Snow melt and hydrometric observations – perspectives for flood forecasting

In the second of a series of articles on how the mountains of Scotland influence our approach to monitoring and flood forecasting, Dr. Andrew Black writes about the role of hydrometric observations.

Figure 1. The River Feshie peak flow of 18 December 2018 is ranked 2nd equal in the Feshie Bridge gauged record commencing 1992. The water on the rising hydrograph was visibly filled with ice and resembled a drink of coke! Photo: Julian Scott

“Continually rising global temperatures over the coming century might easily be associated with a lessening in the importance of snowmelt in many parts of the world. However, the ‘Beast from the East’ of 2018 and the wintry conditions as I write this in early 2021 show that snow and its subsequent melt very much remain part of Scottish climate for the time being, especially in mountain and upland landscapes, from where Scotland’s largest river systems flow. It is not uncommon for large floods in Scottish rivers to be caused partly by snowmelt (Figure 1).

This article seeks to examine the potential benefits and pitfalls of using tipping bucket rain gauge (TBR) data from higher elevations for flood forecasting purposes during snowy conditions.

Observations during snowfall

Observations may be direct or indirect. The most direct observations can be obtained from instruments designed specifically for monitoring snow water equivalent or snow pack depth (Figure 2). Weighing gauges and acoustic distance measuring equipment is relatively rare and, even when they are available, they provide information only for one particular point, and may reflect local influences such as aspect combined with snow blow after initial snowfall. Webcams such as those hosted by Traffic Scotland and Winterhighland provide visual assessments which can be useful for assessing conditions, especially the former which include infrared capability for night-time.

Figure 2. Instruments for monitoring lying snow: (left) SEPA’s Snow Scale with adjacent Ott Pluvio2 at Cairngorm Mountain (right) acoustic distance sensor being used for snow depth measurement

By way of indirect observations of snowfall, TBRs are more commonly available, so there are advantages in terms of geographical coverage, but users of the data from these gauges must beware of possible problems arising from under-catch and the timing of melt. A helpful early sign of snowfall is the occurrence of rainfall at low altitudes while adjacent higher-altitude TBRs record minimal or zero precipitation. Reliable air temperature data helps in interpretation. A good example is shown in Figure 3: intense rainfall at low elevation was confirmed by eye-witnesses, while adjacent gauges on higher ground registered no tips.

Figure 3. Time series of TBR accumulations in the Feshie catchment allow differences between low and high-elevation sites to be identified and snowmelt anticipated. A rainfall of 41.8 mm over 10 hours on 15/12/2018 at low elevation without corresponding receipts being recorded at higher elevation pointed in real time to the possibility of substantial accumulations of snow, and duly led to substantial melt on the morning of 18/12/2018

Observations during melt events

Problems may arise with use of TBR data during snowmelt events owing to possible under-catch of snow, due to aerodynamic effects and also the timing of recorded gauge tips. However, some gauges may provide an insight into imminent high runoff on at least some occasions. The same Feshie snowmelt event of December 2018 is illustrated in Figure 4 showing the lag between TBR accumulations at a range of elevations from 260 – 900 m OD and river levels at a similarly wide range of elevations.

While it may be interesting to examine the lag time in various ways between TBR accumulations and river level responses, one gauge in Figure 4 stands out – an ARG100 gauge at 520 m which was situated on the east side of a small hummock in the lee of prevailing winds – not normally best practice for gauge siting (and since decommissioned). This gauge began to record at the same time as other gauges, but had recorded 33.2 mm in the first 12 hours of the event, by which time none of the other 5 gauges had recorded more than 9.2 mm. Similar tendencies are found in other snowmelt events, suggesting the 520 m gauge may be able to offer some ability to indicate the onset of a major snowmelt event: the gauge possibly benefits from being especially exposed, allowing faster melt than in other gauges, while also giving a crude indication of the amount of snow in the catchment.

Figure 4. TBR accumulations and river level response, Feshie catchment, 17-19 December 2018.

It must be more typical to regard protruding rain gauges as a problem leading to under-recording of snowfall amounts and unduly early recording of melt compared with the majority of the snowpack. Figure 5 illustrates a set of gauges in melting snow at the Talla gauge comparison site at 425 m OD. At the time of the visit, none of the above-ground gauges contained any snow, save for the gauge buried in the pit, with its rim at ground level. The above-ground gauges clearly provided a misleading characterisation of the melt process – of course, the gauges are rain gauges not snowmelt gauges.

