The importance of proper hydrology in the forest cover‐water yield debate: commentary on Ellison <i>et al</i>. (2012) <i>Global Change Biology</i>,<i> 18, 806–820</i>

In the review article of Ellison et al. (2012) in Global Change Biology, the authors highlight the important debate on whether forests are consumers or suppliers of water. They advocate the idea that we should shift from the local 'demand side' view to the larger scale 'supply side' view. They do so because they argue, based on the review of several recent papers in global hydrology, that forest cover is of major importance for the occurrence of terrestrial precipitation P and that its removal will lead to severe reductions in P. While we in principal agree with these statements, the numbers on moisture recycling presented in their paper are for a certain part based on a misconception of the hydrological cycle creating the impression that forests play a more important role in the hydrological cycle than they actually do. Their general reasoning is that total continental evaporation E (ET in their paper) can be assumed to be equal to green water transpired by plants and trees and thus they neglect all-important nonproductive evaporation fluxes. Subsequently, some parts of the article give the impression that they assume that all this water returns to the continents and none of if it is transported back to the ocean. Moreover, their numbers about evaporative potential, which in reality has a significant spatial distribution, are simply based on the present-day global distribution of land cover, where forests are overrepresented in the tropics (F.A.O., 2010). We will argue in this article that these assumptions are too simplistic and lead to a view that exaggerates the role of forest in the hydrological cycle. The aim of this article is also to ensure that the all-important forest cover-water yield debate is based on state-of-the-art information and the correct application of the hydrological cycle. In any representation of the global hydrological cycle there is atmospheric water transport from the oceans to the continent (i.e. all terrestrial surface including lakes and rivers). This net transport can be calculated (1) by the moisture budget based on winds and moisture, (2) E−P from the ocean and (3) P−E from the continent (Trenberth et al., 2011). Ellison et al. (2012), table 1b) make the much too simplistic assumption that all continental evaporation recycles (returns to the continent). However, there is a substantial amount of the continental evaporation that ends up as precipitation on the oceans. Figure 1, based on an a posteriori water vapour tracking method, which used ERA-Interim reanalyses (Dee et al., 2011), shows that 57% (Ec/E, evaporation that recycles/total evaporation) of the continental evaporation actually recycles and thus 43% (Eo/E, evaporation that return to the ocean/total evaporation) precipitates on the oceans (van der Ent et al., 2010). It should be noted that the use of other periods, other datasets and other atmospheric water tracking methods (see e.g. Bosilovich et al., 2002; Dirmeyer et al., 2009; Goessling & Reick, 2011) will lead to slightly different numbers. The potential effect of conversion from forest to e.g. cropland, grass, new trees, or urban areas is fairly well investigated empirically through paired-catchment studies, where one catchment is left undisturbed and the other receives land cover or land use change (see Brown et al., 2005 for a review). In this respect, Zhang et al. (2001) studied 250 catchments worldwide and their results suggest that grass (pasture) evaporates one-third less than forest cover given a P of 2000 mm yr−1, but about the same amount when P is only 400 mm yr−1. There are also several studies which suggest that young forest which emerges some years after deforestation, transpires more than mature forest (e.g. Forrester et al., 2010). Furthermore, case studies in urban areas indicate that the reduction in E can vary between a minor reduction up to a major reduction by 80% in the extreme case of a fully paved parking lot (Zevenbergen et al., 2010). The reason why evaporation is not fully reduced by paving is due to the fact that not all evaporation is caused by transpiration (green water) as Ellison et al. (2012) seem to assume in some parts of their paper, but also by evaporation from the wet surface (ground interception). The total evaporation flux includes evaporation from transpiration, from the soil, from open water, and from interception. The latter is often overlooked, but can be up to 40% of P in a temperate climate (Gerrits et al., 2010). A study in Portugal investigating the hydrological response of a small catchment after burning found prefire canopy interception to account for almost 50% of P, to which they attributed the significantly increased streamflow (1.6 times) after burning (Stoof et al., 2012). However, on a larger scale the reduction in interception of course leads to a reduction in moisture recycling. The main point we want to make is that the actual reduction in E after deforestation depends on the new land cover replacing the original forest. Ellison et al. (2012, table 1a) acknowledge this and calculate the evaporative potential of different land cover types. However, we argue that this may not be performed on the basis of global land cover data because these land cover types are not equally distributed over the globe. In other words, evaporative potential is very much dependent on the climate zone in which the land cover change takes place and should be a function thereof rather than a single number. Moreover, the evaporative potential of different land cover