Thames into Tyburn

The case for filling in rivers

Many of the world’s largest cities are organised around rivers - rivers that, as Samuel Hughes recently noted, are today mostly empty. The same cities also often have extremely high land valuations, and stretched local government finances. 

Historically, governments often employed a simple solution to these problems - they paved over rivers and then sold the additional land to reduce pressure on housing prices and improve their own financial position. Hence the disappearance of the Tyburn, Fleet and Walbrook. 

We can employ the same solution today. 

We estimate a total cost to reclaim the relevant land with matching sewer capacity within London of £3.13bn. The underlying land value plus a subset of agglomeration, less reductions in amenities, is worth £48bn, resulting in a cost benefit ratio of 6.5:1.¹ The river’s pre-existing volume before entering London is relatively small (it increases in width by a factor of 10 while flowing through), and South East England has been in need of a new reservoir for some time, so the reservoir could just be positioned to capture the existing flow. Any environmental issues would be small, and relatively easy to mitigate. Although the calculations will be undertaken for London, we expect that a roughly similar logic should hold for most large cities around the world.  

Reclaiming the river

We’ll follow a similar strategy as we previously modeled for Dogger Bank: build a breakwater at the points where we wish to interrupt the river’s flow, pump the necessary water out, replace the sewage and drainage with tunnels then re-fill in. Thankfully, the scale will be vastly smaller. The Thames is only 70m wide and 6m deep when it enters London; even when it leaves it barely exceeds 750m with a depth of 17.5m. Assuming that the cost per unit distance of a wall goes as the square of depth - as deeper walls must also be made thicker - this gives a cost of £150,000 for the wall at the upper end of the Thames and £14.1mn at the lower end. Pumping the water out would cost £4.8m. 

Of course, the river currently serves a role both for sewerage and storm drainage - so replacement drainage capacity would have to be built. The river is currently sufficient for this purpose despite also carrying normal flow, so would also be sufficient with strictly less volume. How much? The wettest day in London in the past 25 years involved 42mm of rain; if 70% of water is ultimately diverted to the Thames, this would require 46m m³ of storage plus throughput in a day.² Low-grit water generally shouldn’t move faster than 5.5m/s; a desired maximum system capacity of 80mm of rain in a day gives a required tunnel radius of 7.7m. 

Table 1 (first two columns adapted from Lin and Tsai 2009. Ouyang (2003) was omitted as it modelled no diameter or depth relevant to our calculations)

From here, we can then use existing estimates for the cost of cut and cover sewers. As no digging will be necessary, this does of course strictly overstate total costs. Construction inflation has been roughly double general inflation, and this correction will be applied. This gives a median cost estimate for sewerage capacity of £2.72bn.

In order for the construction to be practical - due to the need to continue potential storm drainage during construction - it would likely in practice be necessary to do a scheme similar to separately draining each half of the river and constructing there. Again assuming wall costs go as the square of depth, this leads to an additional expense of £389mn.³ This gives a total cost to the project of £3.13bn. 

Benefits

We find a £48bn gain from new real estate generation and agglomeration externalities. First, direct land gains. Given the dimensions discussed above, the river will have an area of 21km² within London, assuming a power-law function of river width to distance. The average London residential land valuation is £21mn/hectare (inflation-adjusted). This land will be disproportionately in the lower-average-value east of the city, and on the outskirts, relative to its course; however, other geometric considerations raise the value substantially.⁴

However, riverside properties do command price premiums that would presumably vanish. A home in the South East commands a 46% higher valuation in general if on a waterfront; riverside properties command an average 76% gain vs an average 80% across the entire country, suggesting a 44% gain. Assuming that the valuation premium declines by 50% per property from the waterside, this gives a total 88% gain in neighbouring property values. Under an average property depth of 20m (typical for homes), this gives a total loss of value equivalent to the loss of 2.4km² of land, giving an overall land value of £39bn.

This will also, in addition to releasing a large amount of land, enable substantial agglomeration by reducing inter-city transportation distances. Transiting the river is costly, both leading to a Tube network that connects north and south of the river infrequently and less foot traffic than would otherwise be the case. We won’t model this, but this would also raise the overall benefits substantially. 

We can, however, model the benefits from increasing population density this way. If this land has equal population density to the remainder of the city, then this yields a 1.1% increase in total population, which under an agglomeration elasticity of 0.046 and a 3.5% discount rate yields present value benefits of £9.0bn given London’s GDP of £520bn. 

