Skip to main content

Mapping the World's Sunflowers

field of sunflowers

While not often mentioned alongside major commodity crops like wheat, maize, rice, or soybean; sunflower seeds are major sources of oil and meal and make up a significant portion of global food use. Understanding how these important crops are faring throughout their growing cycle is vital in ensuring market transparency and availability for exporting and importing countries.

 

A team from NASA Harvest, NASA's Global Agriculture and Food Security Consortium is developing methods to improve how the world monitors global sunflower production with special attention to Ukraine and Russia which are the largest producers and exporters of sunflower.

 

Traditional crop monitoring systems involve labor and financially intensive methods like ground and phone surveys. These can be prohibitively expensive for many countries to undergo while also being difficult to conduct in regions experiencing conflict, natural disasters, or disease outbreak. The use of remote sensing, or the observation of the Earth’s surface using Earth observation (EO) satellites) can overcome these challenges and allow governments, markets, and humanitarian organizations to better understand the location, crop type, health, and yield of the world’s cropland. NASA Harvest has integrated EO data into a variety of models and open access platforms, including ARYA and AGMET.

 

Most agricultural monitoring models are trained on specific areas of the globe. Transferring these models to other regions can be difficult as it often requires new datasets and retraining of the model. Given the importance of monitoring global agricultural production. There has been increased interest in developing generalizable models that can be trained in one region and successfully applied in other regions.

 

Developing these models can be difficult as the spectral properties of agricultural vegetation can vary from year to year and in different parts of the world depending on factors like climate, weather, and environment. The Harvest team was able to overcome these challenges however by exploiting a particular characteristic of sunflowers­ (the same characteristic that gives them their name) and observed it using Synthetic Aperture Radar (SAR),commonly referred to as radar. Sunflowers demonstrate directional behavior where their large flower-head faces east permanently when fully grown.

 

Figure. Directional behavior as observed in the sunflower (Helianthus annus) (top) due to the large flower head facing east permanently in comparison to the maize crop (bottom), for descending and ascending pass.
Figure. Directional behavior as observed in the sunflower (Helianthus annus) (top) due to the large flower head facing east permanently in comparison to the maize crop (bottom), for descending and ascending pass.

 

The European Space Agency's Sentinel 1 remote sensing satellite actively sends radar waves towards the surface of the Earth while it orbits the planet. These radar waves bounce off objects on the surface of the Earth back up to the satellite sensor where Sentinel 1 records the amount and direction of reflected radar waves. Given the directional behavior of sunflowers, the recorded radar waves will vary in a predictable pattern as they're collected at different times of the day and thus the Harvest team's model can determine where sunflowers fields are located.

 

The model developed by the Harvest team is one such model. It was trained on field data collected in 2022 from sites across Ukraine. It was then tested for accuracy across space and time. To do this, the team fed the model with satellite imagery from previous years and from different countries. 

 

Sunflower fields were then predicted across Ukraine for the years 2018-2020 as well as in Hungary, France, Russia, and the United States for the year 2018. An accuracy assessment found the model was able to predict sunflower field location with high accuracy across the given years and geographic regions, showing the successful generalizability of the model. Going further, the NASA Harvest team used the predicted cropland to create accurate estimates of sunflower planted area. These results show the promise of new models that can be generalized for a wide variety of regions and time periods, opening up new avenues of agricultural monitoring.

 

“When we talk about SAR, the only capability that is generally discussed is its cloud-free, day-night imaging capability. However, the unique way SAR sensors interact with crops compared to optical sensors is rarely mentioned.” said NASA Harvest’s Abdul Qadir, lead author of this research.

 

“This model provides a unique capability to detect SAR signal interactions with sunflower flower-head, which is not possible with optical sensors. The model is now operational and has been used to assess changes in sunflower crops in Ukraine without relying on field labels during the Russian aggression against Ukraine. We are now expanding this model to map sunflowers in major sunflower-producing countries across the globe, including Russia. This year, we are also partnering with Kernel, the leader in the sunflower oil market, to provide sunflower estimates in Ukraine.

 

This open access research is presented in the journal Remote Sensing of the Environment, A generalized model for mapping sunflower areas using Sentinel-1 SAR data.

News Date
May 15, 2024