Abstract
Abstract. Increasing the accuracy of crop yield estimates may allow improvements in the whole crop production chain, allowing farmers to better plan for harvest, and for insurers to better understand risks of production, to name a few advantages. To perform their predictions, most current machine learning models use NDVI data, which can be hard to use, due to the presence of clouds and their shadows in acquired images, and due to the absence of reliable crop masks for large areas, especially in developing countries. In this paper, we present a deep learning model able to perform pre-season and in-season predictions for five different crops. Our model uses crop calendars, easy-to-obtain remote sensing data and weather forecast information to provide accurate yield estimates.
Highlights
In 2050, the world’s population is expected to reach 9.7 billion (DESA, Affairs), it represents approximately 2 billion more people in the 30 years
The results suggest the combination of satellite and climate may lead to better results, in practice these results can be only achieved if a reliable crop calendar and annotated farm locations are available
Different from previous work, which leverage satellite data for direct farm observation for yield prediction, we employ weather and soil data, which is computationally cheaper to process than large satellite images
Summary
In 2050, the world’s population is expected to reach 9.7 billion (DESA, Affairs), it represents approximately 2 billion more people in the 30 years. There are two main drawbacks when using NDVI to predict yield: (i) the planting should be executed prior to NDVI acquisition, and (ii) for large regions it is hard to obtain a reliable crop calendar definition and to determine where each crop was planted (i.e., crop mask) especially in developing countries Another important approach for yield prediction corresponds to the utilization of Crop Simulation Models (CSM) such as DSSAT, WOFOST, PCSE, and APSIM (Jin et al, 2018). The proposed model utilizes customized weather data according to region-specific crop calendar without changing the model feature set. In this way, we do not need multiple models for different regions, improving model prediction power and scalability.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
