Afforestation impacts on terrestrial hydrology insignificant compared to climate change in Great Britain

It is important to understand the relative effects of afforestation and climate on the hydrology of river catchments. This manuscript considers the effects of both afforestation and climate change in detail for catchments in Great Britain and concludes that afforestation impacts are insignificant compared to climate change. The modelling work seems to have been carried out well and the manuscript is generally well written. However what I feel is missing is a comparison with measured data on the effect of afforestation, which is needed in order to check the results are realistic and that the conclusion is valid. The authors even state on L71 “LSMs should therefore quantify projected hydrological changes whilst modellers determine if outputs are realistic”. My experience suggest the conclusions are valid but it is important to demonstrate that this true.

JULES LSM model simulations were carried out for 51 river catchments in the UK a range of different afforestation and climate scenarios. My understanding is that the baseline model was calibrated and validated against streamflow and soil moisture data. Then afforestation and climate simulations are carried out. The results show that “Afforestation across Great Britain moderately reduces average river flow by 0.17 % PPPoA over the year” where PPPoA is the per percentage point of afforestation. Is this value realistic? In my view all the results and conclusions rely on the model producing a realistic value here and there does not seem to have been any attempt to validate this value.  I appreciate obtaining good data for this is difficult, but there are UK studies available. In the wetter parts of the country there are sites at Coalburn, Plynlimon, Balquhidder and in the drier parts at Blackwood, Alice Holt, Thetford Forest and Clipstone Forest. A lot of these consider a change from grassland to coniferous forests rather than broadleaf forests considered here. Also in many cases there are point scale measurements rather than changes in catchment river flows, but they will give an idea if the results are roughly correct. For example in the Coalburn catchment (Birkinshaw et al. 2014), which has already been cited, a change from grassland to a mature conifer forest (90% of the catchment) has produced a reduction in streamflow of around 350mm for an average river flow of around 900mm, which by my calculation reduces average river flow by 0.35 % PPPoA.

It could also be argued that there is no validation that the change in climate is also producing realistic results in the model. But I am not sure how this can be checked expect that the current validation against existing streamflow measurements covers “known floods and drought events”.

Specific Comments

L85. I note that in sections 3 and 4 that the sub-sections correspond to the 3 research questions. I might be worth highlighting this here as it does make the manuscript easier to read and I did not realise to start with.

L161. “JULES runs at a numerical timestep of 30 minutes”. This is clear, but what is the timestep of the meteorological input data and the streamflow data (I might have missed this)?

L188. There should be a comparison with other hydrological models of UK catchments here, for example Lane et al. (2019) and Lees et al.  (2021). These use NSE as an objective function rather than KGE, but I note in the supplementary material that NSE values are calculated. Also are you using hourly or daily discharge for the comparison?

L199. Which Dee catchment? There are at least two rivers called Dee in the UK

L217 “twenty UKCP18 river basin boundaries”. Why are there 20 here but in L196 there were 51?

L233. “Proportional Influence of Afforestation Compared to Climate”. You talk about changing precipitation, temperature, and CO2 (i.e three variables) but there are also changes in afforestation (so a fourth variable). I got confused and had to go back and read the section again.

L288-292. This could do with being rewritten. The authors define soil evaporation “Simulated soil evaporation, including both evaporation from the soil surface and plant transpiration” but then consider transpiration separately but not soil surface evaporation separately.

L308 Table 1. How are “Soil moisture” and “Canopy Storage calculated”. Is this the average over the simulation period? The data in the “future” columns is discussed in section 3.3, can the caption be changed to make this clear, when I first read this I got confused about why these results had suddenly appeared.

L312 Table 2. The correlations seems to be calculated for the percentage change. I would be interested in calculating the correlation for the absolute change. As I am sure the authors are aware many catchments in Scotland and North West England and Wales have precipitation totals that are around 3 times those in South East England. So if they have a similar percentage change they will have an absolute change that will be very different,  which suggests there will be a significant influence due to the location for absolute change.

