Configure as global-30x30 geo-agent app

Replace template (CA / PAD-US) with a global 30×30 / Target 3 reproduction
kit inspired by Fajardo et al. 2026 (Nat Commun, doi:10.1038/s41467-026-71860-8).

- layers-input.json: global view, grouped collections for WDPA, LandMark IPLC,
  WWF ecoregions, IUCN richness, NCP biodiversity + irrecoverable carbon,
  GHS-POP, and Overture admin boundaries
- system-prompt.md: 30×30 framing with explicit "what we don't have" section
  listing the missing layers (KBA, AOH, MOSAIKS HDI, livelihood datasets)
- coverage.md: figure-by-figure reproducibility table against the paper
- k8s/: rename slug padus → global-30x30
- Include paper + supplementary PDFs (CC-BY 4.0) for reference

Author: cboettig

README.md

@@ -1,113 +1,33 @@
-# geo-agent-template
+# global-30x30 — geo-agent app
 
-A GitHub template for deploying an AI-powered interactive map app.
-Users describe in plain language what datasets to show; the app uses an LLM agent with map tools and SQL access to visualize and analyze cloud-native geospatial data.
+A geo-agent configuration for exploring the spatial and social context of the **Kunming-Montreal Global Biodiversity Framework Target 3** ("30×30") — the commitment that at least 30% of terrestrial, inland water, marine and coastal areas be effectively conserved by 2030.
 
-**No JavaScript to write.** The core modules (map, chat, agent, tools) are loaded from CDN. You configure which data to show via three small files.
+Inspired by the analysis in Fajardo et al. 2026, *Social implications of the 30×30 global conservation target* (Nature Communications, [doi:10.1038/s41467-026-71860-8](https://doi.org/10.1038/s41467-026-71860-8)). See [`coverage.md`](coverage.md) for which of that paper's analyses can and cannot be reproduced from the current STAC catalog.
 
-**Full documentation:** [boettiger-lab.github.io/geo-agent/docs](https://boettiger-lab.github.io/geo-agent/docs/)
+## Layers
+
+- **WDPA December 2025** — current global protected areas
+- **LandMark IPLC v202509** — Indigenous and community-mapped lands
+- **WWF Terrestrial Ecoregions (2017)** — biodiversity representation framing
+- **IUCN Species Richness (2025)** — vertebrate richness rasters
+- **NCP — Biodiversity & Natural Habitat** (Chaplin-Kramer 2019) and **Irrecoverable / Vulnerable Carbon** (Noon 2022)
+- **GHS-POP 2020** — gridded global population
+- **Overture countries / regions** — admin context
 
 ## Repository structure
 
 ```
 index.html          ← HTML shell — loads core JS/CSS from CDN
-layers-input.json   ← which STAC collections to show + LLM settings
-system-prompt.md    ← LLM system prompt (customize per app)
+layers-input.json   ← STAC collections + LLM settings + welcome examples
+system-prompt.md    ← LLM system prompt (30×30 framing + caveats)
 k8s/                ← Kubernetes deployment manifests (optional)
 ```
 
-## Quick start
-
-### 1. Create your repo from this template
-
-Click **"Use this template"** on GitHub → **"Create a new repository"**.
-
-### 2. Choose your datasets
-
-Browse the available STAC catalog:
-
-```
-https://radiantearth.github.io/stac-browser/#/external/s3-west.nrp-nautilus.io/public-data/stac/catalog.json
-```
-
-Edit `layers-input.json` — set your collections and adjust the default map view.
-See the [configuration reference](https://boettiger-lab.github.io/geo-agent/docs/guide/configuration) for all fields.
-
-### 3. Edit `system-prompt.md`
-
-Describe the domain, what users are likely to ask, and include SQL examples relevant to your datasets.
-
-### 4. Deploy
-
-#### Option A: GitHub Pages (no server needed)
-
-The `llm` block in `layers-input.json` is already enabled. Each visitor enters their own API key (e.g. from [OpenRouter](https://openrouter.ai)) in the in-app settings panel — keys are stored in the browser only, never on the server.
-
-1. Enable GitHub Pages in your repo: Settings → Pages → Source → **GitHub Actions**
-2. Add `.github/workflows/gh-pages.yml`:
-
-```yaml
-name: Deploy to GitHub Pages
-on:
-  push:
-    branches: [main]
-permissions:
-  contents: read
-  pages: write
-  id-token: write
-jobs:
-  deploy:
-    runs-on: ubuntu-latest
-    environment:
-      name: github-pages
-      url: ${{ steps.deployment.outputs.page_url }}
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/configure-pages@v5
-      - uses: actions/upload-pages-artifact@v3
-        with:
-          path: .
-      - id: deployment
-        uses: actions/deploy-pages@v4
-```
-
-3. Push — the workflow deploys automatically on changes to `main`.
-
-#### Option B: Kubernetes
-
-API keys are injected server-side via a ConfigMap + Kubernetes secrets — no user-facing key entry.
-
-1. Delete the `llm` block from `layers-input.json` (the server-injected `config.json` takes precedence anyway)
-2. Replace the git clone URL in `k8s/deployment.yaml` with your repo URL
-3. Replace the slug `calenviroscreen` throughout `k8s/` with your app name
-4. Set your hostname in `k8s/ingress.yaml`
-5. Create the required secret:
-
-```bash
-kubectl create secret generic llm-proxy-secrets \
-  --from-literal=proxy-key=YOUR_PROXY_KEY
-```
-
-6. Deploy:
-
-```bash
-kubectl apply -f k8s/
-kubectl rollout status deployment/my-app
-```
-
-After pushing changes, redeploy: `kubectl rollout restart deployment/my-app`
-
-See the [deployment guide](https://boettiger-lab.github.io/geo-agent/docs/guide/deployment) for full details on all options including Hugging Face Spaces.
+This is a fork of [`boettiger-lab/geo-agent-template`](https://github.com/boettiger-lab/geo-agent-template); the JavaScript runtime is loaded from CDN. See the [geo-agent docs](https://boettiger-lab.github.io/geo-agent/) for the full configuration reference.
 
