Can seasonal hydrological forecasts inform local decisions and actions? A decision-making activity

.


Introduction
There has been a recent shift away from the conventional linear model of science, where research is carried out within the scientific community with the expectation that users will be able to access and apply the information, towards coproduction and stakeholder-led initiatives that bring together scientists and decision-makers to frame and deliver 'actionable research' (Asrar et al., 2012;Lemos et al., 2012;Meadow et al., 2015).Regular and clear communication between scientists and policymakers and practitioners in workshops, focus groups, consultations and interviews, and through the development of games, activities and interactive media, is imperative for ensuring that projects deliver impact outside of the academic environment.
Here, we share findings from an activity that explored the use of seasonal hydrological forecasts † for local decision-making.
This was conducted as part of an IMPREX (IMproving PRedictions and management of hydrological Extremes) stakeholder focus group for the West Thames, UK (van den Hurk et al., 2016;IMPREX, 2018), co-organised by the University of Reading (UoR), UK Environment Agency (EA) and supported by the European Centre for Medium Range Weather Forecasts (ECMWF).
Seasonal hydrological forecasts (SHF) have the ability to predict principal changes in the hydrological environment such as river flows and groundwater levels weeks or months in advance.This has potential to benefit humanitarian action and economic decision-making e.g., to provide early warning of potential flood and drought events, assist with water quality monitoring and ensure optimal management and use of water resources for public water supply, agriculture and industry (Chiew et al., 2003;Arnal et al., 2017;Li et al., 2017;Meißner et al., 2017;Turner et al., 2017).SHF systems covering a range of spatial scales have been developed -Hydrological Outlook UK forecasts at a national level (Prudhomme et al., 2017;CEH, 2018), while the Copernicus European and Global Flood Awareness Systems (EFAS and GloFAS) provide operational forecasts over larger scales (JRC, 2018ab).Recent research has demonstrated improvements in SHF quality † , including increased accuracy out to 4 months for high flow events during the winter in Europe (Arnal et al., 2018;Emerton et al., 2018).
There is growing interest about SHF amongst policy-makers and practitioners, however, in many cases, there is limited information about whether SHF products are actually being used.Research output has focused largely on technical system development and improvements to forecast skill † (see review by Yuan et al., 2015), with relatively fewer studies exploring how users engage with and apply SHF to inform decisions (see Crochemore et al., 2015;Viel et al., 2016).Many seasonal forecasting studies, including those investigating the application of seasonal meteorological forecasts † (which provide information about future weather variables, rather than hydrology more specifically), have identified forecast uncertainty † , whereby forecast skill and sharpness † decreases with increasing lead time † (Wood and Lettenmaier, 2008; Soares and Dessai,    2015), and how this uncertainty can be communicated effectively as key barriers to use (Arnal et al., 2016;Vaughan et al., 2016).Non-technical factors including the level of knowledge and training required to interpret and apply SHF information 3 effectively (Bolson et al., 2013;Soares and Dessai, 2016), the visualisation, format and compatibility of the information provided (Fry et al., 2017;Soares et al., 2018) and the level of communication between different users in the water sector and between research developers and practitioners (Golding et al., 2017), have all been found to act as both barriers and enablers depending on the user group in question.
The potential for SHF to meet the needs of the water sector is recognised by a host of UK environmental organisations including the Environment Agency, Met.Office and research centres (see Prudhomme et al., 2017).The West Thames specifically is underlain by a slowly-responding, largely groundwater-driven hydrogeological system (Mackay et al., 2015) meaning that there is potential for extreme hydrological events such as the drought of 2010-12 (Bell et al., 2013) and winter floods of 2013-14 (Neumann et al., 2018) to be detected weeks or months in advance.It also has a dense population and high demands for water which require effective long-term management of resources for public drinking supply, industry, agriculture and waste water treatment (further details about the West Thames can be found in Sect.2.2).The value of using SHF in the West Thames is of particular interest to the Environment Agency, however information on the level of understanding, uptake and application is currently unknown.We therefore aimed to develop a clearer understanding about how different professional water sector usersbroadly forecasters, groundwater hydrologists and water resource managersare currently engaging with SHF in the West Thames using a decision-making activity.
In the context of flood science communication with experts, real-time activities such as simulation exercises (that imitate real-world processes and behaviours) or roleplay (where participants engage with real-world scenarios but take on personas and positionalities that differ from their own) are known to be effective when engaging with stakeholders who bring a range of scientific ideas and perspectives to the table (McEwen et al., 2014).Such activities encourage participants to apply their knowledge to realistic situations and to reflect on issues and the perspectives of other stakeholders (Pavey and Donoghue, 2003, p.7).They are also valuable for understanding decision-making processes e.g., for environmental hazards and conflicting community views (Harrison 2002), for capacity building in response to new water legislation (Farolfi et al., 2004) and for understanding climate forecasts and decision-making (Ishikawa et al., 2011).Our decision-making activity provided an interactive and entertaining platform that encouraged participants to engage with real-world scenarios whilst fostering discussions about the barriers and enablers to use of SHF.Using three activity stages, participants were provided with sets of progressively confident and locally tailored SHF for the next 3 to 4 months.The SHF were produced using output from operational systems including Hydrological Outlook UK and the European Flood Awareness System (EFAS), and hypothetical forecasts generated through scientific research (see Neumann et al., 2018).Participants were asked to play in real-time i.e., as if receiving the forecasts on the day for the next 3 to 4 months.They did not know in advance whether the SHF's represented a flood, drought or business-as-usual scenario and had to use their knowledge and experiences to make informed decisions based on the maps, hydrographs † , tables and text provided.In reality, all 3 sets of SHF represented the same time period: winter 2013-14 (a period of extensive flooding nationwide that occurred at the end of two years of drought conditions in the UK).
Between December 2013 and February 2014 the West Thames experienced extreme flooding from fluvial and groundwater sources which had knock-on impacts for local water quality, sewage treatment and water resource managementopening up discussions for all participants.Given that issues relating to flood and drought risk, water quality and water resource management in the West Thames are generally managed by local and regional-area authorities (Thames Water, 2010), the activity focused on whether SHF can be used to support decision-making at the local level.To the best of our knowledge, this scale of practical application has yet to be explored, we suspect mainly due to the lower skill of seasonal meteorological forecasts in Europe, particularly with respect to precipitation which is a key variable of interest for hydrology (Arribas et al., 2010;Doblas-Reyes et al., 2013).A brief overview of the focus group is provided in section 2, the full activity set-up is detailed in section 3, and the findings and the discussion are presented in sections 4 and 5.