L325 Figure 4. In winter canopy evaporation increases as expected (Page et al. 2020). But what is driving the massive increase in soil evaporation. Soil evaporation “includes both evaporation from the soil surface and plant transpiration” but L291 says transpiration decreases with afforestation. This implies that increasing the forest is producing a massive increase in evaporation from the soil surface in winter. This does not make sense to me as a mature forest will have very little soil surface evaporation in winter. So either the model is doing something strange or I have misunderstood the results.

L340 Figure 5 In the bottom panel the runoff increases slightly and the canopy evaporation decreases slightly for 100% afforestation compared to 50% afforestation. The changes are small but I was wondering why this was happening.

L360 Figure 6. I feel it would be easier to interpret if the y axis had the same scales for Top Flows, Median Flows and Low Flows

L406-L407 “Average simulated river flow, compared to present, drops at a slightly lower rate of -0.12 % PPPoA (ρ = -0.82) [Table1]. Runoff decreases with afforestation at a comparable rate to present (-0.27 %; 1.84 mm yr-1 PPPoA)”. I do not remember seeing the difference between “river flow” and “runoff” defined. Can this be added somewhere and then explained why they are getting different results. Also Table 1 has runoff but no river flow. where is the value of -0.12% in Table 1?

L421 Discussion. The discussion is interesting but my personal feeling is that it is on the long side (I struggled to concentrate whilst reading it) and there are bits that are not completely relevant.

L432 “suggesting potential reductions in water yield can be directly estimated from the areal extent of woodland planted rather than its location”. This follows from an earlier point this may be true for the percentage reductions but does the location affect the absolute reductions?

L466 “In reality, tree root depths would be much deeper” is this due to using on a 3m deep soil column in JULES? Maybe make this clear.

L485 “Therefore in reality, afforestation may have a muted influence on streamflow in these regions with roots accessing the deeper groundwater (Roberts and Rosier, 2005).” I do not understand this. If forests can access deeper groundwater then it might have a greater influence on streamflow in the longer term as it can transpire water even when there is a meteorological drought.

L513 “afforestation both decreases and increases streamflow”. This needs explaining

L520-L527 Could this bit be removed? It is all a bit vague I do not really understand the point it is making.

L560 where is Figure 8?

L580 “with climate…”. Is this bit a repeat from the previous section?

L585 “By applying atmospheric changes across the whole country, variations in landcover, topography and soil type are insufficient to substantially alter the hydrological response.” Has this been shown for topography and soil type?

References

Birkinshaw, S. J., Bathurst, J. C., & Robinson, M. (2014). 45 years of non-stationary hydrology over a forest plantation growth cycle, Coalburn catchment, Northern England. Journal of Hydrology, 519, 559-573.

Lane, R. A., Coxon, G., Freer, J. E., Wagener, T., Johnes, P. J., Bloomfield, J. P., ... & Reaney, S. M. (2019). Benchmarking the predictive capability of hydrological models for river flow and flood peak predictions across over 1000 catchments in Great Britain. Hydrology and Earth System Sciences, 23(10), 4011-4032.

Lees, T., Buechel, M., Anderson, B., Slater, L., Reece, S., Coxon, G., & Dadson, S. J. (2021). Benchmarking data-driven rainfall–runoff models in Great Britain: a comparison of long short-term memory (LSTM)-based models with four lumped conceptual models. Hydrology and Earth System Sciences, 25(10), 5517-5534.

Page, T., Chappell, N. A., Beven, K. J., Hankin, B., & Kretzschmar, A. (2020). Assessing the significance of wet‐canopy evaporation from forests during extreme rainfall events for flood mitigation in mountainous regions of the United Kingdom. Hydrological Processes, 34(24), 4740-4754.

The study evaluates the potential hydrological changes under the possible afforestation regime alongside climate changes. Therefore, the results suggested that afforestation has an insignificant impact on the hydrology. The conclusion is not well supported by the modelling results since the overall model performance remains unsatisfactory. Model performance in most of the study sites was not satisfactory (in terms of NSE>0.5, which is more commonly used than KGE), which may require more detailed calibration and validation before future climate scenario is carried out. Further, some hydrography showing the changes might be useful for the presentation since the study evaluates hydrology.

