Potential for Improved Hydrological Functioning and Stream Baseflows through Forest Landscape Restoration in the Tropics

1. The Tropics: a forest landscape restoration ‘hot spot’

In response to the loss and degradation of the world’s forests and soils, forest restoration is being scaled up rapidly for the UN Decade on Ecosystem Restoration [1,2,3,4]. A total of 350 million hectares of forest are to be restored by 2030 according to pledges made through the Bonn Challenge alone; other initiatives are similarly ambitious [5]. Tropical forests assume a central position in pledged restoration efforts, as the humid tropics have the greatest potential for tree growth and carbon sequestration [6,7]. Further, the dry tropics are amongst the most vulnerable regions worldwide in terms of vegetation loss, soil degradation, and the number of people living on degraded and deforested land [3,8,9,10,11]. Looking ahead, major increases in tropical forest cover will be required to address the combined needs for restoring biodiversity, capturing carbon to mitigate global warming, increasing resilience to climatic change/variability, as well as supplying high-quality water to ever-growing urban populations [12,13,14,15,16].

An adequate and reliable supply of clean freshwater is essential for domestic uses, food production, energy security, and provisioning of aquatic and terrestrial ecosystems and their biodiversity [14,16,17,18]. The major changes in forest cover proposed within the global framework of Forest Landscape Restoration (FLR) [5] will have important consequences for the integrity of flow of water through landscapes, affecting water availability at multiple scales [19,20,21]. We call attention, however, to the discrepancy between the repeatedly expressed need for quantitative and objective information on the hydrological impacts of reforestation and FLR in the tropics [19,22,23,24] and pertinent findings that are summarised to date [25,26,27,28].

In this paper, we address this discrepancy by exploring the evidence and sketching out the scope for restoring hydrological functioning of degraded environments and improving water availability via FLR in the tropics. We draw both from recent advances and older notable works, while keeping a ‘real-world’ perspective that recognises that most areas targeted for restoration have experienced some soil degradation that affects hillslope and catchment hydrological functioning [29,30,31]. We further emphasise the importance of stream baseflow recovery as an important indicator of FLR success. Here, baseflow is the sustained flow of water in streams and rivers that is maintained by infiltrated precipitation between rainfall events, typically moving towards the stream via various underground pathways [see Table 1 for a definition of terms; 32]. Hydrological functioning refers to streamflow response to rainfall, including the timing and magnitude of seasonal baseflows as well as the flow peaks associated with distinct precipitation events [33,34].

2. Forests and streamflow: a seemingly settled debate

One long-established view of the hydrological role of forests is that their complex of trees, understorey vegetation, surface litter, roots and soil acts as a ‘sponge’ absorbing rainfall during wet periods and releasing the stored water subsequently during dry periods [30,35,36]. Following forest removal, this ‘sponge effect’ tends to diminish or may be lost altogether, causing springs and streams in seasonally dry climates to desiccate in dry periods; meanwhile, flooding is typically exacerbated during periods with high rainfall because of increased surface runoff [37,38,39,40].

The ‘forest as sponge’ metaphor and related thinking came under serious scrutiny after Bosch & Hewlett [41] summarised the changes in annual streamflow totals (“water yield”) associated with vegetation change (deforestation or afforestation) for 94 so-called “paired-catchment” studies conducted mostly under temperate climate conditions. Although the variation in results was deemed “extreme”, Bosch & Hewlett [41] concluded that “no experiments in deliberately reducing [vegetation] cover caused reductions in [water] yield, nor have any deliberate increases in cover caused increases in yield”. In a companion review of the flood-mitigating capacity of forests, Hewlett [42] concluded that an undisturbed forest cover generally moderated peak discharges and stormflow as the ‘sponge’ metaphor implied, although the effect decreased as the size of the rainfall event and catchment wetness level increased. Further, the influence of forest on the magnitude of the largest events (“floods”) was marginal [42]. In short: the ‘forest sponge’ was seen to have limitations, breaking down for extreme rainfall events and very wet soils.

