Within the GDPFS, established by WMO, GPCs-LRF form a network of centres responsible for seasonal forecast provision. Most of the GPCs-LRF also function as National Meteorological and Hydrological Services (NMHSs) and typically have more than one designation within the GDPFS mechanism, e.g. Regional Climate Centres (RCCs).
The WMO manual on GDPFS (WMO 2019a) is the single source of technical regulations for all meteorological centres designated by WMO. Its function is to ensure the consistency and standardisation of the data, information and production practices of centres. The manual lists a number of requirements that centres must fulfil in terms of the products (e.g. the minimum variables to display) and activities provided (e.g. presenting the forecasts on a website). However, the manual does not discuss any procedures in place to ensure that these standards are met in the long run to maintain the forecast quality, and does not specify any visualisation requirements that centres must meet. In addition, the GPC-LRF network can be unclear for those unfamiliar with the GDPFS mechanism. Nevertheless, a recently published WMO report (2020) is a positive undertaking towards increasing the transparency of the WMO architecture for seasonal forecasts.
This section provides a detailed description of the architecture of seasonal forecast provision by GPCs-LRF, based on background information from the literature and findings from the interviews conducted in our study. Whilst the results are presented anonymously, direct quotes from interview participants are clearly indicated.
Global Producing Centres for Long-Range Forecasts and the Lead Centre
A network of seasonal forecast producers
In the last decades, the number of institutions producing and providing seasonal forecasts and the use of this type of climate information have been increasing. This served as motivation to establish a network of accredited centres that regularly provide global predictions for seasonal timescales. Since 2006, WMO has designated a total of 14 GPCs-LRF across the globe (Beijing, CMCC, CPTEC Brazil, ECMWF, Exeter, Melbourne, Montreal, Moscow, Offenbach, Pretoria, Seoul, Tokyo, Toulouse and Washington), with the task of producing global seasonal forecasts that are available to all WMO member states (Table 1).
GPCs-LRF develop and run their own climate models, producing global seasonal forecasts in the form of data and graphical products according to the mandatory requirements set in the GDPFS manual (WMO 2019a). The minimum requirements that must be fulfilled include producing ensemble forecasts for 2-m (surface) temperature, total precipitation and sea surface temperature, with further recommended variables of 500-hPa height, 850-hPa temperature, 850-hPa wind and mean sea-level pressure. These forecasts must be issued monthly and have a global coverage and lead time of 0–4 months. Furthermore, GPCs-LRF are required to perform skill assessments, according to the standard verification measures defined in the GDPFS manual, and present the forecast products on a dedicated webpage.
Within the GDPFS framework, the LC-LRFMME was also designated to support the GPC-LRF activities and widely disseminate the forecasts produced by GPCs-LRF, ensuring a consistent and uniform visualisation of the results. The LC-LRFMME’s responsibilities include collecting the global seasonal forecasts from all GPCs-LRF, presenting them in a common format and developing MME forecasts. Whilst each GPC-LRF is responsible for the verification of its produced forecasts, the LC-LRFMME is responsible for the standardised verification of the forecast systems, and for maintaining an archive of the GPC-LRF and MME forecasts. The forecast products displayed on the LC-LRFMME website include monthly and seasonal mean anomalies from individual GPCs-LRF, and a range of deterministic and probabilistic MME forecasts produced using different methods (e.g. regular multiple regression).
Interview results: insights into the centres’ role
Most centres mentioned that they had already been producing seasonal forecasts for several years prior to their GPC-LRF designation, thus becoming a GPC-LRF “was not a big step, […] but the driving motivation was putting together data to improve forecasts” (GPC-LRF interviewee). Another important motivation for organisations to become GPCs-LRF was the possibility to be part of a wider community, the GPC-LRF network. Being part of this network was seen by some participants as an accreditation for following good practices and an opportunity to increase the visibility of their organisation. Some participants argued that the WMO affiliation also helped them establish regional collaborations, since their activities and products are often centred on their corresponding region despite offering global forecasts.
The central role of LC-LRFMME within this network was evidenced from the interviews. By providing all the forecasts in a uniform way on its website, the LC-LRFMME provides centralised access to forecast products that allows for easy comparison of the forecasts. Some interviewees considered that “there’s not a strong drive to actually produce better graphics and improve [their] own web output”, and that the role of the LC-LRFMME is “a good alternative for having a standardisation, and that’s the main purpose of the Lead Centre: to collect all the data and put all the information out in the same format” (GPC-LRF interviewee). By contrast, other centres endeavoured to improve the interface and visualisation of these forecasts based on user feedback, as they saw the need to also provide this information directly to their users, rather than only through the LC-LRFMME route. This appears to be highly dependent on the individual centre’s requirements and perception of their users’ needs.
