mekong
Watch priority areas
Register mines, factories, rivers and water intakes. Check each available Sentinel-2 acquisition; retain date, bands and source identifiers.
03 / DETECT
Early signals. Connected evidence.
From satellite observations to community tests and river stations.
DELIVERABLE #4 · SATELLITE AND AI SURVEILLANCE
A closer look at activity on the ground.
Sepon mining areaLoading scene metadata…
10 m Sentinel-2 pixels · Zoom enlarges existing detail
Vegetation index (NDVI)
−10+1
Masked pixels show the true-colour image beneath.Read the imagery and analysis correctly ＋
True-colour imagery uses Sentinel-2 red, green and blue bands at 10 m. The vegetation index is calculated from red and near-infrared reflectance; it describes vegetation response, not contaminant concentration. Cloud masks are imperfect, and only selected valid land and water classes enter the index. One date cannot establish expansion, a trend or a pollution event.
These sites demonstrate image interpretation. No non-compliance or contamination is alleged. Dissolved metals, pathogens and discharge legality require field evidence, laboratory analysis and records.
Sources: Copernicus Sentinel-2 products · Cloud and scene classification · Element 84 Earth Search.
PROPOSED SURVEILLANCE MECHANISM
Observe → review → verify.
Keep usable pixels
Mask cloud, shadow and missing data. Compare equivalent seasons and flag areas without a usable observation for follow-up.
Find candidate changes
Propose AI segmentation for excavation and industrial footprints, then compare dates for disturbance and vegetation change. Calibrate water-colour indicators locally.
Check the explanation
An analyst reviews images, uncertainty, source records and permits. Reject cloud artifacts and seasonal changes; record the reason.
Send a focused field task
Give the community, station or mobile laboratory team a location, questions and a sampling plan. Confirm contamination before assigning a cause.
Improve with evidence
Return verified labels to the model. Track false alarms, missed events, usable-image delays and time from screening to confirmation.
DELIVERABLE #5 · COMMUNITY TESTING NETWORK
Local tests. A shared picture.
✓ Routine! Re-test! Priority review? No recent data
Generalized river geography · Natural Earth · Illustrative testing points
No points match this screening state. Choose another filter.
Traffic-light screening rules ＋
These are illustrative workflow thresholds, not Lao regulatory limits or a water-safety certification. Results apply only to the displayed sample and parameters.
- Priority review: a reviewed E. coli detection or turbidity above 20 NTU.
- Re-test: a complete sample awaiting review, pH outside 6.5–8.5, or turbidity above 5 NTU.
- Routine: all three results available, reviewed, within these example ranges and no more than seven days old.
- No recent data: incomplete, invalid or older results. A previously reviewed priority signal stays red until resolved.
A routine result does not demonstrate that all microbial and chemical hazards have been assessed. WHO drinking-water context.
COMMUNITY → QUALITY REVIEW → SHARED MAP
Make every test traceable.
The proposed network pairs trained volunteers with district teams: record the location, sample time, kit, method and quality checks; validate the result; publish a clear status and next step. Confirmatory laboratory results close the loop.
DELIVERABLE #6 · MAINSTREAM–TRIBUTARY MONITORING
One river network. Connected observations.
■ Mainstream● TributaryReportingReview signalOffline
Concept station positions on generalized river lines · Natural Earth
No stations match these filters.
COMPARE SELECTED STATIONS
Follow the same parameter.
Last available sample for each matching station. Offline readings remain historical; they are excluded from current summaries.
How the proposed network works ＋
Place stations on priority mainstream reaches and tributaries, with paired observations around selected confluences. Combine frequent sensor measurements with routine laboratory sampling, shared calibration protocols and one timestamped data hub.
For this demonstration, an online station with turbidity above 40 NTU receives a review flag. This is an illustrative operational trigger, not a legal limit. Turbidity can increase naturally after rain; it does not identify a polluter. Compare flow, rainfall and lab results before attribution.
River lines are geographically real but generalized. Stations are proposed examples snapped to these lines, not official MRC assets. The map does not calculate flow connectivity, arrival times or pollutant transport.