Intermediate tree cover can maximize groundwater recharge in the seasonally dry tropics

Water scarcity contributes to the poverty of around one-third of the world’s people. Despite many benefits, tree planting in dry regions is often discouraged by concerns that trees reduce water availability. Yet relevant studies from the tropics are scarce, and the impacts of intermediate tree cover remain unexplored. We developed and tested an optimum tree cover theory in which groundwater recharge is maximized at an intermediate tree density. Below this optimal tree density the benefits from any additional trees on water percolation exceed their extra water use, leading to increased groundwater recharge, while above the optimum the opposite occurs. Our results, based on groundwater budgets calibrated with measurements of drainage and transpiration in a cultivated woodland in West Africa, demonstrate that groundwater recharge was maximised at intermediate tree densities. In contrast to the prevailing view, we therefore find that moderate tree cover can increase groundwater recharge, and that tree planting and various tree management options can improve groundwater resources. We evaluate the necessary conditions for these results to hold and suggest that they are likely to be common in the seasonally dry tropics, offering potential for widespread tree establishment and increased benefits for hundreds of millions of people.


Supplementary Figures and Legends S1-S9
In all these simulations 50 % of transpired water is assumed to originate from at least 1.5 m belowground.    Upscaling of point measurements to the plot level can give rise to potential sources of error.
Here, we would like to discuss a couple of such sources of error in order to evaluate the validity of our main conclusions.
A first potential source of error can originate when scaling up soil water drainage measured at 1.5 m soil depth to the plot level. In the simplest nearest-neighbour version of our spatial model we consider that drainage at 1.5 m depends only on the nearest tree to a given point. We believe this is a reasonable approximation in the context we are considering since our data does not indicate any additive effect in which the model can be influenced by the overall number of trees ( Supplementary Fig. S7). We recognize that our spatial replication is limited to detect such effect and that this could be more pronounced in other situations. Therefore, we assessed the implications of an additive version of the model in which drainage at 1.5 m depends not only on the nearest tree to a given point but on distance to all trees. These simulations also yield unimodal patterns but the maximum ground water recharge is 17 mm year -1 higher and broader and occurs at higher tree densities ( Supplementary Fig. S8). Notably the optimum canopy cover increased from 7% to 43% in the additive model, assuming 50% water uptake below 1.5 m.
Whether the nearest-neighbour model, the additive model, or some combination of these, is the best reflection of reality, and whatever depth the trees draw most of their moisture from, our main conclusions remain unaffected: there will be a non-zero tree cover value that maximizes ground water recharge.
The upscaling of sap flow from point measurements to the whole-tree level, and from the individual tree to the stand level are additional potential sources of error. Our sensitivity analysis on the effect of potential overestimation errors in tree water use derived from not adequately accounting for radial variation in sap velocity shows that assuming a potential overestimation error of 100% in our original estimates results in more groundwater recharge under the optimal canopy cover, and also in an increase in the tree density at which this optimum occurs ( Supplementary Fig. S9). Upscaling from the tree to the stand level also can lead to further sources of error. When scaling up sap flow it is important to select a representative sample of trees to monitor, whose range in size covers the actual range in size of the stand trees 7 , and we have done that. Stand transpiration under the different tree density scenarios was calculated based on the relationship found between canopy area and daily sap flow per individual tree, and assuming that all trees in the stand had the same size (average, large and small trees scenarios were considered). We admit that this assumption is not fully realistic. Indeed, stand transpiration will vary according to spatial variability in tree water use due to a number of factors such as tree age, density and variation in soil moisture and depth 8 . Most likely tree transpiration will not always increase linearly with increasing tree density as we have assumed, but the increase per added tree will be less once tree root systems and canopies start overlapping due to competition and reduced leaf area per tree. All the above mentioned potential sources of error linked to sap flow upscaling most likely lead to overestimations of tree water use in our models. Hence, even if these sources of error might affect the specific value of canopy cover at which groundwater recharge is maximized, our main conclusion that an intermediate tree cover for optimal groundwater recharge exists will not change. Further research on the relationships between stand structure and water use is needed to be able to improve estimates of groundwater recharge from various types of tree cover.
-Supplementary Information References