Aims of the focus group
The focus group was developed in collaboration with the Environment Agency and in line with the objectives of the IMPREX project.The aims were to:  Introduce and discuss current SHF projects, products and initiatives for the UK and Europe.
 Engage with participants' experiences and knowledge of using SHF.
 Learn how SHF are being applied in the West Thames and recognise how different users in the water sector approach and apply SHF information for decision-making.
 Identify limitations and barriers to use.
 Identify future opportunities for SHF application and research.These aims were delivered through a series of 4 interactive sessions designed to actively engage participants to share their knowledge and experiences of SHF, and short presentations that introduced the main topics surrounding SHF and informed participants about current SHF projects and developments in the scientific research.While this paper focuses on the decisionmaking activity (interactive session 2), discussions from the other sessions are also presented where relevant.An outline of the focus group programme is provided in Supplement 1 and a full report of the activities is available; see Neumann et al., (2017).

Physical geography
The West Thames refers to the non-tidal portion of the Thames river basin † , from its source in the Cotswolds in the west of England to 230 km downstream at Teddington Lock in west London (Fig. 1).It covers an area of 9,857 sq.km (the Thames basin is 16,980 sq.km) and comprises 10 river catchments † that are the tributaries † that feed directly into the River Thames (Fig. 1).The western catchments are predominantly rural; land-use is a mix of agriculture and woodland with rolling hills and wide, flat floodplains (elevation up to 350 m asl).Towards the centre and east, the region becomes increasingly urbanised, encompassing the towns of Reading, Slough and outskirts of Greater London (elevation 4 m asl at Teddington Lock).
Lithology † varies markedly across the West Thames.Catchments overlaying the Cotswolds (upstream) and the Chilterns (middle sections) are dominated by chalk and limestone aquifers † with high baseflow † , while a band of less-permeable clays and mudstones separates these two areas.Sandstones, mudstones and clays are also prevalent towards London (downstream) these catchments have higher levels of surface runoff † and can exhibit a more flashy † response to storm events (Bloomfield et al., 2009;EA, 2009).

Water demands, risk and managementwhy the West Thames is of interest
The West Thames is a highly pressured environment -15 million people and a substantial part of the UK's economy relies directly on its water supply (EA, 2015).There are more than 2,000 licensed abstraction points in the chalk aquifers and superficial alluvium and river terrace gravel deposits.90% of abstractions are for public water supply, the remaining provides water for agriculture, aquaculture and industry (Thames Water, 2010).There are 12,000 registered waste-water discharge points; pollution from sewage treatment works, transport and urban areas affects more than 45% of rivers, water bodies and aquifers, largely towards London.Diffuse pollution and sedimentation from agricultural and forestry practice is the main contributor to poor water quality in the upper catchments, especially during times of high rainfall (EA, 2015).
Urbanisation and land-use change in combination with more varied rainfall patterns has seen the region affected by a number of extreme drought and flood events in recent years (EA, 2009;Parry et al., 2015, Muchan et al., 2015).Across the Thames Basin, 200,000 properties are at risk from a 1:100 † year fluvial flood, with 10,000 at risk from a 1:5 † year event (EA, 2009).Low and high river flows also pose risks to navigation and management of the canal network which is highly important for recreation, local-living and the economy (Wells and Davis, 2016).

