Researchers across Africa convene in Mombasa to address climate change and soil health challenges

Date: December 12, 2024 Author: John Recha (ILRI), Silvana Summa (ISRIC)

LSC-IS team joins AICCRA training on Climate-Smart Agriculture and Soil Health in Eastern and Southern Africa

John Recha, Research Scientist and LSC-IS project manager at ILRI.

The Accelerating Impacts of CGIAR Climate Research for Africa (AICCRA) is collaborating with the Center for International Forestry Research and World Agroforestry (CIFOR-ICRAF) and various regional development and research organisations - Association for Strengthening Agricultural Research in Eastern and Central Africa (ASARECA), Centre for Coordination of Agricultural Research and Development for Southern Africa (CCARDESA) and Intergovernmental Authority on Development (IGAD) - and the DeSIRA-funded Land, Soil, and Crop Information Services (LSC-IS) project, to enhance the knowledge capacity of agricultural professionals in Central, Eastern and Southern Africa.

Training highlights

As part of this collaboration, a regional training on climate-smart agriculture, soil fertility management, and soil health monitoring in Eastern and Southern Africa was held in Mombasa, Kenya, on 23-27 September 2024. The LSC-IS project team members led the design and delivery of the training. Participants included members from the Ethiopian Institute of Agriculture Research (EIAR), Kenya Agricultural & Livestock Research Organisation (KALRO), Rwanda Agriculture and Animal Resources Development Board (RAB), International Union for Conservation of Nature (IUCN), ASARECA, International Livestock Research Institute (ILRI), and ICRAF. A total of 40 participants attended from 15 countries: Botswana, Djibouti, Ethiopia, Kenya, Lesotho, Madagascar, Malawi, Mozambique, Namibia, Somalia, South Africa, South Sudan, Sudan, Uganda, and Zimbabwe.

Training objectives

The event aimed to equip participants with comprehensive knowledge and practical skills in climate-smart agriculture, soil health, and fertility management. Additionally, it focused on integrating advanced monitoring techniques, policy advocacy, and gender considerations to effectively promote sustainable agricultural practices at both national and regional levels. The four main objectives were:

Fundamentals of climate-smart-agriculture: Provide participants with a solid grounding of the fundamental concepts of climate-smart-agriculture.
Skills in soil health and fertility mapping: Equip participants with the skills necessary for soil health and fertility mapping, monitoring, and integrated management approaches tailored to different agroecological regions.
Advanced techniques in soil monitoring: Train participants in geospatial soil mapping techniques and the use of artificial intelligence (AI) and machine learning for advanced soil health monitoring.
Policy and gender integration: Facilitate discussions on policy and gender integration to scale climate-smart agriculture practices and soil health management into national and regional agricultural policies supported by appropriate legislation and funding mechanisms.

Step-by-step insights into DSM training

A transformative training session on Digital Soil Mapping (DSM) was conducted to explore advanced techniques for sustainable land management and agriculture. Led by Dr. Ermias Betemariam, a Land Health Scientist at ICRAF, the session equipped participants with practical skills to utilise the Random Forest (RF) machine-learning model for predicting soil organic carbon. This training brought together soil science enthusiasts, researchers, and practitioners with the aim of clarifying the integration of spatial data, environmental covariates, and machine-learning models for accurate soil mapping. The hands-on format of the session allowed participants not only to learn the theoretical foundations of DSM but also to gain practical experience using the RF model.

Below is a step-by-step overview of the training:

Step 1 - Loading installed packages: Participants started by loading essential R packages required for the exercise. This included libraries for spatial analysis, data manipulation, and machine-learning operations.
Step 2 - Setting up the working environment: The working directory was configured to include datasets for the exercise. Participants familiarised themselves with their workspace, ensuring they could efficiently navigate and manage files.
Step 3 - Calling point data: Spatial point data, representing soil sampling locations, was introduced. Participants learned how to load and inspect these data points as part of the foundational dataset.
Step 4 - Incorporating raster covariates: Raster covariates, encompassing environmental datasets such as elevation, rainfall, and temperature, were added to the project. These covariates are essential predictors in DSM.
Step 5 - Visualising data: Participants visualised both point data and raster layers to assess the completeness of coverage. This step underscored the importance of ensuring alignment between the point and covariate datasets.
Step 6 - Stacking rasters: To streamline processing, raster covariates of uniform resolution and extent were stacked into a single object. This facilitated efficient analysis and modelling.
Step 7 - Fitting random forest model: The highlight of the session was running and fitting the RF model. Using point data as the dependent variable and raster covariates as predictors, participants trained the model to predict soil organic carbon.
Step 8 - Model validation: Validation, both internal (e.g., cross-validation) and external (e.g., independent dataset testing), was conducted to assess the model’s reliability and accuracy.
Step 9 - Plotting prediction results: Participants plotted prediction outputs, creating visual representations of soil organic carbon distribution.
Step 10 - Mapping confidence intervals: The final step involved mapping the standard error or confidence intervals of the predictions, giving participants insights into the precision of their maps.

Key outcomes and takeaways

The training concluded with the production of a predictive map showcasing soil organic carbon distribution. The final map, similar to the one below, demonstrated the power of DSM and RF in providing actionable insights for land management.

A typical map produced by a random forest model

Participants left the training with enhanced capabilities to integrate advanced data analytics in their soil mapping projects. The hands-on experience proved to be invaluable in demonstrating how machine-learning models, like RF, can revolutionise soil science, particularly in improving the precision of soil health monitoring.

As digital tools continue to reshape landscape management, training events like this are pivotal in building the technical expertise required to scale solutions for sustainable agriculture and ecosystem restoration.

Looking ahead

This training session marked the beginning of a series of capacity-strengthening efforts focused on leveraging digital tools and integrated techniques to address challenges in soil health and land management. Future training opportunities will be broader and include topics such as integrated soil fertility management, soil and water conservation, and other topics central to advancing climate-smart agriculture.

These training sessions are designed to address the soil health challenges faced by Ethiopia, with a population of over 126 million); Kenya, with a population of approximately 54 million; and Rwanda, with a population of around 13 million. The overarching goal is to strengthen food and nutrition security in these countries by equipping stakeholders with the tools and knowledge to manage soils sustainably, adapt to climate change, and boost agricultural productivity.

Stay tuned for updates on these events and innovations aimed at building resilience and achieving sustainability in agriculture across the region. Together, we can create a future where soil health supports food security and economic prosperity!

Share