Hamilton & King [43] were among the first to realise the implications of these findings for the tropics—tropical forest conversion in particular. Seeing that forest removal led to increased streamflow, they surmised that trees might be more appropriately labelled ‘pumps’, raising water from the soil profile and returning it back to the atmosphere. Further, they concluded from the fact that the flows associated with the largest storm events were not affected much by the presence or absence of forest cover: “Major floods occur due to too much precipitation falling in too short a time or over too long a time, beyond the capacity of the soil mantle to store it, or the stream channel to handle it”. In several provocative articles targeting the “four M’s of myth, misunderstanding, misinformation, and misinterpretation” regarding the hydrological role of tropical forests—including the “myth of the forest sponge”—Hamilton [44,45] called for “greater accuracy and realism”. In his footsteps, many subsequent 'hydrological myth-busters' highlighted the high water consumption of trees and forests and their inability to prevent extreme flooding [23,46,47,48,49]. The value of a ‘good’ forest cover for maintaining other aspects of hydrological functioning and for providing high-quality water was largely neglected in these publications.

Both Hamilton & King [43] and Hewlett [42] did recognise that the impact of ‘deforestation’ on the hydrological functioning of a catchment could be significantly altered if widespread soil degradation were associated with forest clearance, either during (e.g., through the use of heavy machinery) or after (e.g, via accelerated surface erosion or mass wasting). However, these pioneers were not aware of any experimental evidence supporting such intuitions. In the words of Hamilton [44]: “Suggestions implying tropical reforestation or afforestation of non-forested lands, including extensive grasslands, will cause higher well levels, renewed spring flows, and increased low flows in streams are not supported by evidence from temperate zone research that indicates the reverse”. Likewise, although Hamilton & King [43] and Bruijnzeel [50] acknowledged various anecdotal reports of renewed springs and more reliable streamflow following tropical forestation, sound scientific data from the region were lacking at the time [25,51].

The initial conclusions of Bosch & Hewlett [41] have been broadly echoed by successive reviews of the gradually expanding global literature on land-cover change effects on annual water yield, with the strongest relative changes in annual yields following the gain or loss of forest cover noted for sub-humid rainfall conditions [52,53,54,55,56,57,58,59,60]. Similar conclusions about changes in annual water yield (i.e. lower yields under forest), were reached for the much smaller humid tropical dataset [22,27,28,31,61].

In view of the seemingly overwhelming evidence that ‘more forest implies less total streamflow (lower yield)’, evidence demonstrating improved hydrological functioning and water availability achieved through forestation in the tropics is generally considered exceptional [27,28,31,47]. However, based on new work, we now contend that the debate regarding forest cover and streamflow is still alive. The dominating view that forestation can only reduce streamflow and water availability, diminishes opportunities to restore the hydrological functioning of degraded landscapes and supply environmental and societal needs. Rather, our interpretation of a growing body of evidence indicates that the opportunities for FLR to improve hydrology and water availability can be large and important in specific geographical settings.

3. Critical considerations regarding FLR and streamflow

Before detailing our perspective, we note five aspects that are important for judging the hydrological effectiveness of FLR, but are often neglected or inadequately emphasised in reviews and compilations [52,55,56,57,58,59,60]:

• Catchment water yield, or total annual streamflow, differs from the fraction of streamflow that is useful to ecosystems or humans. The common focus on total streamflow [23,55,56,57,58,59,60] neglects the importance of sustaining a stable ‘baseflow’ between rainstorms. Flow stability provides the water to support aquatic ecosystem functioning and daily human water needs. The other streamflow component, ‘stormflow’, is typically less useful for humans as it is often laden with sediment, and can be destructive because of flooding and siltation of reservoirs, irrigation channels and river beds [62,63,64]. In seasonally dry areas of the tropics, reliable dry-season baseflows are critical for supporting ecological systems and a host of human water uses [16,30,65].
• Water use by vegetation is only one element influencing streamflow changes – and it is not always the most important. Depending on catchment morphology (e.g., steep slopes with narrow valley-bottoms versus gentle topography with wide valley-bottoms), the presence of free-draining or poorly drained soils, depth of soil above the bedrock, permeability of the whole regolith, and surface conditions (e.g., related to level of degradation), the stormflow component of total streamflow may be large or small [66,67,68]. Where large changes in stormflow following forest removal or addition occur [38,62,69,70,71], conclusions that changes in vegetation water use alone produce the observed changes in “total water yield” are bound to be erroneous [57,60,68]. For instance, large-scale soil conservation works (terracing, check dams) and vegetation restoration on the Chinese Loess Plateau have produced large reductions in total water and sediment yields [68,72]. However, the vast majority of these decreases reflect reductions in stormflow owing to greater infiltration and storage in the soil profile; conversely, dry-season baseflows gradually increased with time, stabilising once the vegetation cover reached 60–70% [68].
• The influence of land degradation and the subsequent recovery of soil hydrological processes should be considered in assessing streamflow changes related to FLR. Global inventories of land degradation and soil erosion indicate extensive areas where hydrological functioning is likely to be affected adversely [1,8,9]—largely via increased surface runoff during storms and possibly reduced groundwater recharge, both related to changes in soil infiltration capacity [Figure 1B]. However, none of the catchment studies referenced in the cited global literature reviews above [52,53,54,55,56,57,58,59,60] consider degraded soil conditions [22,28,31]. Such summaries are, therefore, not fully representative of typical FLR situations where the effects of soil degradation on the partitioning of rainfall into surface runoff, infiltration and soil water storage cannot be ignored [25,27].
• A distinction is needed between varying hydro-climatic regimes when assessing the hydrological potential of FLR. Relative changes in water yield are more pronounced under drier (and sunnier) conditions [56,57]. This response is reflected in the global finding of Hou et al. [60] who reported a much larger change in annual water yield per unit forest cover change for 59 catchments undergoing forestation than for 197 catchments experiencing forest loss. Closer scrutiny of the data revealed that the difference between the two groups of catchments primarily reflected drier overall climatic conditions in the catchments receiving forestation [60]. This finding has two important implications: (a) the global dataset on the impact of forestation on water yield is biased; and (b) reductions in water yield after forestation under more humid conditions (e.g., in the equatorial tropics) may be smaller than suggested by the average values presented by global reviews [55,57,60].
• Changes in hydrological response related to vegetation gains following losses are often non-linear and hysteretic. In the context of restoration, impact assessments should recognise the existence of several potential positive ‘feedbacks’ on hydrological processes at various scales, including ‘trade-offs’ between changes in vegetation water use and infiltration after foresting degraded land [25,73,74]. Also, the ability of vegetation to capture ‘occult’ precipitation (fog and low cloud) in specific coastal and montane settings should eventually recover [75,76,77,78]. Finally, the potential for moisture recycling, transport and convergence at various scales increases, thereby affecting patterns of precipitation [20,79,80].

4. ‘Pumps’ and ‘sponges’: a paradigm of hydrological ‘trade-offs’

To frame our message, we return to the analogy that well-developed forest ecosystems function both as ‘pumps’ and ‘sponges’. This conceptualization implies that vegetation has a drying effect as it intercepts rainfall and takes up water from the soil, subsequently releasing moisture back to the atmosphere via transpiration (i.e., the ‘pump’ side of things). Meanwhile, the ‘sponge’ effect pertains to the underlying soil absorbing, retaining, and moderating the passage of water through the catchment. If rainfall exceeds the soil's infiltration capacity and ponding occurs, surface runoff is generated that contributes to stormflow. Infiltrating water can be stored in the soil or transferred underground to the groundwater reservoir or move towards the stream system through subsurface flow pathways at variable rates and depths (faster and shallower during rain). Baseflows, which are derived from deeper subsurface flows and groundwater, represent the equilibrium between water losses through ET, gains through infiltration, and the balance with storage [81; Figure 1A].

As long as the soil’s infiltration capacity is more or less maintained, removing forest cover tends to increase baseflows at a level largely proportional to the changes in vegetation water use. This response occurs because the associated changes in stormflow are generally small under non-degraded conditions [31,82,83]; and peak discharges for forested and non-forested lands typically converge as amounts of rainfall and soil wetness increase [28,42,63]. However, where substantial surface degradation (e.g., through loss of soil organic matter, increased soil compaction/consolidation or crusting) has occurred, stormflows and peak stream discharges can increase substantially [62,69,70,71,84,85,86].

The extra surface runoff caused by reduced infiltration does not replenish soil moisture reserves or the groundwater that sustains baseflow, and is thus, effectively lost to the catchment ecosystem. In cases of greatly reduced soil infiltrability (e.g., due to heavy crusting or compaction), losses via surface runoff can lead to marked reductions in groundwater recharge and dry-season baseflow, compared with the prior forested situation, despite the post-forest vegetation using less water (Figure 1B). In other words, the ‘sponge’ effect is lost [25,29,30,31,73,87].

