“A little bit obsessed with the weather”: Leveraging Australian farmers’ online weather practices to inform the design of climate services

ABSTRACT

Farmers’ local knowledge represents an important yet underutilised resource in climate adaptation. The emergence of online region-specific climate projections presents an opportunity to leverage farmers expertise in reading and responding to short-term weather forecasts as a design input into longer-term climate services. This paper leverages insights gained through a detailed exploration of existing everyday weather application (app) practices from 25 Australian farmers across different commodities. Through the lens of Social Practice Theory, the paper details how farmers chose, accessed and utilised online weather information in decision-making. Farmers accessed between one and six weather apps daily, with perceived accuracy the largest determinant of adoption. The paper provides detailed knowledge of farmers’ practices accessing online weather information and based on these practices, recommends four considerations for the design of multi-decadal online climate services including: (1) Leveraging tacit knowledge and existing practices of tinkering and appropriation. (2) Supporting the triangulation and comparison practices common to use of weather apps. (3) Setting expectations regarding the perceived accuracy of climate projections. (4) Ensuring climate information is available and salient when climate-relevant decisions are made. Through these considerations, the paper aims to benefit end-users of climate services by ensuring climate information responds to user needs and fits within existing practices.

KEYWORD:

• Weather
• climate services
• social practice theory
• farmers
• usability

1. Introduction

Addressing wicked problems such as climate adaptation requires understanding and supporting decision-making in agriculture (Park et al., 2012; Rose, 2014). Weather and seasonal climate forecasts represent a critical informational input into on-farm decision-making that are considered and interpreted against rich accumulated tacit knowledge and intuition regarding one’s land, crop, soil and local factors (Crane et al., 2011; Eastwood et al., 2012; Mousumi et al., 2023; von Diest et al., 2020). A strong demand exists for accurate agricultural weather forecasts (Klemm & McPherson, 2017; Schneider & Wiener, 2009). Some farmers are experts in accessing, synthesising and operationalising multiple online weather-related information sources as one of many inputs into farm management decisions (Kernecker et al., 2021; Lacoste & Kragt, 2018; O’Grady et al., 2021). Others utilise or depend on intermediaries such as advisors, agronomists (Haigh et al., 2015) or more targeted agro-climatic decision support tools (Fraisse et al., 2006; Takle et al., 2014) to operationalise climate information in decision-making. Yet, despite the demand for – and centrality of – weather and climate information as data inputs into agricultural decision-making (Klemm & McPherson, 2017; Schneider & Wiener, 2009), there is limited understanding of exactly how, when and what apps or weather services farmers access as part of their everyday practices and how this knowledge translates into on-farm decisions (Darbyshire et al., 2020; Kusunose & Mahmood, 2016; Lacoste & Kragt, 2018).

Long-term climate projections are evolving rapidly, reflective of substantial national investment in meteorology and climate science. Over USD$56 billion was spent on Weather and Climate Information Services (WCIS) in 2014/2015 (O’Grady et al., 2021). Advances in climate modelling and computational power have catalysed the development of a rapidly expanding range of online tools offering multi-decadal climate projections for agriculture.¹ These include shorter term tools integrating crop and climate models to assist seasonal planning (Dainelli et al., 2022), and longer term interactive multi-decadal climate projections, with user-selectable emissions scenarios, future time periods and agricultural commodities (Webb et al., 2023). These tools share many user-interface similarities with more familiar online weather products (e.g. geographic specificity of forecast/projection, map-view options, comparison to historic data and customisability of input parameters), yet they inform entirely different contextual decisions. Equally, online climate projections are only recently emerging, whereas online weather forecasts have been commonplace for decades. Weather forecasts for a given region may be obtained from numerous online sources, but (at present) few online sources of long-term climate information for a given region exist.

To date, despite substantial investments, the overall adoption and impact of long-term climate services for farmers remains relatively modest (Findlater et al., 2021). Human-centred design and the incorporation of farmer knowledge are recognised as integral in addressing this problem by designing more salient, usable and actionable climate services for farmers (Findlater et al., 2021; Fleming et al., 2021), thereby maximising their potential to inform decisions and improve climate change adaptation in the agricultural sector (Webb et al., 2023).

We contend the rapidly emerging nature of online climate tools creates an opportunity to understand how farmers’ knowledge and practices, concerning short-term weather and seasonal forecasts, might be operationalised as a design input for online multi-decadal climate services to increase relevance and adoption. Incorporating users as central to the design of climate services involves recognising farmers “ … not as mere recipients of [climate] adaptation knowledge from upstream sources [but] actively involved in adaptation processes as producers and holders of knowledge” (Klocker et al., 2018). Farmers’ local knowledge is an important source of information on climate adaptation strategies (Labeyrie et al., 2021). We suggest that farmers’ substantial local weather knowledge (Klocker et al., 2018; Takle et al., 2014) and demonstrable proficiency in accessing and operationalising online weather information (Lacoste & Kragt, 2018) constitutes an important yet under-utilised resource for the design of long-term climate services.

AIM: This paper explores farmers’ online weather practices as a potential user-centred design input for online multi-decadal climate tools for agriculture, making two related contributions: Contribution 1: We provide a thus-far novel account of Australian farmers’ practices accessing and interpreting of short-term weather information. Contribution 2: We apply this understanding to the design of long-term climate projections, suggesting ways in which climate services might cater to existing habits and expectations. Using social practice theory (SPT) we detail 25 Australian farmers’ practices interacting with online weather sources. We highlight potential facilitators and challenges of integrating future climate projections as a normative component of farmers’ practice.

2. Background

2.1. Defining weather and climate services

“Climate services” describes production and delivery of weather and climate information with an aim to inform decision-making (Findlater et al., 2021; Vaughan et al., 2018). The term covers a range of outcomes, where a meta-review of 101 deployments of climate services included the provision of weather/climate information on multiple temporal scales, including historic, short-term, seasonal and multi-decadal weather/climate information (Vaughan et al., 2018). Figure 1 shows the temporal breadth of weather/climate products. In this paper, we consider “weather” as forecasts up to 28-days in the future and “climate” as greater than 28-day outlooks, up to multi-decadal future climate projections.

Figure 1. Temporal breadth of weather/climate products. Adapted from (Lacoste & Kragt, 2018).