types has different seasonal patterns (see e.g. Teuling et al., 2010). Besides the experimental studies mentioned above, several models exist that can estimate the effect of land cover change on E globally. Using a Global Climate Model (GCM) Goessling & Reick (2011) studied the extreme case of nonevaporating continents. Their results showed a severe reduction in continental precipitation (in July, 54% average over all continents and 95% in Europe), which they attributed to reduced moisture recycling as well as circulation changes and local coupling (i.e. P is affected locally via the thermal structure of the atmosphere). However, this does not mean that this is the Earth's fate if it would be completely deforested, as E would never completely disappear. In fact, the crude green planet vs. desert world experiments by Kleidon et al. (2000) suggest that the desert world evaporates significantly less than the green planet, but that the desert world E still is about one-third of the green planet E due to ground interception and soil evaporation. A more realistic comparison is to try to estimate how large the evaporation flux would be if it was not for the human induced land use and land cover changes. Figure 2a is based on two runs with the dynamic vegetation model LPJmL. One run is with actual land use and land cover, which is mainly based on the Ramankutty & Foley (1999) dataset, but additional sources are used as well (see Bondeau et al., 2007 for details). The other LPJmL run is with natural (=potential) land cover based on climate input data only (see Nikoli, 2011 for details). Figure 2a shows that there are some areas where the normal evaporation is significantly lower than in the situation with natural land cover. Most notably this is the case in parts of South-America and Africa. On the other hand, in India, due to the large-scale irrigation, E has actually increased compared to natural vegetation. Please note that Gordon et al. (2005), fig. 4 arrive at similar findings with a different model. Figure 2b goes one step further and gives a first order approximation of the effect that a return to natural vegetation would have on precipitation. It seems that returning to a state of natural land cover would increase P in West-Africa, but decrease P in Southeast Asia. From these results it can thus be concluded that conversion from forest to cropland can even lead to an increase in E and P. Furthermore, we like to stress that reality is probably not as simple as Fig. 2b is suggesting, because evaporation is a water as well as an energy flux and can thus lead to changes in the thermal structure of the atmosphere, wind patterns and cloud cover. Besides, there are other changes associated with land use and land cover changes: such as changes in albedo, aerosols, and emitted particles, all of which can influence precipitation (see e.g. Kleidon et al., 2000; Werth & Avissar, 2002; Bonan, 2008; Pitman et al., 2009; Goessling & Reick, 2011). As a conclusion, Ellison et al. (2012) initiated an important shift in thinking of forests as water suppliers, instead of mere water users. We have nuanced the numbers they presented, but still find that forests are important water suppliers. The question remains how this knowledge could be implemented in policy making. From a water and land management perspective it would be ideal if we were able to predict exactly the effect on the regional as well as the hydrological cycle after deforestation. Such an attempt was for example made by Werth & Avissar (2002) for Amazonian deforestation, but unfortunately different climate models are likely to provide different answers leading to large uncertainty (Pitman et al., 2009). From a moisture recycling perspective it also matters a lot where the deforestation takes place. Based on the analysis of van der Ent et al. (2010) it can be seen that deforestation in the Amazon or the Congo is more likely to have a downwind continental effect than deforestation in eastern Canada or Indonesia. In this light, Keys et al. (2012) recently proposed the 'precipitationshed' as a tool to identify and analyse possible positive and negative effects on P from changes in E due to land cover changes. In future study, this can possibly be related to the 'water footprint' and water-pricing strategies (Berger & Finkbeiner, 2012). However, although dynamic vegetation models like LPJmL exist, the change in E due to land cover change is far from evident and more understanding can be gained if we start separating E in transpiration, interception, etc. The effect on P is an even bigger scientific question; we like to make a case for efforts aimed at gaining fundamental understanding of different biogeophysical and biogeochemical effects of removing forest cover and their consequences for the hydrological cycle at different scales (see also Bonan, 2008). In our opinion, the uncertainty involved in estimating the effects of deforestation is even more reason to be extremely reserved with further deforestation. On top of that, we agree with Ellison et al. (2012) that there are many other (ecological) reasons to make a case for preserving global forest cover. This work is supported by the Division for Earth and Life Sciences (ALW) of the Netherlands Organization for Scientific Research (NWO). The authors thank Holger Hoff, Katharina Waha and Jens Heinke from the Potsdam Institute of Climate Impact Research (PIK) for sharing and providing background to the LPJmL data that was used to prepare Fig. 2. The authors also thank all three reviewers; in particular Helge Goessling, who helped to significantly improve the manuscript.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.