Who took part?
SHF have the potential for wide-ranging application and it was important to capture the different perspectives of the West Thames water sector.The organisers agreed that the focus group would work well with a relatively small number of participants (up to 12) so that all perspectives could be heard.Based on discussions held between the organisers, individuals from local organisations working in established (i.e.long-term/ permanent/ leadership) roles relevant to SHF in the West Thames were invited; many, but not all participants had previously collaborated with the University of Reading and/ or Environment Agency.
In some cases, an invitee was unable to attend due to prior commitments or because they had a colleague who they felt would be a better fit for the focus group.A total of 17 participants were invited from 6 organisations -12 accepted and 11 took part on the day.They were responsible for flood and drought forecasting (F x 3), groundwater modelling and hydrogeology (GH x 2), navigation (N x 1), water resource and reservoir management (WR x 2), public water supply (WS x 2) and waste water modelling and operations (WW x 1).They represented 5 organisations: 2 non-departmental public bodies (sponsored by Government Agencies), 2 science and research centres, 1 water service company and 1 not-for-profit organisation (Table 1).

Current engagement with SHF
By inviting local stakeholders we ensured that participants represented a range of different water sector personas and were familiar with the West Thames environment.We did not assume that participants had any prior knowledge of SHF and invitees were encouraged to attend even if they were unfamiliar with the concept as this would be an important indicator of the state of play in the West Thames (invite poster see Supplement 1).
All 11 focus group participants were familiar with the concept of seasonal hydrological forecasting and 10 regularly used SHF in their everyday job (according to results from interactive session 1 -'What are seasonal hydrological forecasts?').
Using post-its, participants noted that Hydrological Outlook UK (CEH, 2018) and the associated raw forecasts from the analogue, hydrological and meteorological models (produced by the UK Met Office, Centre for Ecology and Hydrology, British Geological Survey, Environment Agency, Natural Resources Wales, Scottish Environment Protection Agency and Rivers Agency Northern Ireland) were the main sources of SHF information currently being used, primarily for flood and drought outlook, groundwater monitoring and river flow projection purposes.Scientific research, operational planning and sharing of information with other organisations in the water sector were also listed as reasons for engaging with SHF.It' is important to note that no prior definitions or information were provided and no restrictions or guidance was placed on what participants should write down.This suggests that many in the water sector are using SHF to obtain an insight about whether the upcoming season will be drier or wetter than normal, but that they also believe SHF potentially have the capability to forecast possible flood and drought risk, which could be used to support decision-making and provide better preparedness.
This is an encouraging starting point, although many participants noted that this potential is not currently being realised due to the uncertainty and coarse spatio-temporal resolution of SHF e.g., Hydrological Outlook UK forecasts are only published monthly for the main UK river basins.

Background
Our activity was inspired by the success of previous decision-making activities and games run by the HEPEX (Hydrological Ensemble Prediction EXperiment) Community (e.g.Ramos et al., 2013;Crochemore et al., 2015;Arnal et al., 2016).The aim was to better understand how different water sector users in the West Thames interpret and act on SHF by providing them with hydrological context, maps and forecasts for the region.The activity was designed for the West Thames so that we could capture the relationship between local stakeholders and the environment in which they work.