We appreciate but do not explore nuances within this spectrum (Qian, 2017) or the interactions between climate and weather, e.g. climate change contributing to the occurrence of more extreme weather (Ebi et al., 2021). Rather, we note that users are unlikely to hold rigid definitions for these factors (Fleming & Vanclay, 2010; Houser, 2018). In accordance with our human-centred framing, our interest is how farmers self-describe their use of the weather/climate information they access (Guido et al., 2021) and the timescales of decisions for which this information is salient.

2.2. Climate services as decision support in agriculture

A UK study found that almost half (49%) of farmers use some form of decision support in their operation (Rose et al., 2016). Decision support technology is evolving apace with advances in climate information, where short-term and seasonal information is progressively integrated into detailed agro-climate decision support tools which incorporate crop models, soil information, farmer-input data and can generate detailed actions or recommendations (Dainelli et al., 2022; Fraisse et al., 2006; Han et al., 2019; Rupnik et al., 2019). Yet equally, such detailed climate-focused decision support tools (at present) predominantly support flagship commodities such as wheat (Dainelli et al., 2022) and corn (Prokopy et al., 2017), and farmers continue to access online weather and climate information independently from decision support tools (Lacoste & Kragt, 2018; O’Grady et al., 2021).

2.2.1. Co-design of climate services

Climate services represent an important tool with potential for informing climate adaptation decisions (Clarkson et al., 2022) and overcoming apprehension and distrust concerning the drivers of future climate (Arbuckle et al., 2015; Petersen-Rockney, 2022). Yet various authors note issues with the delivery of climate services which constrains this potential (Born et al., 2021; Findlater et al., 2021). Despite important exceptions (Dainelli et al., 2022; Webb et al., 2023), the development of new climate services is more typically driven by advances in forecasting science or data availability than motivated by user’ needs or co-designed with users (Born et al., 2021; Feldman & Ingram, 2009). A focus for climate services research is on the evaluation of their impact on decision-making (Vaughan & Dessai, 2014), yet relatively less literature seeks to understand the details of when and how climate services are accessed as part of farmers’ practices and if/how they inform farmers’ decision-making processes (Findlater et al., 2021; Lu et al., 2021; Vincent et al., 2018). Given interaction similarities between new online multi-decadal climate projections and existing online weather sources (Webb et al., 2023), this paper seeks to contribute to the user-centred design of climate services by identifying relevant practices of farmers’ accessing short-term and seasonal weather information.

2.2.2. What do farmers do with weather and climate information?

Weather and climate information often represent a fundamental input into daily, weekly and seasonal agricultural decision-making (O’Grady et al., 2021), alongside farmers’ tacit knowledge of land, soil, crop and their micro-climate to inform farm management decisions (Crane et al., 2010; Takle et al., 2014). The “Climate based decision cycle for corn” details how and what weather and climate information is integrated into seasonal decisions around corn production, including lead times and requirements for climatic information throughout the year (Takle et al., 2014). Studies find farmers’ weather requirements change with different decisions and depend on multiple factors including crop stage (O’Grady et al., 2021) and farmer values and objectives (Mousumi et al., 2023). The perceived value of seasonal forecasts to end-users depends on forecast skill, resolution, industry and external factors including market conditions (Darbyshire et al., 2020). Klemm and McPherson (2017) detail user desires for online weather and climate products finding producer requests include direct and derived forecast products, such as total rainfall and consecutive dry days, information on uncertainty, and comparisons to previous years (Klemm & McPherson, 2017).

Despite research on user requirements for weather/climate information (Klemm & McPherson, 2017; O’Grady et al., 2021), factors affecting the perceived value of this information (Darbyshire et al., 2020), and farmers’ utilisation of climate information on a seasonal basis (Takle et al., 2014), little literature exists on how farmers access and utilise weather and climate information in farm management decisions (Lacoste & Kragt, 2018). An exception is Lacoste and Kragt (2018) who provide a comprehensive overview of West Australian farmers’ everyday use of online weather products, finding farmers typically access 4–5 separate online weather products on at least a daily basis. This paper aims to address this knowledge gap by contributing a practice-based lens to farmers’ use of weather and climate information. In doing so, we outline a new application area for SPT (Contribution 1).

2.2.3. Motivation and theory: Studying practice as a conduit to adoption

Our research is motivated by: (1) The growing recognition of the importance of local knowledge and experience as a design input into agricultural products and services; that climate adaptation should be “learnt in partnership with farmers”, rather than delivered to them (Klocker et al., 2018). (2) A wish to better understand technology use beyond the farm gate; i.e. how farmers “live with technology, rather than simply how they act on it” (Rose et al., 2022). These pursuits are well suited to SPT.

SPT represents a body of social theory with origins in anthropology, which understands human behaviour and actions as the enactment of a series of interlinked social practices (Reckwitz, 2002). SPT recognises practices as the fundamental unit of analysis, as opposed to actors themselves or the social or institutional structures in which they exist (Reckwitz, 2002; Shove et al., 2012). Practices are defined as interconnected and contingent, but equally routinised, mundane and “everyday” (Reckwitz, 2002; Shove, 2010; Shove et al., 2012). SPT focuses on the reproduction of practices in routines, habits and performances within the setting of users’ everyday life (Shove, 2012), which allows more context than focusing on moments of individual decision-making (Hargreaves, 2011). The theory is widely applied to understand human behaviour (and ways of changing it) in multiple contexts including understanding the drivers of household energy and water use (Strengers & Maller, 2012), sustainable purchasing and consumption (Hoolohan & Browne, 2020) and adoption of agricultural innovations (Phisanbut et al., 2021)

In terms of agricultural innovation, SPT provides a more nuanced and user-focused approach to understanding adoption relative to linear theories of adoption. The Diffusion of Innovation Theory, for example, tends to: (a) assume users as homogenous, (b) assume users adopt technology to maximise utility, which is not always the case and (b) treat adoption as a static-state end-goal, rather than a fluid and sometimes temporary condition (MacVaugh & Schiavone, 2010). Whereas practice-based explorations of adoption find adoption to be a social process involving overlapping effects and influences; farmers may trial new technologies and may appropriate or modify technologies to suit their needs (Klerkx et al., 2019; McGrath et al., 2023). The practice of on-farm innovation involves “tinkering” (i.e. appropriating or innovating with materials to suit farmer needs (Higgins et al., 2017)), drawing upon embedded tacit knowledge of land, water, weather, crop and soil (Kaiser & Burger, 2022). Acquired digital tools and services in this practice represent “new materials in the fibre of farming which alter how other elements are brought together over time and space in performances” (Abdulai et al., 2023). Recent applications of SPT also hint at SPT as a useful lens to understand – and potentially affect – adoption practices. For instance, observing farmers’ practices de-worming livestock to inform understanding how to motivate adoption of best-practice techniques (Bellet, 2018). Understanding herd recording practices in small-holder dairy farming informed work towards the adoption of better recording practices (Phisanbut et al., 2021). SPT identifies centrality of trust in decision around smart farming technology (Jakku et al., 2019). We note difficulties faced regarding farmers’ (sometimes) limited adoption of climate services, yet proficiency in use and utilisation of online weather information (Lacoste & Kragt, 2018). This paper, however, examines farmers’ online weather practices, and from this new knowledge, extrapolates design considerations relevant to the adoption of long-term climate projections (Contribution 2).

