The geographical focus of the project is Cư M’gar and Krông Búk Districts of Đắk Lắk Province, and Lâm Hà and Bảo Lâm Districts of Lâm Đồng Province, Central Highland, Vietnam (Figure 1). The two provinces were selected based on their diverse climate and farming systems. While both is influenced by a tropical monsoon climate with distinct rainy and dry seasons, Đắk Lắk Province is characterized by a drier, more arid environment compared to Lâm Đồng, which is more humid and experiences greater amounts of rainfall. Coffee production is highly intensified and coffee is often intercropped with black pepper (Kahsay et al., 2022).

With changing weather and climate, coffee farmers in Vietnam often adapt practices such as water management, crop diversification and intercropping (Thi and Chaovanapoonphol, 2014; Nguyen and Nguyen, 2019). A recent study in the region found that farmers with access to internet were more likely to take adaptation measures (e.g., water and crop management practices) and obtain higher coffee yields (Kahsay et al., 2022). With Vietnam has a high number of mobile internet users (70% of the population in 2018) (World Bank Group, 2019) and the Ministry of Agriculture and Rural Development promotes the use of ICT for agriculture information distribution (Hoang, 2020), smartphone applications have the potential to be an effective agricultural advisory tool.

The ThIRST model was developed within the framework of the project “Improved Business through Seasonal Forecasting for Coffee in Vietnam.” The project was funded by the Nordic Climate Facility and implemented through cooperation between University of Copenhagen (UCPH), the Alliance of Bioversity and CIAT, Sustainable Management Services Vietnam (SMS), and Real Time Analytics (RTA). The project aimed to transform and re-orientate agricultural development under climate change with the development of a tailored seasonal climate forecasting system and a set of associated decision support tools for coffee farmers. A key output of the project was a mobile app which provides advisory on different aspects of coffee farming such as fertilization, pests and disease management, intercropping, irrigation, pruning, mulching, and registration of farm practices. An important component of the app was the provision of seasonal forecasts together with associated agronomic advisory based on the forecast and current phenological period. The ThIRST model was one of the advisory tools incorporated in rtWork app available for both Android https://play.google.com/store/apps/details?id=vn.rta.rtwork and iOS https://apps.apple.com/vn/app/rtwork/id1097767529, and in two languages: Vietnamese and English.

User needs were gathered in 2019 via a baseline survey of 400 farmers, stakeholder consultation (SMS) and expert interviews (agronomists). These 400 coffee farmers were randomly selected from two districts of Đắk Lắk (n = 200) and two districts of Lâm Đồng (n = 200). For each household, one main coffee farmer and one secondary farmer (opposite gender) were interviewed. The survey asked them about their socio-economic characteristics, coffee farming systems and practices, use of ICT, access to information and farming advisory, adaptation practices, and personal attitudes toward trust and risks.

Subsequently, the sampled farmers were introduced to and trained in the use of the rtWork app, however, installation of the app on their phone was optional. In 2022, an endline survey was carried out to assess the project’s impacts. Out of 400 sampled farmers, 237 participated in the endline. The endline survey gathered similar information to the baseline with the addition of farmers’ habit on using rtWork app and the advisory modules in the app. These data provided valuable information on the usability and applicability of ThIRST model.

According to the baseline survey of 400 participants, the irrigation needs among farmers varied greatly across the regions; in Bao Lam district only 20 of 100 surveyed farmers irrigated their coffee, while all farmers in the Dak Lak Province depended on irrigation. This was confirmed during a workshop conducted in October 2019, where experts expressed that irrigation and water resources are more of a concern among farmers in more arid areas such as Dak Lak province, rather than in less arid areas in Lam Dong province. This also matched expectations based on the aridity index of the central highlands (Figure 1). Workshop participants also stated that there had been unpredicted changes in weather patterns in recent years and so the need for weather forecast and advisory is pressing. 87% (347/400) of the surveyed farmers have experienced changes in rainfall patterns and in the timing of the dry and rainy season; all of which affects irrigation practices. The most common coping strategy to these changes was to “irrigate more,” as mentioned by 150 farmers of the 230 farmers who at least carried out one coping strategy. Meanwhile, only a very few farmers (11/400–3%) applied any kind of soil feel test to determine when to irrigate. Furthermore, when asked about the most important adaptation practices to reduce the amount of ground water used for irrigation, the most common answer was: “more accurate and timely weather forecasts to prevent having to irrigate just before it rains”.