-## Local development
+## Local preview
 
 ```bash
 python -m http.server 8000
-# Open http://localhost:8000 — enter your API key in the settings panel
+# Open http://localhost:8000 — enter an API key in the settings panel
 ```
-
-## More resources
-
-- [Configuration reference](https://boettiger-lab.github.io/geo-agent/docs/guide/configuration) — all `layers-input.json` fields with examples
-- [Deployment guide](https://boettiger-lab.github.io/geo-agent/docs/guide/deployment) — GitHub Pages, Hugging Face Spaces, Kubernetes
-- [Core library](https://github.com/boettiger-lab/geo-agent) — source code for the map, chat, and agent modules

coverage.md

@@ -0,0 +1,39 @@
+# Reproducibility coverage
+
+This app's layers are a partial reproduction kit for Fajardo et al. 2026, *Social implications of the 30×30 global conservation target* (Nature Communications, [doi:10.1038/s41467-026-71860-8](https://doi.org/10.1038/s41467-026-71860-8)). The table below tracks what we can and cannot reproduce against the paper's reported figures.
+
+Datasets we currently lack are tracked in [boettiger-lab/data-workflows#161](https://github.com/boettiger-lab/data-workflows/issues/161) — see also #140 (Theobald gHM), #157 (KBA + IUCN AOH ranges).
+
+## Reproducibility status by figure / table
+
+| Paper artefact | What it shows | Reproducible? | Blocker |
+|---|---|:---:|---|
+| **Headline: 396 M in existing PAs, 1.15 B within 10 km** | Resident + buffer population for current PA/OECM network | 🟡 Partial | Have WDPA + GHS-POP, missing WD-OECM. Numbers will differ by year (we have 2020 not 2023). |
+| **Fig 1 — scenario maps** | Three Target-3 scenarios overlaid with current PAs | ❌ | Scenario rasters not hosted; the `prioritizr` runs that produced them require species AOH, KBA, NCP, HMI we don't have |
+| **Fig 2a — resident + neighbouring population by scenario** | Population in new PAs vs buffer | ❌ | Same scenario rasters |
+| **Fig 2b — HDI breakdown by scenario** | % of scenario population in Low / Medium / High / Very High HDI | ❌ | MOSAIKS gridded HDI not in catalog |
+| **Fig 2c — livelihood overlap (wild harvest / farm / livestock)** | % of scenario area / population overlapping each livelihood class | ❌ | Wells (wild harvesting), Lesiv/Mehrabi (farm size), GLPS v5 (livestock) all missing |
+| **Fig 3 — continental HDI × population** | Per-continent comparison of scenario populations and HDI | ❌ | Both blockers above |
+| **Suppl Table 1 — population in each scenario** | Per-scenario population totals + % of global | 🟡 Partial | Row for "Existing PAs" only; scenario rows need the rasters |
+| **Suppl Table 2 — population in scenarios by continent** | Continent-level totals | 🟡 Partial | Existing-PA row only |
+| **Suppl Table 3 — population × HDI** | All scenarios × HDI groups | ❌ | MOSAIKS HDI |
+| **Suppl Table 4 — livelihood overlap** | Wild harvesting / farm area / livestock × scenario | ❌ | Livelihood layers |
+| **Suppl Fig 1 — % global area in scenarios + buffers** | Area accounting | 🟡 Partial | Existing-PA portion only |
+| **Suppl Fig 2 / 3 — spatial overlap between scenarios** | Venn / overlap maps | ❌ | Scenario rasters |
+| **Suppl Fig 5 — ITT scenario distribution** | 100-replicate ITT map | 🟡 Partial | Have LandMark IPLC as a proxy, not the ICCA Registry + HMI-masked composite the paper uses |
+
+## What we can reproduce today
+
+These analyses are doable with the layers currently configured in `layers-input.json`:
+
+- **Global PA coverage**: `% of land area inside a designated PA` — `wdpa-december-2025`. Will land short of the paper's 17.2% figure because WD-OECM is missing.
+- **Population inside / near PAs**: `ghs-pop-2020` × `wdpa-december-2025`. Comparable to the paper's first row.
+- **Country-level PA coverage**: WDPA `ISO3` × Overture countries.
+- **Ecoregion representation**: `wwf-ecoregions-2017` × `wdpa-december-2025`. % of each ecoregion inside the PA network — speaks to the biodiversity-representation argument of the paper without reproducing the `prioritizr` optimization.
+- **Biodiversity value of PAs vs unprotected**: `iucn-richness-2025` (combined or per-taxon) × WDPA. Aggregated richness only — not the per-species coverage check the paper performs.
+- **Carbon stocks inside / outside PAs**: `irrecoverable-carbon` × WDPA. Closest analogue to one of the paper's 10 NCP layers (vulnerable ecosystem carbon).
+- **IPLC × PA overlap**: `landmark-iplc-poly` × `wdpa-december-2025`. Proxy for the ITT lens.
+
+## Things we should *not* fabricate
+
+If a user asks for HDI breakdowns, wild-harvesting estimates, farm-size or livestock-area overlaps, or scenario-specific results: say plainly we don't host those layers and point at the source datasets in [data-workflows#161](https://github.com/boettiger-lab/data-workflows/issues/161). Do not substitute unrelated layers (e.g. don't approximate HDI with SVI — SVI is US-only and a different construct).