Figure 5. Preferential melt through TBRs at the Talla Water gauge comparison site: above-ground SBS500, Casella and ARG100 gauges have lost all their funnel snow to melt, while the snow pack above a pit-installed ARG100 remains in-tact (snow partially cleared for the photo).
Andrew is a Reader in Hydrology at the University of Dundee. He is a hydrologist with 30 years experience, mostly in Scotland’s rivers. Having started his career looking at statistical methods for estimating flood risks he has more recently developed expertise and strong interests in applying field measurement techniques to address important questions in hydrology.

Concluding remarks

The Feshie network of TBRs captures data from an unusually large range of altitudes, from 260 – 900 m OD, and provides an abundance of data with which to examine snowmelt events. This brief review illustrates how, by extending from the valley bottom up to some of the highest ground, it is possible to obtain a much fuller, if inevitably imperfect, portrayal of snow-related fluxes from the Cairngorms, which would be of value to a flood forecasting duty officer.

TBRs protruding above the ground surface may be unrepresentative of the general timing of melt in a snowpack, but might on some occasions be of value in giving early warning and an approximate snow water equivalent for an imminent snowmelt flood.”


Research sponsors and collaborators at Wildland Limited, Cairngorm Gliding Club, Borders Forest Trust, Newcastle University/Environmental Measurements Ltd and Michael Pollock.

Posted in Flood, Hydrology, Hydrometry | Leave a comment

How do Scotland’s mountains affect its rainfall?

In the first of a series of articles on how the mountains of Scotland influence our approach to monitoring and flood forecasting, Mike Kendon writes about their role on rainfall.

“Much of Scotland’s landscape is dominated by its mountains. Although relatively modest in height, reaching typically around 1000 metres in elevation, they nevertheless have a dramatic influence on rainfall.

Mike Kendon is a climate information scientist at the Met Office National Climate Information Centre. He studied an MSc in Hydrology at Imperial College, London and joined the Met Office in 2009. He helps develop and maintain systems to monitor the UK’s climate based on observations from the surface network of weather stations, and is the lead author for the Met Office’s annual state of the UK Climate reports (Kendon et al., 2020)

Due to its northerly latitude, Scotland is located close to the North Atlantic jet stream and therefore lies in the path of low pressure systems and associated rain-bearing fronts. As the westerly airflow is forced to rise over this barrier, low-level clouds are formed over the mountains. When precipitation from higher level ‘seeder’ clouds fall through the lower level ‘feeder’ cloud, the droplets collide and coalesce. This mechanism, known as the ‘seeder-feeder’ effect, is responsible for greatly enhanced orographic precipitation.

As a consequence, the West Highlands receive rainfall in abundance. Annual average rainfall may exceed 3500mm per year, and there may be a ‘day of rain’ (exceeding 1mm) on average almost two days out of every three. However, this orographic effect is highly localised and rainfall rapidly diminishes from west to east across Scotland, so much so that parts of the east coast (such as Angus, Fife and East Lothian) receive less than 700mm – a five-fold reduction – and these areas are among the driest parts of the UK (figure 1).

Figure 1: Annual average rainfall for Scotland based on the period 1981-2010 (source: HadUK-Grid 1km resolution dataset, Hollis et al. 2019)

Scotland’s rainfall is not evenly distributed through the year. The wettest months are from October to March, associated with the normal seasonal southerly shift in the Atlantic storm track – and it is not uncommon for 50mm to fall widely across the West Highlands in a day – locally totals sometimes exceed 100mm. The driest and more settled months tend to occur in late spring and early summer.

The seasonal influence on rainfall tends to reduce moving eastward, as this area increasingly falls within the rain-shadow of the mountains. In eastern Scotland, there may be a greater proportion of convective, rather than frontal rainfall, and rainfall totals on the east coast are more evenly distributed through the year. Figure 2 shows monthly average rainfall totals moving west to east – Glendessary, West Highlands (annual average 3508mm), Gairlochy, in the Great Glen (2168mm), Dalwhinnie, Central Highlands (1304mm), Braemar, Grampians (932mm) and Inverbervie, east coast (683mm).