Overview of the set-up
The set-up of the activity (illustrated in Fig. 2) followed the structure: Choose groups > Define the Objectives > Background Participants divided themselves into 3 groups based on their area of expertise and where they felt they could best contribute to the discussions.There were 3 flood and drought 'forecasters' and 2 'groundwater hydrologists'.The remaining participants (navigation, water resource and reservoir management, public water supply and waste water operations) grouped themselves as 'water resource managers'.While the results and discussions focus on these 3 broad groups, individual perspectives are also included to capture the variety of water sector personas present.There were also 3 research facilitators and 3 note-takers whose role it was to capture and record the key discussion points.
Groups were first provided with background context to the West Thames to set the scene, followed by 3 sets of progressively confident SHF for the next 3 to 4 months (Stages 1 -3).Stage 1 forecasts were from Hydrological Outlook UK, Stage 2 from EFAS-Seasonal (European Flood Awareness System) and Stage 3 were 'improved' output from EFAS-Seasonal (Fig. 2 and Sect.3.4).Participants were asked to discuss the information presented in their groups and make informed decisions about each of the 10 West Thames catchments (Fig. 1, Sect.3.3.2).All groups were provided with exactly the same information and discussion was encouraged.The activity took around 2 hours and timings were only loosely controlled.
SHF at all 3 stages of the activity represented the same time perioddating from 1 November 2013 to 28 February 2014 (or 31 January 2014 for Hydrological Outlook UK which only extends to 3 months (CEH, 2018)).These dates captured a period of severe and widespread river and groundwater flooding in the West Thames (Huntingford et al., 2014;Kendon and McCarthy, 2015;Muchan et al., 2015).Participants did not know the dates of the forecasts, nor were they informed whether the situation being forecasted was a high flow (flood), low flow (drought) or a business-as-usual scenario.Dates were removed from all information, and streamflow and groundwater level units were removed from the Stage 2 and Stage 3 EFAS hydrographs although exceedance thresholds were provided for context.The decision to remove units was advised by the Environment Agency.The concern was that participants familiar with average and high-flow values for specific catchments would deduce that the SHF must represent the 2013/14 floods, which would bias their decision-making based on their previous experience and memories.No information on forecast skill or quality was given and participants were asked to treat all information as being 'current' i.e., as if receiving the SHF today, for the next 3 -4 months to create a realistic forecasting scenario.

Recording the decisions
In real life, a user's decision process can encompass a range of possible actions and associated consequences (Crochemore et al., 2015).Decisions can be controlled for by providing participants with a set of options to choose from e.g., to deploy temporary flood defences or notthe consequences of which usually determine the outcome of a game or activity.In this case, participants were asked to select from a broad range of colour-coded options (Table 2) but specific decisions were not defined as these had the potential to differ greatly between participants and might prompt unrealistic answers.At each Stage, the colour-coded options were discussed by the 3 groups, simulating conversations that could happen in real-life but it was stressed that the colour chosen was to be representative of what an individual participant, or their organisation, would do with the SHF information in each catchment.This was recorded on an A1 map using coloured sticky dots marked with the participant's initials (n ~ 110 dots per map (11 participants, 10 catchments)) (Fig. 3).In cases where participants were not familiar with all catchments, or did not feel able to make an informed decision they did not place a dot.It was important to gather a written record explaining how and why the decisions were reached and so participants were also asked to complete an A4 empathy map at each Stage (Fig. 4).Originally designed as a collaborative tool to be used in business and marketing, empathy maps aim to gain a deeper understanding about an external user's experiences and decisions (Gray, 2017).Here, we adapted the traditional use by asking individuals to reflect on their own decisions based on their real-life experiences and discussions with other group members.This allowed us to capture individuals thought processes, influences, discussions and the potential risks and gains associated with their decision (Fig. 4).By combining the information recorded on empathy maps for each group, we also gathered an overview of the shared understanding between forecasters, groundwater hydrologists and water-resource managers and how their SHF needs and expectations match and differ when it comes to decision-making.

Table 2: Colour codes and corresponding action or decision to be take
Decision to be made or action to be taken Ignore the SHF information: Wait for the more skilful forecasts with shorter lead times (e.g. a 7 -10 day forecast).
Look at the SHF information: Decide there is no notable risk and do nothing at this point.
Look at the SHF information: Discuss or pass the information on to relevant colleagues / departments in your organisation and agree to keep an eye on the situation.
Look at the SHF information: Discuss or pass the information on to relevant colleagues / departments in your organisation but also external partnersactively request further information about the situation or seek advice on possible actions.
Look at the SHF information: Decide to implement or set in motion action(s) in a catchment e.g., to help with drought preparedness, early warning, repairs or maintenance to flood defences etc.

Background context
Groups were given information about the West Thames catchment characteristics and 'current' hydrological conditions (units and dates removed) to place the upcoming SHF into context and aid interpretation.

Catchment characteristicsdriving factors, risks and opportunities
5 maps (Supplement 2) that provided a visual representation and a numerical breakdown of the characteristic differences between each catchment were given to participants:  Hydrogeology †dominant geological type (sandstone, chalk, clay)  Elevationminimum, maximum and mean elevation (metres a.s.l)  Slopeminimum, maximum and standard deviation of slope angle (degrees)  Landcoverdominant land-use (urban, woodland, agricultural, semi-natural)  Flood risk -Flood warning and flood alert areas and an indication of 'urban flood risk' Participants were asked to discuss and identify the key differences between catchments and highlight the associated risks and opportunities.As some participants were more familiar with specific areas/catchments based on their day job, the maps provided a wider view of where catchment characteristics differ across the West Thames region.