3. Methods

Farmers’ everyday use of weather and climate information was drawn from semi-structured interviews with 25 Australian farmers.

3.1. Recruitment

Sampling aimed to recruit farmers from around Australia and across a range of commodity and production types without imposing strict quotas for participation. Farmers were recruited by (1) an external agricultural network of advisors, growers and researchers called FarmLink, (2) outreach activities concerned with the development of a climate services² platform including webinars, demonstrations and other activities, (3) researchers’ relevant professional contacts used to identify potential participants. These factors explain the variance in commodities represented in the sample (Table 1). Ethical clearance for the research was sought and approved prior to sampling (CSIRO Ethics Clearance number 207/22). Ethics documentation including project information, participation requirements, conditions and risks of participation and informed consent documentation were sent via invitational emails to prospective participants. No strict sampling criteria were set, but we sought individuals who were involved in substantial farming operations and did not attempt to approach hobby farmers or community gardeners, which are not included in the sample.

Table 1. Participants by farm and commodity type and their stated utilisation of weather and climate information.

ID	Family farm/Corporation	Main commodity(s)	Short-term forecast	Seasonal forecast*	Long-term climate projections*	Num. online weather sources consulted regularly
F1	Corporation	Almonds	y	y		4
F2	Corporation	Almonds	y		y	2
F3	Corporation	Pork	y	y		2
F4	Family	Wheat, barley	y			4
F5	Family	Sheep, beef	y	y		5
F6	Family	Dairy	y	y		3
F7	Family	Macadamia	y	y		4
F8	Family	Sheep, fodder crop	y	y		4
F9	Family	Table grapes	y	y	y	2
F10	Family	Table grapes	y	y		4
F11	Family	Cattle	y	y		1
F12	Family	Sugar cane	y			2
F13	Family	Broadacre grain	y	y		3
F14	Corporation	Sheep, beef, fodder crop	y	y		4
F15	Family	Sheep, fodder crop, aquaculture	y	y	y	3
F16	Family	Sheep, some beef	y	y		1
F17	Corporation	Wine grape	y	y	y	3
F18	Family-run Corporation	Beef, sheep, fodder crop	y			2
F19	Family	Broadacre grain, some sheep	y	y		5
F20	Family	Sheep, fodder crop	y			2
F21	Family	Dairy, beef	y	y		3
F22	Family	Beef	y	y		3
F23	Family	Dairy	y	y		3
F25	Family	Sheep, crop, agro-forestry	y	y		3
TOTAL			25	20	4	Median = 3 Average = 2.96

*Blank = no data, ‘n’ = participant stated they do not.

3.2. Participants

Twenty-five farmers took part in the interviews which were conducted between May and October 2021. These included 12 livestock dominant, 4 broadacre dominant, 6 tree/vine crops and 3 dairy farmers. Six participants worked as employees on corporate farms, while 19 owner-managed family operations (Table 1). Farms were located in six of Australia’s eight states and territories, including New South Wales (6), Victoria (4), South Australia (4), Western Australia (4) Queensland (3), Tasmania (2) and two farmers whose corporate farms ran operations in multiple states. Farms existed in diverse bioclimates including cool temperate, Mediterranean, cold semi-arid, and humid-subtropical regions, among others. In some instances, a spouse/partner also involved in farming contributed to the interview.

3.3. Interviews

The interviews were conducted in 2021, were between 21 and 68 minutes in length (mean: 42 minutes) and were semi-structured in nature, allowing points of interest to be explored alongside a common question list. Interviews sought to gather wide ranging information on farmers’ perceptions of climate risk, adaptation practices, use of weather and climate information as well as gather initial feedback on a prototype of an online climate service product: Climate Services for Agriculture (renamed: My Climate View³ subsequent to this series of interviews). Questions were intentionally broad and open-ended in nature, e.g. “how do you use climate information to make farm decisions?”, “how resilient do you feel your farm and the farming community in the region is?” and follow-up questions were used to explore points of interest. All interviews were audio recorded with participants’ consent and transcribed professionally.

3.4. Analysis

The analysis conducted for this paper represents a post-hoc analysis of an existing sample. Analysis was carried out using NVivo qualitative analysis software using a two-step process. Initially, demographic data and responses to questions related to weather and climate were extracted and tabulated (Table 1). Thematic analysis was then used to identify emergent themes within this frame. The thematic analysis followed Braun and Clarke (2006), namely, familiarisation, initial coding, collation of codes into themes, review and refinement of themes (Braun & Clarke, 2006). Familiarisation involved reading and re-reading all transcripts. Coding entailed the creation of 65 individual codes related to products used (e.g. “nowcasts-radar” “seasonal outlook”), practices (e.g. “accessing”, “choosing apps”, “triangulating of sources”), and human values and sentiment (e.g. “trust”, “reassurance”, “personal predictions”). Collation resulted in the merging of codes into themes resulting in rich descriptions of farmers’ practices with weather and climate information. A second iteration further refined themes and consolidated themes into sub-headings of three key practice constituents: accessing, choosing, and operationalising (refer to Results).

3.5. Limitations

The findings are drawn from Australian farmers with at least basic computer literacy and good access to internet and digital technologies. This may not be the case in all contexts and as such the findings should not be considered widely generalisable. Our categorisation of online weather practices (accessing, choosing, operationalising) should not be considered exhaustive, rather it considers the preliminary outline of three key practices concerned with online weather information and associated use. Future work could further explore the nuances of farmers’ online weather practices and seek to generalise findings more widely.

