From flood forecasting to fish stocks, Owen Lewis reveals how the Environment Agency is using modelling in its day-to-day activities
The Environment Agency (EA) relies on environmental modelling to assist with most of its actions to improve the environment.
So, what do we really mean by modelling? A Google search unearths a mind-boggling array of answers. A truly inspiring definition remains evasive, however. The best one so far is: “modelling allows us to take a glimpse at the future through fragments of truth”.
Being able to take a glimpse at the future is essential to the EA. Models are integral to the agency’s success. Each day it uses models to make decisions that have an impact on people and the environment. Modelling of hydrology and hydraulics enables the EA to confidently inform people and businesses about the risks of flooding for their particularly location.
Modelling water chemistry and ecology helps the agency, water companies, farmers and local communities to agree on the measures that need to be put in place if there are to be noticeable improvements in the health of wildlife in the rivers of England and Wales.
25k and counting
EA staff have access to more than 250 different types of modelling tools. These tools help them make decisions at various levels, from using global climate data to predict how changes to our climate could alter the variety and abundance of the nation’s migratory fish stocks, to the models used to indicate when to close the Thames Barrier (pictured above). Looking at all the environmental media across all scenarios, it is likely that the EA and its partners have about 25,000 models in use.
The agency has a blend of the old and new. Many of its older models continue to stand up to the test of time. The water quality planning (WQP) mass balance model, for example, has been around since the early 1990s and uses Monte Carlo simulation – computer algorithms that use repeated random sampling to produce results – to predict the impact of individual water industry discharges on river quality.
The EA uses this information to set the environmental headroom that water companies must operate within. Every five years, as part of water-company planning cycles, the WQP is used in conjunction with its cousin SIMCAT (simulation of catchments) to agree £20 billion of environmental improvement schemes. This process is now in the fifth planning cycle and, thanks in no small part to the evidence that modelling has provided, over those 25 years there has been a remarkable and sustained improvement in the chemical composition of rivers.
Experienced agency staff say that it has taken two decades of continual improvement to get the models to a level where they are a trusted and established part of the way the EA operates. But sometimes the agency does not have the luxury of time, and recently has had to respond quickly to new challenges.
The Water Framework Directive, for instance, inspired the EA to take a step change and, as well as evolving water chemistry models so that they can cope with diffuse pollution, it is now placing ecological modelling at the heart of its planning decisions.
The EA is now engaged in making some significant changes to its models. Examples include the fish health model, the Fisheries Classification Scheme (FCS2), which is being enhanced to increase its ability to identify environmental impacts, and the models that predict the health of plants and algae.
These models are providing a new perspective of the environment, taking the EA away from the black and white world of water chemistry towards amassing evidence to show how interconnected the environment is, how decisions taken to manage flood risk can impact on the ecology of the same river many miles away.
Truly integrated catchment planning is an established part of the agency’s modus operandi, but the need to be constantly making the right call for the catchment, all who live and work within it and the wider environment, on a 24/7 basis, is a gargantuan challenge.
The local touch
“Rubbish in, rubbish out” is one of the most common phrases spoken by a modeller. The phrase is used with reference to the data a model needs in order to make a prediction. A model can be fed all the data in the world, yet still produce a rubbish answer.
Time after time, EA modellers find out that the best people to seek advice from about fine-tuning models are within the local communities. But do they really want to get involved in modelling exercises? The EA has consistently found that the answer is yes.
Whether that is because the Big Society ethos promoted by the government has really taken hold, or because there has been an increase in environmental awareness, the EA is not sure, but it realises that the potential benefits of engaging local communities is huge.
One of the many examples is the Peterborough visualisation model, created by Peterborough City Council, capitalising on strong community interest. The model will help the local population to monitor environmental performance and identify sustainable solutions to solve local problems.
It works by combining an impressive range of environmental performance data, covering energy, water, transport and waste systems, gathered from businesses and verified by local members of the public. As modellers, we need to accept graciously our blind spots and listen to the voice of local communities.
