What we do?

We deliver location-specific daily ET estimates.

The FineET project provisions a scalable and publicly-accessible digital ecosystem of integrated data and model ensembles to generate high-precision evapotranspiration in support of effective daily decision-making at the site-level. FineET is funded by the National Institute of Food and Agriculture.


What is ET?

Evaporation refers to water evaporated from the land surface including soil, wetlands, and ocean – and even the leaf surface of a plant. Transpiration happens when plants release water from their leaves during the chemical and biological changes that occur as the plant undergoes photosynthesis and converts carbon dioxide into oxygen.

  • Evapotranspiration is a main component of the water cycle and is important in crop maintenance.
  • Evapotranspiration is responsible for 15% of the atmosphere’s water vapor.
  • Evapotranspiration is derived from a combination of two water processes: evaporation and transpiration.

How does daily evapotranspiration estimate relate to agriculture?

If you can estimate evapotranspiration rates, you will be able to determine whether you need to irrigate or not. If plants do not receive enough water, the plants will fight to conserve as much as water they can and this is reflected in the reduced evapotranspiration rate.


The FineET project encorporates several services that are aligned to support farm-level decision-making and planning precise water irrigation.

Inputs. Generating ET values requires certain pre-processing of input features such as  developing a cloud removal model that will remove cloud coverage from satellite imagery and accurately predict the land area underneath. Cloud coverage in satellite imagery can make gathering accurate information difficult, so removing cloud coverage can improve the quality of the information being gathered. These models use image-to-image translation Generative Adversarial Networks with U-Net and ResNet architectures.

Machine Learning solution. Our goal is to to provide daily ET estimates at high-spatial resolution of 30m. We leverage super-resolution GAN models to overcome low-spatial frequency of MODIS sensor’s thermal reflectance values and low temporal resolution of LANDSAT 8 sensor capturing Earth image once in two weeks. Our final product is an heterogeneous empirical model that incorporates the process-based model, Simplified Surface Energy Balance (SSEBOP)  for calculating evapotranspiration with Machine learning based algorithms. The model with fit the data and produce ET estimates using CNN and Random Forests. 

ET Data Exploration and Visualization

FineET provides interactive and flexible data exploration to users with the most recent ET estimates for a particular area. Automatically generated daily ET estimates are available at the spatial resolution of 30m. To generate ET estimates at a high spatiotemporal resolution, FineET harnesses Deep Learning based spatiotemporal data imputation over massive satellite imageries that are acquisitioned from multiple satellite systems from NASA, USGS, and ESA. These models use image-to-image translation Generative Adversarial Networks, U-Net, and ResNet architectures.

Advanced Cyberinfrastructure for Modelers

FineET is a scientific cyber infrastructure designed for modelers. The synthesized data with high spatiotemporal resolution is available for modelers and are ready-to-use for scalable testing and execution. A containerized distributed cluster is available with the streamlined data access to the critical satellite images and weather observations. Modelers are allowed to monitor their computation using our web-based job dashboard.

Education and Training

FineET provides a rich set of training materials for the researchers, students, and stakeholders in the areas of ecology, agriculture, and data science. You can learn how to use FineET, what to retrieve from FineET, and how it works from a series of short video clips, on-line workshops, and academic publications.



MODIS/Terra Land Surface Temperature/Emissivity Global Data

MOD11A1 and MOD11A2 datasets provides daily and average 8-day per pixel Land Surface Temperature and emissivity at 1 kilometer spatial resolution across US region.

National Land Cover Database

The National Land Cover Database (NLCD) provides nationwide data on land cover and land cover change at a 30m resolution with a 16 thematic class.

NLCD 2016
Landsat USA

Landsat 8 Surface Reflectance

Available at 30m resolution for R,G,B bands and at 100m for two thermal infrared (TIR) bands processed to orthorectified brightness temperature. Landsat satellite provide atmospherically corrected Earth images in every two weeks with multispectral and thermal data. Dataset is available from April 2013 till Present.

Landsat 7 Surface Reflectance

This dataset is also the atmospherically corrected surface reflectance from the Landsat 7 ETM+ sensor. These images contain 4 visible and near-infrared (VNIR) bands and 2 short-wave infrared (SWIR) bands and one thermal infrared (TIR) band processed to orthorectified brightness temperature. The TIR band is collected with a resolution of 60m/pixel. The satellite timespan is from January 1999 to Present.

Landsat 5 Surface Reflectance

This dataset is the atmospherically corrected surface reflectance from the Landsat 5 ETM sensor. The TIR band was collected originally with a resolution of 120m/pixel has been resampled using cubic convolution to 30m. The dataset is available from March 1984 – May 2012.


GridMET provides daily high-spatial resolution (~4-km, 1/24th degree) surface meteorological data covering the contiguous US from 1979-present. These data can provide important inputs for ecological, agricultural, and hydrological models. It captures Primary Climate Variables such as max/min temperature, precipitation accumulation, downward surface shortwave radiation, wind-velocity, humidity (max and min relative humidity and specific humidity) along with Derived variable Reference evapotranspiration.

gridMET max temperature

NOAA Temperature Map

Soumi National Polar-Orbiting Patnership (NOAA)

This dataset includes observations from World Meteorological Organization. The observations recorded are daily maximum and minimum temperature, temperature at the time of observation, snowfall and snow depth and other  meteorological elements. These station-based measurements are collected from over 90,000 land-based stations worldwide.


AmeriFlux is a network of PI-managed sites measuring ecosystem CO2, water, and energy fluxes in North, Central and South America. Established to connect research on field sites representing major climate and ecological biomes, including tundra, grasslands, savanna, crops, and conifer, deciduous, and tropical forests.

Soil Maps

Produced by The National Cooperative Soil Survey (NCSS), provides access to the largest natural resource information system in the world.