Sri Lanka decision proposal
Tell us what to change or add.
Archived decisions
Open a decision below. Check the area, period, inputs, result and proposed next step.
Use Change example to try different values. Use Data needed to check the fields and sources. The land-use change decision has its own comparison views.
Record your corrections below, then download your answers or print them.
Compare published rainfall departures for the same period and climate baseline.
Published official data · District cumulative rainfall departure
The published cumulative rainfall departure for Matale is -0.4% for 2026-01-01 to 2026-11-04, against 1981–2010.
Next step. Use the dated district summary as background; check local observations for site decisions.
Saved Table 01 has 23 district summaries. Kegalle is NA. Table 02 and its weekly map are marked NOT AVAILABLE.
For a directly published departure, retain the published percentage. For matched totals only: departure = 100 × (observed − normal) / normal, where normal > 0. Percent of normal = 100 × observed / normal; it is a different measure.
District cumulative figures do not measure rainfall at a Knuckles site or establish a drought. The weekly table marked NOT AVAILABLE is excluded.
Calculation appendix
Identify the dated low-rainfall runs supported by the supplied daily series.
Validation required
Longest observed dry spell (days) = 3.
2025-06-01 records 4 mm; this does not meet the stated example dry-day rule.
Next step. Review the dated daily sequence and any gaps.
The saved official table is one three-hour snapshot. A continuous daily rainfall series is still required for a real dry-spell analysis.
Count a run only across distinct consecutive observed days at one gauge using one daily convention and cutoff. Trace and missing days break the known run unless a documented measurement rule supplies a usable value.
This is a known observed run, not a drought diagnosis or warning threshold. A three-hour snapshot cannot replace a daily history.
Check received records against an explicit observation schedule and quality review.
The source reports 43444 at 2026-11-10T17:30. A schedule and QC delivery log are needed to calculate completeness.
Next step. Obtain the records listed below.
The saved table identifies 24 stations and their published report times. It does not provide a complete delivery log or source QC flags.
Completeness = 100 × distinct valid / scheduled observations in the stated window. Age = assessment time − latest valid time in a confirmed common time zone. Compare age with the supplied reporting interval.
One public snapshot does not establish how many readings were scheduled, received or quality-checked.
Check the monthly rainfall series and method evidence needed before a separate SPI calculation.
350 of 360 stated monthly observations are usable (97.222%). No SPI value is calculated.
No long monthly station series is present in the saved capture. The published drought and anomaly products do not supply that missing series.
Reported coverage = 100 × usable / expected months. Verify the expected monthly sequence, accumulation period, station changes, missing-month treatment and fitted distribution separately. This screen does not calculate SPI.
WMO No. 1090 describes long monthly records and the effect of missing data. A coverage percentage alone does not validate SPI or establish drought.
Compare the same temperature statistic and observation interval before investigating a difference.
Published instantaneous temperature is 24 °C. A matching reference has not been established.
The saved station table supplies instantaneous air temperature, not a daily maximum/minimum history.
Temperature difference = observed − matching reference in °C. Match station, statistic, observation interval, units and source definitions. Do not calculate a relative percentage from Celsius values.
An instantaneous reading is not a daily maximum or minimum. A difference does not identify a sensor fault or establish a climate trend.
Sri Lanka source match. Check the area and time covered by a warning.
Match the affected area and selected time to the Department’s warning. Check the warning ID, original level, issue time and validity window. Use Sri Lanka time (UTC+05:30).
Department of Meteorology, Sri Lanka · Weather Advisory System. Warnings endpoint, response fields and time format. CAP 1.2 format reference.
Confirm the duty officer, observation evidence and escalation procedure in Institution review.
Check the original warning product and accepted observations before a duty-officer review.
This case identifies the bulletin and evidence needed for a duty-officer warning review. It does not issue a warning.
The official site publishes advisories and states that warning API access is available on request. No API key or operational warning feed is used here.
Check bulletin identity, issue and validity, replacement history, accepted observation references, affected area, applicable procedure and duty responsibility. No hazard threshold is inferred.
This prepares a review note. It never issues or approves a warning.
Measure errors for matching forecasts and observations at one stated lead.
Verified published formula · Matched-pair RMSE
Matching observations = 3; Root mean square error (mm) = 4.655; Average absolute error (mm) = 4.333; Mean forecast minus observation (mm) = 1 Positive mean error means wetter forecasts on average; negative means drier forecasts..
Forecast 48 mm; matching observation 42 mm. Forecast minus observed is 6 mm.
Next step. Review the matched sample scores and its coverage.
The saved observations do not contain matching archived daily point forecasts. Issued seasonal category probabilities cannot be substituted for rainfall amounts.
Error = forecast − observed. For complete matching pairs: RMSE = sqrt(mean(error²)), MAE = mean(abs(error)), bias = mean(error). Match station, variable, accumulation, lead, time zone and observation QC before pooling.
Scores describe the selected sample. They do not establish future forecast reliability or authorise an operational change.
Compare a defined seasonal event with archived probabilities and verifying rainfall.