The person living in the house with the stream at the end of the garden might not have a working knowledge of the agency’s two-dimensional flood simulation model, but they will have a much better knowledge of how close the water got to their back door last time, which is vital knowledge.
“A model is never right, it’s just less wrong than the alternative” is another classic from the modellers’ phrase book. But the homeowner with a stream at the bottom of their garden can rightly expect to receive an accurate and timely warning before the next flood occurs.
The summer floods of 2007, and the flooding in Cumbria in 2009, have thrust flood forecasting into the spotlight. Since these events, the EA has been working to ensure it provides local authorities, emergency services, the media and local communities with the right information at the right time.
It has formed closer ties with the Met Office and, in 2009, the agency opened a joint Flood Forecasting Centre where experts in hydrometeorology – a new skill, linking weather expertise with river expertise – monitor the development of approaching areas of low pressure as they move in from the Atlantic, which, coupled with knowledge of existing river and groundwater levels, can help to make early predictions of flooding.
In the same way that financial institutions were put under stress tests this summer, in March 2011 the EA worked with local authorities and fellow emergency responders in a national-scale flooding drill called the Exercise Watermark.
The only alternative
Often there is no choice but to use a model. It is impossible for the agency to gather evidence on something that will not occur until next week, so the use of a model is clearly justifiable.
Likewise, modelling is a sensible alternative to some of the logistical hurdles of monitoring. For example, the alternative to groundwater modelling is to pepper the countryside with boreholes so that the EA can track the level and movement of groundwater. If we did that Kent, Surrey and Sussex would have more holes in them than a pincushion.
Often overlooked is the role that models can play in reducing the costs of gathering observed evidence. Modelling and monitoring programmes should be designed together, but balance is key.
Certainly the data a model derives is not normally as accurate as the data gathered from monitoring programmes, but the savings to the taxpayer are significant. And why would you not want to do something in cheaper way if it can make you more, not less, effective?
Do people believe modelling is a cheaper and sometimes a more effective approach? Not always. Trust is hard won and hard to maintain. The EA puts significant effort into ensuring the benefits of modelling are fully maximised and that its modelling is rigorously managed. If it isn’t, then that is when things can go wrong.
For the EA, a wrong decision based on dubious modelling outputs can tarnish its reputation, consume significant resources and create overly cautious attitudes in the long run. The organisation has been speaking to a wide range of businesses and organisations to see how they manage their models.
These include how Lloyds of London is implementing new legislation on the stringent management of financial models, and the way the British Geological Survey and the Met Office ensure their models are trustworthy.
The agency also shares best practice with the US Environmental Protection Agency. Each of its 250 types of models is now under a model management framework, supported by data custodianship principles that other government departments are now becoming interested in.
Known knowns, and known unknowns
Probably – and that word is used deliberately – the biggest challenge modellers face is to get better at communicating the level of confidence they have in their predictions. They already use probabilistic (probability of an event happening), or “ensemble”, modelling techniques in their day-to-day business.
The ensemble approach has been a fundamental part of the quality modelling technique for decades. It is now applied to flood modelling and it is the accepted way of modelling steady changes to the world’s climate.
The way this works, in a nutshell, is by deliberately firing thousands and thousands of random variations in the data at a model so that the probability of one particular outcome is set against a large range of other outcomes.
The sheer amount of data being exchanged can sometimes challenge the agency’s computing infrastructure, something which cloud technology – virtual IT infrastructure – may be able to help with, but the bigger problem for the EA is how to convey confidence to the public, to key decision-makers and even to its own staff.
Being open and honest about the limitations of modelling is the intention; baffling people with science is not. The uncertainty of modelling should be recognised, but should never be a bottleneck to action. This is just as true for decisions the EA takes on environmental improvements as it is for flood forecasting.
The EA can only solve the conundrum by building on the dialogue it is already having with partners and local communities about the level of confidence that people expect to be reached before they are willing to take action.