Current hydrological situation
To help set the scene with respect to initial conditions i.e., the 'current' levels of water contained in the soil, groundwater, rivers and reservoirs, groups were provided with information from the Hydrological Summary (NRFA, 2013) for the last month, past season and past year (October 2013, June to September 2013 and November 2012 to October 2013 with dates removed).The Hydrological Summary (Supplement 3) focuses on rainfall, river flows, groundwater levels and reservoir stocks and places the events of each month, and the conditions at the end of the month into a historical context.In the real world, decision-makers are already prepared with this information, thus providing evidence about whether hydrological conditions were wet, dry or normal at the point of receiving the forecasts was an important piece of information for the participants to consider.

Stage 1 -Hydrological Outlook UK
The first set of SHF information provided to participants was the Hydrological Outlook UK (from 1 November 2013 to 31 January 2014, with dates removed) (CEH, 2013).This provided regional information for the next 3 months with reference to normal conditions for precipitation, temperature, river flows and groundwater levels.Hydrological Outlook UK uses observations, ensemble models and expert judgement (CEH, 2018) to produce the seasonal forecasts.Information is publicly available and consists of text, graphs, tables and regional maps (examples are shown in Fig. 5 and the full set of forecasts provided to participants are in Supplement 4).

Stage 2 -EFAS-Seasonal
EFAS-Seasonal (European Flood Awareness System) is an operational system that monitors and forecasts streamflow † across Europe, with the potential to predict higher than normal streamflow events up to 2 months ahead in an operational capacity, and up to 7 months in practice (JRC, 2018a;Arnal et al., 2018).It runs on a 5 km x 5 km grid and uses the LISFLOOD hydrological model (Van der Knijff et al., 2010;Alfieri et al., 2014).Seasonal ensemble † meteorological forecasts from the ECMWF's 'System 4' operational meteorological forecasting system (Molteni et al., 2011) are used as input to LISFLOOD, from which seasonal ensemble hydrological forecasts are generated on the first day of each month (see Arnal et al., 2018 for details).For the activity, SHF were produced from 1 November 2013 out to 4 months to focus on the period of extreme stormy weather and flooding experienced.As EFAS-Seasonal is designed to run at the scale of large river basins (i.e. the whole Thames basin), GIS shapefiles were used to extract forecast information for the 10 West Thames catchments using Python v3.5.This provided more locally-tailored forecasts compared with Hydrological Outlook UK (Stage 1).

month groundwater outlook
To ascertain whether participants had a preference for how SHF information is presented, the Stage 2 forecasts were presented as both hydrographs and choropleth † maps (Fig. 6).Ensemble hydrographs for streamflow (m 3 /s -1 ) and groundwater levels (mm) indicated the predicted trajectory of the hydrological conditions for the next 4 months in each of the 10 catchments (n.b. the greater the spread, the more uncertain the forecast) (Fig. 6a).Units and dates were removed, however exceedance thresholds † , based on daily observed streamflow and groundwater records between 1994 and 2014 for each of the catchments, were provided for context (EA, 2017;NRFA, 2017).Q50 (median) indicated average streamflow and groundwater conditions for the catchment.Q10 (90 th percentile), indicated high streamflow / high groundwater level conditions -90% of all recorded observations over the previous 20 year period fell below this line.
The choropleth maps showed the maximum probability that the full forecast ensemble for a catchment exceeded the Q10 (90 th percentile) threshold in a given month (Fig. 6b), thus providing a snapshot of the probability of potentially extreme conditions at catchment level.The full set of EFAS-Seasonal SHF provided to participants can be found in Supplement 5.

Stage 3 -'Improved' EFAS-Seasonal
Stage 3 followed the exact same set-up and provided the same style output (Fig. 7a, b) as Stage 2the only difference being that the seasonal meteorological forecasts used as input to LISFLOOD were taken from a set of atmospheric relaxation experiments † conducted as part of a scientific study in the West Thames (see Neumann et al., 2018) rather than the operational seasonal meteorological forecasts from 'System 4'.
Atmospheric relaxation experiments were conducted by the ECMWF in late 2014 after the extreme weather and flooding (Rodwell et al., 2015).The aim was to recreate the atmospheric conditions that prevailed between November 2013 and February 2014, so that ECWMF could better understand how weather anomalies across the globe contributed to the flooding experienced in the West Thames (Neumann et al., 2018).The SHF at Stage 3 represented near 'perfect' forecasts as they were produced once the floods had happened and the weather conditions were known.The hydrographs are thus much sharper and more accurate than those presented to the participants at Stage 2 (Fig. 7, Supplement 6).It is important to note that this is not something that can be achieved by operational systems currently, but does represent the theoretical upper level of forecast skill that may be available to water sector users in the future.4. Results