Figure 2: Monthly 1981-2010 long term average rainfall for five Scottish stations.

The broad west-east contrast shows the most obvious influence of Scotland’s mountains on rainfall, but in reality the picture is much more complex. For example, in an easterly airstream the pattern may be reversed, with the wettest weather across the Grampians and the West Highlands falling into the rain-shadow. Scotland’s complex coastline also influences its rainfall patterns; for example coastal fringes of the West Highlands and Islands may be much drier than mountainous areas a few miles inland. Across upland areas, much of the precipitation falls as snow during the winter months, and lying snow often lasts well into the summer months. Over the mountain summits, snow may fall (if not lie) at almost any time of year.

Understanding Scotland’s rainfall distribution is important for many sectors – e.g. water resources, flood risk management – agriculture, ecology and hydro-power. A network of several hundred rain-gauges across the country provides data to help us monitor rainfall across Scotland, and understand trends, variability and extremes. Annual precipitation is, in general, dominated by a large annual variability about a relatively stable long-term mean, but with an increase in the latest few decades (Figure 3). Understanding these patterns in rainfall and how these may be changing as our climate changes, remains an important area of ongoing research for the Met Office and other research institutions. But for Scotland in particular, what makes monitoring precipitation so fascinating is the complexity of the climate across relatively short distances, and this complexity is in large part due to the influence of the mountains.”

Figure 3: Annual rainfall for Scotland, 1862 to 2019 (source: HadUK-Grid 1km resolution dataset, Hollis et al. 2019)


Hollis, D, McCarthy, MP, Kendon, M, Legg, T, Simpson, I. HadUK‐Grid—A new UK dataset of gridded climate observations. Geosci Data J. 2019; 6: 151– 159.

Kendon, M., McCarthy, M., Jevrejeva, S., Matthews, A., Sparks, T. and Garforth, J. (2020), State of the UK Climate 2019. Int J Climatol, 40: 1-69.

Posted in Climate, Hydrometeorology | 1 Comment

Operational forecasting and hazard early warning systems: Call for Abstracts at #vEGU21!

With a couple of weeks to go before the EGU abstract deadline (13th Jan. 2021, 13.00 CET), it’s time to consider submitting that abstract! If you working in the field of operational forecasting and warning for hazards then consider this unique international session EGU session.

This interactive session aims to bridge the gap between science and practice in operational forecasting for different water-related natural hazards. Operational (early) warning systems are the result of progress and innovations in the science of forecasting. New opportunities have risen in physically based modelling, coupling meteorological and hydrological forecasts, ensemble forecasting and real time control. Often, the sharing of knowledge and experience about developments are limited to the particular field (e.g. flood forecasting or landslide warnings) for which the operational system is used.

The focus of this session will be on bringing the expertise from different fields together as well as exploring differences, similarities, problems and solutions between forecasting systems for varying natural hazards. Real-world case studies of system implementations – configured at local, regional and national scales – will be presented, including trans-boundary issues. An operational warning system can include, for example, monitoring of data, analysing data, making forecasts, giving warning signals and suggesting response measures.

Contributions are welcome from both scientists and practitioners who are involved in developing operational forecasting and/or management systems for water-related natural or man-made hazards, such as flood, drought, tsunami, landslide, hurricane, hydropower, pollution etc.

This session (HS4.4/NH9.13) will be held during 19 – 30th April 2021 at the General Assembly of the European Geosciences Union in Austria (Vienna) as a virtual conference.

Posted in Conference, EGU, Forecasting, Natural Hazards | 1 Comment

New coastal flood forecasting system for Eilean Siar goes live!

This week sees the launch of two new Flood Warning schemes by SEPA. One is a scheme addressing fluvial flooding in Aberfoyle, in the southern Highlands. The other, discussed here, is for the Outer Hebrides (also known as Eilean Siar in Gaelic). The scheme has been developed over the past year or so and presented a number of forecasting challenges. 

Eilean Siar/ Western Isles

The Outer Hebrides consist of around 60 islands, of which 15 are inhabited. Many of the islands are linked by causeways. The coastline is complex, with many inlets, small islands and sand bars. The weather can be extreme (for the UK), with the west coast fully exposed to Atlantic storms. Travel around the islands is often severely restricted due to flooding of causeways by high water levels or wave action.