4. Results

Through the thematic analysis process described above, we report results according to three overarching practices concerned with online weather and climate, namely (1) accessing, (2) choosing, and (3) operationalising. Accessing relates to the process of checking weather sources including what sources are accessed. Selecting relates to the practice of selecting which sources (apps) to incorporate in the daily accessing ritual, trialling new apps and sideling ones which are difficult to use or not performing. Operationalising relates to the practice of interpreting or triangulating weather/climate information alongside other sources of information and local knowledge and using this (combined) information in decision-making.

4.1. Accessing weather information

All 25 participants described looking at weather information as an ingrained habit, sometimes a pre-occupation, where almost all participants accessed short-term weather forecasts on at least a daily basis, sometimes multiple times per day. Participants were, with very few exceptions, experts at locating and synthesising weather information, self-describing as being “obsessed” (F15), “addicted” (F8), looking at forecasts “all the time” (F6), “constantly” (F13, F18), “daily, hourly sometimes” (F19), and accessing an average of three different sources of online weather (Table 1).

Every morning while I’m having breakfast at half past 4, 5 o’clock, I’ll have the phone there and I’ll just go to the Elders app and the BoM app. (F12)

Having this conversation is highlighting to me that we are a little bit obsessed with the weather now. […] It allows us to make those decisions on both the daily, weekly, monthly obviously, and even on the annual conversation about how we conduct business and management. (F15)

Table 1 shows the number of weather apps participants reported using (between 1 and 5, median 3). Importantly, most participants had previously consulted many more weather sources than they reported they currently used. For example, F22 mentioned having reviewed “hundreds” of weather sources over the years. F8 listed several websites he used at the time, before adding “ … then we use a few others”. F7, F10 and F17 also noted there were too many to list in full:

… [We’ve] probably got about six or seven different weather apps on our phones. (F17)

There’ll be so many, we use Weather Zone, Elders – oh jeez … What’s that Swedish one? (F7)

These findings highlight the satisfaction and assurance gained by accessing multiple sources for weather information. These findings do not, however, detail exactly how many independent forecasts each participant accessed, given many weather websites pull data from central weather bureaus or third-party sources. We discuss this further below.

4.2. Variety in sources and contexts

Participants consulted a variety of sources and types of weather information beyond forecasts, including from international meteorological agencies or websites (F4, F6, F7, F8, F10, F19, F22), weather model outputs and/or charts (F22), third-party commentary on weather models or forecasts (F3, F7, F19), participants’ own weather instruments such as weather stations or evaporation pans (F1, F2, F9, F15, F17, F18), weather information from grower groups or personal contacts, such as networks of weather stations or soil moisture probes (F1, F11, F13, F18) and social media (F19, F20).

Several participants additionally received weather information passively, as well as actively seeking it. F18 subscribed to a service which automatically emailed the rainfall totals for chosen locations each day. F16 spoke of the “various things that pop up on your telephone”, F19 mentioned social media as a common source of weather information, while F17 reported subscribing to “ … all the feeds” as well as receiving notification alerts from their weather stations: “Yeah, it’s pretty much daily, we’ve got alerts coming through, for example, I was woken up at two o’clock last night, because it was a frost alert watch coming through” (F17).

4.3. Comparison of sources of weather information

Participants described their everyday practices of comparing forecasts, typically drawing upon multiple sources of weather and multiple weather products (e.g. radar, short-term, medium term or wind estimates), to build a best guess personal prediction of how a day or week’s weather may play out. Participants often provided detailed accounts, of these processes, demonstrative of their enthusiasm and expertise:

Yeah, for long-term, I would use the BoM first and then for a medium term, I’ll use an app I’ve got, a subscription to ACCU-weather. And then obviously WillyWeather and then sort of on a daily basis […] our own data that we collect on farm because we have our own evaporation pan on site so we collect our own weather information daily and record that. We have our own private weather stations; we also use the NRM network of weather stations. […] The [anonymised] Research Centre and the [anonymised] Irrigation Trust; both also measure evaporation. They record that on their own website so sometimes you cross reference and look at their data as well just to see. (F1)

BoM is, probably, the one I use the most. And then when I’m looking for commentary stuff, I’ll then go to the – these, the weather guys that I follow. […] You put three or four different sources together and make a judgement on your own and then – and then make a decision. (F3)

All but two participants employed some form of comparison of different weather websites or information sources:

I use a host of – personally I use about nine different models from around the world, and Australia and the Bureau [of Meteorology]. And trying to use those models to predict what’s going to happen in the next week, and particularly the next month; coming up to summer and harvesting […] I’ve been doing it for 30 years; looking at the weather, collecting the weather data. So I’ve got a fair handle on when rain’s coming and how to interpret the forecast rates and what-not. (F22)

Even F11, who rarely or never sought weather information online himself, consulted trusted people who he felt had done this sort of triangulation or background research already. Including his daughter and another advisor:

Bob [Surname] is at [Department] in [Town]. He has been very good. I used to ring him up and say, ‘Well what’s going on Bob?’ And he would say, ‘No, this is what their predictions are’, and that sort of thing. And I would follow that to a tee. (F11)

4.4. Seasonal forecasts (28 day to 3 months)

The majority, 20 of the 25 participants (refer Table 1), accessed seasonal forecasts. Farmers frequently spoke of accessing monthly or seasonal forecasts in the same sentences as their practices of accessing short-term forecasts, utilising different weather products for different forecast requirements.

We read all the stuff on the BoM, talking about seasonal forecasts […] get your own information on what the sea temperature’s doing over in the Indian Ocean, and look at those indexes and make up your own mind for your own region. (F1)

The ABC weather map is good, and the Elders weather map is good […] With Elders we can actually plot almost a cattle breeding cycle along where their long-range forecast comes in. (F25)

Two key differences between the practice of sourcing seasonal forecasting versus shorter term forecasting was: (1) the tendency for seasonal forecasts to be used in conjunction with – or obtained as part of – advice from advisors, agronomists, consultants, grower groups or other third party information sources, and (2) a lower level of confidence in the accuracy of seasonal versus short-term forecasts, where seasonal forecasts were still closely monitored but approached cautiously: “if they say there’s going to be dry next – drier than average for the next three months, I wouldn’t sell 3000 sheep because of that” (F3). This said, confidence in seasonal forecast increased with the extremity of the forecast: “ … when everything’s pointing to a La Nina, or an El Nino, especially in El Nino, we certainly do [pay attention] […] we look at all the different indicators, we find when they’re all pointing to that, El Nino is pretty reliably coming” (F14).

