Banana (Musa spp.) is the world’s fourth most important food crop and of critical importance to the food security and income generation of more than 70 million Africans. There are many types of genetically distinct banana types, including sweet bananas and various types of cooking bananas (aka Plantains). East African Highland bananas (EAHB) of the triploid AAB group are threatened by a bacterial disease (Xanthomonas wilt, BXW) spreading in Uganda, Kenya, Tanzania, Rwanda and Burundi. Sweet bananas of the AAA triploid group, including the Cavendish cultivar (the most globally consumed banana) are also grown in many African countries for domestic markets and for export. Cavendish bananas around the world are seriously threatened by fungal Fusarium wilt (aka Panama Disease), a fungicide resistant soil disease. Discovery of Fusarium in Mozambique in 2013 has sparked concern that the disease may spread to other East African countries.
Various types of banana are unique to Africa, and these can be eaten fresh, cooked, fried and processed. In East Africa, EAHB are used to make matoke – commonly referred to as cooking bananas. The EAHB is a major source of income for many farmers, generating up to 1500 USD per hectare per year from fresh banana sales. EAHB are threatened by BXW, a bacterial disease that was first reported in Ethiopia (Yirgou and Bradbury, 1968), it had spread to Uganda and from there to Tanzania in 2005, and to Kenya and Burundi in 2006 (Carter et al. 2010). The disease causes loss both through death of the plant and rotting of the fruit. The leaves gradually turn yellow and start looking lifeless as if they were melting under intense heat. They eventually turn brown and die. Internal symptoms revealed by doing a cross-section of an infected pseudostem are yellow-orange streaking of the vascular tissues and the presence of yellow bacterial ooze, which can also be seen from any other infected plant part. Based on estimates of the Ugandan government, BXW caused yield losses of up to US$75 million in 2006 with a projected overall economic loss of $2-$8 billion in the next 10 years. BXW is not treated with pesticides, instead the recommended treatment includes removing male buds, which are the entryway for new infections, destroying infected plants, using clean field tools and clean planting material, as well as not using banana remains from unknown sources as mulches. Once fully developed, our platform will integrate remote sensing and mobile technology to both gather data from farmers regarding infections, to alert neighboring farmers, and to improve containment practices by improving communications among farmers and banana processors.
We propose a Banana Scouting Platform (BSP), a smart and scalable GIS communication platform that will monitor multiple pests in banana cultivars across Africa. Integrating historical data, satellite data, and mobile apps we will provide nearly real time surveillance of infected banana farms. Mobile communication channels (text, voice and email) will alert and educate farmers on best practices for prevention, treatment and containment. In the future, machine learning models will be developed to predict disease spread, develop disease resistant cultivars and improve practices. To the best of our knowledge, remote sensing system for banana is not currently available. Phase 1 of this project will be dedicated to developing a ‘proof of concept’:
Milestone #1: Accurate identification of banana farms from satellite images
Milestone #2: Accurate identification of BXW and other banana infections from high resolution satellite images and/or from mobile phones pictures
Milestone #3: Introduction of GIS/mobile tools to farmers, agronomists, and policy makers
Tzvi Aviv successfully utilized soil, weather and remote sensing data to successfully predict soy yields in North-American farms (Aviv, 2017). Deep learning neural networks have been used to identify coconut plantations and to predict yields (Jayashree et al. 2015). Similar to coconuts, the distinct morphology of banana canopy and the large size of leaves suggest that banana farms in East Africa can be identified using publicly available RGB satellite images at 10-30 m/pixel resolution. Identification of BXW or Fusarium infections will likely require higher resolution and multispectral imaging. Mobile technology is used successfully to educate low income farmers in India (http://www.icrisat.org/). The increasing penetration of mobile phones in Sub-Saharan Africa suggests that mobile communications channels will be adequate in banana farms. Computer vision technology is rapidly evolving, and Prof. Graham Taylor, a pioneer in deep learning computer vision is on our advisory board. Given the reported BXW damage in Uganda, the world’s second largest producer of bananas, we believe that our solution should be developed first in this country. To this end, we established a strategic alliance with the Bassiouni Group, a global development firm with agriculture partners in Uganda and other countries in Africa. We have inked provisional agreement to develop our solution in partnership with Jakana Foods Ltd. a leading African Agro-processor of bananas on our solution.
If you have any questions