L: Baleshare causeway 2020. R: 2014 event in Stornoway

A forecasting system was developed for the islands in a project supported by coastal modellers at JBA Consulting, for use in SEPA’s Delft FEWS flood forecasting platform. It involves still water level (SWL) calculations for 44 forecast points, and automated look up tables to calculate overtopping rates, these rates being generated from offline modelling. In partnership with the Met Office, tidal surge, wind and wave short term forecasts are added to long term astronomical tide series to determine the values at the forecast points, which are compared to warning thresholds derived during the project.

Schematisation of the forecasting system. Image from JBA Consulting

The scheme was launched on 10 November 2020 and will deliver flood warnings to emergency responders and members of the public. The current forecasting outlook suggests it’s likely to produce forecasts of exceedance over the coming days.

Image from FEWS Scotland showing possible threshold exceedance using ensemble level forecast

Messages will be issued via SEPA’s Floodline service, as appropriate. Forecasts will also inform the coastal risk assessment for the forecasting service’s Flood Guidance Statement.

Launch of this new system is another milestone in our aim of working with the Met Office to extend flood warning lead times and understanding uncertainties in forecasts, through the use of ensemble forecasting.

Posted in Uncategorized | Tagged , , | Leave a comment

Exploring an impact-based approach to flood forecasting

As we’ve previously reported on here, many parts of the North and West Highlands of Scotland can be spared major flooding impacts even when significant rainfall (100 to 200 mm) can fall.

In October 2018, although the town of Oban was impacted by localised flooding, many areas were spared more property flooding. Likewise in March 2015, a hydrologically significant event across much of the Highlands again only led to minor property flooding. However, it’s often the road and rail infrastructure that is most significantly impacted by these types of events as we recently observed.

A frontal wave affecting parts of western Scotland overnight on the 12 and 13 September brought with it strong winds and orographically enhanced rainfall across the mountains.  Quite widely 50 to 60 mm of rain was recorded with peaks of 80 to 120 mm recorded across Skye and Lochaber.  The forecast (see below) was for widespread minor flooding to occur, but the guidance also suggesting a very low likelihood of more significant impacts.

The Area of Concern as issued by the Scottish Flood Forecasting Service on Friday 11 September

During the Sunday morning reports started to materialise of major disruption across the transport network. Impacts included flooding and disruption on the Crainlarich to Fort William rail line at Bridge of Orchy and flooding of main trunk roads including the A890 and A82 in Skye and Lochaber and the A8 in Renfrewshire.

Attempting to link our forecasting capabilities with the transport receptors is the focus of a new area of work for us.  In partnership with Transport Scotland, we’re now exploring ways of developing an impact-based forecasting approach for the Trunk Road network in Scotland. 

The research will look at how best to predict the impacts of flooding across the vulnerable parts of the transport network and explore whether empirical links are sufficient or whether these links need to be developed through pre-simulated scenarios as set out in the recent research on towards improved surface water flood forecasts in Scotland.

Posted in Forecasting, Partnerships, Research, Surface Water | Leave a comment

Operational Forecasting Across the Globe: Sharing Science at EGU20

For several years the ‘operational forecasting and warning systems‘ PICO session has been a popular feature of the EGU hydrological sciences program.  However, with restrictions on travel and meetings in place due to Covid-19, the whole of the 2020 Assembly took on a new form – virtual. #shareEGU20 was a week long event taking place over the first week of May and freely available for anyone to join. Over 18,000 abstracts were published with 63% being presented virtually through various online displays, all of which were supported by over 200,000 discussion point or messages being posted by the virtual delegates.

As for the operational forecasting session, alongside the abstracts, 12 displays were presented with 120 people joining the discussion. To help with the facilitating, the conveners connected via video conference between New Zealand, Sweden, the Netherlands and Scotland to steer the presenters and discussion.

The diverse range of operational forecasting systems that were presented, included the implementation of an impact-based forecasting system in Barbados (University Corporation for Atmospheric Research, US) to experiences of co-developing flood forecasting in West Africa (Jafet Andersson, SMHI, Sweden).

