CVCE — Atlas of the Global Clean Industrial Base

01 — Global and Country Atlases of the Clean Industrial Base

★ Flagship · 13 case studies

Global Atlas of the Clean Industrial Base

World map trade flows, value chain timeline, product treemap, capability radar, sankey decomposition, PC scatter, and firm activity — all in one panel view across 13 country × technology cases.

Case Studies

SolarIndia NuclearJapan NuclearKorea BatteriesJapan BatteriesKorea BatteriesIndia BatteriesBrazil TransmissionJapan TransmissionKorea TransmissionVietnam WindJapan WindVietnam WindIndia

Full Value Chain · 10 techs

Clean Industrial Base

Brazil

All 10 technologies. Trade map, capability radar, SHAP gap analysis, firm activity layer, sankey, dendrite, and PC scatter for Brazil across the full clean energy value chain.

Technologies Covered

Solar · Nuclear · Batteries · Transmission · Wind · Biofuel · Electrolyzers · Geothermal · Heat Pumps · Magnets

Full Value Chain · 10 techs

Clean Industrial Base

India

All 10 technologies. Trade map, capability radar, SHAP gap analysis, firm activity layer, sankey, dendrite, and PC scatter for India across the full clean energy value chain.

Technologies Covered

Solar · Nuclear · Batteries · Transmission · Wind · Biofuel · Electrolyzers · Geothermal · Heat Pumps · Magnets

Additional Country Atlases & Methods

Full Value Chain · 10 techs

Vietnam

All 10 technologies across the clean energy value chain for Vietnam — trade map, radar, firm activity, sankey, dendrite, and PC scatter.

Raw Materials · Chile & Brazil

Raw Material Atlases

Critical materials supply — copper, lithium, cobalt, iron ore. Multi-tech with upstream/processed filters.

IO Structure · Leontief

Input-Output Structure

EXIOBASE-based Leontief network, Solar × India. 393 nodes, power-law sizing, role-based palette.

06 — NZIPL Team Claude Kit

Shared knowledge base maintained by Gilberto García-Vázquez. Loads automatically into Claude Code sessions — data source conventions, API patterns, design tokens, and hard-won gotchas so the team doesn't rediscover the same issues independently.

github: Gilix/nzipl-claude-kit

NZIPL Claude Kit

Clone the repo and open with Claude Code — CLAUDE.md loads automatically. Covers the full Lab toolkit: OEC and DataMexico API patterns, sprint structure, deliverable templates, design system, and running discoveries log.

Skills (invoke via /skill-name)

/nzipl-design /naics-hs-matching /gather-cleantech-greenfield /gather-ev-greenfield /audit-naics-hs /enrich-fdi

Presentation → CLAUDE.md →

Known Gotchas

OEC API: HS6 ID padding — strip leading digits: str(id)[-6:]

OEC Cloudflare: Requires browser User-Agent header or gets 403

DataMexico /data: Silently drops manufacturing sectors 31-33 — use /stats/rca

BACI 2022: HS6 870839 (braking systems) missing — fall back to baci_a_02

FDI Markets: Investment totals include multi-phase projects — press releases cite initial phase only

Design system: Kit uses CICE theme (#56a360) — different from Atlas (#073309)

Data Sources & Auth

OEC: Bearer token + X-API-KEY + token param + User-Agent

DataMexico: economia.gob.mx/datamexico/api/ (legacy api.datamexico.org deprecated)

BACI / CVCE: ~/Documents/R/data/BACI/ — outside repo

BNEF: Pre-compiled JSON — no API

OSM: earth-osm — transmission lines, substations, generators

Glossary Highlights

Play · Play Cards · Constraint Maps
Binding constraint · Gating requirement
Ubiquity · Diversity · ECI · PCI

10 — Pipeline Commands & Data Scope

Reference for rebuilding atlases, rendering panels, and understanding the data cross-walk. All commands run from the project root.

