Heating & Cooling Degree Days for Any Address - No API Key

Give it addresses (or lat/lng) and a base temperature, and it returns heating degree days (HDD) and cooling degree days (CDD) for any period back to 1981, month by month, from NASA POWER. No API key, worldwide. Built for energy managers, M&V and ESG consultants, facility teams and anyone normalising energy bills for weather. Optionally fits kWh against degree days for you.

Example inputs

1. One address, US convention (65 F base), full year

{ "locations": ["350 5th Ave, New York, NY"], "startDate": "2024-01-01", "endDate": "2024-12-31", "unit": "F", "baseTemperature": 65, "coolingBaseTemperature": 65 }

2. Several sites in Celsius, UK/EU style bases (15.5 C heating, 22 C cooling, whole numbers)

{ "locations": ["51.5074,-0.1278", "48.8566,2.3522", "Berlin, Germany"], "unit": "C", "baseTemperature": 16, "coolingBaseTemperature": 22, "startDate": "2023-01-01", "endDate": "2023-12-31" }

3. Weather-normalise a building: regression of monthly kWh against HDD

{ "locations": ["40.7128,-74.0060"], "startDate": "2024-01-01", "endDate": "2024-12-31",
  "consumption": [ {"period":"2024-01","kwh":18200}, {"period":"2024-02","kwh":16900}, {"period":"2024-03","kwh":14100}, {"period":"2024-04","kwh":11800}, {"period":"2024-05","kwh":9800}, {"period":"2024-06","kwh":10400} ] }

Input guide

Field	What it does
locations	Addresses, place names, or "lat,lng". Coordinates skip geocoding.
startDate, endDate	YYYY-MM-DD. Days not yet published by NASA (the last few days) show up as daysMissing.
unit	F or C. Degree days come out in that unit.
baseTemperature	Heating balance point (whole number). coolingBaseTemperature is the cooling one.
includeMonthly / includeDaily	Add a month-by-month array / the daily values to each row.
consumption	Monthly kWh to regress (needs 4+ complete months). Add "location" to target a site when you list several.
regressionUseCooling	Fit baseload + a x HDD + b x CDD (needs 6+ months).

Output

One location-period row per location (one regression-fit row per fitted model, and error rows that are never charged).

Field	Meaning
resolvedAs, lat, lng, geocoder	What the address turned into
hdd, cdd	Totals for the whole period, in the chosen unit
monthly[]	month, hdd, cdd, meanTemp, daysWithData, daysMissing
daysWithData, daysMissing	Completeness of the period
regression row	baseloadKwhPerMonth, kwhPerHdd, kwhPerCdd, r2, cvRmsePct, nmbePct, meetsAshrae14Monthly, fitted[]

Sample row (trimmed): {"type":"location-period","location":"40.7128,-74.0060","unit":"F","baseHeating":65,"baseCooling":65,"hdd":4301,"cdd":890.4,"daysWithData":366,"daysMissing":0}

Use it with

• Google Maps scraper or a CRM export of building addresses -> this Actor -> weather-normalised energy benchmarking.
• Utility-bill extractors -> consumption -> this Actor's regression -> baseload and weather sensitivity per site.

FAQ

Where does the weather come from? NASA POWER (MERRA-2 reanalysis), the daily mean 2 m air temperature of the grid cell around your point (about 50 km wide). It is free, key-less and NASA open data. It is not a weather station. How close is it to degreedays.net or a station? Usually within a few percent for annual totals, but a grid cell smooths out urban heat and hills, and a daily-mean method differs slightly from min/max or hourly methods. Use it for screening and portfolios, and verify with station data where an audit requires it. Why not Open-Meteo? Its free tier is non-commercial, so this Actor does not use it. Geocoding? OpenStreetMap Nominatim at 1 request per second with Photon as a fallback; (c) OpenStreetMap contributors, ODbL. Pass lat,lng to avoid it. Is the regression M&V-grade? It is a monthly least-squares screening fit and reports ASHRAE Guideline 14 style CV(RMSE) and NMBE, but a formal M&V report needs your own validation.

Limits

Grid-cell temperature, not station data. Daily mean method only. Up to 15 years per location per run. Latest days may be missing. Monthly regression only.

Price

Pay per event: location-period $0.002 per location, regression-fit $0.01 per fitted model. Example: 50 sites for a year = $0.10, plus $0.01 for each site you regress. Failed locations are free.