Speight et al

Display on ‘Reviewing operational and near operational progress in surface water flood forecasting for urban areas’ by Linda Speight (University of Reading)

Co-convener Céline Cattoën-Gilbert (Hydrological Forecasting Scientist at NIWA) offered her highlights of the session: “Advances are being made in real-time inundation forecasting at national, regional and urban scales. For example, Linda Speight (University of Reading) presented an overview of the latest science, tools and approaches available for monitoring and forecasting flood impacts to inform future developments for the Scottish Environment Protection Agency (SEPA). With empirical-based threshold scenarios, hydrological forecasting chains linked to pre-simulated impact scenario for national or regional scales, or linked to real-time hydrodynamic simulation at the urban scale, there is not a one size fits all solution – a reflection of different decision-makers’ needs. The work presented by Alessandro Masoero (CIMA Research Foundation) follows the last approach by building an early warning system in Guyana at the national scale with real-time hydraulic inundation forecasts at selected locations. The adoption of real-time inundation forecasting approaches will require strong interdisciplinary efforts to support informed decision-making.”

As for the presentation on surface water flood forecasting, the delegate discussion focused on: whether there was a focus on a means of communicating the forecasts; on how pre-simulated impact scenarios compensate for the absence of real time modelling; and whether there were attempts to link these to crowd sourcing of information. More on this research and how scientists are tackling surface water prediction in this Conversation article from December 2019.


World-wide hydrological predictions using the HYPE model by SMHI as presented by Pechlivanidis (SMHI)

Ilias Pechlivanidis (Scientific Leader in Forecasts of Water Variables at SMHI) and another co-convener, said “a big effort is currently  given on setting up early warning and climate services at the large scale with the objective to address local needs. Among the various interesting studies, Katie Smith (CEH) presented a proof-of-concept global hydrological outlook system developed by many international partners to increase resilience to hydro-climatic risks. The system is setup for the WMO under the HydroSOS topic to provide an overview of historical and forecast hydrological status. My co-author Jafet Andersson presented a number of operational hydrological forecasting services with applications in European, Niger river basin and global domain. The services provide information at medium-range, sub-seasonal and seasonal time horizons, whilst strong user engagement was followed in the co-design of the services addressing further the local user needs.”

CEH water resources

The UK Water Resources Portal as presented by Lucy Barker (UKCEH)

Current talk of a potential drought in the UK links well to the presentation by Lucy Barker (UKCEH) on dynamic real-time hydrological status monitoring in the UK. Lucy’s presentation highlighted how understanding the current hydro-meteorological situation is critical to manage extreme events and water resources, illustrating how a new portal has been developed to enable dynamic, interactive, real-time access to hydro-meteorological data. This article highlights how the tool is providing a hydrological outlook for the current situation. There were other abstracts that were contributed and all of these and the displays for this session are available on the EGU20 pages.

Yes, I’m sure we’ll all admit to missing sharing stories and capabilities in the more familiar surroundings of the Austria Center Vienna, but has #shareEGU20 set a new benchmark in conferencing?

Posted in Communications, Conference, EGU, Forecasting, Hydrometeorology, Research | Leave a comment

Royal Meteorology Society Award for the Scottish Flood Forecasting Service

Scientists from Met Office and SEPA working within the forecasting service have been recognised at the Royal Meteorological Society Awards for their work on the ‘Surface Water Flood forecasting in Urban Communities’ project. They, along with their colleagues from The James Hutton Institute, CEH Wallingford and CPAESS – UCAR, USA, received the Innovation Award which is based around innovation in meteorology, with a particular focus on business and/or public impact. It recognises people, projects or programmes within the academic, scientific or business communities who have made significant contributions to educating, informing or motivating organisations in their response to meteorological challenges. The full list of award winners can be seen here.

Example output from the surface water impact model for Glasgow

We reported on the project in a number of blog posts around the time of the initial project in 2014. A probabilistic surface water impact forecasting tool was developed for central Glasgow and used operationally during that summer’s Commonwealth Games. The model is still in use by the forecasting service, and the principles behind it have since been applied elsewhere in the UK.


Posted in Partnerships, Research, Risk communication, Surface Water, Uncategorized | Leave a comment

Scottish Flood Forecasting Service Annual Report

This week the Scottish Flood Forecasting Service (SFFS) publishes its report for 2017-2018.

