Helping Smallholder Crop Farmers Adapt to Climate Change: Co-Design of a Seasonal Climate Forecasting Tool

Abstract

The largest proportion of farmers in South Africa are smallholder crop farmers with limited access to technology, finances, or information, and who rely on rain-fed agriculture highly susceptible to climate variability. Crop yields have already been negatively impacted by climate change and are predicted to worsen by 2050. Seasonal climate forecast (SCFs) can provide insights into medium term meteorological conditions to help farmers adapt their farming inputs to manage risk and support their food security. This study adopted an ergonomics/human factors (E/HF) approach to co-designing an SCF tool to assist smallholder farmers to access, explore, understand, and facilitate the extraction of actionable information for their farming decision-making. The co-design process was undertaken in five steps, across three districts in the Eastern Cape Province, South Africa: (1) observations of 9 farmers; (2) interviews with 16 farmers; (3) prototype online platform; (4) focus group discussions with 8 groups; (5) user-testing of revised prototype with 9 farmers and 43 extension officers. Despite the positive impressions from farmers and extension officers some user interface issues were identified during user-testing that need to be amended during the second iteration of the prototype.

Introduction

Malnourishment and food insecurity are on the rise in South Africa equating to 4.7 million malnourished South Africans and 5.3 million reporting severe food insecurity (FAO, 2023). The largest proportion of farmers in South Africa are smallholder crop farmers with limited access to technology, finances or information, who rely largely on rain-fed agriculture which is susceptible to climate variability. Smallholder crop farmers have limited access to land, produce food predominantly for their own household, a low probability that they may have small surpluses that can be sold for commercial gain, and mainly use family or shared labor with neighbors (Carelsen et al., 2021).

Crop yields have already been negatively impacted by climate change in Southern Africa (Ray et al., 2019) and this is predicted to worsen by 2050 under current climate change predictions (Holleman et al., 2020). Despite these warning signs, South African smallholder crop farmers have yet to fully develop resilience to climate change. To address this growing food insecurity, farmers need more usable precision agriculture methods to boost their yields and output. One possibility is targeted seasonal climate forecasts (SCFs). SCFs provide insights into medium term meteorological conditions to enable farmers to adapt their farming inputs to manage risk and support their food security (Hansen, 2005). SCFs provide estimates of average seasonal conditions over a specified period, typically 1 to 3 months (Weisheimer & Palmer, 2014). While they cannot predict specific meteorological events (such as days when rain will fall or how much rain will fall in a particular event), SCFs indicate whether the upcoming season is likely to be wetter, drier, hotter, or cooler compared to normal. In South Africa, SCF information is either produced in-country (e.g., by the South African Weather Service for official dissemination), or is produced globally (e.g., European Centre for Medium-Range Weather Forecasts or International Research Institute for Climate and Society). The format of the SCF information is general, aimed at experts or scientists, and not easily understood by laypeople, especially people with limited background education.

SCFs may be particularly useful in the agricultural sector where they could be used to influence a farmer’s decisions about when and what to plant, what supplementary inputs to purchase or use, what yields to expect, and when to expect harvesting timing (Hansen, 2005). Despite these potential benefits, SCFs have rarely been used for agricultural decision-making. Where SCFs have been used they have been shown to improve food production (Ebhuoma, 2022; Hansen et al., 2011). Although the benefits of SCFs are clear, historically they have not often been easily understood or adopted in the agricultural sector (Alexander & Block, 2022). Farmers encounter challenges in accessing, interpreting, applying, and acting on the forecasts due to the poor match between the design of the forecast tools and the needs of the users (Hansen et al., 2022). In South Africa, Ebhuoma (2022) found that the uptake of SCFs was low due to the lack of access to SCFs (especially in local languages), difficulties in understanding probabilistic forecasts, distrust in science, and the spatial scale of the SCFs was too large.

This study reports on the co-production of an SCF tool for South African smallholder farmers. The WMO (2023) recommends strengthening collaborations between the forecast scientists, social scientists, and the forecast users to determine what to communicate and how to reveal actionable information for people using forecast information. This follows on from our earlier work identifying the needs of smallholder and commercial farmers (Thatcher et al., in press). For this study we adopted a participatory ergonomics/human factors (E/HF) approach to co-designing an SCF tool to assist farmers to access, explore, understand, and facilitate the extraction of actionable information for their decision-making.

Methods

The co-design process was undertaken in five steps, across three districts in the Eastern Cape Province, South Africa (Figure 1). The Eastern Cape is the poorest province in South Africa with an average household income of US$15,772/annum, with 80% of households earning less than US$650 per month (Stats SA, 2022).

Figure 1.

Data collection areas in the Eastern Cape.