Figure 1. Partitioning of precipitation into evapotranspiration, (near-)surface runoff, and groundwater recharge under (A) mature forest, (B) degraded land, and (C) natural regrowth to illustrate the ‘infiltration – evapotranspiration trade-off’ mechanism governing groundwater recharge and dry-season flows [adapted from 88]. Green arrows: evapotranspiration; brown arrows: surface runoff; light blue arrows: subsurface stormflow; dark blue arrows: groundwater recharge. Arrow sizes indicate relative magnitude of the respective fluxes for the three land covers.

Figure 1. Partitioning of precipitation into evapotranspiration, (near-)surface runoff, and groundwater recharge under (A) mature forest, (B) degraded land, and (C) natural regrowth to illustrate the ‘infiltration – evapotranspiration trade-off’ mechanism governing groundwater recharge and dry-season flows [adapted from 88]. Green arrows: evapotranspiration; brown arrows: surface runoff; light blue arrows: subsurface stormflow; dark blue arrows: groundwater recharge. Arrow sizes indicate relative magnitude of the respective fluxes for the three land covers.

A recent spatially distributed hydrological modelling study by Peña-Arancibia et al. [81] assessed the ‘trade-off’ between changes in the ‘pump’ (ET) and ‘sponge’ (infiltration) functions after forest conversion to pasture across the tropics, both with and without imposed soil degradation (i.e., infiltration reduced by 50%). Overall, simply replacing forests by non-degraded pasture increased annual water yield by 18% because of the lower water use of the grass (reduced pumping). The greatest relative increases were found in water-limited regions; smaller changes were derived for the rainy humid equatorial regions where higher cloudiness limits ET. Annual water yields rose to 26% when forest conversion was associated with soil degradation, reflecting increases in surface runoff due to reduced infiltration. However, for nearly one-fifth of all grid cells (19%), a reduction in stream baseflow was inferred during one or more of the driest months in the soil degradation scenario, despite the lower water use of the pasture. In short, the change in infiltration had a greater effect on baseflows than did the change in ET (i.e. the ‘trade-off’ between the two was negative; cf. Figure 1B) [81]. It is worth noting that many of the areas for which decreases in dry-season flows were predicted coincide with areas targeted by the Atlas of Forest and Landscape Restoration Opportunities [89]. Note also that any potentially negative changes in precipitation following widespread forest conversion [90] were not included in the modelling.

In general, major improvements in soil infiltration capacity following the re-establishment of a good vegetation cover on degraded soils have been observed [74,91,92,93]. Vegetation maturation is also associated with the redevelopment of biologically mediated soil macropores, root channels, and animal burrows – some of which are typically lost during repeated slash-and-burn cycles, annual cropping, and grazing [92,94,95,96]. Connected networks of macropores act as ‘preferential pathways’ guiding infiltrating rainwater (typically within a day) through the root zone once a critical soil moisture storage threshold value is exceeded, usually during times of ample rainfall [97]. This mechanism promotes the deep subsurface flows and groundwater recharge that contribute to baseflow [30,68,98].

Improved macropore flow can also explain why stormflow responses of reforested headwater catchments can be reduced even under extreme rainfalls compared with nearby areas with greater degradation, as shown in various locations such as Panamá, the Philippines, Mediterranean SE France, and South Korea [62,99,100,101]. Indeed, the largest decreases in storm runoff – and therefore the largest gains in infiltrated rainfall – have been observed after foresting heavily degraded areas under intense rainfall [62,100,101,102]. In such cases, increased macropore flow combines with other processes resulting from forestation, notably increased water storage potential in the soil due to pre-storm water use by the vegetation [99,100].

Given such trade-offs, it is premature to conclude that the ‘localized’ flood-moderating effect of reforesting degraded headwaters is limited to small or intermediate rainfall events only, as suggested by Marshall et al. [23] and others investigating catchments with limited soil degradation [19,63]. However, it would be equally premature to extend such findings uncritically to much larger scales (e.g., large river basins) and expect upland forest restoration to eliminate all downstream flooding [103]. Large-scale flooding typically results from extensive and persistent rainfall fields of long duration and/or high intensity, often occurring when soils have been wetted up by previous rains [42,63,104,105]. Flood risk relates to the interplay of a host of additional factors as well, including degree of urbanization and hard surfaces, floodplain occupancy, wetland conversion, presence of storage reservoirs and other infrastructural works (e.g., dikes and embankments), etc. [103,106,107,108,109].

