A recent study conducted by researchers at North Carolina State University (NC State) has made significant advancements in the realm of agricultural productivity by combining satellite imagery with machine learning technology to accelerate and enhance the accuracy of rice crop modelling. This innovative approach holds particular relevance for global decision-makers involved in rice cultivation, a staple food source for over half of the world's population.
Conducted with a focused analysis on Bangladesh, the study underscores the country's prominent position as the third-largest rice producer globally while also spotlighting its vulnerability to climate change impacts. Bangladesh is notably at risk of flooding, which has a detrimental effect on rice crops and exacerbates food insecurity. Varun Tiwari, a doctoral student at NC State and lead author of the study, elaborated on the limitations of traditional crop monitoring methods that rely heavily on field data collection. He stated, "In order to estimate crop productivity, people in Bangladesh use field data. They physically go to the field, harvest a crop and then interview the farmer, and then build a report on that. It is a time-consuming and labor-intensive process."
The reliance on limited field samples has hindered accurate national-level crop assessments, posing challenges for decisions regarding exports, imports, and crop pricing while impairing long-term strategic planning essential for climate resilience and agricultural adaptation. The study proposes a robust solution through the application of time series satellite imagery, capturing multiple images of the same geographical area at regular intervals. This method assesses essential factors such as vegetation growth, crop water content, and soil conditions in combination with existing field data.
Through this data fusion, researchers trained their machine learning model to generate more precise estimates of rice productivity over a two-decade span, from 2002 to 2021. Tiwari explained the potential implications of their findings: "With this model, we can see for instance that one area is doing well and another area is not doing as well as it needs to. If we have a highly productive area, we can decide to build more storage capacity in that area or invest more in transportation there."
Initial results from the model are promising, showing accuracy levels between 90-92 percent with a margin of uncertainty of around 2 percent. Tiwari noted that as the model is further developed, it could be adapted for diverse crops across various landscapes, indicating its versatility and scalability. "If we can get similar datasets from other regions, we can apply this same framework there. Whether it's the U.S., India or an African country, we want this method to be reproducible," he stated.
This research has been bolstered by collaborations with multiple stakeholders, including the U.S. Department of Agriculture, the International Maize and Wheat Improvement Center, and the Bangladesh Rice Research Institute, to utilise the best scientific methods for producing actionable insights in agricultural policymaking. The study has garnered funding from the Bill and Melinda Gates Foundation alongside USAID, as part of broader initiatives to enhance agricultural resilience in South Asia.
The findings are published in the paper titled "Advancing Food Security: Rice Yield Estimation Framework using Time-Series Satellite Data & Machine Learning" in the journal PLOS ONE. The collaborative research team includes experts from various fields, further enriching the study's conclusions and implications for global agricultural practices.
As the world wrestles with the dual challenges of food security and climate change, innovative applications of technology in agriculture, such as those developed at NC State, may play a crucial role in shaping sustainable future practices.
Source: Noah Wire Services