In 2021, for the first time ever, seven universities across Sri Lanka contributed to a major food flow mapping exercise across the country’s nine provinces, as part of the Colombo Resilient and Sustainable City Region Food System (CRFS) project. This article explains the methodology and shares lessons that will help stakeholders elsewhere who want to understand where food comes from, how it moves between rural, peri-urban and urban areas, and how supply chains are vulnerable to climate events and other crises.

Securing high-level government support 

The CRFS project in Colombo was initiated in 2020, and an initial ‘Rapid Scan’ of existing and secondary information identified a need for stakeholders at every stage in supply chains to adapt to multiple climate-related hazards, including extreme rainfall and floods, droughts and extreme temperatures, and sea level rise. Some food production areas face threats from multiple hazards. 

As they planned to delve further into the vulnerabilities, members of the core team acknowledged that they would benefit from greater understanding of food flows across the city region, according to Professor Buddhi Marambe of the University of Peradeniya, who led the research. Their discussions crystallized when the Rapid Scan findings were presented to the then-Secretary of the Ministry of Agriculture (MOA). The Secretary suggested that the food flow mapping should cover the whole country rather than just the Colombo CRFS, because similar consequences of extreme climate events and other crises, such as COVID-19, are experienced everywhere. In all locations, it is always the poor who are hardest hit.

For Professor Marambe, securing the MOA buy-in was the first, critical step in initiating the study – and that engagement continued through four consecutive Secretaries during the research period. 

‘Without the support and encouragement of the MOA, the project would not have advanced’, he said. ‘It led to more support extended by the staff of the government departments and private sector at the provincial and district levels, in providing required information and facilitating the process. The MOA buy-in also gave the legitimacy for the respondents to take part in the interviews.’

Building on previous work 

With funding secured from FAO, the next step was to agree on which commodities to study. It was important to include both crop- and animal-based commodities that have are important components of local diets, and for which areas of cultivation and distribution networks could be detected. The final selection was influenced by previous, limited food mapping in Sri Lanka. 

‘Previous work by FAO, RUAF, and IWMI [the CGIAR International Water Management Institute] provided a basis for us. We wanted to provide continuity with what had been done before, rather than start from square one,’ said Professor Marambe.  

The seven selected commodities were rice, maize, banana, beans, potatoes, chicken meat, and marine fish. 

De-centralised research teams 

The agriculture faculties at seven of Sri Lanka’s public universities were engaged to carry out data collection. Senior academic staff were appointed as regional coordinators, and recent graduates were hired as sub-coordinators to work with the student enumerators on the ground.

In total, 120 student-enumerators conducted survey questionnaires across the provinces.  The enumerators were carefully selected for their English language skills, as they would have to translate responses from the local language (Sinhala or Tamil) into English. Only second or third year students were selected. This was because final year students would soon leave the university system, so it might not be possible for them to follow up with respondents if needed. 

The decision to appoint regional coordinators and enumerators turned out to be judicious, as it enabled the fieldwork to continue while COVID-19 travel restrictions between provinces were in place.  

‘The regional coordinators and enumerators could still travel within their provinces. If the study had been run centrally, using a team based in Kandy, it would have been game over from day one,’ said Professor Marambe. ‘I would recommend others to take a decentralised approach, because in our times anything can happen.’ 

One questionnaire for all 

Each survey questionnaire consisted of 28 questions for the respondent, organised into six sections by role in the supply chain: farmer/fisher; collector/transporter; miller/warehouse operator/processor; wholesaler; retailer; consumer. This meant that, where a respondent played multiple roles, the survey could capture their experiences in each. Any roles that were not relevant could be skipped. A final section was provided for the enumerator to input their own reflections on the survey. 

A total of 1942 survey questionnaires were completed. The respondents were selected using snowball sampling, starting from the main markets, which are in the centre of the supply chain. From there, the researchers identified respondents up-stream as far as the producers, and downstream to the retailers. 

While snowball sampling allowed for a breadth of responses, it did not allow the researchers to follow individual consignments from the farm to the consumer. Professor Marambe acknowledged that this was a methodological drawback that led to a certain generality in the findings. 

‘In-depth studies should ideally focus on individual consignments of selected commodities in order to increase precision of the outcome. This would be tedious and time-consuming considering the coverage of the whole country. However, such studies would further increase understanding the issues and lead to more pragmatic solutions to strengthen the food flow.’ 

The enumerators visited 90% of the respondents twice – once during the Maha (major) cultivating season and once during the Yala (minor) season. The second data collection, funded by IWMI, was expected to show seasonal variations in the food flows. 

Troubleshooting in real time

At the end of each week, the enumerators inputted survey data into Microsoft Excel spreadsheets. The team held weekly meetings to review progress and troubleshoot any issues. 

One such issue concerned the seasonal data. While there were considerable differences in quantities of maize and paddy produced in the Maha and Yala seasons, for other commodities the difference was less marked. The percentage of losses was the same, and there were no discernible differences in partners. Once this became apparent, the researchers pooled the data from the two seasons to strengthen the evidence base.

Another problem was the quality of global positioning system (GPS) data obtained by the enumerators, who had to rely on their mobile phones for the coordinates of survey locations. When the coordinates were overlaid on departmental maps, the central research team realised that many were inaccurate. 

Consequently, the central research team and enumerators worked together to pinpoint the precise places. In some cases, the researchers visited interview locations themselves (when COVID-19 restrictions allowed) with specialist GPS equipment, which was not available during the initial surveys because multiple teams were working in different places, at the same time. 

Multiple methods for robust data

The research team supplemented the surveys with several other data collection methods: focus group discussions, expert consultation with government officials and private sector actors, interviews with farmer organisations, and analysis of documentary sources. 

The expert consultations were particularly helpful for bringing to light issues that were missed in the survey design. Moreover, the additional methods allowed the team to verify any outlier findings from the surveys, so that they could be either validated or discounted. They could also build up data from multiple sources to build up a more detailed picture.  

‘When we discussed about climate change, especially in terms of losses, a lot of vague answers came out of the questionnaire.  Even with the key informant interviews and focus group discussions, the messages were unclear so we had hard time coming to conclusions,’ explained Professor Marambe. ‘Other secondary data helped us understanding the general toll in each district.’ 

A solid base for forthcoming work

The research team based their analysis on the Excel files, into which were entered data from all sources. Technical staff from the geographical information systems (GIS) team of the Department of Agriculture (one of the main operating arms of the MOA) have used the databases to produce a series of maps showing the food flows. 

While the extent of the study, and the methods used, meant that the findings were somewhat general, the intention was to establish an initial database on the selected commodities for others to build upon in the future. To this end, the primary databases have been lodged with the Natural Resource Management Center (NRMC) of the Department of Agriculture. 

‘Nobody wants to speak to me asking permission for data access’, said Professor Marambe. ‘If, at any given moment, they want to produce some maps using it, they can do so.’  

Interested in the main findings of the Sri Lanka food flow mapping? Read all about them here. 

---Written by Jess Halliday (PhD), Chief Executive, RUAF CIC