How ddg-64/waterqualityprediction_cn works
This repository builds probabilistic water-quality forecasts for Chinese monitoring stations. Its main workflow acquires remote-sensing inputs and combines them with station observations and yearbook indicators; processing prepares features, splitting creates horizon-specific samples, and training produces quantile forecasts that evaluation scores and visualizes. The graph preserves documented 3-day and 14-day prediction paths and the four-modality model. Some processing and splitting links are documented rather than visible in sampled code; the README is cited for those relationships.
Main entry points: run_pipeline.py Processing runner · run_download.py Download runner · evaluate.py Evaluation reports · settings.py GEE setup
Components 23
Data acquisition
- Download runner
data_download/run_download.py - GEE setup
config/settings.py - CAMS download
data_download/download_cams.py - ERA5 download
data_download/download_era5.py - Dynamic World download
data_download/download_dw.py - Population download
data_download/download_population.py
Feature processing
- Processing runner
data_processing/run_pipeline.py - ERA5 features
data_processing/step1_era5_features.py - CAMS alignment
data_processing/step3_cams_align.py - Land-use fusion
data_processing/landuse_image_fusion.py - Feature merge
data_processing/step6_merge_all.py - Clean and combine
data_processing/step7_clean_combine.py - Climate assignment
data_processing/assign_climate_zone.py - Population and yearbook
data_processing/step9_statistical_yearbook.py
Forecast modeling
- Horizon data splitting
data_splitting - Quantile model
models/model.py - Training losses
models/losses.py - Evaluation metrics
models/metrics.py
Training and evaluation
- Dataset loaders
training/dataset.py - Model training
training/train.py - Evaluation reports
training/evaluate.py
Other components
- Researcher
- Google Earth Engine
Connections 32
- Researcher → Download runner: runs downloads
README.md - Download runner → CAMS download: dispatches
data_download/run_download.py - Download runner → ERA5 download: dispatches
data_download/run_download.py - Download runner → Dynamic World download: dispatches
data_download/run_download.py - Download runner → Population download: dispatches
data_download/run_download.py - CAMS download → GEE setup: uses setup
data_download/download_cams.py - ERA5 download → GEE setup: uses setup
data_download/download_era5.py - Dynamic World download → GEE setup: uses setup
data_download/download_dw.py - Population download → GEE setup: uses setup
data_download/download_population.py - CAMS download → Google Earth Engine: queries
data_download/download_cams.py - ERA5 download → Google Earth Engine: queries
data_download/download_era5.py - Dynamic World download → Google Earth Engine: queries
data_download/download_dw.py - Population download → Google Earth Engine: queries
- Researcher → Processing runner: runs processing
README.md - Processing runner → ERA5 features: calls
data_processing/run_pipeline.py - Processing runner → CAMS alignment: calls
data_processing/run_pipeline.py - Processing runner → Land-use fusion: runs fusion
README.md - Processing runner → Feature merge: calls
data_processing/run_pipeline.py - Processing runner → Clean and combine: dispatches
data_processing/run_pipeline.py - Processing runner → Climate assignment: dispatches
data_processing/run_pipeline.py - Processing runner → Population and yearbook: dispatches
data_processing/run_pipeline.py - Feature merge → Clean and combine: feeds features
README.md - Clean and combine → Population and yearbook: enriches data
README.md - Researcher → Horizon data splitting: prepares horizons
README.md - Horizon data splitting → Dataset loaders: writes split inputs
README.md - Dataset loaders → Model training: builds loaders
training/dataset.py - Model training → Quantile model: trains
training/train.py - Model training → Training losses: computes loss
training/train.py - Researcher → Evaluation reports: evaluates forecasts
README.md - Evaluation reports → Dataset loaders: loads splits
training/evaluate.py - Evaluation reports → Quantile model: loads forecasts
training/evaluate.py - Evaluation reports → Evaluation metrics: computes scores
training/evaluate.py