5. Improved baseflows: a measure for assessing hydrological success of FLR

Annual streamflow totals associated with forested catchments are usually lower than those for non-forested catchments due to the generally higher water use of trees [57,83,110] and to the propensity of the whole ‘forest complex’ to limit the generation of storm flows [19,54]. However, the influence of tree cover on baseflow generation remains enigmatic [19,25,27,111]. Yet, as stated earlier, baseflows have a particular practical importance, especially in seasonally dry climates [14,16,30]. As such, we argue that recovery of baseflows is a more suitable indicator of FLR success in hydrological terms, along with reductions in soil loss and stream sediment yields [31,68,72,112,113,114,115]. In the following subsections we explore the potential benefits of FLR with respect to baseflows.

5.1. Conditions where FLR decreases or increases baseflow

Depending on the relative changes in both vegetation water use and infiltration related to forest restoration, baseflow will be reduced where losses through increased ET exceed the gains related to improved infiltration [83,116]. The net effect may also be near-neutral [117,118], or positive where infiltration gains override evaporation losses [27,73] (Figure 1C). For large areas, the conditions and locations where FLR may be most effective are best shown through modelling studies. For example, Peña-Arancibia et al. [81] found the largest negative hydrological changes following deforestation and soil degradation in areas having strong seasonality in precipitation, a high rainfall surplus over ET during wet months, and deep soils. Upon restoration of the soil infiltration capacity after forestation, the largest absolute gains in baseflows may therefore be expected in highly degraded areas under seasonally high rainfall, and with sufficiently deep soils that are capable of storing the extra infiltrating water [25,27].

Using the same modelling approach as Peña-Arancibia et al. [81], but in an inverse manner (i.e., forest addition instead of removal), areas in mainland and maritime SE Asia (including Papua New Guinea), SE China, NE India, parts of West Africa, SE Brazil, and Central America were identified to have the greatest likelihood of increased baseflows following forestation (Figure 2) [119]. Out of a total of 3,554 grid cells with a climatic potential for sustaining forest vegetation (1⁰ resolution, ~100 km), a positive influence on baseflows was predicted for 340 cells (~10%). These ‘bright spots’ were again those with a high seasonal rainfall surplus, sufficiently deep soils, and significant initial surface degradation. A near-neutral change in baseflow (<2 mm decline per year) was predicted for an additional 292 grid cells (8%). Again, potential changes in precipitation after forestation [20] were not considered in the modelling.

At the other end of the spectrum, the largest decreases in baseflows have been reported after planting fast-growing exotic tree species (often pines or eucalypts) in areas where average rainfall is insufficient to support (evergreen) forest naturally, and grassland or shrubland is the assumed natural baseline vegetation [57,83]. However, there are indications of gradually diminishing tree water use – and therefore partial streamflow recovery – as these planted forests mature [120,121,122], although evidence from the tropics remains particularly sparse [58,123].

Where baseflows are reduced following forest restoration or afforestation, the higher water use of the vegetation may be viewed as a cost of achieving other, non-hydrological benefits [124,125]. Examples of the latter include carbon sequestration, biodiversity and habitat improvement, protection against erosion, improved streamwater quality, etc. Further, the water that is released to the atmosphere through ET is not ‘lost’ per se, as it will contribute to precipitation elsewhere [20,126]; see below for a more detailed discussion.

5.2. Examples where forest restoration has increased baseflows in degraded tropical areas

Over the last two decades an increasing number of studies have reported increases in dry-season flows following forestation of degraded land in the tropics across a range of catchment scales (from a few hectares up to 7,325 km²). This list includes studies in the lowlands of the Philippines [100,127], monsoonal SW India [73], and South China [128], as well as the wet mountains of Costa Rica, Atlantic Brazil, and Puerto Rico [78,129,130]. Most of these locations had a large seasonal rainfall surplus and sufficiently deep soils; these conditions were identified by Peña-Arancibia et al. [81] and Bruijnzeel [119] as being conducive to increased baseflow after forest restoration. In all cases infiltration improved. For the wet montane cases, the increasing capture of cloud water (fog) as the trees gained height and exposure to the prevailing winds likely contributed to improved dry-season flow as well [77,131,132].