Procedures

• Step 1: Observations. Observations with smallholder farmers occurred in various locations in the Eastern Cape (Raymond Mhlaba, Joe Gqabi, and Alfred Nzo districts) for 1 day while the farmer went about their daily farming activities. Informal discussions took place with the farmers while they engaged in their daily activities asking about the types and timing of their farming choices and decisions, how their farming activities varied during the year, and what they would be able to do if they had advance warning about hotter/colder or wetter/drier seasons ahead. All discussions took place in the farmers’ native language, isiXhosa. Observations were recorded through comprehensive hand-written notes and photo-ethnography.

• Step 2: Interviews. Semi-structured interviews were conducted with smallholder farmers in Raymond Mhlaba, Joe Gqabi, and Alfred Nzo districts of the Eastern Cape. Questions covered the farmers’ seasonal forecasting needs, the range of farming actions they could take based on SCF information, and how they would be able to access SCF information. Interviews were conducted in isiXhosa and were audio-recorded. Recordings were transcribed and translated into English for analysis.

• Step 3: Prototype Development. Three different prototypes of an SCF forecasting tool were developed based on the information from Step 1 and Step 2. The three prototypes were: an entirely graphical interface; an entirely narrative interface; and interactive interface that incorporated both graphical and narrative elements.

• Step 4: Focus Groups Discussions (FGDs). The three prototype formats were presented to groups of smallholder farmers in the Bizana, Nqanqarhu, Tlokeng, and eMaxesibeni municipalities. The interpretability, understanding, accessibility, and farmers’ preferences for the different formats were discussed in the focus group discussions. The sessions were conducted predominantly in isiXhosa (some SCF terms required extensive discussion in English and isiXhosa to ensure that the interpretations were correct). The discussions were audio recorded, transcribed, and translated into English for analysis.

• Step 5: User-Testing. The integrated/interactive format prototype was developed further to create an online tool that was linked to actual SCF data. The revised prototype was presented to a combination of smallholder farmers and extension officers from the Bizana and Nqanqarhu municipalities. They farmers and extension officers then spent time individually interacting with the online tool on their smartphones or laptops. This was followed by focus group discussions where the participants discussed what they liked about the prototype, what they would like changed, what they understood/misunderstood, and how to get the SCF information from the prototype to the smallholder farmers. These discussions took place predominantly in English. The discussions were audio recorded and transcribed for analysis.

Samples

Given the wide range of agricultural practices, a decision was made to focus on crop farmers for the development phase of the SCF tool.

• Step 1: Observations. Observations took place on the land of 9 smallholder farmers.

• Step 2: Interviews. Semi-structured interviews were conducted with 16 smallholder farmers.

• Step 4: Focus Groups Discussions. Eight focus groups were held totaling 93 smallholder farmers and 3 extension officers. Extension officers are professionals employed by the provincial Departments of Agriculture, Land Reform, and Rural Development to act as the link between farmers and the latest research, information, and best practices.

• Step 5: User-Testing. The two user-testing sessions included 9 smallholder farmers and 46 extension officers. The reason for the shift to a sample of predominantly extension officers is explained in the results.

Analyses

The ethnographic observational data, interviews, and focus group discussions were analyzed using thematic content analysis for each data collection step (Clarke & Braun, 2017).

Results

Steps 1 and 2: Observations and Interviews

The smallholder farmers were particularly enthusiastic about receiving any additional information that would help improve their likelihood of success. Even if the information may be of marginal value, the smallholder farmers still expressed an interest. For a smallholder farmer, a poor crop yield could be the difference between being able to feed their family and going hungry (and in extreme cases even starvation). The smallholder farmers expressed a surprisingly wide range of decisions that could be made with SCF information. These options ranged from the timing (e.g., field preparation, planting, the length of growing season, and harvesting), to seed choices, crop mixes, fertilizer purchases, when to ensure that labor is available (e.g., for field preparation, planting, and harvesting), and when to invest in insurance against the possibility of crop failure.

When considering what SCF information they would like, the smallholder farmers expressed a particular need for: (1) spatial specificity (i.e., being able to locate their specific farm rather than a general map of the whole country or province); (2) historical data to be reminded about what is “normal” for their farm; (3) the historical reliability of the SCF (e.g., how often did it make a correct/incorrect prediction in the past; (4) being able to choose the forecast period (e.g., will there be good rains in the preparation phase, will it stop raining when I need to plant, will there be good rain and warmth during the growing season, will the rains stop when I need to harvest); (5) depending on the crops being farmed and the types of outcomes from different crop yields the farmers wanted to be able to select thresholds (e.g., if the rain is above a certain threshold the yield would be good, marginal, poor, or catastrophic); (6) access to a range of meteorological measures such as rainfall, temperature, wind, hail, frost, and extreme events.