Catchment differences -"Hydrogeology is the driving factor of risks and opportunities"
All groups recognised spatial variability between the catchments and general consensus was that hydrogeology was the most important factor determining flood risk, drought risk and water availability in the West Thames (Supplement 2).All groups were interested in the persistence, hydrological memory and slower response of the groundwater-driven catchments upstream (e.g. the Evenlode, Thame and South Chilterns and Kennet) as these provided the greatest opportunity for water supply but also increased risk of local groundwater flooding and widespread fluvial flooding further downstream.Forecasters also highlighted the risks posed by impermeable catchments (e.g. the Cherwell and Lower Thames) that have a more flashy response to rainfall.Water resource managers stated that upstream reservoirs were at increased risk of pollution (from agriculture), whilst dry weather (drought) was a greater issue towards London.

Current hydrological situation -"Normal"
Hydrological Summary placed the 'current' hydrological conditions for river flows, groundwater levels and reservoir stocks within the 'normal' range (Supplement 3).Maps indicated that rainfall was below average over the past season but above average the previous month.All groups were happy with the current hydrological situation (no risks currently) although water resource managers stated that rainfall deficiency in the background should be kept in mind due to future drought potential.

Participant responses from Stages 1 -3
The findings from each stage of the activity are presented below.At no point did participants ignore the SHF information (no black stickers were placed on the maps), which matched previous discussions about organisations current use of SHF (Sect.

2.3.2).
Colour-coded decisions made by all participants (calculated by counting the stickers on the A1 catchment maps) are represented as pie charts.An accompanying bar chart details the breakdown of choices made by each participant and their specific role in the water sector (Fig. 8 a-c).Quotes and information in the text are taken from discussions recorded on the day and empathy mapsthese are presented for the 3 groups (forecasters, groundwater hydrologists and water resource managers).

Stage 1 -Hydrological Outlook UK
General consensus was for normal or above-normal conditions over the next 3 months, however the information was "too vague to be actionable".Forecasters and groundwater hydrologists were more likely to discuss the situation with colleagues and keep an eye on the situation (green / blue) although there was some disagreement about the level of risk.Those involved in water resources, water supply, navigation and waste-water operations (water resource managers) identified no risks requiring action (blue) (Fig. 8a).

Discussion
Our decision-making activity was designed to help understand how different water sector users engage with and act on SHF at a local level.The SHF for the three activity stages represented an extreme flood event between November 2013 and February 2014.There was clear evidence that more confident (sharper) and locally tailored forecasts led to increased levels of decision and action, although water sector users did not respond uniformly.Forecasters and groundwater hydrologists were most likely to inform other organisations, request further information about the situation and implement action, while water resource managers more consistently adopted a 'watch and wait' approach.In this section, the results are discussed in more detail and the findings are placed into the wider context of policy, practice and next steps based on discussions captured during the focus group.

