Project Overview: Predicting Mitigation and Adaptation on Tomato Disease using
machine Learning and climate data
Problem Statement
In Africa, tomato production was estimated at 21.5 million tonnes, with Egypt being the largest
producer followed by Nigeria, Morocco, Tunisia, Algeria and Cameroon (FAOSTAT 2019).
Despite the economic importance of this valuable crop, its production is vulnerable to some
physiological, biotic and abiotic stressors.
Every year, the agricultural sector loses about 20% to 30% of crops from pathogens, animals,
and weeds (Oerke and Dehne 2004; Flood 2010). Pathogens alone, account for approximately
12.5% of crop losses, worldwide (Oerke 2006). Farmers in developing countries experience
higher production losses than their counterparts in developed countries (Hughes and Salathe
2015). Further, scientific predictions have indicated that abiotic factors such as climate change
exacerbate disease incidence leading to changes in suitable production areas (FAO. 2018). In
fact, warmer temperatures and prolonged wet seasons, in particular, may create favourable
conditions for certain pathogens, which could lead to larger-scale food crop damage (FAO.
2018). This will, subsequently, result in increased yield losses and consequently, food and
nutrition insecurities.
Terna et al. (2016) observed the occurrence of tomato diseases in major growing areas of Benue
4786 M. C. MOTHAPO ET AL. State, Nigeria reached 100% between July 2012 and 2013,
when the rainfall was very high. During the wettest month, high relative humidity increases the
prevalence of fungal, bacterial and several other plant pathogens. Similarly, the pathogen causing
tomato late blight develops when relative humidity is over 90% and the mean temperature is
between 16 C and 26 C (Zhang et al. 2002).
Tomato blight, a significant threat to tomato production worldwide, causes substantial yield
losses. Early detection and effective management strategies are crucial to mitigate its impact.
This project aims to develop a machine learning-based system to accurately detect tomato blight
at early stages, enabling timely interventions and reducing crop losses. Additionally, the project
will explore how machine learning can inform adaptation strategies to changing climatic
conditions that influence blight prevalence.
Project Objectives
1. Early Disease Detection: Develop a robust machine learning model capable of
accurately identifying tomato blight symptoms in images of tomato plants at early stages
of infection.
2. Disease Severity Assessment: Quantify disease severity based on image analysis to
inform treatment decisions and monitor disease progression.
3. Climate Data Integration: Incorporate historical and projected climate data to analyze
the correlation between weather patterns and blight outbreaks.
4. Mitigation Strategy Development: Utilize machine learning to optimize disease
management practices, such as fungicide application timing and dosage.
5. Adaptation Strategy Recommendations: Provide insights into potential adaptation
measures, such as crop variety selection, planting schedules, and cultivation techniques,
to address changing climatic conditions and reduce blight susceptibility.
Methodology
Image Acquisition: Collect a diverse dataset of tomato plant images, including healthy
plants and those infected with blight at various stages.
Data Preprocessing: Prepare images for machine learning by resizing, normalizing, and
augmenting them to improve model performance.
Model Development: Train and evaluate various machine learning algorithms (e.g.,
convolutional neural networks, support vector machines) to classify images as healthy or
infected and assess disease severity.
Climate Data Analysis: Analyze historical weather data to identify correlations between
meteorological factors and blight outbreaks.
Model Integration: Combine the disease detection model with climate data to predict
blight risk and inform mitigation and adaptation strategies.
Evaluation: Assess the performance of the developed system in terms of accuracy,
precision, recall, and F1-score.
Expected Outcomes
A high-performance machine learning model for early tomato blight detection.
A tool for assessing disease severity to guide treatment decisions.
Insights into the relationship between climate factors and blight outbreaks.
Optimized disease management strategies based on machine learning predictions.
Recommendations for adapting tomato cultivation practices to changing climate
conditions.
Potential Impact
By providing early warning systems and data-driven recommendations, this project can
significantly contribute to reducing tomato losses due to climate change among other factors. It
can also support sustainable agriculture by promoting climate-resilient practices
Reference
[Link]
%20use%20of%20geospatial%20and%20remote%20sensing_2022.pdf?
sequence=1&isAllowed=y