4.5. Future climate projections (>12 months)

Only four participants reported having accessed longer term climate projections. There was no established practice around the use of climate projections. Sources of long-term climate information accessed include passive sources such as government department updates and the “rural press” (F15). F2 and F9 both reported looking at climate “10 years down the track” (F2) and accessing “long-term sources” (F9) but did not elaborate which. Only F17, a viticulturalist, spoke of seeking out different online sources:

We’ve been looking at the Climate Atlas that the wine industry do […] so that’s also prepping us into the future. So the recent past and then what’s going to happen after 2041 and then again out to 2070 and under sort of different RCP profiles too. […] what’s the shape of our landscape into the future under those different scenarios? (F17)

These findings highlight the pronounced divide between the ingrained, informed and habitualised practices of accessing short and medium term (seasonal) forecasts, yet almost non-existent use of long-term climate projections.

4.6. Choosing sources of weather information

The process of choosing, appraising, trialling, discarding or adopting, a source of online weather information (short-term, seasonal or both) emerged as a form of sub-practice in its own right. For many farmers the process of choosing weather sources was ongoing, fluid, constantly evolving, and even social in nature. F22 described the progression of this practice over time:

… over the last 30 years, I think I’ve narrowed it down. I’ve looked at probably hundreds of sites over the years, and probably now, like I said before, [narrowed] down to seven to 10 sites I have a fair bit of faith in. (F22)

F6 and F15 described the social dimensions of choosing weather sources. F6 described how their choice of websites evolved through comparison and discussions with others on the land. Similarly, F15 compared weather sources with their partner, who used a different combination of apps to her:

Over time you get a feel for which one you believe in the most, and you follow it a bit more. […] Somebody will be talking to another farmer and they’ll say, “Well on this, it looks like we’re going to get 25 mm by the end of the weekend”, [imitates second farmer] “All right, I’ll check what mine is saying”. (F6)

I usually compare three [weather] sites and just look at where we’re going, and my husband and I both look at different ones. [Asks husband] Which one do you look at the most now? He’s going to come across, hang on. (F15)

4.7. Perceived accuracy, usability and familiarity

Perceived accuracy, ease of use/usability and familiarity emerged as interconnected determinants of whether a source of weather was adopted. Farmers’ use of the term “accuracy” referred primarily to their own experiences of forecasts eventuating, rather than any more technical measure of accuracy.

Perceived accuracy was the largest determinant of adoption, non-adoption and dis-adoption of weather apps. F13 described in detail their practice of choosing what weather sources to use and the centrality of accuracy to this practice:

If you find one, you use it for a period of time, whether it’s days, weeks or months until you have your idea of whether it’s accurate or not, or how accurate it is. If it’s reasonable, well, you say “righto, I’ll add that to the favourites list” and we’ll have a look at it every now and again and if it’s better than the one I’m using then I’ve got no problems in changing […] If they’re wildly inaccurate or are not reliable, or whatever, you just don’t use them anymore. (F13)

F6 considered a new weather source must “prove itself” against the existing benchmark for accuracy for it to be adopted as a regular source of information: “If this is as good over a week as the BOM is over the four [days] at the moment, people will use it all the time. It’s got to prove itself first”. (F6). F8 and F2 both discussed the importance of accuracy in terms of the lost time or money resulting from an inaccurate forecast:

That’s probably one of the big things for us, is having that reliable estimate, I guess, because if we’re going to have a big rain event, we want to know beforehand so we can prepare. If it’s not going to – if nothing’s going to eventuate out of it, then, yeah, obviously, we don’t want to be spraying. It’s just an extra cost. […] That’s one of the big [things] – is knowing that it’s going to be accurate. It’s based on the accuracy. (F2)

Only two of the 25 participants reported paying for weather services (F15 and F23), where the monetary investment was clearly related to perceived accuracy of the product. F5 who did not currently pay for weather information, mentioned a willingness to pay for a weather product if it could be proven to be accurate:

Its $150 per year […] obviously, it’s user pays, but that sort of data becomes invaluable if you’re trying to predict your rainfall. (F15)

They’re all free apps [that I use]. But as a consumer, or customer or however you want to look at it, if there was accuracy around a 3-month outlook, I’d pay for it, for sure. (F5)

Some nuances exist around accuracy as a determinant of app choice, where F17’s choice of weather source related to perceptions of the forecasting process which they felt underpinned accuracy. F14’s choice of seasonal forecasting source related the website’s openness about the limits of its accuracy.

Usability and familiarity emerged as important, yet clearly secondary factors compared to accuracy. F7 explained how ease of use can be overcome if a weather product is perceived to be accurate: “Look, I think if something’s accurate, you can cope with learning how to use it. I think accuracy is the biggest pain point” (F7).

Usability and familiarity were influential in a choice between weather sources of similar perceived accuracy (F1, F7, F9, F12, F13, F15). For instance, F7 and F15 utilised information they knew was sourced directly from the Bureau of Meteorology (perceived to be trusted and accurate) but preferred interpretation of this data through alternative avenues due to their greater perceived usability and comprehension.

We don’t often go straight to the BoM; we find the BoM a bit clunky […] everyone else was just using what the BoM says in the first place and presenting it a bit – we find the presentation easier and quicker on the other site, a bit more user friendly. (F7)

So I subscribe [to] an Eastern States guy, and he gives us the [Indian Ocean] Dipole and gives us what the medium term and long-term is, and it all comes from the Bureau obviously, that’s where most people get their data from. But he puts it in a way that you can actually [use it]– it gives you predicted maps and all that sort of stuff with rainfall and significant events. (F15)

Ease of use/usability intersected with familiarity, where through the practice of using a given website or app regularly, the app’s familiarity increases which in turn serves to reinforce ease of use:

INTERVIEWER: “Why do you prefer those ones?” PARTICIPANT: “Ease of use, I suppose. The app’s on the phone and yeah, you just get used to going to the same thing and you know where you can look up your radar and different things like that”. (F20)

4.8. Operationalising weather information

Weather is one of many inputs into any farm decision alongside knowledge of markets, available resources, labour, crop and soil considerations (Schneider & Wiener, 2009; Takle et al., 2014). Hence, instead of attempting to detail how weather information affects decisions, which may be limiting, we focus on the processes through which weather information is interpreted and operationalised. That is, parsed though farmers’ own local and/or tacit knowledge systems or triangulated with readings from their own technology and third-party information to help inform decisions.