Operational SHF systems can support decision-making and uncertainty is expected
Throughout the focus group, participants expressed positively the potential for SHF to deliver better preparedness and earlywarning of flood and drought events, and the benefits associated with more consistent management of water resources, whilst recognising that low skill and coarse resolution are current barriers to use (see also Soares and Dessai, 2015;Soares and Dessai 2016;Vaughan et al., 2016;Soares et al., 2018).These benefits and barriers were demonstrated during the activity as participants increased their level of decision-making in response to the more confident and locally-tailored forecasts presented: Stage 1 Hydrological Outlook UK > Stage 2 EFAS-Seasonal > Stage 3 'Improved' EFAS-Seasonal.
Hydrological Outlook UK is the first operational SHF system for the UK and was the product that participants were most familiar with, likely due to its partnership set-up (Prudhomme et al., 2017).All groups indicated that the regional focus of the maps i.e., the whole Thames basin, and lack of resolution and certainty as to the trajectory of the upcoming hydrological conditions, limited their ability to make informed decisions.No participants however, ignored or dismissed the information despite there being no perceived risk.All agreed that on a day-to-day basis, Hydrological Outlook UK serves as a useful outlook tool when supplemented with additional sources of information including water situation reports (UK Gov., 2018) and other hydro-meteorological forecasts.As of 2017, exactly how the water sector uses Hydrological Outlook UK in practice had yet to be assessed (Bell et al., 2017) and here we provide a first step towards answering this question.
Stage 2 (EFAS-Seasonal) also represented an operational forecasting system designed to run at the scale of the whole Thames basin akin to Hydrological Outlook UK.The forecasts however, were presented at a catchment level on a month by month basis to provide a more localised outlook.This finer spatio-temporal resolution allowed participants to supplement the SHF with their knowledge of local hydrogeology and other risk factors to identify those catchments where attention would likely be most needed.This led to increased levels of communication within organisations, even though the overall hydrological outlook was very similar to that observed at Stage 1 (uncertain but with indication towards normal-high flows).
The use of large scale (regional or global) operational forecasting products that trigger worthwhile actions at the local level has been demonstrated at shorter lead-times (e.g. de Perez et al., 2016).While the development of higher resolution seasonal meteorological forecasts and better representation of the coupled system and initial conditions are expected to lead to improvements in SHF (Lewis et al., 2015;Bell et al., 2017;Arnal et al., 2018), we pose the open question about whether operational systems such as Hydrological Outlook UK already have the potential to support better communication and decision-making if they could be presented at a more local scale?This would require careful communication of the uncertainty, reliability and skill of the forecast, and how to do this effectively is a topic of current interest in meteorological and hydrological forecasting (e.g.Ramos et al., 2013;Vaughan et al., 2016;Fry et al., 2017).Although communicating uncertainty was not a specific focus of our activity, one key message from the focus group was that "uncertainty is expected" with SHF and water sector users would engage with a local forecast, even if they chose not to act on it.As pointed out by Viel et al., (2016), "low skill" is not the same as "no skill" and a SHF which may have minimal value from the perspective of a scientific researcher, can sometimes elicit significant interest from the view of a water sector user who is familiar with the area.Importantly, it should also be noted that although no measures of forecast skill and quality were included in our activity, participants only expressed a need to verify the quality of the forecasts at Stage 3. In discussions as to why this was the case, the forecasters and groundwater hydrologists stated that holding internal briefings and increasing awareness of 'at risk' catchments are suitable low-cost actions when dealing with a SHF that indicates some degree of risk, even if the information is uncertain and unverified.At Stage 3, to obtain such a confident SHF was well-beyond current operational standards, thus its reliability was questioned.Participants did agree however, that even in the absence of information on forecast quality, a sharper, more confident forecast that indicated high potential flood risk would be more likely to provoke a response than a dispersive one, even if the maximum of the forecast ensemble indicated values of comparable magnitude in both cases.

Interactions with SHF are user-specific and should be tailored accordingly
The manner in which users approached and used SHF differed markedly depending on the perceived severity of the flood event; the responsibilities and risk appetite of an organisation; and the local knowledge and experiences possessed by the individual (see also Kirchoff et al., 2013;Golding et al., 2017).Forecasters and groundwater hydrologists displayed the lowest risk appetite, admitting that they were likely to err on the side of caution to avoid negative media impacts, economic damages and loss of trust by the public.While a flood event is less of an immediate issue for water resource managers, secondary effects relating to closure of canals (navigation), turbidity and sewer surcharge (waste water operations) did invoke action where there was potential to impact on the public.Participants were notably proactive where they'd had previous experience of extreme events e.g.forecasters analogies with the 2007 floods (Chatterton et al., 2010), or had been witness to poor management, e.g. the waste water operations manager recognised high potential for groundwater flooding and sewer surcharge at one month's lead time in the Evenlode, Cherwell and Colne (Fig. 7).
"Based on previous operational issues, I'd advise pre-emptive actions such as the cleaning and maintenance of pumping stations for these catchments" -Wastewater operations manager at Stage 2 / 3.This highlights the value of retaining institutional memory where possible (see also McEwen et al., 2012) and being aware of organisations or individuals pre-determined positions or perceived self-interests which may largely be founded on previous experiences (Ishikawa et al., 2011).
It' is important to note that while this activity focused on a flood event, decisions made by the groups would almost certainly have differed if the SHF had indicated drought conditions.The impacts of drought have the potential to affect larger areas, for longer (Bloomfield and Marchant, 2013), notably with respect to agriculture (Li et al., 2017), reservoir management (Turner et al., 2017) and navigation (Meißner et al., 2017).The difference in response between water sector users supports the notion that tailoring SHF information to specific user-groups will improve uptake and ability to inform decision-making (Jones et al., 2015;Lorenz et al., 2015;Vaughan et al., 2016;Soares et al., 2018); an area currently being explored by the IMPREX Risk Outlook (IMPREX, 2018).