The conclusion is questionable based on the results of the study. For example, most of the changes in the broadleaf forest area are less than 10 per cent across the study sites. Although the scenario might be realistic under the conditions mentioned in the manuscript, however, not enough to support the conclusion since the hydrological change is not expected to be significant under the level of land cover change. Contradictory results are also referred to in the literature reviews, (L6: Many studies suggest afforestation can reduce overall streamflow). I suggest that results from the relevant studies should be discussed in the manuscript. Further, I believe that per percentage point of afforestation (PPPoA) is not a good indicator to evaluate the hydrological changes. I would suggest the model be simulated under selected paired-experiment sites (at least some examples should be carried out in the manuscript), which better considers the effects of afforestation.

Various indicators are selected to evaluate the results, but most of them are not well defined in the methods sections or elsewhere in the manuscript (e.g. ANOVA/Kruskal). Terms such as high flow, low flow, summer, and winter should also be defined. The flow duration curve (FDC) is mentioned but rarely explained in the result section.

Some information about the model setup is missing. For example, it is unclear what temporal resolution is the results based on (hourly, daily or monthly), which is important for model performance evaluation. Also, the period of model simulations is not well described in the method section (e.g. L272 is unclear). The landcover data CAMELS-GB is not mentioned in the manuscript but is shown in the supplementary table.

Specific Comments

L107, L131 & L161 The land cover types are not consistent here, Non-default JULES land cover is referred to in L107 (acid grassland, arable and horticultural areas, heather, heather grassland, improved grassland and neutral grassland), however, how to turn it into 8 JULES land cover types are not mentioned.

L128 I assume the percentage of changes was compared between the CEH 2000 landcover map (the year 2000), and scenarios in the year 2050 (but the selection of dataset was not well described, also the year 2050 is not pointed out clearly).

L159 temporal resolution of CHESS-met is not mentioned in the manuscript, which should be 1 hour.

L162 Broadleaf tree (and other PFTs) setup: what is the parameter set used in this study? I would believe default parameters are used, however, should be mentioned somewhere in the manuscript.

L183 The Kling-Gupta Efficiency Measure (KGE) formula should be described here if it is used for assessment. NSE (also found in the supplementary document) should be a more common indicator for hydrological evaluation.

L186 This should be the results rather than the method. Further, the temporal resolution (hourly, daily or monthly?) of the results is not defined.

L196 The location of 51 study sites should be pointed out here or in the supplementary document.

L198 Abbreviation as UKBN2 should be explained when it is first referred to in the manuscript.

L202 “that processes simulated are more faithful at JULES’ spatial (1 km2) and temporal (30 minutes)” the statement needs more explanation.

L219 Season should be defined here although I believe it follows spring MAM, summer JJA …

L220 & L227 Theil-Sen slope estimator and Spearman’s rank correlation coefficient formula should be explained if it is relevant for evaluating the results.

L246 ten years (using 2000-2001): should be a typo here.

L264 To clarify, The CHESS-SCAPE dataset (RCP 8.5) should be referred to in the first place.

Table 2 Using percentages to describe the change of FDC might be not that meaningful.

L345 “As the climate changes, land cover is also expected to change.” It seems not to change in the three scenarios used in this study.

L348 ANOVA is not explained in the manuscript.

L371 “Rising temperatures appreciably alter many of JULES’ hydrological parameterisations.” Should affect JULES hydrological processes, not parameterisation.

L383 Kruskal-Wallis (KW) test is not explained in the manuscript.

L385 High flow/low flow are not defined in the manuscript.

L437 “This is also seen in observational studies where a 1 % increase in upstream afforestation area does not detectably change streamflow” This could not conclude that” Afforestation impacts on terrestrial hydrology are insignificant.”

L441 “Alternatively, the large epistemic uncertainty within JULES means that highly sensitive hydrological parameters are not included that would lead to diverging regional afforestation responses.” However, this part should be important to evaluate the hydrological regime.

L560 Figure 8 is missing.

L585 “By applying atmospheric changes across the whole country, variations in landcover, topography and soil type are insufficient to substantially alter the hydrological response” This is not real, soil parameters could considerably affect the hydrological regime, and it is not changed in this study before coming to this conclusion.

L666 “Future research should use fully coupled land surface–atmosphere LSMs” I disagree that JULES has already reached its limitation here, there is more room to be improved for better model performance.