Verified published component · Binary Brier score and skill component
Matching observations = 3; Brier score (0–1) = 0.137; Reference Brier score (0–1) = 0.336; Brier skill = 0.593 Smaller Brier scores mean smaller probability errors. Positive skill means better than the supplied benchmark on these examples; it is not forecast accuracy..
The supplied 1,040 mm exceeds the example 900 mm cutoff. The squared probability error is 0.09.
Saved seasonal outlooks are source products. They do not include an archive of matched hindcasts and seasonal observations.
For the predefined event, y = 1 when rainfall exceeds the stated cutoff, otherwise 0. Brier score = mean((p − y)²). Brier skill = 1 − score/reference score for matching cases and a positive reference score. Category and benchmark choices are local assumptions.
A published Brier formula does not verify a local category threshold, benchmark or forecast. No complete seasonal calibration assessment is claimed.
Prepare a site-survey worklist from coverage, exposure and operating evidence.
The site-survey checklist identifies coverage, exposure, access and operating evidence for Rambukoluwa demonstration candidate. It does not rank or select a site.
The public station table supplies station identities. It does not establish a coverage-gap study or candidate-site survey.
Review variable-specific siting and exposure against the cited standard using a surveyed site and operating records. No generic distance score or automatic preferred site is generated.
Complete records support a site review. They do not select or authorise a station.
Review due dates, service impact, spares and technician capacity before proposing work.
The task is 29 days overdue. Review service impact, spares and technician availability before proposing work.
Instrument serial numbers, calibration certificates, defects, spares and technician bookings are not in the saved public capture.
For open work, days overdue = max(0, assessment − due date). Compare required effort with available effort in the stated window and detect overlapping bookings for the same resource. Retain completion evidence for closed work.
Overdue age is not a calibration result or a safety priority. A proposed booking does not assign a technician.
Identify the station-history evidence needed before assessing a series break.
Retain the dated station change and both metadata versions. A series-break assessment needs the original and comparison observations; no adjustment is calculated here.
The saved station snapshot has no relocation history, parallel series or statistical break assessment.
Use dated station metadata and the original and comparison series to investigate an apparent break. Preserve originals and any documented adjustment separately. This screen does not homogenize observations.
A metadata change is a reason to investigate, not proof of a climatic or non-climatic shift.
Match crop exposure with weather for the same place and crop period.
Published cumulative departure 16.5% for Jan–4 Nov 2026. This is background rainfall context. Crop stage, planting area and matching local weather are needed before preparing advice.
Saved district cumulative rainfall and the agromet bulletin can provide context. Crop locations, planting dates, stages and matching local weather are still required.
Join crop, stage and planted-area records to suitable weather and forecasts for the same place and period. Use the cited crop protocol to prepare a specialist review; no advice is inferred from a cumulative rainfall percentage alone.
District cumulative rainfall is background context, not a measurement of crop stress or a field-specific forecast.
Check daily weather variables, units and station conditions before an evapotranspiration calculation.
Published snapshot: 24 °C, humidity 92%, wind 0 km/h. Daily extrema, radiation, humidity convention and reference-site wind metadata remain separate inputs. No ETo is calculated.
The saved snapshot supplies temperature, humidity and wind. It does not supply daily temperature extrema, solar radiation, wind height or reference-site metadata.
Check the daily temperature, humidity, radiation and wind definitions required by the selected reference-evapotranspiration method. Wind unit conversion alone does not provide a 2 m wind observation. This screen checks inputs; it does not calculate ETo.
A three-hour observation is not a daily weather dataset. Missing radiation, wind height or daily extrema must remain missing.
Check the marine area, forecast validity and vessel context before a specialist review.
The archived product states 20–30 km/h normal wind and slight or moderate seas. 60–70 km/h temporarily during thundershowers. Check issue, validity and vessel context before a marine specialist review.
The preserved official Marine-page capture supplies the published wind range, sea-state wording and issue time. It does not supply numerical wave heights or vessel-specific clearance.
Retain the original marine product, area, issue time and validity. Keep normal wind, temporary thunderstorm gusts and sea-state terms separate. Compare the intended activity against the appropriate marine protocol.
This is an archived source-review example, not live vessel advice or a new warning. Inland field records do not establish marine conditions.
Reconcile required report delivery and recorded outage intervals for an aerodrome service.
Recorded outage union is 3 hours; time availability is 98.214%. On-time report delivery is not available. The two measures answer different questions.
The public weather capture is not an aerodrome service log. Report schedules, deadlines, outages and operational impact records must be obtained.
Time availability = 100 × (service-window hours − union of outage hours) / service-window hours. Merge overlapping outage intervals. On-time delivery = 100 × distinct qualifying reports / scheduled reports. Match the agreed report schedule and deadline.
Availability and delivery metrics do not establish aviation meteorological compliance or flight safety.
Match an authorised lightning source with the exposure frame used for awareness planning.
Obtain the dated lightning source and a matching exposure frame before reporting a share.
The official site publishes lightning advisories. The saved package has no historical strike dataset or matched household exposure frame.