The interviewees specified that GPCs-LRF have a formal monthly interaction with the LC-LRFMME, although in many cases this is an automated job of data transfer. When it comes to interactions between GPCs-LRF, some formal channels are established, such as through WMO’s Expert Team on Operational Climate Prediction Systems, although one interviewee suggested that collaboration “is more based on personal initiative and contacts”. Other occasional collaborative opportunities include joint research projects involving various GPCs-LRF.
In brief, GPCs-LRF see this network as “an opportunity to collaborate with other institutions that work at the same level on developing seasonal forecasts and also opening up the data to WMO users” (GPC-LRF interviewee). The LC-LRFMME appears to act as a substitute for the lack of widely shared standards on the visualisation and presentation of seasonal forecasts.
Seasonal forecast products
When discussing the value of global seasonal forecasts, the interviewees suggested that users from different parts of the world have different interests in certain climate conditions depending on their region. A default regional interest thus emerges naturally in all GPCs-LRF, with some centres mentioning that they develop customised products for their region, considering regional needs or characteristic climatological phenomena (e.g. Asian summer monsoon). Some GPC-LRF representatives went further than this to suggest that they lacked the necessary knowledge and experience for providing forecast information to users in other regions, since this information is region-specific. Thus, information provided by RCCs and Regional Climate Outlook Forums (RCOFs) to their local region may be more useful or relevant in some cases. Many interviewees saw the tailoring and provision of information for users in other countries or regions as the responsibility of the corresponding NMHSs.
Even though seasonal forecast provision is fundamentally driven by the need to provide salient regional information, the interviewees highlighted their global value and pointed to a recently developed WMO product, namely the Global Seasonal Climate Update (GSCU).Footnote 3The operational GSCU product was created through collaboration of a WMO task team, the LC-LRFMME, the GPCs-LRF and IRI (WMO 2020), and is available on the LC-LRFMME website.Footnote 4It provides a summary of the current climate conditions and global forecasts for temperature, precipitation and other indices for the upcoming season, based on the MME forecast of LC-LRFMME. The report is sent to all the GPCs-LRF for their feedback before its monthly issue.
The regional and global aspects of seasonal forecasts are not necessarily seen as detached but rather as complementary by some interviewees. An interesting suggestion by an interviewee was to produce a global forecast for the rainy season onset. Although most of the users would be interested in the regional forecast, such a map could be used for global food emergency planning and security, amongst other purposes.
The conversation in the interviews further addressed the climate variables provided by the GPC-LRF forecasts, considering also variables other than those required by guidance. In this respect, a few participants emphasised the importance of focusing on variables that might optimise the skill. For example, large-scale atmospheric variables, such as geopotential height at 500 hPa, may provide a good indication of large-scale circulation patterns, which are generally associated with higher skill (Lledó et al. 2020); however, those are not necessarily the variables typically requested by users. There were suggestions by the interviewees for using these global circulation patterns as a proxy for the local precipitation and temperature. Indeed, some users downscale information on regional temperature and precipitation from the global circulation data (e.g. as suggested in Ramon et al. 2021). In addition, the issue of forecast quality has been central in scientific discussions and, hence, the forecast quality measures always accompany the GPC-LRF forecast maps.
The visual representation of the seasonal forecasts was also discussed in our interviews. Although guidelines for visual representation are not provided by WMO, some of the participants stated that visual standardisation could promote consistency within the GPC-LRF infrastructure. A suggestion was to follow the good practices applied by some centres, such as the visual communication approach of the Australian Bureau of Meteorology,Footnote 5suggested by one participant as a good practice. On the other hand, other interview participants thought that any standardisation of the visual presentation would be difficult or unnecessary for a global product, since some aspects (e.g. how maps are centred) can vary depending on the region of origin of the key audience.
However, the visualisation and other aspects of these seasonal forecast products are not constant, but are continuously evolving and improving, according to the interviewees. GPCs-LRF are going through frequent redesign processes to improve their websites, visual representation of their products or their climate models. In some cases, product redesign is guided by user feedback, in an attempt to make these products more user-friendly. The available resources, and financial or technical support for making such improvements vary significantly from country to country, and depend mainly on national and institutional contributions, as revealed in the interviews. This means that, in some cases, GPCs-LRF can only focus on the most urgent needs when it comes to product or model improvements.