However, smallholder farmers, expressed problems with being able to access SCFs on a regular basis. Most smallholder farmers did not own smartphones, even if they did own a smartphone, data packages were viewed as expensive, and internet connectivity was poor (the areas are geographically mountainous and remote). Instead, they suggested that SCF information should be available on the radio. This request arose from the availability of radio weather forecasts which broadcast regularly. This option posed problems for the research team, as SCFs are only updated monthly (rather than hourly in the case of weather forecasts). This would mean that the same information would need to be broadcast multiple times, or a specific time would need to be found to broadcast detailed SCF information. Questions on how to address the accessibility problem were therefore built into subsequent steps.

Steps 3 and 4: Prototypes and FGDs

During the FGDs it was overwhelmingly evident that the integrated, interactive format was preferred over the purely narrative or purely graphical formats, despite literacy and language concerns (i.e., the interactive, integrated format and the narrative format were each only in English). Farmers also experienced issues with graph literacy (i.e., correctly understanding, interpreting, and using graphs). The farmers also did not consistently interpret percentages correctly within the context of an SCF (e.g., what does a 45% chance of being wetter than normal mean?). This suggested that either a comprehensive education initiative or some intermediary mechanism was required (e.g., improved interface design). A comprehensive education initiative was not deemed as feasible by the research team.

When discussing the accessibility problem, the farmers expressed their preference for a trusted intermediary, such as an extension officer, to help them interpret, translate, and accurately convey SCF information. They preferred extension officers over recognized scientists for two reasons. First, there was considerable distrust in scientists following the Covid-19 pandemic. They perceived that medical scientists were responsible for determining lockdowns, mask mandates, and vaccine mandates and yet Covid-19 infections in rural areas were minimal and serious complications were rare. Second, they wanted the information to be conveyed by someone who was known by the community. If the SCF was later found to be incorrect they wanted to have a familiar face that they could talk to in order to complain or in order to understand what had happened.

Steps 5: Revised Prototype and User-Testing

As discussed in the results for Step 4, a mixture of extension officers and smallholder farmers were included for the user-testing. A small group of smallholder farmers was retained to see how they would respond to the revised SCF prototype tool. Examples of one of the interface features of the revised prototype is shown in Figure 2.

Figure 2.

Threshold choice and forecast output.

Extension officers and farmers found the prototype easy to understand, useful, and efficient. They felt that the prototype could be used to decide when to prepare the soil for planting, when to plant, what crops to plant, what cultivars to choose, and whether to buy compost or fungicides. The main difficulty expressed by participants was accurately interpreting the meaning of the percentages (see Figure 2). User-testing also uncovered a number of usability issues with the revised prototype tool including the tool not being optimized for a smartphone, the map background was too faint, map was still not at a fine-enough resolution to identify their farm location, there were too few place names to identify their farm location, there was a need to know the historical extremes, and the tool was slow to update after changes (although admittedly the tool was not designed for 30+ people to access at the same time).

Discussion

This research project was recently profiled for Barton et al.’s (2025) work on lessons learned for community ergonomics work. It is therefore worthwhile to first reflect on the importance lessons from the design process. First, was the lesson of building trust with the community. This meant meeting the smallholder farmers in their own space (i.e., on their “farms”), listening to their concerns, and adapting the design process to incorporate their concerns. Second, we needed to acknowledge that boundaries may change. We initially thought that we would design a tool to be used by the smallholder farmers. As the design process progressed, we realized that this boundary would need to be expanded to included trusted intermediaries. Third, we needed to acknowledge that engaging with stakeholders involves translational work. Not all scientific terms are universally understood and not all terms are easily translatable. This is an ongoing process. Finally, the design process is one of continuous improvement, rather than the delivery of a finally polished end-product. The usability issues identified during Step 5: user-testing will need to be addressed in the next iteration of the prototype. Perhaps the most significant challenge will be learning how to prepare the trusted intermediaries, agricultural extension officers, to convey the forecasts to the farmers in a thorough, accurate, and honest manner. This is important to build long-lasting, trusting relationships between the climate scientists, the agricultural extension officers, and the smallholder farmers.

There are also several limitations. The prototype only contained rainfall forecasts, and it is obvious that this needs to be expanded to include temperature forecasts, and with climate change, possibly also the likelihood of extreme events. The issue of translating the tool into local languages must also be considered, but this may be difficult given that many of the terms (like climate, weather, skill, and accuracy) have specific scientific meanings that don’t always have specific local language equivalents. Finally, further work is required to ensure that the graphical formats can be consistently and accurately interpreted. Work has begun to understand how extension officers and smallholder farmers interpret odds-ratios and graphical representations of uncertainty such as heat maps, color-coding, and Receiver Operating Characteristic (ROC) curves.