CT-ACE Explorer

Method

Near-real-time tracking of global land anthropogenic CO2 emissions constrained by satellite NO2 observations.

Overview

Carbon-Transformer outputs served as regional monitoring evidence.

Carbon-transformer Platform converts monthly 0.25 degree emission rasters into administrative-region statistics, trend diagnostics, anomaly checks, hotspot mining, and API-ready analysis products.

Satellite constraint Satellite NO2 provides near-real-time combustion signals for anthropogenic CO2 inference.
Ancillary predictors Nighttime lights, meteorology, land fields, temporal features, and historical emission states.

Serving Pipeline

From CT-ACE raster files to interactive analysis

  1. 01 Index monthly GeoTIFF files by month and product mode.
  2. 02 Load GADM administrative metadata from the prebuilt JSON index.
  3. 03 Rasterize selected administrative geometries to the CT-ACE grid.
  4. 04 Aggregate monthly totals, MoM changes, and YoY changes.
  5. 05 Expose query APIs for Explorer, Carbon Bot, data export, and reproducible downstream analysis.

Carbon Bot Agent

A RAG + Tool Agent for questions that need evidence and data.

Carbon Bot retrieves Carbon-Transformer method notes, CT-ACE data descriptions, NO2-CO2 mechanism background, and Sichuan 2022 event materials, then combines the evidence with structured emission-analysis tools for anomaly diagnosis, hotspot validation, and mechanism-aware interpretation.

Carbon Bot
  • Context awareUses the current Explorer region, month, and metric.
  • Tool routingHandles emission lookup, time series, anomaly detection, trend plots, and hotspots.
  • Anomaly diagnosisCombines multiple detectors to summarize severity, direction, supporting scores, and possible uncertainty.
  • Top-K hotspot validationChecks whether highlighted hotspots are structurally credible and explains why emissions may be high in each cell.
  • Cited answersReturns retrieval citations, tool-call traces, and plot links when available.
  • LLM optionalRuns locally by default, with OpenAI-compatible summarization when configured.