Finally, some participants had interesting perspectives about the future of seasonal forecast products. As stated by a GPC-LRF interviewee, “the future is true Web Map Service, like GIS [geographic information system] data. So in a sense, we will probably move away from the graphical products and be more data providers […], providing data in a format that many apps on a cellphone or computer can easily display”.
In brief, a climate information product has a broad meaning in the context of seasonal forecasts, from customised regional information to a global seasonal outlook. Similarly, the visual presentation of these products has been receiving different attention. We recognised an interesting tension between providing the mandatory variables, those that users indeed need and those that show high skill: “The idea is basically to construct products that are sensible and meet the needs of the users, but also have some skill” (GPC-LRF interviewee).
Audience and increasing demand for seasonal forecasts
The main audience of the GPC-LRF products are NMHSs of the 187 WMO members in six regions, covering the entire globe. Besides NMHSs, GPCs-LRF provide information to their respective country’s governmental authorities, for example, to the ministries of agriculture and energy, or to the environmental department. Often, the forecasts enable NMHSs to produce information about regional seasonal prospects. Other target audiences of the climate data and forecasts produced by the GPCs-LRF include RCCs, which are responsible for providing regional products (such as seasonal forecasts) to support regional and national climate activities and needs, although their mandate and role often overlap those of GPCs-LRF (Rapp et al. 2011; WMO 2011), and RCOFs, which are platforms bringing together climate experts and stakeholders.
Interview results: increasing demand from new audiences
When discussing the profile of their users, the interviewees identified specific sectors interested in seasonal forecasts, such as the water, energy (including power companies), transport and insurance sectors. Participants also reported their interaction with contingency planners and emergency responders. There is an evident link between the type of sectors engaged and the nature of climate emergency in certain regions or countries. For example, GPCs-LRF interact with water supply bodies and reservoir managers, or agricultural community experts and farmers’ associations in regions that have recently suffered from droughts, such as South Africa. In the UK and other regions affected by European windstorms, seasonal forecasts can provide support to transportation planners during winter. Another example is the 2019–2020 bushfires in Australia, which led to a significant increase in requests for monthly and seasonal briefings, and tailoring of the services they provide, according to one interviewee. “I think the [2019–2020] fires were quite a game changer. If [users] came to the Bureau in the past, 10 years ago, and said: ‘how can you help me?’, we would probably just point them at the website, but we do a lot more now” (interviewee from GPC-LRF Melbourne). Besides, other governmental institutions and departments dealing with topics not typically connected with weather and climate issues (e.g. trade) have also shown an interest in global seasonal forecasts. As stated during several interviews, some GPCs-LRF are directly “called by [their national] government to participate in meetings for specific sectors” (GPC-LRF interviewee). We could not expect, however, the survey participants to always clearly distinguish their role as a GPC-LRF or NMHS.
Most GPCs-LRF have some interaction with the users of their forecasts through a support service, bulletins, newsletters or media briefings. Some centres also organise occasional training sessions. These trainings sometimes involve intermediaries, and as a GPC-LRF interviewee specified, they do not tend to train users directly, but rather provide training to advisers who then interact with users.
Some GPCs-LRF also collaborate with commercial users who purchase seasonal forecast products, such as airports or renewable energy companies. These collaborations often include training on the use of seasonal forecasts. However, some participants pointed out that the nature of the forecast is the limiting factor for its broader use, noticing a “mismatch between user expectations and what the seasonal forecast can provide. People might want to know when to have their party in the garden or something, […] on a particular day” (GPC-LRF interviewee), which is something that a seasonal forecast cannot provide with skill beyond what a naïve climatological estimate can do. Finally, whilst some participants stressed that they were continuously trying to reach more potential users, others pointed out limitations: “we also have limited resources, we can not consult every single user” (GPC-LRF interviewee).
In brief, although NMHSs present the key audience of the GPC-LRFs, most of the interviewees observed an increasing interest in seasonal forecasts and tailored information, particularly from sectors strongly dependent on climate, such as agriculture, water management and energy, but also from other sectors. However, managing the users’ expectations was mentioned as an issue, as they can have high expectations on the accuracy, whilst the forecast skill may be too low to be considered useful: “forecasts with limited or modest skill can be only used with success by sophisticated users” (GPC-LRF interviewee).