4.9. Interpreting forecasts using tacit knowledge and technology

Several farmers described their practice of interpreting weather forecasts or observations through their own local knowledge, local measurements (F3, F4, F6, F7, F12, F13, F14, F15, F19) or alongside various technologies on their farm (F1, F12, F13, F18) to base actions. This process involved factoring in knowledge of how their local microclimate differed from that of the nearest online weather station or forecast area, and hence what a given local weather forecast might mean for their farm specifically:

Own region is quite sparse, in the sense that if you’re taking it [weather observations] from [name of town] BoM data- well that’s 80 kms away from us. And the further north-east you head, the greater the climate variability. So, yeah, you can use it I think to the point where it’s a predictor, but you’re probably going to put your own outliers and factors across it based on your experiences. (F19)

… we just go look at the local records at the airport, which is only 23 km away from us. There is a slight difference, we’re a little bit drier and a little bit hotter, but we know our tolerance there. (F15)

F1, F12, F13 and F18 utilised weather information alongside or through measurements taken with their own instruments to inform farm management decisions. When spraying for pests, F12 cross-referenced BoM wind data against their own hand-held anemometers:

I get on the BoM [app] when I’m spraying – spraying herbicides and that on the sugarcane. The problem with that is all the BoM data is from one site in [town]. And what’s happening in [town] isn’t necessarily what’s happening at my place […] we have those little hand-held Kestrel units, three to four hundred dollars, and you can get the information right at the site. (F12)

F18 utilised software which utilised and processed multiple variables including local weather taken from their own weather stations to determine and manage heat stress among cattle at their feedlot:

We use a thing called Cattle Heat Load Toolbox. Every day, we’re monitoring a whole lot of trigger points based on our heat load index. (F18)

Less technical innovations were equally important. F1 and F13 both spoke of far less technical, but equally indispensable technology which they used alongside (or sometimes in replacement of) weather data from online sources, to complement their rich on-farm knowledge:

The most important source of climate I have is my evaporation pan on my farm. […] So, as soon as the trees have flowered and got leaves on, which was a month ago, we’re measuring evaporation and looking at climate data on a daily basis, recording it and then running our model over that to know how much water to put in the profile every single day. The evaporation takes into account everything that’s going on in the climate and the wind direction, speed, global radiation, the whole lot accounted for with the pan. (F1)

One of them [visiting researchers] said, ‘Do you keep accurate rainfall records?’ And I said, ‘Nup’, and I walked over to the back of my ute, pulled out my soil probe; I said, ‘There’s my rain gauge’. I don’t care how much rain I write on the wall; if it’s not measurable with this thing, it’s either gone down the creek or it’s gone somewhere else, I can’t use it. (F13)

These findings highlight how knowing what weather may be coming is important, yet equally important is being able to determine the effect of different weather, e.g. exactly how much rainfall has soaked in versus run-off (F13) or how much irrigation to apply based on a given weather pattern (F1).

4.10. Temporality in weather practices and relationship with technology

Farmers described how and why their practices of accessing and acting upon weather information fluctuated over time. Farmers’ tacit knowledge of land, crop, water and weather in their location and differences in crop stages affected the temporality of when – and for what purpose – forecasts were accessed. For example, the use of weather information increased from daily to (up to) multiple times per hour, when making on-farm decisions during rainfall or flood events (F15, F18), if participants were concerned about fire risk in their area (F17, F24) or when they were planning and conducting spraying (F2, F7, F9, F12, F19, F24). Similarly, a given daily forecast could determine what other weather products would be required throughout the day, e.g. needing to conduct in-situ weather readings while spraying (F12) or accessing rain radar services if a rain event is possible (F19, F20).

[We access online weather information] daily. Hourly sometimes. Even today, we’re having to do some spraying because we’ve had a string of rainfall events in the past month. So it’s quite windy today. Do we do it? Or do we not do it? There’s all those factors that we have to take into consideration. (F19)

[Use of weather information] depends on the season. Definitely … well, not so much this time of year, because we’re pretty low fire risk, but when it’s a borderline fire risk I’d certainly be checking, wind, humidity, temperature, just, I guess make sure what’s happening outside. In terms of spraying, yeah, daily, or a few times a day. (F24)

The use of weather information for many also fluctuated seasonally, and again, based on tacit knowledge of the decisions necessary in that season. For example, annual and tree croppers followed both weather and climate information more closely during the growing season than at other times of the year in order to be responsive to potential extreme heat or cold events:

What I’m looking for is depending on the time of year. If I’m looking out for frosts […] or if I’m looking for extreme heat, I’m looking a week ahead to see how much extreme heat we’re going to get. Yeah. So, it’s probably that five to six months you’re always looking out for those extreme weather conditions. Winter we don’t look at it so much. (F9)

Seasonal forecasts contributed to within-season decisions when accessed during the growing season and next season decisions when accessed out of the growing season, including whether and what to plant, fly treatment, stock sales or purchase decisions, how much silage to prepare, vine canopy maintenance among many others.

4.11. Use of climate projections in forward planning

Climate projections did not feature strongly in forward planning, given only four of the 25 participants reported having accessed long-term climate projections prior to the interview. Three of these four participants grew grape or tree crops, which require a greater degree of forward planning relative to seasonal crops: “You need 3 years to get crop off a vine. Can’t just change vines each year” (F9).

Longer term climate was clearly a factor in some participants’ minds, when considering succession planning or longer-term business planning, but climate data was generally not accessed for this purpose. F2 mentioned not considering long-term climate “ … unless we’re making huge decisions, based off our whole operation going forward” (F2). F19 was consistent with several farmers who noted the importance of future climate in long-term planning but had not yet accessed climate projections:

We have children who are interested in farming. So that succession is very important […] How do we invest our money? Do we look for higher rainfall areas if we’re buying properties? All that kind of stuff definitely is part of our thinking. (F19)

These findings highlight the importance of future climate for longer term decisions, but equally that these decisions take place outside of shorter term or seasonal farm planning cycles. Hence, future climate projections (if they are used) are not considered in the same decision context as weather or seasonal information.

