Lack of data compromises research on climate change

Researchers across low- and middle-income countries are set to contribute towards the creation, expansion and maintenance of inclusive climate datasets in a project spearheaded by the Lacuna Fund.

The initiative will provide resources to research teams across Africa, Latin America and South and Southeast Asia for dataset creation and expansion in the climate and energy as well as climate and health fields.

The climate-centred programme will be implemented in collaboration with Climate Change AI (CCAI), a global organisation working at the intersection of climate change and machine learning.

CCAI will provide a mentorship platform connecting potential applicants with experts working in the relevant topic areas.

During a workshop hosted on 17 May, Jennifer Pratt-Miles, the director of the Lacuna Fund, mentioned that, in the training and evaluating of artificial intelligence (AI) and machine learning models for social good, there were gaps in the availability of data and, in some cases, data was completely missing for certain populations and geographies.

In cases where datasets existed, key information was missing, leading to biases and inaccuracy.

The Lacuna Fund, therefore, through its initiatives, aimed to provide resources to multidisciplinary data scientists’ teams in low- and middle-income contexts to fill the gaps in machine learning datasets needed to solve urgent problems in their communities.

The main project outcome was to fund the creation of open and accessible datasets for machine-learning that will enable equitable climate outcomes within the specified regions, including Africa.

The research teams with experience in data science, climate, health and energy will focus on collecting and releasing new data, annotating and releasing existing data, expanding existing datasets and increasing usability as well as linking and harmonising existing datasets.

During the workshop, research teams were challenged to uphold locally driven approaches in the creation and expansion of datasets useful in future AI applications because community engagement was paramount in generated climate outcomes.

Datasets are critical

According to the Lacuna Fund secretariat, despite contributing the least towards climate change, people in low- and middle-income countries are the most affected.

Additionally, these communities lack crucial information required for mitigation and adaptation strategies to climate change and variability.

Machine learning, which holds enormous potential to advance many sectors such as agriculture to understand, mitigate and adapt to climate change, is affected by lack of home-grown data.

“In low- and middle-income contexts globally, the effective use of machine learning is hampered by a lack of ground truth data accessible to all.”

For example, global models of the impact of rising precipitation on malaria incidence are blatantly incorrect in some parts of the world because they are missing local data.

“A lack of data about how extreme heat events are affecting human health is preventing policymakers from preparing their communities to respond to these events,” the secretariat mentioned.

“Artificial intelligence and remote sensing could map energy infrastructure needs and enable us to effectively deploy renewable energy globally, but many parts of the world that would benefit most from such technologies do not currently have the data to power them.”

Participation of African researchers

During a meeting with University World News, Dr Winston Ojenge, a senior research fellow and head of the digital economy programme at the African Centre for Technology Studies, stated that climate issues have had a cross-cutting effect on different sectors in Africa, from health to agriculture and energy. Due to this factor, participation of African researchers as a priority would help boost the creation of climate-based data relevant to the African context.

He stated that, in Kenya, as in other African countries, where rain-fed agriculture is the basis of a majority of livelihoods, the influence of climate change and variability was greater. As such, collecting and analysing of data would help in developing AI mitigation and adaptation measures.

“Africa’s data-based research is compromised due to limited contextual datasets. Climate is one area that will affect environments, economies, health care, energy and food security more in coming years,” he said.

“Modelling climate is of utmost importance and this is done best using big data, AI and machine learning. African researchers must obtain their own data,” he added.

“Climate similarly affects health and health economics in many ways: the thriving of infectious diseases depends on climates, epidemics rates of movement depend on seasons whose severity is affected by climate change. As such, twin data of health and climate data must be analysed.”