index.html

@@ -4,7 +4,7 @@
 <head>
     <meta charset="UTF-8">
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    <title>CA Protected Lands</title>
+    <title>Global 30×30 Explorer</title>
 
     <!--
         Geo-Chat core modules loaded from CDN.

k8s/configmap.yaml

@@ -1,15 +1,15 @@
 # LLM config template — NO SECRETS.
 #
 # The upstream LLM proxy key lives only in the nginx config (see
-# `padus-nginx` below), which injects the Authorization header
+# `global-30x30-nginx` below), which injects the Authorization header
 # server-side via `proxy_set_header`. The browser only sees a relative
 # `/api/llm` endpoint and never holds the key.
 #
 # See ../../geo-agent-ops/SECRETS_EXPOSURE.md for the migration plan.
 apiVersion: v1
 kind: ConfigMap
 metadata:
-  name: padus-config
+  name: global-30x30-config
   namespace: biodiversity
 data:
   config.template.json: |
@@ -81,7 +81,7 @@ data:
 apiVersion: v1
 kind: ConfigMap
 metadata:
-  name: padus-nginx
+  name: global-30x30-nginx
   namespace: biodiversity
 data:
   nginx.conf.template: |

k8s/deployment.yaml

@@ -1,10 +1,10 @@
 apiVersion: apps/v1
 kind: Deployment
 metadata:
-  name: padus
+  name: global-30x30
   namespace: biodiversity
   labels:
-    app: padus
+    app: global-30x30
 spec:
   replicas: 1
   strategy:
@@ -14,11 +14,11 @@ spec:
       maxSurge: 1
   selector:
     matchLabels:
-      app: padus
+      app: global-30x30
   template:
     metadata:
       labels:
-        app: padus
+        app: global-30x30
     spec:
       initContainers:
       - name: git-clone
@@ -30,7 +30,7 @@ spec:
           # Clone YOUR app repo (the GitHub repo you created from the template).
           # Update this URL to point to your own GitHub repo:
           # Update this URL to point to your own GitHub repo:
-          git clone --depth 1 https://github.com/boettiger-lab/geo-agent-template.git /tmp/repo
+          git clone --depth 1 https://github.com/boettiger-lab/global-30x30.git /tmp/repo
           cp /tmp/repo/index.html /usr/share/nginx/html/
           cp /tmp/repo/layers-input.json /usr/share/nginx/html/
           cp /tmp/repo/system-prompt.md /usr/share/nginx/html/
@@ -103,9 +103,9 @@ spec:
         emptyDir: {}
       - name: config-template
         configMap:
-          name: padus-config
+          name: global-30x30-config
       - name: nginx-template
         configMap:
-          name: padus-nginx
+          name: global-30x30-nginx
       - name: nginx-conf-d
         emptyDir: {}

k8s/ingress.yaml

@@ -1,7 +1,7 @@
 apiVersion: networking.k8s.io/v1
 kind: Ingress
 metadata:
-  name: padus-ingress
+  name: global-30x30-ingress
   namespace: biodiversity
   annotations:
     haproxy-ingress.github.io/cors-enable: "true"
@@ -16,15 +16,15 @@ spec:
   ingressClassName: haproxy
   tls:
     - hosts:
-        - padus.nrp-nautilus.io
+        - global-30x30.nrp-nautilus.io
   rules:
-    - host: padus.nrp-nautilus.io
+    - host: global-30x30.nrp-nautilus.io
       http:
         paths:
           - path: /
             pathType: Prefix
             backend:
               service:
-                name: padus
+                name: global-30x30
                 port:
                   number: 80

k8s/service.yaml

@@ -1,14 +1,14 @@
 apiVersion: v1
 kind: Service
 metadata:
-  name: padus
+  name: global-30x30
   namespace: biodiversity
   labels:
-    app: padus
+    app: global-30x30
 spec:
   type: ClusterIP
   selector:
-    app: padus
+    app: global-30x30
   ports:
     - protocol: TCP
       port: 80