DEVELOPING A MACHINE LEARNING SYSTEM TO DETECT WATER STRESS ON MAISE FARMS

Agriculture plays a vital role in the economies of many countries. In Tanzania, a developing country, agriculture represents almost 30% of the country’s GDP, with three-quarters of the country’s workforce involved in this sector.

However, productivity is low, and progress over the past two decades has been modest – with expectations that crop productivity in maise and wheat will fall due to high post-harvest losses, below-average rainfall, pests, and high input prices. At the same time, the Tanzanian population is expected to grow by almost 70 million people between 2020 and 2050, causing the food demand to double within the same time frame.

One of the most challenging tasks of our generation is to meet the 2030 Agenda for Sustainable Development’s second goal: to end hunger, achieve food security, improve nutrition, and promote sustainable agriculture. To reach this, agricultural processes must be optimised, and innovative farming methods must be developed to guarantee food supply since arable land and environmental resources have almost reached the limits of sustainability.

Water is the most valuable resource on our planet. Agriculture consumes 80-90% of the freshwater used by humans worldwide, and about two-thirds of this freshwater is required for crop irrigation. With an expected increase of 1-2.5 degrees Celsius in the annual mean temperature over the next 50 years, agriculture will be tremendously impacted by increased evapotranspiration and greater demand for water for crop irrigation. It is thus essential to reduce the amount of water used per unit yield by reducing yield loss and the amount of water used for irrigation. Early detection and monitoring of plant responses to water stress in crops are mandatory to achieve this.

Remote sensing offers the opportunity to acquire high spatial, spectral, and temporal resolution data as input for precision agriculture which promises the great potential to close the yield gap by optimising food production using the proper management practices at the right place and the right time while keeping the consumption of resources at an environmentally sustainable level.

FAR-LeaF’s Dr Anna Msigwa’s research includes agricultural intensification activities that aim to increase the productivity of maise farming, which would benefit ecosystems and human wellbeing. The project uses advanced agricultural technologies that will enable farmers, especially smallholder farmers, to double their production and yields, improving the well-being of many people and providing a springboard for remarkable economic growth.

The study aims to develop a machine learning-based crop water stress mapping system using thermal infrared multi-/Hyperspectral remote sensing imagery. She has installed a weather station that measures climate parameters such as rainfall, temperature, wind speed, humidity, and solar radiation for monitoring weather on an experimental farm field. As well as soil moisture to monitor crop water stress using IoT devices.

She has been assessing plant response to water stress using thermal infrared multi/Hyperspectral remote sensing imageries to develop a machine learning-based crop water stress detection system in the Kikuletwa catchment, a sub-basin of the Pangani basin that covers 6077km2 within the Arusha, Manyara and Kilimanjaro regions.

Rainfall within the catchment is bimodal, with long rains (Masika) from March to June and short rains (vuli) from November to December.

The project uses advanced agricultural technologies that will enable farmers, especially smallholder farmers, to double their production and yields, improving the well-being of many people and providing a springboard for remarkable economic growth.


Farmers have been reporting a shift in rainfall patterns. Irrigated agriculture is mainly practised in the highlands and lowlands along the Kikuletwa River. The main crops in the area are bananas, coffee, and maise. Dr Msigwa has been focusing on maise.

One of the projected outcomes of the study is a web application for water stress detection. The results will give an overview of the management practices, soil types and characteristics, and up-to-date spatial and temporal rainfall variability in the Kikuletwa catchment. This will lead to maps about management practices, soil types and characteristics, and spatial and temporal rainfall variability in the Kikuletwa catchment.

The project works closely with the Local and District Governments of Hai, Tanzania, which are essential in ensuring that the research findings are integrated into their policies and guidelines. But also, the extension Officers are responsible for connecting the farmers to the researcher and communicating the research findings to farmers in the region. Dr Msigwa's research strives to provide actionable insights for farmers, enabling them to make informed decisions regarding irrigation, crop management, and resource allocation. Ultimately, this approach has the potential to bridge the yield gap and enhance food production while minimising the environmental impact.

The developed web application for water stress detection and comprehensive maps will equip stakeholders with valuable information for decision-making, enabling them to implement effective management strategies tailored to the local conditions. The collaborative effort between researchers, government entities, and farmers will lead to sustainable agriculture and lays the foundation for long-term food security and economic prosperity in Tanzania.

Heidi Sonnekus | FAR-LeaF Program

 

The Future Africa Research Leader Fellowship (FAR-LeaF) is a fellowship programme, focussed on developing transdisciplinary research and leadership skills, to address the complex, inter-linked challenges of health, well-being, and environmental risks in Africa.