Opportunities for interaction
Interaction through RCOFs
RCOFs serve as an important platform for providing seasonal forecasts. These forums bring together national, regional and international climate experts, policymakers and other stakeholders within a region to provide consensus-based, user-relevant climate predictions (Ogallo et al. 2008; WMO 2016). Since their launch in 1997, they have spread to 19 regions of the world (Gerlak et al. 2018). Prior to each meeting, the key participants (including GPCs-LRF, NMHSs, RCC and other regional/climate prediction centres) conduct the preparatory work to contribute to the provision of the seasonal outlook, aiming to reach a consensus on the message that is going to be provided for the next season for a specific region.
Interview results: more opportunities for interaction
Most of the interview participants recognised RCOFs as important venues for interaction amongst regional experts, such as staff members of meteorological services, but also for interactions between seasonal forecast providers and users. These regional events also provide an opportunity for collaboration between GPCs-LRF, since representatives of GPCs-LRF from other regions are sometimes invited, according to some interviewees. These events may host training sessions on the application of climate prediction products, providing added value to all participants and promoting learning. Often, RCOFs consult different sources of information and different models.
Another important opportunity for collaboration and face-to-face interaction for GPCs-LRF is the WMO Expert Team meetings, which take place every 2 years and where GPC-LRF representatives participate. These meetings discuss the work of the GPC-LRF network, the latest developments and challenges related to seasonal forecasts, applications from new GPC-LRF candidates and future activities.
In brief, the collaboration between GPCs-LRF has a strong regional character, with RCOFs and similar events providing an important opportunity for networking, joint learning, but also consensus-building for regional seasonal outlooks. A more global collaboration is achieved through the WMO Expert Team.
One of the topics covered in the interviews was the key challenges related to the production and provision of seasonal forecasts. Most of the interviewees pointed at the forecast skill as the key issue: “It’s really down to basic predictability, more than thinking about how we package the forecasts and how we get them out. If you have more skill, all those downstream aspects would improve.” (GPC-LRF interviewee). The lack of skill issue was present when discussing individual climate model prediction, but it also seems to affect MME forecasts. The key challenges described in the interviews could be grouped under the following three topics:
Improving climate models and forecast provision. Climate models are continuously improved to include a better representation of physical climate processes. Model improvements also involve increasing resolution, which would require increased computational power and data storage; however, many institutions face a lack of resources for these improvements. When it comes to forecast provision, there is an evident interest in the community for moving towards cloud computing solutions, which poses new challenges. Another challenge mentioned was the provision of seamless predictions consistently over multiple temporal scales. “I think it would generate a lot more interest just simply being able to provide information at all timescales” (GPC-LRF interviewee).
Producing more skilful ensemble forecasts. The majority of the institutions interviewed expressed a strong interest in participating in MME forecasts. The provision of these forecasts by the LC-LRFMME was seen as an important motivation for some institutions to seek a GPC-LRF designation. However, producing MME forecasts requires certain consistencies in how the models are run; for example, the contributing models should have a common hindcast period. In addition, producing an MME requires coordination between the participating institutions. An earlier monthly distribution of the Lead Centre’s MME forecast was suggested by some participants, but that would need to accommodate the operational schedule of the 14 GPCs-LRF. Furthermore, participating in multiple MME initiatives at the same time poses a further challenge for centres, since they are required to format their data according to the specifications of each initiative.
Providing useful and usable products. Generating products that are useful and usable would require more substantive interaction between providers and users, as well as building local competencies. However, the provision of local data from global models is challenging, since precise information for the entire globe at the seasonal timescale using a single model is “scientifically not possible”, as explained by a participant. “For each region, [the model] will have different behaviour; a model which works fine over Europe might not work fine over South America and vice versa” (GPC-LRF interviewee). Another challenge is developing sectoral applications, as this requires consideration and combination of different types of information to produce useful indicators and information for the sector. Some GPCs-LRF went a step forward, and their team includes experts who have direct conversations with various sectors and stakeholders. Finally, some participants suggested that, to build user capacity, the users need to be provided with sufficient data so that they can calculate their own forecast product. This would increase the independence of users, and hence enhance local capacities. Freeing climate data is a topic of an ongoing discussion and change within the seasonal forecasting community, as discussed in the next section.
In brief, insufficient forecast skill often appears as an inherent aspect of seasonal forecasting. Suggestions for moving the forecasting practice forward included blending timescales and providing seamless information in a single product, from several weeks to decades ahead. Finally, the interviewees recognised that providing usable seasonal forecasts requires closer interaction and a coproduction process with users.