Universal Model for Multi-Factor Forecasting and Estimation of ETo
Description
Evapotranspiration is a pivotal process in the ecosystem, and accurately forecasting short-term daily reference crop evapotranspiration (ETo) is of paramount importance for real-time irrigation decision-making and regional water resource allocation. In the provided "research" folder, there are four subfolders. The "data" folder contains weather forecast data and historical observation data for 100 sites involved in our research, available for readers to use and debug. The "Quantitative transformation model" folder offers deep learning models we have trained to transform qualitative data into quantitative data. These models can convert the quantitative data from weather forecasts—specifically weather conditions and wind speed ratings—into specific values of radiation and wind speed. Model inputs need to include the latitude and longitude of the location being used, the day number for conversion, and the quantitative data. The "Estimating model" folder includes simulation models provided based on high-precision meteorological data. Inputs E1-E5 correspond respectively to temperature, temperature and radiation, temperature and wind speed, temperature and coordinates with day number, and temperature along with all aforementioned inputs. Among these, E5, which combines all inputs, has the highest model accuracy. Taking GRU as an example, its RMSE is between 0.13-0.14 mm/d, R2 ranges from 0.98-0.99, MAE is between 0.08-0.09, and NRMSE is within 0.018-0.019. Lastly, the "Forecasting model" folder provides corresponding prediction models for 1, 4, 7, and 15 days. Since our study area is China, all the training data used are sourced from regions within China, with specific models incorporating information such as latitude and longitude. Therefore, these models cannot be directly applied to other regions. The obtained ETo (Evapotranspiration) data can be utilized according to your needs in remote sensing systems or crop models for applications in irrigation management and smart agricultural production. This tailored approach ensures that the data is most effective within the specified geographical context, facilitating precise and efficient agricultural practices.