DAILY ESTIMATOR
Use the measurements you have. Explore the estimate, its interval, and the model’s response.
Enter temperature, date, and location to begin. Add radiation, humidity, and wind when available.
The same available inputs, evaluated by different estimators. Agreement is not independent validation.
Both temperatures change together; other inputs stay fixed. A sensitivity check, not a climate forecast.
ET₀ describes a well-watered short-grass reference. Crop coefficients, rainfall, soil water, and irrigation efficiency are needed to assess irrigation requirements.
WEATHER SERIES
Upload a CSV, inspect the estimates and intervals, and download every result.
Each row may contain a different measurement pattern.
Up to 10,000 rows / 10 MB. Sent securely to this server for calculation; not stored by this app.
30 observed Aswan days using temperature and wind. Radiation and humidity are withheld to demonstrate the TW model.
Download CSV templateYour series visualization will appear here.
Series shading shows daily 90% calibration intervals, not joint limits for a season or irrigation demand.
GEOGRAPHY AND TRANSFER
Follow the global weather coverage, then inspect real held-out and later-year results.
Blue circles: global grid locations. Rings: held-out grid locations. Teal triangles: observed stations. A map symbol shows where data were evaluated, not a guarantee of local accuracy. Base map: Natural Earth.
GLOBAL LOCATION
Mean daily errors against calculated grid-weather FAO-56 ET₀, 2023–2024. T is temperature only; TR also includes radiation; TW also includes wind.
Download a temperature–radiation exampleMonthly mean of actual daily test outputs. Black: calculated reference. Blue: global mixed TR model. Dashed teal: original v2 TR model. This is a historical comparison, not a future forecast.
The expanded radiation-assisted model improves mean station RMSE in the Japanese and Brazilian tests. The temperature-only model becomes less accurate there. Gridded weather has different local variability and exposure from station measurements; more locations alone do not ensure better transfer.
Tongxiang and Xiamen are NASA POWER grid cases. Complete independent station weather was not obtained for those cities. Their inclusion supports testing across Chinese weather conditions, not a claim of observed station validation.
ABOUT ET₀ ATLAS
Daily atmospheric reference water demand estimated from incomplete weather observations.
Date, latitude, elevation, and minimum/maximum temperature are required. Eight models cover the additional measurements supplied: radiation (R), mean relative humidity (H), and wind (W).
The expanded v3 model combines 41 Arizona/German station locations with 41 global NASA POWER grid locations for training in 2015–2020. Separate 2021–2022 data calibrate the intervals. Nineteen global locations, including Tongxiang and Xiamen, were withheld. The original v2 station models remain available in the method selector.
With complete weather, selected-model mode calculates FAO-56 Penman–Monteith directly.
60 gridded locations · 41 countries · six continentsNASA POWER broadens weather coverage. These are gridded case studies, not observed weather stations.
17 new station tests · 11,044 complete daysJapan, the Netherlands and Brazil provide independent station records for 2023–2024. Coverage does not imply equal accuracy everywhere.
50 stations · Arizona & GermanyLater years and unseen spatial groups test temporal and geographic transfer.
20 additional US stationsIndependent networks test different climates.
6 Egyptian stations · 7,171 complete daysObserved EMA/SUSIE weather from 2004–2010; time and wind-height sensitivities are documented.
All targets are calculated ET₀. Regional performance and interval coverage differ substantially. This is not validation against measured field evapotranspiration.
THE REFERENCE EQUATION
TMean of Tmin and Tmax, °C
RₙNet radiation, MJ/m²/day
u₂Wind at 2 m, m/s
eₛ − eₐVapour pressure deficit, kPa
ΔSaturation vapour-pressure slope
γPsychrometric constant
Daily soil heat flux G = 0; reference albedo = 0.23. Actual vapour pressure uses mean RH and the average saturation vapour pressures at Tmin and Tmax. Wind is converted to 2 m when another height is supplied. Clear-sky radiation ratios are capped at one; reported ET₀ is bounded below by zero.
COMPACT MODE
ET₀ = max(0, baseline + 5 × Σ coefficient × term). The baseline is zero for direct equations or missing-data PM where marked. Complete weather uses direct FAO-56. Frozen coefficients are displayed to six significant digits.
HS is Hargreaves–Samani ET₀; Rₐ is extraterrestrial radiation from latitude and day of year. Scaling uses the units in the input form.
The 90% interval describes empirical error against the calculated reference in held-out calibration years. Coverage is not guaranteed at new locations. Weather ranges and proximity to development data are checked; latitude is reported separately.
Full-input PM has zero emulation error against itself. This does not mean the weather measurements or physical evapotranspiration have zero uncertainty.
AZMET · DWD · CONUS-AgWeather · Egyptian EMA/SUSIE data
The research package supplies code, data provenance, exact model datasets, equations, and sensitivity analyses. Future-climate experiments are separate analyses; this calculator is not a weather forecast.
Weather inputs are sent over HTTPS to this application server for calculation. This app does not save the submitted weather or send it to a separate prediction service. The hosting provider may retain normal connection metadata.