report β, 95% CI, and Cohen's d
+- **Log scale (Poisson/lognormal):** Back-transform: exp(β) = multiplicative factor
+- **Example:** β_BACI = -0.48 on log scale → impact sites had exp(−0.48) = 0.62× the abundance of control sites after disturbance, i.e. a 38% reduction

package/skills/ecological-impact-assessment/resources/fragmentation-metrics-reference.md

@@ -0,0 +1,86 @@
+# Landscape Fragmentation Metrics Reference
+
+Computed with `landscapemetrics` (R) or FRAGSTATS. Always report the spatial resolution used.
+
+## Patch-Level Metrics
+
+| Metric | Abbreviation | Unit | Description |
+|--------|-------------|------|-------------|
+| Patch area | AREA | ha | Area of each patch |
+| Perimeter | PERIM | m | Total perimeter of patch |
+| Perimeter-area ratio | PARA | m/ha | Shape complexity; higher = more irregular |
+| Shape index | SHAPE | — | PERIM / (4√AREA); 1 = square; higher = complex |
+| Fractal dimension index | FRAC | — | 1–2; higher = more complex shape |
+| Core area | CORE | ha | Area > edge distance from patch edge |
+| Core area index | CAI | % | % of patch that is core area |
+| Proximity index | PROX | — | Area/distance to nearest same-class patch |
+
+## Class-Level Metrics
+
+| Metric | Abbreviation | Unit | Description |
+|--------|-------------|------|-------------|
+| Total class area | CA | ha | Sum of all patch areas |
+| Number of patches | NP | count | Count of patches |
+| Patch density | PD | n/100 ha | NP per 100 ha landscape |
+| Largest patch index | LPI | % | % of landscape in largest patch |
+| Mean patch area | AREA_MN | ha | Average patch size |
+| Mean patch shape index | SHAPE_MN | — | Average patch shape complexity |
+| Edge density | ED | m/ha | Total edge length per ha |
+| Mean nearest-neighbour distance | ENN_MN | m | Isolation between patches |
+| Aggregation index | AI | % | 0–100; 100 = fully aggregated |
+| Splitting index | SPLIT | — | Fragmentation; 1 = unfragmented |
+| Effective mesh size | MESH | ha | Probability two random points in same patch × landscape area |
+| Landscape shape index | LSI | — | Total edge / min edge for same total area |
+
+## Connectivity Metrics
+
+| Metric | Description | R package |
+|--------|-------------|----------|
+| COHESION | Patch cohesion index; physical connectedness | landscapemetrics |
+| CONNECT | Functional connectivity based on distance threshold | landscapemetrics |
+| IIC | Integral Index of Connectivity (graph-based) | conefor / igraph |
+| PC | Probability of Connectivity | conefor |
+| dPC, dIIC | Importance of each patch to overall connectivity | conefor |
+
+## Key Metrics for Impact Assessment
+
+Recommended minimum set for before/after comparison:
+
+1. **CA** — total habitat area (has it declined?)
+2. **NP** — number of patches (has fragmentation increased?)
+3. **LPI** — largest patch index (is the main mass intact?)
+4. **ED** — edge density (are edge effects increasing?)
+5. **MESH** — effective mesh size (functional fragmentation — most comprehensive single metric)
+6. **COHESION** — physical connectivity of patches
+
+## R Code Template
+
+```r
+library(terra)
+library(landscapemetrics)
+
+lc <- rast("landcover.tif")
+habitat_class <- 3  # class code for the habitat of interest
+
+# Patch-level
+patch_metrics <- calculate_lsm(lc, what = c("lsm_p_area", "lsm_p_shape", "lsm_p_core"),
+                                 class = habitat_class)
+
+# Class-level
+class_metrics <- calculate_lsm(lc, what = c("lsm_c_ca", "lsm_c_np", "lsm_c_lpi",
+                                              "lsm_c_ed", "lsm_c_mesh", "lsm_c_cohesion"),
+                                 class = habitat_class)
+print(class_metrics)
+```
+
+## Interpreting MESH (Effective Mesh Size)
+
+MESH (m_eff) = Σ(Ai²) / A_total
+
+where Ai = area of patch i, A_total = total landscape area.
+
+- **High MESH** = landscape dominated by large, connected patches → low functional fragmentation
+- **Low MESH** = landscape cut into many small patches → high fragmentation
+- MESH is the **most recommended single metric** for summarising fragmentation because it is sensitive to both patch size and patch number and is interpretable in area units (ha).
+
+Example: MESH drops from 5,000 ha to 800 ha → a person starting at a random point is now 6× more likely to be confined to a smaller patch.