Communication is both a barrier and enabler to decision-making
Communication is one of the most frequently identified barriers when it comes to uptake and use of seasonal meteorological and hydrological forecasts (Soares and Dessai, 2015;Vaughan et al., 2016;Golding et al., 2017;Soares et al., 2018).
Discussions captured during the focus group and indicated on some empathy maps identified two key communication barriers in the West Thames: 1) between water sector users themselves and how they interpret and communicate SHF information and 2) a disconnect between scientists developing the forecasts and those involved in policy, practice and decision-making.
All groups said they felt better able to interpret and communicate the messages when presented with a range of complementary forms of SHF information including maps, hydrographs and text, with maps being of particular value.This supports findings by Lorenz et al., (2015) who identified clear differences in users' comprehension of and preference for visualisations of climate information.Mapping information was also found to be important in the survey by Vaughan et al., (2016), while numerical representations were preferred over text and graphics in the study by Soares et al., (2018).Many participants said they would feel better prepared and able to discuss upcoming hydrological conditions if SHF information was visualised in a variety of ways and regular engagement was made a routine part of their job (see Sect 5.4).
A number of participants also felt that scientific improvements and developments to SHF are not being adequately communicated to those involved in policy and practice.General consensus was that knowledge exchange events and information sharing services through projects such as IMPREX are an excellent way of addressing this disconnect.
Presentations during the focus group shared findings from other projects including the 'European Provision Of Regional Impacts Assessments on Seasonal and Decadal Timescales' (EUPORIAS) (Met Office, 2016), the 'End-to-end Demonstrator for improved decision-making in the water sector in Europe' (EDgE), 'Service for Water Indicators in Climate Change

Learning outcomes and future considerations
Encouragingly, we identified that SHF are being used and participants agreed that the decision-making activity was an entertaining platform for fostering discussions which complemented their everyday work and general understanding of SHF.
From the participants' perspective, learning outcomes included knowing more about the ongoing scientific developments in SHF and a better understanding of how different organisations in the West Thames water sector are using SHF.Many also expressed that the activity and focus group discussions enhanced their ability to think about possible decisions and actions that may be taken in the future.As the activity developers, we found that the group discussions stimulated participant's motivations and interests more so than would have been achieved by asking participants to engage on an individual basis.We also advocate the use of empathy maps or others forms of obtaining a written record of participant thought-processes in addition to their decision choices.
Our activity was designed to provide a first insight into the current state of play regarding SHF in the West Thames.
Although 11 participants was a small sample size, they represented an important and well-balanced mix of water sector decision-makers in the West Thames.The only exception was the agricultural sector who could not attend and thus it would be interesting to capture this perspective with ongoing research (e.g.Li et al., 2017).We also recognise the possibility that those who took part had a vested interest in SHF, however, we did encourage participants to attend even where they had no background knowledge or experience of SHF.Finally, we advocate that others conducting a similar activity may wish to consider whether participant interpretation can be subconsciously influenced by the information provided.For example, flood risk maps were provided as part of the background context, but may have inadvertently led participants to consider the upcoming forecasts with respect to high-flow event.Likewise, there is potential that the 3 month SHF (Stage 1) may have been interpreted differently to the 4 month forecasts (Stage 2 and Stage 3) and we do not know the degree to which individuals may have been swayed to place a particular colour on the map based on the conversations they had with their group members (and how big an influence such conversations play in real life).Discussions with the participants at the end of the activity with respect to these points would have been helpful.

Conclusions
Key findings were that engagement is user-specific and SHF have the potential to be more useful if they could be presented at a scale which matches that employed in decision-making.The ability to interpret messages is aided by complementary forms of SHF visualisation that provide a wider overview of the upcoming hydrological outlook, with maps being of particular value.However, improved communication between scientists, providers and users is required to ensure that users are kept up to date with developments.We conclude that the current level of understanding in the West Thames provides an excellent basis upon which to incorporate future developments of operational forecasts and for facilitating communication and decision-making between water sector partners.

Fig 1 :
Fig 1: Location and lithology of the West Thames and its 10 main river catchments

Fig. 5 :
Fig. 5: UK 3 month outlook maps from November 2013 (colours based on the percentile range of historical observed values).(a) Regional river flow forecasts created from climate forecasts.(b) Groundwater level forecasts at 25 UK boreholes created from climate forecasts (CEH, 2013).
Fig. 6: 4 month hydrological forecasts from EFAS-Seasonal (Stage 2).(a) Ensemble hydrographs for streamflow (light blue) and groundwater levels (dark blue) for the Lower Thames (LT) catchment.Exceedance thresholds (based on records from 1994 -2014) are shown as Q10 = dashed line and Q50 = dotted line.(b) Choropleth map shows the maximum probability that the full hydrograph ensemble for a catchment exceeds the Q10 streamflow threshold in a given month.

"
Fig 8: Summary of decisions and actions taken by different water sector personas based on (a) Hydrological Outlook UK; (b) EFAS-Seasonal; (c) 'Improved' EFAS-Seasonal.Blueno notable risk, greendiscuss internally; yellowdiscuss externally and seek advice; redimplement action.Refer toTable 12 for full colour code descriptors.