Cimilada

Cimilada (Somali for "the weather") is a free, public weather forecasting dashboard for Somalia, in English and Somali: a 7-day hazard outlook in WMO impact-based colours with CAP 1.2 alerts, forecast maps from NOAA GFS, ECMWF IFS and ECMWF AIFS, two-week ensemble town forecasts, GloFAS river flow, port wave outlooks, SWALIM station water balance and the Deyr seasonal outlook.

Live: https://cimilada.github.io/

The page is a single static HTML file with its data built in. A GitHub Actions workflow refetches the forecasts every morning (07:40 Mogadishu time), rebuilds the page and deploys it to GitHub Pages. If a required source fails, the run stops and the last good page stays online.

Contact

Questions, corrections and Somali translation fixes: [email protected]

Pipeline

Step	Script	Runs
Forecast grid, satellite clouds, cyclones, terrain, imagery	fetch_wx.py	daily, CI
Town ensemble, heat, dust, GloFAS rivers, port waves	fetch_fc.py	daily, CI
Inject data into the template	build.py	daily, CI
SWALIM station data and seasonal products	scrape.py → process.py → fetch_v2.py → process_v2.py	locally, then commit site.json and site2.json

Run everything locally with ./refresh_forecasts.sh (needs pip install -r requirements.txt).

Data and licences

This is an unofficial, non-commercial project. Official warnings for Somalia come from the Somali Disaster Management Agency (SoDMA) and FAO SWALIM.

• FAO SWALIM ClimSeries station data and FRRIMS river levels. Owned by FAO SWALIM.
• CHIRPS v3 and CHIRPS3-GEFS, Climate Hazards Center, UC Santa Barbara.
• NOAA GFS (public domain); ECMWF IFS, AIFS and SEAS5 (CC BY 4.0), via Open-Meteo (free for non-commercial use).
• GloFAS and CAMS, Copernicus Emergency Management and Atmosphere Monitoring Services.
• Meteosat imagery, EUMETSAT (free for personal, research and educational use).
• IBTrACS v4, NOAA NCEI; GDACS (EC JRC, UN OCHA).
• Sentinel-2 cloudless 2016 by EOX IT Services (CC BY 4.0); AWS Terrain Tiles.

Ideas for the weather layers follow God's Eye View (MIT).

Visitor and feature analytics

The page can count visits and feature use with GoatCounter (free for non-commercial sites, no cookies, no personal data) and/or Umami Cloud (free up to 100k events a month). Nothing is loaded unless configured. Set repository variables under Settings → Secrets and variables → Actions → Variables, then re-run the workflow:

Variable	Example	Effect
GOATCOUNTER_CODE	cimilada	loads GoatCounter for https://cimilada.goatcounter.com
UMAMI_WEBSITE_ID	a1b2c3d4-…	loads Umami Cloud
SITE_URL	https://cimilada.github.io/	address used in "Copy link to this view"

Tracked events (one count per action per 30 s, at most 80 per visit): map layer, field, model, unit, season, station, overlays, 2D/3D view, timeline play, point forecast, towns and town tabs, hazard cards, CAP alerts, raw data and CSV copies, share links, section links and outbound links.

Custom address

The site is served from GitHub Pages. Free ways to give it a "cimilada" address:

1. cimilada.github.io: create a free GitHub organization named cimilada, transfer this repository to it and rename it cimilada.github.io. No DNS needed, HTTPS automatic.
2. cimilada.dpdns.org / cimilada.qzz.io via DigitalPlat FreeDomain: needs an account with real contact details and an external DNS host (e.g. Cloudflare). Then add GitHub Pages A records (185.199.108–111.153) and set the custom domain in Settings → Pages.
3. cimilada.eu.org via nic.eu.org: free and permanent, manual approval that can take weeks; needs nameservers first.
4. cimilada.com (or .org) costs about US$10 a year and is the most dependable.

Search engines and icons

build.py (with PAGES=1) writes the public site with the metadata that jekyll-seo-tag and jekyll-sitemap generate for GitHub Pages sites, following Google Search Central:

• one address per language, / (English) and /so/ (Somali), linked with hreflang and a canonical URL; the language switch moves between them
• title, description, Open Graph and Twitter cards per language
• schema.org JSON-LD: WebSite, Organization, WebApplication and Dataset (the Dataset entry makes the forecasts eligible for Google Dataset Search)
• robots.txt, sitemap.xml (daily lastmod, language alternates), a <noscript> summary
• favicon.ico, SVG and PNG icons, Apple touch icon and site.webmanifest, so phones can install the site like an app

To verify ownership in Google Search Console or Bing Webmaster Tools, choose the "HTML tag" method and save the code as the repository variable GOOGLE_SITE_VERIFICATION or BING_SITE_VERIFICATION, then re-run the workflow and submit sitemap.xml.

Fires, flood water and daily imagery (NASA)

fetch_wx.py adds three NASA layers, all keyless and refreshed daily:

• Fires: active fire detections from NASA FIRMS (VIIRS 375 m on NOAA-21, NOAA-20 and S-NPP; MODIS 1 km), last 7 days, from the open global CSV files. The map shows 24 h, 48 h or 7 days; the Fires hazard card and statistics count detections inside Somalia only.
• Flood water: the VIIRS 2-day flood product from NASA GIBS, reduced to its flood classes.
• Latest satellite image: VIIRS NOAA-21 true colour for the newest complete day.

We acknowledge the use of data and imagery from NASA's Fire Information for Resource Management System (FIRMS), part of NASA's Earth Science Data and Information System (ESDIS).

Vegetation, sharing and offline use

• Vegetation: VIIRS NOAA-20 8-day NDVI from NASA GIBS, the greenness index FEWS NET uses to follow drought and pasture; fetch_wx.py keeps the newest composite with full coverage.
• Share on WhatsApp: the hazard summary and each town's 7-day forecast can be sent as plain text (English or Somali) with a link back to the page.
• Offline: on the public site a service worker (sw.js, written by build.py) keeps each page a visitor opens, so it still loads without a connection and says how old the forecast is. Only small files are precached, so a first visit downloads nothing extra.
• Desert locust layers were investigated (FAO Locust Hub) but its services now need a token and its open extract ends in 2020, so the page links to FAO Locust Watch instead.

Weather now (after Google Weather)

A town card modelled on the features in Google's How Google Weather works:

• Current conditions: temperature, feels-like, sky, humidity, wind and gusts, cloud, UV index (WHO bands), sunrise and sunset. "Now" is the hourly forecast for the viewer's current hour in Somalia, so it stays right through the day.
• Next 12 hours: hour-by-hour chance and amount of rain with a plain summary. Somalia has no weather radar, so this comes from forecast models (Google's nowcast also uses radar).
• Hourly (24 h) and 10-day strips.
• Air quality: US AQI with the EPA category and health advice; PM2.5, PM10, dust, ozone (CAMS).
• Pollen: not shown; no open pollen forecast covers Somalia.
• Record and unusual temperatures: the week's forecast highs against 1991–2025 NASA POWER normals and records for the time of year, corrected for each town's model bias over the past month (fetch_fc.py stages now and clim; data/fc/clim.json is fetched once).
• "What's the weather where you are?": visitors' reports arrive as Umami events (weather-report, e.g. Bosaso-dusty).