Back

Reference ET Specification

freeET computes reference ET itself, every day, for any ZIP or postal code in the USA and Canada. It uses the ASCE Standardized Penman-Monteith method, the same one FAO-56 defines and irrigation science has used for 30 years, fed by seven weather inputs for that exact point. A second, independent source is printed beside the number as a check. Past days come from observed weather; today comes from forecast weather and says so.

Download the One-Page PDF

What Goes In: Seven Inputs, Every Day

InputWhat it does
High and low temperatureSets how much moisture the air can hold
HumiditySays how much it already holds; the gap is the pull
WindCarries moist air away; measured at 10 m, adjusted to 2 m
Solar radiationThe energy that evaporates water; the largest driver
ElevationAir pressure, and the most sun physically possible
Latitude and dateDay length and the sun's angle

All weather inputs come from one request per place: 31 days back, 16 days ahead, in the place's own time zone. A day with any input missing is not computed. No guessing, no filling gaps.

Solar Radiation, the Energy Input

What it is. The sun's energy reaching a flat patch of ground over the whole day, in megajoules per square metre. One unit evaporates about 0.4 mm of water. At 45°N a clear June day brings 25 to 28; the same day under heavy cloud, 8 to 10; a December day, 3 to 6. In most places for most of the year it decides more of the ET than any other input.

Where it comes from. Open-Meteo's daily shortwave radiation: the weather model's own estimate of sun at the surface, built from its cloud forecast and, for past days, its analysis of the day. It is model output, not a measurement. No pyranometer, weather station or satellite retrieval is in the chain today.

What we do with it.

  1. 1Cap it at clear sky. We work out the most sun possible for that latitude, date and elevation. A value above it is a data error and is clipped.
  2. 2Take off what grass reflects. About a quarter bounces back; 77% is absorbed and drives evaporation.
  3. 3Use the cloudiness twice. Actual sun against clear-sky sun says how cloudy the day was, which also sets how much heat the ground radiates away at night. Clear days lose more.
  4. 4Net energy drives the number. Sun in, minus heat out, becomes the radiation share of the day's ET; wind and dry air add the rest.
~0.18 mm
of ET per unit of sun on a 21 °C day
3.4 mm
clear vs overcast June day, same heat, humidity and wind
10 to 20%
typical forecast error in cloud; why today is tagged

What it means. Two days with the same high, humidity and wind can differ by half their ET on cloud alone, and a controller that runs on temperature never sees it. Past days are better than forecasts but still modelled; expect a few percent bias and 5 to 10% day-to-day scatter against a good pyranometer. Shade, slopes and lake-effect cloud below the model's grid are not resolved. A station at the site would replace the sun input with a measurement under the same method; that is planned, not built.

Where the Data Comes From

Open-MeteoWeather in
Supplies all seven inputs. National weather-model output (NOAA, ECMWF, DWD, Canada's GEM), interpolated to the point. Covers the USA and Canada.
gridMETMeasured check
Independent check, never the number. A 4 km daily weather grid from the Climatology Lab, University of California Merced. Contiguous US, one to two days behind.

How We Tell Time

  • Observed. Any day before today, computed from the record of what happened. Solid bar on screen.
  • Forecast. Today and later, computed from forecast weather. Dashed bar and a tag until the day ends. The two are never averaged together.
  • “So far today” adds up this morning's hourly ET to the current hour, as progress toward the forecast.

How We Know It Is Right

  • The code reproduces the FAO-56 handbook's own worked example (Bangkok, 6 July: 5.72 mm/day) within 1%, plus its examples for humidity, wind and sun angle. These tests run on every build.
  • Live at Traverse City, MI, in October 2026: the week's total within 7 to 10% of gridMET's, the expected gap between two weather models feeding one method.
  • Next step: a side-by-side against a pyranometer-equipped reference station (CIMIS, AgriMet, Enviroweather).
Reference ET is for a short, well-watered lawn. Plants use a share of it; the watering step applies a plant factor the way the EPA WaterSense water budget does.
Download the One-Page PDF
FAO-56ASCE-EWRI 2005gridMETOpen-MeteoAbout freeET

Rev 1, October 8, 2026. Harmony Analytica · freeET.app. Results are recommendations, not a guarantee of plant health or water use.