Similar increases in baseflow have also been observed under various drier, non-tropical conditions (but deep soils), such as in Texas after heavy grazing ceased [133], in montane Ethiopia after implementing soil conservation measures coupled with a judicious use of native trees [134], and in the world’s largest afforestation experiment: the Loess Plateau in China [68,135]. For the latter, the increased winter baseflows (during the cold dry season) some 20 years after the widespread introduction of soil conservation measures and vegetation is somewhat puzzling in that the establishment of the shrubs and trees under the prevailing low rainfall has produced a dry layer within the soil that hampers deep percolation and groundwater recharge [136,137,138]. The application of stable isotopes suggests recharge in the area takes place through macropores during periods of high rainfall [139]. Although such macropores are found mostly beneath shrubs and trees [96,140,141], it is possible that groundwater recharge in the area is linked primarily to places where soil conservation measures favour infiltration (i.e., valley-bottom check dams, terraces with agricultural crops rather than taller and deeper rooted vegetation). Further work is necessary to separate the effects of forestation and soil conservations works on baseflows in the area.

5.3. Recovery times for regaining hydrological functioning and baseflow integrity

Most studies suggest that a period of at least 10–15 years of soil recovery is needed following forestation before the stormflows generated by some of the more intensive rainstorms can be attenuated and a more normal level of hydrological functioning is achieved, including the recovery of baseflows [74,93,142]. This length of time is needed to build the soil humus layer and dense understorey, which together, eliminate the erosive power of rain drops reaching the forest floor, as well as promote infiltration through root network development and slowed down surface runoff [143,144]. Recovery time will also depend on initial soil and vegetation conditions at the time of forestation, rate of vegetation development, and the prevailing rainfall intensities. Large increases in topsoil infiltration capacity associated with forestation will not have much of an effect on surface runoff and stormflow generation if rainfall intensities are mostly low to moderate, as is typically observed at higher elevations in the tropics [145,146,147]. Furthermore, both the intensity of pre-forestation land use (notably grazing) and post-forestation usage (e.g., harvesting of litter and fuelwood, grazing) exert a distinct influence on the rate and magnitude of soil hydraulic recovery [74,148]. Recovery in infiltrability can be stalled in cases of severe soil compaction/consolidation, such as can result from mechanised timber extraction [149,150,151].

As for the net hydrological effect of improved infiltration and increasing water use by vegetation over time, both rainfall interception losses and soil water uptake (transpiration) tend to increase rapidly with increasing leaf biomass during the first two decades of tropical forest regrowth, with transpiration stabilising when the leaf area index reaches ~4 m² m^-2 [152,153]. Baseflow increases (if any) may therefore take a decade or more in humid areas [70,100,128], and even longer in water-limited areas [68]. Again, tree water use in plantations is often seen to decline as they mature, with timing varying greatly between species. For example, work has shown that for eucalypts growing under subtropical conditions, a decline in water use sets in after about five years, whereas for pines it varies between 15 and 25 years [120,121,122]. Similar indications of reduced water use at advanced stages of natural regrowth (>35 years) have been reported under temperate conditions [154,155]. Comparable evidence for semi-mature regrowth from the tropics seems limited to a single case in Central Amazonia where overall ET from ~20-year-old regrowth was still about 20% greater than that from nearby old-growth forest [156]. There is a clear need for new dedicated work documenting the changes in water use (and streamflow) as different types of tropical forests mature [24,58,153].

As much of the tropical world includes seasonally-dry climates [16,157], it is sensible to avoid high tree water use in forestation in such water-limited areas [137,138], also in view of the seemingly more limited opportunities for boosting baseflows through forestation (as opposed to soil and water conservation measures to promote infiltration [134]) compared to wetter areas (Figure 2). Several precautions would help to limit excessive water use by planted vegetation [121,158,159,160,161,162]: (a) choosing species judiciously (e.g., native, slow-growing or deciduous rather than exotic, fast-growing evergreen trees); (b) maintaining a mosaic of vegetations of different ages and types; (c) avoiding short rotations, overly dense planting, and coppicing; and (d) optimising tree cover in accordance with prevailing rainfall and soil conditions.

Figure 2. Modelled increases in annual baseflow (mm/y) after reforesting degraded land across the tropics [modified from [119]; based on work by JL Peña-Arancibia]. See Peña-Arancibia et al. [81] for background on methods, the model and materials.

Figure 2. Modelled increases in annual baseflow (mm/y) after reforesting degraded land across the tropics [modified from [119]; based on work by JL Peña-Arancibia]. See Peña-Arancibia et al. [81] for background on methods, the model and materials.

