El Niño can predict cocoa beans to be harvested two years ahead of schedule

When seasonal rains arrive later in Indonesia, farmers often take it as a sign that it is not wor...

El Niño can predict cocoa beans to be harvested two years ahead of schedule

When seasonal rains arrive later in Indonesia, farmers often take it as a sign that it is not worth investing in fertilizers for their crops. Sometimes they choose not to plant annual crops at all. Usually, they make the right decision, because the late start of the rainy season is usually related to the state of the El Niño Southern Oscillation (ENSO) and insufficient rainfall in the coming months.
The new research published in “Science Reports” shows that ENSO is a weather deformation cycle of warming and cooling along the Pacific Ocean along the equator, and a powerful forecast for up to two years before the cocoa tree is harvested.
This may be good news for smallholder farmers, scientists and the global chocolate industry. The ability to predict the size of the harvest in advance may affect farm investment decisions, improve tropical crop research programs and reduce risks and uncertainties in the chocolate industry.
Researchers say that the same method that combines advanced machine learning with strict short-term data collection on farmer customs and yields can also be applied to other rain-dependent crops, including coffee and olives.
Thomas Oberthür, co-author and business developer of the African Plant Nutrition Institute (APNI) in Morocco, said: “The key innovation of this research is that you can effectively replace weather data with ENSO data.” “Using this method, you can explore anything related to ENSO. Crops with production relations.”
About 80% of the world’s arable land relies on direct rainfall (as opposed to irrigation), which accounts for about 60% of total production. However, in many of these areas, rainfall data is sparse and highly variable, which makes it difficult for scientists, policymakers, and farmer groups to adapt to changes in the weather.
In this study, the researchers used a type of machine learning that does not require weather records from the Indonesian cocoa farms participating in the study.
Instead, they relied on data on fertilizer application, yield, and farm type. They plugged this data into a Bayesian Neural Network (BNN) and found that the ENSO stage predicted 75% of the change in yield.
In other words, in most cases in the study, the sea surface temperature of the Pacific Ocean can accurately predict the harvest of cocoa beans. In some cases, it is possible to make accurate predictions 25 months before harvest.
For starters, it is usually possible to celebrate a model that can accurately predict a 50% change in production. This kind of long-term forecast accuracy of crop yields is rare.
The alliance’s co-author and honorary researcher James Cock said: “This allows us to superimpose different management practices on the farm, such as fertilization systems, and infer effective interventions with high confidence. “International Biodiversity Organization and CIAT. “This is an overall shift to operations research.”
Cock, a plant physiologist, said that although randomized controlled trials (RCTs) are generally considered the gold standard for research, these trials are expensive and therefore usually impossible in developing tropical agricultural regions. The method used here is much cheaper, does not require expensive collection of weather records, and provides useful guidance on how to better manage crops in changing weather.
Data analyst and lead author of the study Ross Chapman (Ross Chapman) explained some of the key advantages of machine learning methods over traditional data analysis methods.
Chapman said: “The BNN model is different from the standard regression model because the algorithm takes input variables (such as sea surface temperature and farm type) and then automatically’learns’ to recognize the response of other variables (such as crop yield),” Chapman said. “The basic process used in the learning process is the same as the process that the human brain learns to recognize objects and patterns from real life. On the contrary, the standard model requires manual supervision of different variables through artificially generated equations.”
Although in the absence of weather data, machine learning may lead to better crop yield predictions, if machine learning models can work properly, scientists (or farmers themselves) still need to accurately collect certain production information and make these Data is readily available.
For the Indonesian cocoa farm in this study, farmers have become part of a best practice training program for a large chocolate company. They track inputs such as fertilizer application, freely share this data for analysis, and keep neat records at the local organized International Plant Nutrition Institute (IPNI) for researchers to use.
In addition, scientists previously divided their farms into ten similar groups with similar topography and soil conditions. The researchers used the harvest, fertilizer application, and yield data from 2013 to 2018 to build a model.
The knowledge gained by cocoa growers gives them confidence in how and when to invest in fertilizers. The agronomic skills acquired by this disadvantaged group can protect them from investment losses, which usually occur under adverse weather conditions.
Thanks to their collaboration with researchers, their knowledge can now be shared in some way with growers of other crops in other parts of the world.
Cork said: “Without the joint efforts of the dedicated farmer IPNI and the strong farmer support organization Community Solutions International, this research would not be possible.” He emphasized the importance of multidisciplinary cooperation and balanced the stakeholder’s efforts. Different needs.
APNI’s Oberthür said that powerful predictive models can benefit farmers and researchers and promote further cooperation.
Obertoor said: “If you are a farmer who collects data at the same time, you need to achieve tangible results.” “This model can provide farmers with useful information and can help incentivize data collection, because farmers will see that they are doing To make a contribution, which brings benefits to their farm.”

suzy@lstchocolatemachine.com

www.lstchocolatemachine.com


Post time: May-06-2021