For a documented matching exposure frame, share = 100 × distinct matched receptors / receptors in that same frame. Count each receptor once. An issued advisory and a historical strike dataset are different source types.
A matched share is not a strike probability or an individual risk rating. This view performs no live spatial intersection.
Compare the same annual statistic at one station with its quality and station-history evidence.
Matching observations = 3; Mean annual temperature (°C) = 25.7; Sample standard deviation (°C) = 0.1.
The supplied 2023 mean temperature is 25.6 °C, with 100% reported daily coverage. No trend or attribution is calculated.
Next step. Review the documented case with the responsible role.
A real annual station series and aggregation/QC history are still required. Annual figures below are labelled demonstration values.
Display comparable annual statistics for one station and definition. Report their mean and sample standard deviation only for complete compatible selected records. No trend significance or climate attribution is calculated.
Three illustrative annual values do not establish a long-term climate trend. The saved three-hour snapshot cannot supply an annual mean.
Compare the same variable, unit and interval from a sensor and an independent reference.
Primary reading: 24 °C. The concurrent independent comparison is incomplete.
Saved station temperatures can populate the primary reading. Independent concurrent readings and instrument/calibration metadata are not in the saved package.
Difference = primary − independent reference only for a matching variable, unit, interval and time convention with instrument and calibration evidence. Temperature differences in °C have no relative percentage.
A difference does not identify a faulty instrument. Preserve both originals and their calibration evidence.
Separate request delay from the permission to release a dataset.
The request is 15 days overdue while open. Recorded release permission: Not recorded. Delay does not grant release rights.
The Department provides a data-contact route. Its internal request, due-date and permission registers are not public source data.
For open requests, days overdue = max(0, assessment − recorded due date). Check permission and scope independently. A late response does not grant release rights.
This prepares a follow-up. It does not transmit a dataset or approve reuse.
Sri Lanka source match. Verify the original warning before sharing.
For a Department warning, check the warning ID, original level, issue time, validity and affected area against the Department’s current record.
Confirm reuse permission and the publication reviewer in Institution review. Public availability does not grant permission.
Check source rights, product version, validity and audience before a publication review.
Original Issued seasonal outlook issued 2026-10-30. Validity as printed: November, December and January (NDJ) 2026/27. SHARED reuse permission, audience and publication review are separate records.
Three saved issued outlooks populate the source records. Conflicting temperature-outlook period labels remain flagged for publisher clarification.
Retain the original product, issue and validity, source version, replacement history, intended audience, review evidence and permitted SHARED use. Public availability alone is not redistribution permission.
The original Department publication is separate from a SHARED publication decision. No publication is authorised or sent by this screen.
Source. Sri Lanka publication.
Data needed. Station and month, SPI accumulation and reference periods, reviewed SPI, monthly rainfall and matching monthly normal.
Method. Compare reviewed SPI for the same month, accumulation period and reference years. Apply the Department’s SPI classes. Separately calculate rainfall as a percentage of normal: 100 × monthly rainfall ÷ matching positive monthly normal.
Worked example. A reviewed 3-month SPI of −1.60 falls in the severely dry band. Separately, 72 mm of monthly rainfall against a 120 mm normal is 60% of normal.
Next step. Ask a climatologist to review comparable stations with dry conditions and verify the inputs before advising users.
Department of Meteorology, Sri Lanka · December 2024 Drought Bulletin. Pages 1–3: Interpretation of Maps, Figure 2 and SPI Classification.
Confirm the matching month, SPI period, reference years and station records in Institution review. The cited bulletin uses 1981–2010 for monthly rainfall normals. Rainfall percentage does not calculate SPI.
Check which existing source fields can supply the decision inputs.
Existing form reference: 678a3e2dcda3b2df
Existing form reference: 4404e8d7fb7938f5
Existing form reference: 34f3d59c2619b006
Existing form reference: 1d5963784380e5d7
Existing form reference: c7953153a5200702
Existing form reference: 2fd8102b7732d899
Existing form reference: 8645ce85db37fa03
Your institution has its own document and image library. Use the same library for reports, photos and other evidence.
Open this screen for Department of MeteorologySee the presentation example
Files added in the prototype stay in this browser. Public sharing is an access preview. Existing public reference links already work online.
Data structure keeps the existing source tree, records, charts and map. Manage structure adds a place to propose changes to definitions.
An entity type means a kind of thing or event, such as a reservoir or an inspection. Drafts stay in this browser. A demo approval does not change real definitions, data, chart links or sharing rights.
User management shows example profiles for each institution. People and Roles and access let you review the proposed account screens.
These are interface previews only. They create no account, send no email and change no access. Use example details in the preview. Institution verification does not grant a real permission.
We currently use public data and published reports where available. Please confirm whether your institution can provide data directly, so we can avoid retyping or reading figures from PDFs.
Do not enter passwords, access keys or private account details.
Use the Download answers button at the top of this page. Share the downloaded file or printed copy through your agreed review channel. Nothing is sent automatically.