When Public Climate Data Becomes Meal Infrastructure: The Data-Continuity Standard MBG Needs

MBG Watch · 2026-09-28

The premise

A meal-safety ledger depends on civic data before it depends on a dashboard.

That is the quieter lesson in this week’s climate-data signals. NASA’s Earth Observatory described a newly developed method for mapping hidden Greenland bedrock, derived from ice-surface signatures and aimed at improving future versions of BedMachine Greenland, a high-resolution terrain dataset. Grist, republishing Yale Environment 360, reported that Germany’s Climate Computing Center is adding storage to preserve backup copies of U.S. climate simulations because scientific data can be lost, blocked, or manipulated for technical or political reasons.

Neither source is about Indonesia’s Makan Bergizi Gratis program. The connection is operational, not symbolic. Once a public meal system depends on heat, haze, rainfall, flood, earthquake, maritime, road, school-status, or health-post signals, those signals become part of the food-safety infrastructure. The question is no longer only whether a meal was cooked and delivered. It is whether the public record can show what the operator knew before the meal moved.

MBG Watch has already argued for adjacent records: forecast provenance, haze measurement-chain gaps, wet-season readiness, local knowledge governance, source authenticity, rehearsal before forecast use, and baseline climate stress testing. This piece names the upstream dependency that sits beneath those ledgers: MBG’s continuity record is only as trustworthy as the data chain it stands on.

What the evidence supports

Indonesia already has several public operational data channels that could plausibly shape a meal day.

BMKG’s open-data portal says it provides data in reusable formats to increase transparency, accountability, and public participation. Its weather-forecast product covers villages and urban wards for three days, with eight forecasts per day, updated twice daily, and with an API endpoint for administrative area codes. Its weather early-warning product is Common Alerting Protocol data, valid down to subdistrict level, updated continuously, with fields for urgency, severity, certainty, affected areas, start time, end time, and release authority. Its earthquake data page lists recent earthquakes, felt earthquakes, and tsunami-potential earthquakes in XML, JSON, and image formats, updated when an event occurs.

BMKG also publishes air-quality information. Its extreme-air-quality warning page describes three-day PM2.5 warnings for unhealthy, very unhealthy, and hazardous levels, updated daily at 08:00 WIB. The same page shows smoke-particulate forecasts, numerical pollutant modelling, and near-real-time PM2.5 monitoring from BMKG’s national sensor network.

BNPB’s disaster incident page presents geospatial information about reported disaster events in Indonesia and says the data is provided as a public service. It also carries a caution that the data shows relative reported locations and is not for legal or civil-engineering purposes. ASMC’s haze hotspot page provides daily hotspot counts and explains an important limitation: hotspots derived from NOAA satellite data may go undetected because of cloud cover or incomplete satellite passes, and reprocessed data may later differ from near-real-time information.

These are exactly the kinds of public signals that can change MBG operations without turning MBG into a climate program:

• heavy rain may change a kitchen’s cooking window, school delivery route, or holding time;
• flood reports may force rerouting, substitute service points, or a no-delivery decision;
• haze and PM2.5 warnings may change outdoor distribution, worker protection, and whether children queue outside;
• earthquake and tsunami information may change school status, kitchen integrity checks, or shelter-feeding arrangements;
• maritime disruption can alter ingredient movement to islands and coastal districts;
• conflicting or stale public signals can create the most dangerous condition of all: an operator believes the decision is evidence-based when the evidence chain has degraded.

BGN’s own Radar MBG announcement points in the same direction, though at a different layer. The August 2026 press release says Radar MBG is intended to let parents, teachers, heads of school, local government, and offices monitor implementation. It says the portal can show which schools receive the program, what the menu is, menu photos, nutrition information, and which SPPG produced the food. A public spot check of the Radar MBG page shows a menu lookup by province, city or district, subdistrict, village, and institution, plus a public feedback channel. That is useful transparency. It is not, on the public page reviewed, a data-continuity record for weather, haze, disaster, route, or school-status decisions.

That absence should be stated carefully. It does not prove that BGN lacks internal continuity procedures. It only means the public-facing transparency layer does not currently show the evidence chain a family, school, district inspector, or parliamentarian would need to reconstruct an environment-disrupted meal day.

What the evidence does not support

The evidence does not support saying MBG has failed because a climate dataset changed or disappeared. It does not support saying BGN must build a climate observatory. It does not support asking kitchen workers, cadres, schools, or parents to fill national data gaps with unprotected informal reporting.

The sharper claim is smaller and more inspectable: when external data becomes part of a stop/go, reroute, substitute, discard, school-online, shelter-feeding, or worker-safety decision, MBG should leave a public data-continuity trace.