The majority of farmers in the baseline survey still relied on word-of-mouth advice from friends, neighbors, and local input suppliers instead of traditional extension services or TV/internet-based advisory (Figure 2). This was further expanded in the WEFOCOS project workshop, in which local experts were asked about the information sources that coffee farmers in the Central Highlands use to make irrigation decision. Experts came to a consensus that in less arid areas, farmers mostly rely on informal information channels such as their own experience and friends’ and neighbors’ experience to define how much to irrigate. In more arid areas, farmers mostly rely on formal information channels such as advice from coffee traders and government extension officers. Regarding the usage of smartphone for agricultural advisory, even though 83% (331/400) of the surveyed coffee farming households had at least one smartphone, only 3% (13/400) stated that they receive information or advice on coffee farming from online sources and smartphone applications. This suggests a large information gap in the delivery of extension services, especially ICT-based extensions, revealing a unique opportunity to take advantage of the high rate of phone ownership to deliver extension to farmers in a more efficient way.

The endline survey was conducted in July 2022, 3 years after the baseline survey. Out of 400 farmers interviewed in baseline, 237 participated in the endline, 156 in Đắk Lắk and 81 in Lâm Đồng Province. The lower number of participants in the baseline survey was due to various reasons such as farmers transitioning to alternative crops, relocating to other places, and more. For instance, coffee farmers in Bảo Lâm district, who participated in the baseline study, had converted from coffee to vegetables in the period leading up to the dissemination of the smartphone application. Thus, these farmers were therefore not included in the endline survey.

Only around 5% (11/237) of the farmers reported using smartphones and the internet to get advice and information on agriculture and coffee farming. The majority of farmers still rely on information from input suppliers as well as their friends and neighbors. This is reflected in the number of farmers who use the app. 48 out of the 237 farmers in the endline survey originally installed rtWork in their smartphone, however, only 10 of them reported to currently have the app in their phone. In addition, very few farmers stated that they would look or had looked for another agricultural advisory app (4 and 5/237). In terms of modules in the app, of the 9 farmers having used ThIRST, 7 of them found it to be useful. When asked about the reasons making ThIRST useful, most of the users stated that it was easy to follow the advice (6/7), the others said that the information was well adapted to their needs (3/7) and the instruction video was quite informative (2/7). 12 out of 13 weather forecast users would find this module useful and 4 out of 12 farmers that used seasonal advisory made changes to their coffee farming. Among the 237 interviewed, around 10% (23/237) of the farmers reported making their irrigation decision based on soil feel test. This indicates an increase from the baseline, despite most farmers still irrigate to stimulate the coffee flowering process regardless of the current or upcoming weather conditions.

Our tool directly brings scientific based recommendations to users through their smartphones and translates the scientific information into actionable insights (Figure 3, English translation in Annex A5). Seasonal forecast information is delivered to its potential users through various channels. In Vietnam, seasonal forecast information is communicated nationally and regionally via radio and television, and via loudspeakers at local level (Coulier et al., 2018). Previous studies have demonstrated that internet access is associated with higher agricultural outputs, higher income, and higher adoption rate of sustainable practices (Kaila and Tarp, 2019; Kahsay et al., 2022; Nguyen et al., 2022; Zheng et al., 2022). However, there is a distinction between climate information access and use. Climate information itself does not relate to information use in management practice decisions. However, it is still unclear how this information is building end-users and farmers’ adaptive capacity with regards to climate change and climate shocks (Nkiaka et al., 2019; Sen et al., 2021). In fact, regardless of the weather forecast, coffee farmers in the Central Highlands perceive irrigation as an insurance to maintain high yield levels, increasing the risk of draining water resources and impacting the environment (D’Haeze et al., 2005a,b; Byrareddy et al., 2021; Tran et al., 2021). Our tool makes seasonal forecast information easily available to farmers and importantly, incorporates farmers in the process of building recommendations. This innovative process aims to increase the usage rate of seasonal forecasts among coffee farmers.