In one example, maintaining an intermediate tree density with scattered tree cover in an agroforestry parkland setting in seasonally dry West Africa had notable hydrological benefits [163]. This approach resulted in improved groundwater recharge, with local increases in infiltration and reductions in surface runoff and soil evaporation. The groundwater recharge in this scenario was found to be five to six times higher compared with areas without trees [163]. In addition, the trees supplied surrounding shallow-rooted crops with moisture through ‘hydraulic lift’ during dry periods [164]. Rey & Garrity [165] documented the rapid expansion of naturally regenerating trees on agricultural fields elsewhere in the West African Sahel, suggesting such farmer-managed natural regeneration to be a low-cost approach to large-scale forest restoration. More work is needed to characterise and develop suitable management strategies for diverse tropical settings (including areas that have been heavily disturbed or grazed, have acidic or salt-affected soils, or very low or very high rainfall, etc.), to restore on-site hydrological functioning with potentially positive impacts on the flow regime of affected rivers and streams [134,165,166,167,168,169].

6. Increased tree cover, moisture recycling, and precipitation

A potentially important positive feedback of FLR is that higher ET associated with maturing trees may contribute to increased rain at downwind locations via moisture recycling and convergence. Considerable advances have been made in the last decade in understanding the global water cycle and the role of trees and forests in this regard [80,126,170,171,172,173,174]. Large-scale studies that track air flows, atmospheric moisture and rainfall show that air passing over large tracts of forest captures more water and produces more rain downwind than does air that passes over sparse vegetation [175]. For example, the high year-round ET from the largely forested Congo River Basin provides >30% of the precipitation falling over large adjacent river basins to the north (e.g., Lake Chad) and south (e.g., Zambezi) [176]. Meanwhile, up to 40% of rainfall in East Africa is estimated to derive from moisture evaporated from large-scale irrigated agriculture in Asia [177]. Likewise, a pan-tropical assessment of the impacts of forest loss between 2003 and 2017 on precipitation demonstrated clear reductions in observed precipitation at scales >50 km, with the greatest declines found at a scale of 200 km, which was the largest scale considered [90].

Recent attention has also been focused on tree cover loss and its contribution to ‘tipping points’, where local climates may reach a threshold and become unable to sustain the existing moisture regime [178,179]. Drought-prone areas are also susceptible to reinforcing feedbacks, where up-wind drought conditions can lead to significant reductions in precipitation [180].

A grand vision for forest restoration aims to reverse these processes, cooling and stabilising the climate while restoring reliable moisture supplies to regions currently facing threats or diminished water availability [181,182,183,184]. Makarieva et al. [79] and Sheil [183] emphasise the potential benefits of increased tree cover, in general, including strategically placed forestations, possibly spanning across borders, to enhance rainfall and water availability downwind through intensified moisture convergence [184,185]. Similarly, promoting the infiltration of rainfall through soil conservation measures in seasonally dry areas can elevate soil moisture during the rainy and immediate post-monsoon periods [134,165,167]. This, in turn, has a favourable impact on air temperatures (reduced) and seasonal precipitation [186,187,188,189].

Given the bi-directional and highly non-linear nature of the underlying relationships, large-scale forestation may affect precipitation and water availability positively through moisture convergence once atmospheric moisture contents are high enough [79,190,191]. There is a need to subject these dependencies to critical evaluation, however. Predictions of the magnitude of the effects of land cover (change) on precipitation vary markedly depending on methodological choices [189]. Further, an unresolved aspect concerns the fraction of the rainfall generated by the large-scale recycling of evaporated moisture that is sufficiently intense to contribute to deep drainage and groundwater recharge (i.e., affecting baseflows). If most of the rain falls at relatively low intensities, much may then be intercepted, evaporated, and/or used by the vegetation during transpiration instead of contributing to groundwater recharge and baseflow [96,98,139]. In the case of the Loess Plateau of China, total precipitation resulting from regional atmospheric moisture convergence increased in some areas following large-scale vegetation restoration, but the amounts of ‘intense’ precipitation (defined locally as >12 mm d^-1) decreased between 2000 and 2015 [192].

Although model predictions of the magnitude of increases in precipitation following large-scale forestation vary depending on model choice and parameterization [189], relatively strong effects are invariably predicted for montane humid tropical locations [20,21,193]. Globally, the fraction of evaporated moisture that precipitates again ‘locally’ (i.e., at a distance <50 km from its source) is estimated at <2%; however, local ‘moisture recycling ratios’ may reach values of 5–7% in tropical mountain areas [80]. Such model predictions are supported by observed increases in the height of the local cloud base following removal of forest in adjacent upwind humid tropical lowlands [194] and by upward / downward movement of the local cloud base reflecting defoliation / regrowth of leaves after forests have been impacted by hurricanes [195].