5. Discussion

Agriculture as performance-adaptation to climate change needs a greater emphasis on farmers’ local knowledge and innovation, e.g. creative adaptive capacities (Crane et al., 2011). So far this paper has explored farmers’ practices with online weather information with a view to identify relevant design implications for climate services. In this section, we unpack the two contributions introduced earlier, namely: (1) new insights into farmers’ practices interacting with online weather and climate information, providing a new application area for SPT and (2) design considerations for multi-decadal online climate services drawn from the knowledge gathered of farmers’ interactions with, understandings of and assumptions regarding shorter term online weather products.

5.1. Contribution 1: New insights into online weather practices

The centrality of weather information in farm decision-making reflects existing knowledge in this space (Klemm & McPherson, 2017; Lacoste & Kragt, 2018; Takle et al., 2014). The findings also build on existing applications of SPT in agriculture (Bellet, 2018; Jakku et al., 2019; Phisanbut et al., 2021) through a specific focus on farmers’ online weather practices. Accessing: Farmers’ enthusiasm, satisfaction and utility gained from the often daily ritual of checking and comparing multiple online weather sources reflects a study of Western Australian grain farmers whose similarly “addicted” respondents self-described accessing 4–5 weather apps daily (Lacoste & Kragt, 2018). Choosing: The process of choosing weather app(s) we observed reflects farmers’ approach to innovation and technology more generally; we found evidence of “tinkering”, trialling and appropriation (Higgins et al., 2017, 2023; Rose et al., 2022). We observed adoption pathways (Montes De Oca Munguia et al., 2021) of weather apps; how apps may be trialled, adopted, or dis-adopted and how engagement may fluctuate over time, e.g. in or out of growing season. Operationalising: Our findings echo findings which describe agricultural decision-making as complex, non-linear and dependent on the triangulation of multiple information and knowledge sources (Klerkx et al., 2019; Rose et al., 2016). The process of operationalisation involved sensemaking from weather forecasts by triangulated weather information against rich tacit knowledge and advice from others before deploying (or not) this combined knowledge in farm decisions.

SPT is a useful lens for understanding micro-level factors influencing farmers’ decision-making and we suggest future work could explore the joint application of SPT and Adoption Pathways analysis. While our pairing of these theories was not deliberate, our findings suggest that SPT might offer useful insights into adoption pathways for agricultural technology (in this case weather apps), in particular, post-adoption use and non-use, which is identified as a knowledge gap (Klerkx et al., 2019; Schewe & Stuart, 2015). Through this process, future work might better explore factors consequential to decisions at different points throughout adoption processes (Montes De Oca Munguia et al., 2021). For example, we found (perceived) accuracy was central to decisions of whether to increase/decrease use or dis-adopt a source of weather information, which reflects the strong demand observed elsewhere for accuracy in weather/climate forecasting for agriculture (Darbyshire et al., 2020; Lacoste & Kragt, 2018; O’Grady et al., 2021).

5.2. Contribution 2: Considerations for the design of online climate services

In contrast to weather and seasonal forecasts, long-term climate projections were rarely or never accessed by participants and lacked any established practice. The existence of a practice around accessing long-term climate information should not be expected, given the less frequent nature of planning decisions requiring them. Rather, it is important for designers to understand why long-term climate information is not accessed at all, and how to foster use, engagement and value in these services in decisions which it may be valuable. Literature highlights a distrust in long-term climate projections (Arbuckle et al., 2015; Petersen-Rockney, 2022), a mismatch between multi-decadal climate projections and typically shorter farm business planning horizons (Moser et al., 2008) and a tendency for farmers to view long-term climate projections as extensions of seasonal forecasts and hence of low value as a (seasonal) decision input (Jagannathan et al., 2023). The marked contrasts in decision context and informational inputs between short-term operational decisions versus long-term business decisions (Schneider & Wiener, 2009) suggests improving climate resilience requires a distinct capability build beyond simply about optimising short-term decisions. However, given similarities in presentation and user-interactions between online multi-decadal climate projections and existing weather apps, we contend there is value in examining farmers’ online weather practices for design considerations relevant to maximising the adoption of future climate projections. We close with four preliminary considerations for the design of multi-decadal climate services based on our participants’ interactions and experiences with short-term and seasonal weather information.

5.2.1. Design consideration 1: Leveraging tacit knowledge, supporting appropriation

Our participants’ online weather practices embody many elements identified in the practice of farming more generally, namely testing, trialling, innovating, appropriating and interpreting against local knowledge (Higgins et al., 2017). We feel it is important for online climate services to provide the same affordances for appropriation, tinkering and incorporation of local knowledge. One means of doing this is allowing customisation of indices. For example, Climate Services for Agriculture provides for user-input and customisation for multiple indices concerning temperature, rainfall variables and self-specification of growing season (Webb et al., 2023). Further work should explore the extent and specifics of customisation desired by end-users and additional ways in which climate services can best incorporate and celebrate farmers’ local and tacit knowledge. Online climate services themselves will play a role in creating tacit knowledge of the land’s responses to more extreme events. The circular interplay between “gut feel”/tacit knowledge and data inputs (e.g. weather or climate projections) may be disrupted where observed events surpass one’s own tacit knowledge, such as in extreme events. Future work should seek to better understand and support the interplay between projections from climate services and lived experience of climate change interpreted through these key technologies. Supporting this interplay assists in the accumulation of tacit knowledge of climatic changes, potentially acting as a path to greater adaptation.

5.2.2. Design consideration 2: Replicating comparison and triangulation in online climate services

The triangulation of multiple online sources of weather information was central to almost all participants’ online weather practice and is reported in other studies of Australian farmers (Lacoste & Kragt, 2018). Less clear, is whether farmers are aware of the data sources behind the short-term forecasts they accessed. Relatively few weather apps issue completely independent forecasts, information may be pulled from national meteorological agencies or third parties, and data provenance information is lacking from many weather apps (O’Grady et al., 2021). We expect that some participants may be accessing the same information but presented in different ways by different weather apps. Further work is required into the importance of data source or independence of forecast to farmers.