That trace matters because public data has known failure modes:

• a forecast can be updated after the kitchen has already committed ingredients;
• a sensor network can be sparse, delayed, offline, or unrepresentative of a village route;
• a satellite-derived haze or hotspot product can miss fires under cloud cover or incomplete passes;
• a disaster incident map can be useful for awareness while still carrying limits on legal or engineering use;
• a dashboard can be authentic and still not show whether the underlying signal was current enough for a food-safety decision;
• a local observation can be true and urgent but unsafe to rely on unless the collector is protected and the escalation authority is clear.

The standard is not perfect data. Perfect data will not exist during the days when MBG most needs continuity: flood, smoke, heat, quake, landslide, power loss, port disruption, or failed connectivity. The standard is an honest record of data condition at the moment a public meal decision was made.

The least-harm standard

MBG’s data-continuity record can be small. It should not expose children, workers, household details, or sensitive facility vulnerabilities. It should show the operating evidence, not personal data.

For any weather, haze, disaster, earthquake, maritime, route, or public-status signal that materially changes a meal day, the public record should show:

1. Source — BMKG, BNPB/BPBD, ASMC, school/district notice, port or road authority, health office, or named local observation channel.
2. Timestamp — when the signal was issued, when it was retrieved, and when the operator acted.
3. Location and coverage — the administrative area, route segment, school, kitchen, or service point affected, expressed no more precisely than needed.
4. Confidence or degraded mode — whether the signal was normal, stale, missing, conflicting, sensor-limited, satellite-limited, connectivity-limited, or awaiting confirmation.
5. Local observation — what was reported locally, by whom in role terms, and how that report was protected from retaliation or overburden.
6. Human authority — the named role, not necessarily the personal name, authorized to hold, reroute, substitute, discard, or resume service.
7. Decision — the actual operating change: delay, reroute, menu substitution, holding-time reset, discard, school handoff, shelter handoff, worker-protection change, or no change.
8. Notice — who was told: school, parents, cadre, health office, local government, SPPG, distributor, or public dashboard.
9. Correction — what changed if the underlying data was later corrected or contradicted.
10. Fallback source — what source was used when the primary signal was unavailable.

This is not an additional bureaucracy for normal days. It is an exception record for days when the environment becomes part of the food-safety clock.

A practical version could be as simple as an incident row:

07:10 WIB — BMKG nowcast showed heavy rain warning for selected subdistricts; BNPB/BPBD flood report pending; school reported access road flooded; SPPG lead and district education liaison delayed delivery by 90 minutes, reset holding-time count, notified school and parents, and used alternate route after road clearance confirmation.

That kind of row is modest. It lets a family see why the meal was late. It lets an inspector see whether holding time was reset. It lets parliament see whether the 2027 budget is buying continuity capacity or only counting nominal meal trays. It lets MBG correct itself when the first signal was wrong.

What MBG Watch will monitor

Through Q4 2026, and before 2027 budget assumptions harden, MBG Watch will watch for five public signals.

First, whether Radar MBG or related BGN dashboards add last-updated time, source fields, degraded-data mode, or correction notices for operating information that families and schools are expected to trust.

Second, whether BGN, BMKG, BNPB/BPBD, education offices, health offices, and local governments name an evidence pathway for disrupted meal days, rather than leaving each school or kitchen to improvise.

Third, whether environmental exceptions are connected to food-safety decisions, not merely to logistics. A flood is not just a route problem if it changes holding time, storage temperature, contamination risk, or discard decisions.

Fourth, whether local observation is treated as protected evidence rather than unpaid risk transfer. Cadres, teachers, drivers, and parents should not become the fallback data infrastructure without authority, training, and protection.

Fifth, whether public reporting distinguishes three different conditions: no incident, incident but no meal change, and incident with meal change. Without that distinction, continuity statistics become too smooth to trust.

What I am uncertain about

I have not found a public BGN document that describes internal degraded-data procedures for MBG kitchens and district operators. Such procedures may exist outside the pages reviewed here. The public claim should therefore stay narrow: the visible transparency layer does not yet show the data-continuity record a climate-disrupted meal day would require.

I am also uncertain how consistently district-level operators already use BMKG, BNPB/BPBD, ASMC, school, road, port, and health-office signals in practice. That is precisely why the record matters. The answer should not have to depend on anecdotes after harm occurs.

MBG does not need to become a climate agency. It needs to show, when public climate and disaster data enters the meal route, that someone knew the source, knew its condition, made a bounded decision, told the people affected, and corrected the record when the evidence changed.