Country & Global Atlases

Full Country Atlas (10 techs)

Rscript scripts/atlas/build_country_atlas.R \ --iso IND --name India Rscript scripts/atlas/build_country_atlas.R \ --iso BRA --name Brazil Rscript scripts/atlas/build_country_atlas.R \ --iso VNM --name Vietnam Rscript scripts/atlas/build_country_atlas.R \ --iso MEX --name Mexico

python3 scripts/render/inline_libs.py \ analysis/atlas/India_Atlas_of_the_Clean_Industrial_Base.html

Outputs to analysis/atlas/{Country}_Atlas_of_the_Clean_Industrial_Base.html · inline_libs.py embeds D3 + topojson + world atlas (offline-capable)

Single-Tech Sheet & Global Atlas

# Single tech × country sheet Rscript scripts/atlas/build_country_atlas.R \ --iso JPN --name Japan \ --tech Nuclear \ --outdir analysis/atlas/sheets python3 scripts/render/inline_libs.py \ analysis/atlas/sheets/Nuclear_Japan_*.html # Global Atlas (13 cases) Rscript scripts/atlas/build_global_atlas.R python3 scripts/render/inline_libs.py \ analysis/atlas/Global_Atlas_of_the_Clean_Industrial_Base.html

Sheets: analysis/atlas/sheets/{Tech}_{Country}_Atlas_*.html
Supported ISOs: IND BRA JPN KOR VNM MEX ESP (extend FLAG_MAP for others)

Module Panel Rendering

All 8 panels for a case

# 8 panels: radar, pc_scatterplot, trade_timeline, # trade_composition, tech_tree, sankey, # trade_map, activity_map, pc_scatter Rscript scripts/render/render_case_panels.R \ --iso IND --tech Solar Rscript scripts/render/render_case_panels.R \ --iso BRA --tech Batteries Rscript scripts/render/render_case_panels.R \ --iso KOR --tech "Nuclear"

Output: analysis/{module}/{Country}/{Tech}_{Country}_{module}.html
tech_tree: analysis/tech_tree/{Tech}_tech_tree.html (country-independent, skip if exists)

Specific panel & activity map refresh

# Single panel type only Rscript scripts/render/render_case_panels.R \ --iso JPN --tech Wind \ --panels activity_map # After expanding S&P firm data: # Rebuild activity maps for all countries for iso in IND BRA JPN KOR VNM MEX ESP; do for tech in Solar Nuclear Batteries ...; do Rscript scripts/render/render_case_panels.R \ --iso $iso --tech "$tech" \ --panels activity_map done done

--panels accepts comma-separated list: radar,sankey,trade_map

Data Pipeline & Cross-Product Concordance

Rebuild firm data (S&P)

# Expand to new countries in FOCAL list, # then rebuild activity maps Rscript scripts/build_data/08d_build_sp_firms.R # Current coverage: BRA IND JPN KOR VNM MEX ESP # 7 countries · 21,832 city-level rows # Geocoded via maps::world.cities

Source: data/sp/sp_company_naics_hs_cleaned.xlsx (gitignored, 62MB)
Output: data/sp/sp_firms_geocoded.parquet

Dataset scope

BACI trade1995–2024 · ~200 countries · HS6 · KUSD

Green dictionary605 HS6 codes × 10 techs · stage/role/shap_category

PC scores155 countries × 10 techs × 2003–2024

SHAP features402 HS6 × tech entries · 5 categories · global avg

EXIOBASE IO9,800×9,800 A-matrix · leontief + direct use shares

S&P firms21,832 rows · 7 countries · geocoded cities

Dendrite edges913 edges × 10 techs · HS6 value chain tree

HS revision windowHS92(1995) · HS96–HS22 non-overlapping

Cross-Product Concordance

HS6 classification hierarchy

HS6 → shap_categoryChemicals / Electronics / Machinery / Metals / Ind.Materials

HS6 → stageUpstream / Midstream / Downstream / Final Product

HS6 → roleRaw Mat. / Processed Mat. / Process Equip. / Product Comp. / Final Product

HS6 → NACE4via naics6_nace crosswalk (08c_build_naics_crosswalk.R)

HS6 → NAICS6S&P firm matching · NZIPL naics-hs-matching skill

HS6 → EXIOBASEexiobase_product · leontief_use_share · direct_use_share

Master file: data/green_dict/green_dictionary.csv (605 rows, 20 columns)

Capability radar pipeline

SHAP weightspc_features.csv → sum shap_mean_z by category → renorm (tech-specific, not country)

Country RCApc_rca.parquet · Balassa formula · country × tech × category × year

ClassificationWP-4 3×3: CRIT GAP / GAP / BUILD / STRENGTH / LOW PRIORITY

SHAP-RCA scatterproduct-level RCA (focal/world share ratio) × SHAP mean|z| per HS code

Macro featuresfdi_z · trade_z · tariff_z · gdp_z · industry_gdp_z · mfg_z (excluded from radar)

Full methodology: capability_radar_methodology.html