For our purposes, however, the actual benefit (or placebo effect) of this practice of comparing multiple forecast is irrelevant, because the practice is widespread and provided substantial satisfaction and reassurance of having the best possible information. In this context, an issue for the perceived legitimacy of long-term climate projections, is that unlike the multitude of short-term weather sources, few sources for locality-specific future climate information exist (to date). We suggest that there may be value in emphasising the range of models incorporated in climate projections to make visible the triangulation of multiple sources already incorporated in online climate services. Future work could explore the ability to click beyond a simplified home page, to access greater visibility of results of the individual climate model runs behind a given projection, allowing users’ to triangulate model outputs independently. This type of feature is available on certain weather apps where one can click past a simple home screen to access comparisons of multiple weather model outputs. Such a feature should certainly not replace the simplicity of existing climate services home pages, given communicating climate modelling is already complex (Daron et al., 2021). Yet allowing for such an affordance caters to highly data-literate participants (e.g. F1, F2, F13, F15, F17, F18, F22), replicates the existing online weather practices we observed, and hence may work similarly to increase ownership, confidence and trust in climate projections, each of which is demonstrably lacking (Arbuckle et al., 2015; Daron et al., 2021).

5.2.3. Design consideration 3: Setting expectations regarding perceived “accuracy”

Perceived accuracy (namely, lived experience of short-term forecasts proving true) was a central determinant of the continued use of a given weather app for our participants. This experience is not possible to replicate with multi-decadal projections which take many years to eventuate. Farmers’ trust in forecasts decreases with the length of forecast window (Lacoste & Kragt, 2018) and the seemingly distant nature of climate projections can demotivate climate action (Moser, 2016). We suggest that it may be beneficial for online climate services to emphasise the strong past accuracy of previous future climate projections, e.g. the good correlation between 2004–2007 model outputs and actual observations (Buis, 2020). Additionally, given the increasingly precise nature of weather forecasts (Lacoste & Kragt, 2018), it is also important to emphasise differences in weather vs climate models outputs and set expectations for accuracy. For example, multi-year projections are outcomes of a range of possible futures, compared to seasonal forecasts which are probabilistic and short-term weather forecasts, which are forecasting specific daily weather events (World Climate Service, 2022). Future work is required to (1) carefully construct accessible and comprehensible messaging which sets expectations of projection accuracy and (2) ensure numerical climate projections do not instil a false sense of precision. For example, specific numbers for long-term warming estimates may be mis-interpreted by novice users as precise forecasts, rather than averages of model outputs. Perceived precision may foster unrealistic expectations, or conversely, scepticism of accuracy (Moernaut et al., 2022).

5.2.4. Design consideration 4: Being there when it matters

The ingrained daily, sometimes hourly, ritual of accessing weather information that we observed is enabled by the continuously evolving nature of online weather information, from twice-daily forecast frequencies to near real-time rain radar updates. This high frequency of user-interaction simply cannot be expected of long-term climate information which is updated far less regularly. Rather, online climate services must be available, front-of-mind and salient at the time when longer term decisions are made (Findlater et al., 2021). The normalisation of weather-related notifications through social media feeds or alerts from technologies for many farmers (F1, F9, F15, F17, F18) highlights a potential role for well-timed nudges such as push notifications. Additionally, literature highlights the importance of embedding decision support systems within the toolboxes of advisors and agronomists (Eastwood et al., 2019; Jakku & Thorburn, 2010). We consider fostering adoption among these parties is key to ensuring long-term climate projections are available and salient to farmers when longer term decisions are made. We envisage a key value of climate services is their role as a boundary object (Ayre et al., 2019; Jakku & Thorburn, 2010) in discussions with advisors, agronomists and other third parties. Their role as a boundary object may serve to increase trust in the services and/or facilitate helpful discussions of preparedness, future climate changes and adaption actions, even if specific projections are not acted upon directly.

Farmers have diverse contexts and objectives. Thus, interest in the long-term viability of their farm may differ for different farmers, such as owner-operators of family farms and employees of corporate owned farms. Sensitivity to specific climatic threat types may also differ depending on their location or climate zone. Our sample (19 owner-managed versus six employees of larger farming enterprises farming in multiple climatic zones – Köppen climate classification) means it is not sensible to speculate on the possible influence of farm ownership or climate zone status on use of multi-decadal climate services, but we suggest this could be a focus for future work. Future work could also further explore the nuances of farmers’ online weather practices and seek to generalise findings more widely.

6. Conclusion

Supporting end-users in climate services delivery requires ensuring climate information responds to user needs and fits within existing practices (Findlater et al., 2021). This paper has documented 25 Australian farmers’ practices using online weather apps, representing a new application area for Social Practice Theory (Contribution 1). We find that farmers are highly proficient and often “a little bit obsessed” with the weather, accessing multiple sources of short-term weather information, which is synthesised against rich local knowledge and operationalised as a constituent in decision-making. Through this in-depth exploration of farmers’ practices and interactions with short-term weather and seasonal forecasts, the paper identified user-centred design opportunities in long-term climate services provision (Contribution 2). These include how long-term climate services could leverage farmers’ tacit knowledge of their land and climate and cater to their short-term weather practices such as triangulating multiple sources. They could clearly differentiate projection processes from short-term weather forecasting and ensure that long-term climate information is front-of-mind and available when longer-term decisions are made.

To maximise the fit of long-term climate services in farmers’ existing practices, there is value in leveraging farmers’ existing proficiencies accessing and operationalising short-term weather forecasts. We recognise that weather and climate information are just some of many considerations in farm decision-making, and further work is warranted around the interactions of these practices with additional informational inputs e.g. economic, market, labour and personal factors but there is also much to be learned from farmers current practices to enable these efforts.

Acknowledgments

The authors would like to thank all participants for their generosity of time and FarmLink for their assistance in sampling for the research. The research has benefited from guidance and support from the CSIRO Drought Resilience Mission. Use of named products does not imply support or promotion and we acknowledge the weather and climate apps mentioned represent only a fraction of the possible apps available online.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/27685241.2023.2296652

Additional information

Funding

This work was funded by the Australian Government through the Department of Agriculture, Forestry and Fisheries’ Future Drought Fund and by CSIRO. There is no grant number.

Notes

¹ For example: Climate Services for Agriculture (Australia): https://climateservicesforag.indraweb.io/. Victoria’s Climate Future Tool (Australia): https://vicfutureclimatetool.indraweb.io/project. Med-Gold (Europe): https://www.med-gold.eu/. Cal-Adapt (US): https://cal-adapt.org/. National Trust Climate Hazards Map (UK): https://national-trust.maps.arcgis.com/.

² Climate Services for Agriculture: https://climateservicesforag.indraweb.io/.

³ www.myclimateview.com.au.