In addition, the tool has made scientific information more actionable for farmers who can adjust and adapt their management practices. Due to a gap in advisory, seasonal forecasts are often perceived as complex scientific information and hard to interpret by various levels of users (Christel et al., 2018). ThIRST has been able to translate and simplify these complex seasonal forecasts into an index that farmers can combine with their knowledge on soil moisture and land slope to make more informed decisions on sustainable irrigation and water use. Thus, the tool bridges the gap between scientific information and practical recommendations.

A more accurate decision support tool would involve the use of soil moisture sensors. Moisture sensors such as Watermark, Acclima and Rain Bird have been used to time irrigation rounds more accurately, thus increasing water use efficiency (Cárdenas-Lailhacar and Dukes, 2012; Haley and Dukes, 2012; Evangelista et al., 2013). According to Haley and Dukes (2012), moisture sensor irrigation controllers reduce 2.5–4 unneeded irrigation events per month for turfgrass, while still maintaining the same quality. However, sensors are expensive for Vietnamese coffee farmers (Sharma, 2019). In the absence of such high technological investment, ThIRST has the potential to fill the information gap by providing inexpensive, easily accessible scientific based recommendations to improve water use efficiency.

There is a gap in what is available and what farmers need (Bell-Pasht and Krechowicz, 2015; Sen et al., 2021). The production and supply of climate services rarely match with users’ needs (McNie, 2012; Bell-Pasht and Krechowicz, 2015; Visscher et al., 2020). While designing a climate service, scientists and researchers would aim to inform adaptation for climate variability and change (Vaughan and Dessai, 2014; Tall, 2015). By utilizing ThIRST, farmers would have the ability to accurately refine irrigation rates, reduce stress on constrained water resources and in turn, reduce associated production costs (Technoserve, 2013; Byrareddy et al., 2021). However, we are aware that the current low use of the app is likely also affected by water for irrigation not being regulated and so far, not being a scarce resource for most farmers. We expect that as water becomes scarce due to extended period of excessive use and as rainfall patterns become even more erratic than today, farmers’ incentives for using tools such as ThIRST and seasonal weather forecasts will increase.

Farmers lack trust in climate information (Sen et al., 2021; Simelton and McCampbell, 2021). Involving farmers during the design, production and evaluation of climate services can improve the trust and uptake of agro-advisory apps (Tall et al., 2014). Involving end users in app co-design increases the feeling of ownership, app usage and data quality (Brofeldt et al., 2018; Theilade et al., 2021). However, app developers tend to not include users/farmers in the development of the app and as a result, compromising the longevity of the app (Simelton and McCampbell, 2021). Because of this, farmers would not have enough trust in the app and therefore, the recommendations become less usable. Our tool addresses a part of this issue, in which it integrates seasonal forecast with farmers’ knowledge, which is known to foster trust, local relevance and use (Tall et al., 2014).

it is also evident that the low number of currently active users of our app and the ThIRST tool means that the app is not having the intended impact. Firstly, even though smartphone use in Vietnam is high and a large proportion of the farmers in our study own a smartphone, most farmers prefer traditional methods of obtaining information from input suppliers and their neighbors, who farmers perceive to understand their production. Secondly, the endline survey showed that the number of farmers using soil feel tests has slightly increased. We believe that the low cost and convenience of this method shows its potential as a valuable tool for famers in water-sensitive areas. The low usage of soil feel tests also indicates that our documentation and guidance in the digital form in the app might not be an effective way to get buy-in from farmers. Thus, there is a strong need for direct training and guidance in the field for coffee farmers. Thirdly, farmers lack trust in climate information (Sen et al., 2021; Simelton and McCampbell, 2021), which very likely also resulted in the low usage of the app and all the modules. One positive sign is that farmers, who used one or more advisory modules, i.e., weather forecast, seasonal advisory, shade tree advice, fertilizer advice, and ThIRST modules, find them useful and even use the advisory to make changes to their management practices. This again emphasizes that it requires direct training and guidance for farmers to strengthen the usage level and make a larger impact. Last but not least, further research should expand on the tool to devise a way to quantitatively measure the water saved.