OR/14/018 Summary and conclusions

Mansour, M M, Hughes, A G. 2014. Land Use, Climate Change and Water Availability: Preliminary modelling of impacts of climate change and land use change on groundwater recharge for England and Wales. British Geological Survey Internal Report, OR/14/018.

Summary

To investigate how land use and climate change can affect potential recharge, 11 RCMs from the FFGWL project have been fed into the recharge model ZOODRM. This has produced potential recharge for the whole of England and Wales for three time slices (2020s, 2050s and 2080s). Allied to this, the historic rainfall and potential evaporation time series has been run for both historic and 'extreme assumed' land use change. The recharge model was run using LCM2000 and LCM2007 datasets as well as three scenarios: all arable, all grass and all forested. A more subtle change in land use was investigated by swapping 50% of one land use for another, e.g. arable to forested. This ensured that land use was modified where such changes are likely to occur, and avoided problems with land use changes in unlikely places, growing crops on mountain tops, for example.

The results have been presented for the Abstraction Reform (AR) catchments (Dee, Ely-Ouse, Hampshire Avon, Stour, Tees, Trent and Derwent) as well as the Thames and results summarised for England and Wales. To investigate variability due to catchment orientation, then two east- west and two north-south strips were also examined. The results have been presented as both difference maps of LTA recharge and box and whisker plots for both the absolute values of recharge and the differences between the modified run and its basecase (historical simulation).

Main conclusions

The output presented in this report is produced using a national-scale model that includes a range of simplifications and inherent assumptions. The results must be discussed, therefore, with these assumptions and simplifications in mind. In addition, the model uses a relatively coarse grid resolution (2 km by 2 km), which means its results are more relevant for water management at a regional scale rather than at local scale.

The pattern for England and Wales is generally increased recharge with significant outliers of greater recharge. However, the results show that generally the 2050s have reduced recharge with the 2080s producing predominately greater recharge. Spatially the most significant changes tend to occur in the west of England and in Wales.

The catchments chosen have a range of sizes and are located in different climate conditions around the country. The response to climate change reflects this with recharge decreasing or increasing depending on the RCM used for the input data and time slice. It has been recognised that considering the variability of RCMs in any recharge study (Holman et al., 2011). For this study, a single climate model has been used to produce 11 different but equally likely futures. This approach has allowed a range of equally plausible futures to be considered (wetter or dryer). However one problematic feature is the relationship of the recharge calculated for the historic simulation 11 RCMs and that produced with observed data. These are different, with the historic simulation typically dryer (lees recharge) than for the observed data which suggests that the future predictions underestimates any increase in recharge.

Examining the plots produced the following generalisations by catchment can be made:

• Dee — lower recharge in general with increasing recharge through the time slices
• Ely-Ouse — very slight increase in recharge which increases through the time slices
• Hampshire Avon — variation depending on the RCM; no significant change across the time slices
• Stour — reduction in recharge
• Tees — reduction in recharge which decreases through time slices
• Thames — variation depending on the RCM; significant outliers with increased recharge in the 2080s
• Trent — variation depending on the RCM; increased recharge through the time slices
• Usk – increased recharge; consistent over time slices

In terms of the results for climate change for the strips — there is greater variability E-W as opposed to N-S. This suggests the influence of Atlantic derived frontal systems and how these may change in the RCMs.

In terms of the effect of land use change then variation due to subtle ‘real changes’ in historic land use (between LCM 2000 and LCM 2007) is small. Extremes of land use change are predicted to result in significant change but these scenarios are very unlikely to be realised. For the Dee, Hampshire Avon, Tees and the Usk the change in recharge for land use change to climate change is comparable with the Ely-Ouse and Trent less and the Stour and England and Wales as a whole greater. This was investigated further by swapping out different land use types, i.e. arable to forested and showed much less variation than for the single land use runs.

The original question that the modelling work was to address relates to the relative changes in recharge related to climate change as opposed to land use change. Taking England and Wales as a whole then the order of change in recharge due to land use variation is: socio-economic land use (LCM2000 w.r.t. LCM2007) is less than spatial replacement whose magnitude of change in recharge is less than wholesale replacement (i.e. all one land use type for England and Wales). Comparing the magnitude of these changes with those resulting from climate change show that variation of recharge related to climate change variation falls in the middle of land use change. However, the variation of recharge due to the use of different RCMs is comparable with the overall variation of land use change, although this is tempered by the underestimation of recharge by the RCMs.

Possible future work

Further work that would help improve the conclusions are an improved understanding of the underlying assumptions regarding the RCMs used by FFGWL. Particularly the change in weather that these predictions incorporate, i.e. does rainfall reduce to the east of the country? This would have implications for understanding the behaviour of some of the catchments. Allied to this then would be an improved representation of drought frequency and the role of 'blocking' in controlling weather systems.

Whilst the work has shown that land use change can produce greater variability than climate change current rates of land use change (i.e. decade to decade) do not result in significant modification of recharge. To properly quantify this, it will be necessary to include land cover scenarios such as those produced by the National Ecosystem Assessment work which may then show change closer to the magnitude observed for the climate change scenarios.

Potential recharge on its own does not give the whole story in terms of the hydrological cycle and the groundwater balance. To address this, the recharge model has to be used in conjunction with a groundwater model, ideally a distributed one. This work should, therefore, be linked to a groundwater balance. Possible solutions to this is linkage with the modelling undertaken by Risk Solutions/HR Wallingford for the AR work, comparison with existing studies of the imapcts of climate change on groundwater, i.e. Marlborough and Berkshire Downs (Jackson et al., 2010) and the work on the Otter Sandstone currently undertaken by AMEC (2013a, b).

The statistical analysis of the results presented here must be treated with caution. This is because small changes in recharge values may result in significant volumes of recharge over a catchment. It would be useful to discuss the impact on the water resources as volume as well as recharge depth after accounting for other processes such as changes in the flow regime in rivers and abstractions. Further work on the results such as presenting monthly averages of potential recharge and comparison between the results from different RCMs would be desirable.

Other work that could be undertaken to benefit the study is a better understanding of the uncertainty in the recharge results. The uncertainty analysis of the undertaken work could be highly complex because of the nature of processes we are dealing with. For example the complexity of weather modelling, the complexity of prediction and representation of the future socio-economic scenarios, and the uncertainty associated with the modelling tools applied. More rigorous sensitivity analysis to the impact of these processes on the estimated volume of water could be useful to address the uncertainty associated with the results. This must include other unforeseen processes such as high intensity events and long drought spells.