Are there going to be any other economic costs from shutting down the river? The portions within London haven’t been used for unloading shipping since the Docklands were turned into the city’s premier financial district in the 1980s; commuters and tourists only take 10,000 journeys per day on the city’s public riverbuses, which even if it generates full surplus equal to the £10 ticket/journey (unlikely given the city’s excellent transport networks) is only a cost of £37mn. Some other economic benefits from the river do exist (such as a pub whose main marketing point is being on an island), but these are likely  to be comparable in size.

This gives net economic benefits of £48bn, or about 2% of UK GDP.

Environmental costs

We’ve already accounted for any local externalities: these will be reflected in property value differences of properties closer versus further from the river. However, some benefits the Thames provides apply across the whole city, or even globally - sequestering carbon directly, alleviating air pollution, providing a space for wildlife, reducing erosion and providing cooling - and so would have been ignored by the above methodology. Will the cost of removing any of these be substantial? No. Direct carbon losses will be small, at 472kg CO2/annum (worth £2500 present value), assuming an intensity of 242g CO2/hectare, as for grassland. Any air pollution effects can be mitigated by ensuring a sufficient fraction of the new land is designated as parks. Biodiversity losses can be easily mitigated by purchasing and rewilding farmland elsewhere; the price of pasture in the North East is only 0.08% of London average land prices above. Although a reservoir might be necessary upstream to catch the Thames’ pre-existing throughput, South East England has been in need of a new reservoir for some time regardless, so this adds no additional required infrastructure on what current population trajectories would regardless suggest. 

Might this cause substantial land losses elsewhere by increasing erosion downstream? The 46% of soft shorelines that erode do so at 0.43m/year - and around 17% of overall UK shorelines are eroding currently. Assuming that the Thames estuary is entirely soft, and using a 100km size, implies that even if this scheme would cause erosion to double and if exclusively arable land was eroded (rather than less valuable pasture, or land unsuitable for agriculture) only 0.02km² of land would be lost per year - worth present value only £1.4mn. 

The cooling effect of rivers is large near a river but dissipates very quickly, entirely vanishing after 500-700m - using the lower end to account for the river’s bends gives an area of 68km², or 3.9% of the city, being raised in temperature by 1 degree during summer. Cooling is just below 10% of UK electricity consumption, and per capita varies very little across the UK. A 1.6-degree average rise in Europe boosts cooling demand by 4.4% - but only 5% of households have air conditioning. To upper bound the total welfare loss, if all buildings affected used air conditioning, electricity is 11.5% of UK carbon emissions, so this would only increase UK aggregate emissions by 0.0037% - under a carbon price of $240 only equivalent to £75m of damages present value. 

Conclusion

The intuitive case for covering rivers follows from the reversibility test - few would suggest returning the now-buried rivers to the surface - and it also seems unlikely that, once the Thames was filled in, there would be any suggestion of excavating and flooding the land again, suggesting that if transition costs are small (which we estimate they are) the transition is worthwhile. General aesthetic benefits barely vary our results also - even if the average Londoner valued the river at £100, this would be worth less than £1bn - in the context of £48bn worth of net land gains. The benefits could also be much larger than we estimate - costs per unit space only increase fairly slowly with building height, and when public authorities capture all of the upside they have often historically been more willing to zone for high density (as with the conversion of the Docklands). This could bring benefits equivalent to much more land reclamation, and enable greatly more agglomeration.

Filling in the Thames would thus have a cost benefit ratio of 6.5:1, even ignoring much of the substantial agglomeration boosts. Environmental costs are very small in comparison to the size of the gains, and so are straightforward to offset. We thus have the opportunity to substantially increase the benefit to society from what are today the least-utilised pieces of land in our most expensive cities.

1. We include reductions in amenities in the costs.

2. The requirement will in practice be slightly lower than this due to existing capacity, lower-bounded at 1.5m m³, plus 1.1m m³ from the new Thames Tideway tunnel. 1.5m m³ comes from treating “130km of railway tunnels” as Tube tunnel-sized, the smallest size (tunnels for mainline trains are usually almost twice as large or more) and ignoring the 2000km of connecting tunnels. 

3. Given that such a barrier would have a much shorter required operating lifespan, it could presumably be made significantly cheaper.

4. The average distance from a point inside a circle to the centre of the circle is ⅔, but the average distance on a line to the centre is only ½, and land values are higher centrally this will understate aggregate land circles. a rectangle overlaying a circle has an average distance from the centre of 0.5, while a circle has average distance internally of ⅔, and as land values are higher centrally this will mean that this understates aggregate land valuations.