At these smaller scales, occult contributions via cloud water (fog) capture by (mostly taller) vegetation can be a crucial additional source of moisture in forests within coastal or montane cloud belts [76]. Such extra inputs assume particular importance in semi-arid regions [75,196,197] where fog has been shown to markedly increase groundwater recharge [198,199], but occult contributions can be substantial under wet conditions as well [76,77,200]. Occult inputs have also been reported to aid tree establishment under conditions where rainfall alone is not sufficient [132,201,202]. Nearly two-thirds of all montane tropical forests experience significant incidence of fog and low cloud [203]. Knowledge of ‘hot spots’ with high fog interception in the tropics may be used in conjunction with knowledge of local moisture recycling patterns to identify suitable areas for enhancing cloud water capture through up-wind FLR projects [76,80,204].

7. RESEARCH NEEDS

Given the extent of tropical forest loss/degradation worldwide, we call for greater involvement of hydrologists and atmospheric scientists in the development and assessment of FLR initiatives [Dib et al., 2023]. There are also opportunities for the involvement of ‘citizen scientists’ to open avenues for data collection that were previously unavailable [205]. More monitoring of the changes in streamflow, vegetation water use, and the soil physical characteristics governing infiltration, plus changes in rainfall (if any) associated with FLR is needed to further improve our understanding of the potential hydrological impacts of forestation initiatives, and to separate these impacts from those of other changes taking place (such as reservoir construction and operation, urbanization, global warming).

Our capacity to measure, monitor, and model environmental change in situ is greater than ever nowadays, including the remote sensing of soil moisture [206] and vegetation water use [207,208]. Further, sophisticated models and observations allow for more accurate descriptions of seasonal changes in dominant atmospheric moisture transport pathways and precipitation source areas, while stable water isotope measurements of precipitation allow tracing of water sources (terrestrial versus oceanic) and the dominant rainfall generating mechanism (convective versus orographic) [209,210,211,212]. However, advances are still needed to elucidate the effect of atmospheric moisture convergence on groundwater recharge and baseflows (as opposed to amounts of total streamflow, which are typically modelled). Doing so will require including rainfall partitioning at the soil surface (into surface runoff and infiltration) in such simulation models. Thus far, model applications have assumed that all precipitation arriving at the soil surface is infiltrating [20,172,185].

8. CONCLUSION

Although a recent ”landmark theme issue” comprised of 20 papers in Philosophical Transactions of the Royal Society B 378 addressed “knowledge gaps that need closing to advance restoration practice” [23], it did not include a dedicated article on the linkage between forest landscape restoration and hydrology (which was their knowledge gap no. 10). Addressing that omission, we argue herein that healthy soils and a reliable supply of high-quality water are crucial for human and ecological well-being, making them essential considerations for any FLR project. This belief is supported by recent field and modelling studies which demonstrate that forest landscape restoration has the potential to restore the hydrological functioning of degraded tropical catchments disrupted by forest loss and disturbance. However, political will and a host of socio-economic factors aside [5], the realization of this potential also depends on various physical factors, including local climate (such as seasonality and rainfall patterns), soil conditions (degree of surface degradation and soil depth), and the choice of vegetation. The timing and extent of hydrological recovery also rely on the initial level of soil and vegetation disturbance and how the emerging vegetation affects the partitioning of precipitation into evaporative losses, surface runoff, and subsurface flow components.

We agree with Marshall et al [23] that “there has never been a more important time to deliver the scientific foundations for effective and long-lasting impacts of forest restoration that meets the needs and priorities of different stakeholders”, particularly in the tropics. The evidence we present in this paper challenges the often-repeated mantra that 'more forest implies less water' by highlighting circumstances where more positive improvements in streamflow dynamics have been brought about via forestation and FLR. Regarding the goals of FLR, we advocate for prioritising the recovery of baseflow, mainly achieved through improving soil infiltration, rather than focusing solely on increasing annual water yields. Restoring this 'high-quality' streamflow component also aligns with the need to address other essential ecosystem services, including carbon sequestration, biodiversity, habitat preservation, soil erosion prevention, and (non-reservoir-associated) human water uses.

[Word count: 6749 words with citations numbered in the text and including table and figure captions]