We wanted all our daily Flood Guidance Statement customers, SFFS partners and supporters to know more about our operational activity and key learnings from April 2017 – September 2018, completed and planned improvements to our products and services and our future aspirations for improved flood forecasting in Scotland.

You can access the report on SEPA’s website here.

The report contains a summary of the period, with three events highlighted for further discussion. We also report on:

  • Customer feedback, which has led to improvements in our risk matrix graphics and descriptions of impacts;
  • Our Science Development Plan, which supports increased forecast lead times and th upgrade of surface water forecasting capabilities;
  • Plans for a public facing flood forecast product, to improve awareness of flood risk in Scotland.

This is intended to be the first in a series of annual reports into the activities of the service.

Posted in Communications, Partnerships, SFFS | Leave a comment

World Meteorological Day 2019

Happy World Meteorological Day!

Saturday 23 March is World Meteorological Day 2019 and this year the World Meteorological Organisation has named the theme – ‘The Sun, the Earth and the Weather’. In the Scottish Flood Forecasting Service (SFFS) the weather services are provided by the Met Office, with hydrology provided by the Scottish Environment Protection Agency (SEPA), so today is a good day to highlight this partnership.

The SFFS has been going since 2011. From the start it has been a working partnership between SEPA and the Met Office. Operationally we work together to produce the daily Flood Guidance Statement, and provide forecast information to aid with the issuing of Flood Alerts and Warnings by SEPA. Additionally, the partnership is active in flood forecasting research, and facilitates the exchange of data between the organisations, as well as in training, development and public awareness of flood risk.

The railway line at Saltcoats in Ayrshire is lashed by the effects of Storm Callum in October. Flood Warnings were in force for this coast. Photo from BBC News.

Over the past year we have issued a Flood Guidance Statement every day, 54 of which have shown heightened flood risk. SEPA has also issued 215 Flood Alerts and 200 Flood Warnings in Scotland during this period, all of which required accurate meteorological input.

The work of the service, as highlighted in all the earlier articles on this site , would be impossible without the high quality meteorological data and expert advice produced continuously every day by the Met Office. For more information on recent weather and the science and history of meteorology, see the podcasts at this link.


Posted in Communications, Forecasting, Weather prediction | Leave a comment

Using the flood risk matrix – two October events

Two widespread flood events affected Scotland in October. One affected a very large area of the West Highlands, and the other, related to Storm Callum, widely impacted England and Wales, but just clipped the south of Scotland. In both cases the flood risk matrix on the Flood Guidance Statement was put to good use to indicate the potential severity even when there was still uncertainty on where the impacts would be.

Mon 8th, Tues 9th October

There was widespread heavy rain over two days, predominantly in the West Highlands as far south as Ayrshire.

Extract from Flood Guidance Statement on Mon 8th Oct

The risk was highlighted at the end of the previous week, with FGS yellow (minor impacts likely) from Friday. Upgraded to significant impacts likely (amber) on Monday for area W of Glasgow due to rain fallen plus forecast. In the event this was downgraded the next day, but there was a widespread event with rainfall in excess of 200mm in some places. Mainly road flooding, landslides etc. The flow return period was probably <5yr in most places, though maybe locally more severe.

Flooding in Oban

This was well forecast with plenty of lead time, though there were few areas at risk and warning schemes in the area most impacted by rainfall and rivers. Amber was based on information available at the time, downgraded when forecasts came down.

Sat 13th October – “Storm Callum”

Widespread heavy rain was forecast to hit northern UK several days in advance. On Tuesday we highlighted significant impacts possible (yellow) for south of Scotland on Saturday. In the event the bulk of the rain was across the border in England and Wales. Some gauges in Scotland recorded around 60-70mm in 36 hours. There were high flows in the south of the country with minor impacts, and Hawick approaching flood warning level.

Extract from Flood Guidance Statement on Thurs 11th Oct

This was well forecast in a UK context and we were correct to highlight possibility of significant impacts, which did occur elsewhere in the UK.

In both events, the Met Office had clear sight for several days that heavy rain was coming, allowing us to give Day 5 notification of potential impacts. We were then able to adjust the likelihood, location and expected level of impacts by using different parts of the matrix as we got closer to the event.

Posted in Uncategorized | 1 Comment