1. Introduction
There is an increasing recognition of the roles of forests and trees in mitigation and adaptation strategies to global climate change [
1,
2]. Tree planting can modify local climate through impacts on temperature, wind speed, and humidity; it can also influence the landscape–scale water balance, cloud cover and albedo, and contribute to global carbon sequestration [
3,
4,
5]. In (sub-) tropical developing countries, re– and afforestation have been included in the portfolio of mitigation efforts as a cost-efficient strategy to reverse the degradation of forests and to increase their atmospheric carbon uptake [
1,
2,
6,
7]. Although trees are central to many climate change adaptation and mitigation strategies, they are vulnerable (particularly in their early growth), to variation in solar radiation, rainfall, and temperature as determinants of potential evapotranspiration (ET
0). Changes in these climatic variables, particularly temperature increases and precipitation shortages leading to higher frequencies of extreme weather events (i.e., severe drought and intense rainfall), are expected to affect tree growth and challenge the sustainable management of forests and tree plantations [
8,
9,
10].
Tree growth sensitivity to drought may be substantial in West African semi-arid zones, where water availability is one of the most limiting factors of plant growth, e.g., [
11], and where extreme drought events are projected to become more frequent [
12,
13,
14]. In this region, extreme drought events are associated with extended dry spells, low air humidity, high atmospheric evaporative demand, and high air temperatures [
15], potentially increasing water stress in trees [
13]. Droughts and dry spells are a major threat to the establishment and early growth of both tropical and temperate tree species [
16,
17,
18], suggesting the need to assess the effects of climate variability on their growth. However, studies quantifying the relationships between these climatic factors and tree growth are lacking for tree plantations in the semi-arid tropics. This information is urgently needed to explore species’ responses to past and current climate and evaluate the effects of climate change on tree growth [
19].
If drought occurrences are exacerbated by global warming, it stands to reason that the sustainability of forests and plantations will largely depend on the physiological adaptations and changes in silvicultural management [
18,
20]. For example, Abdulai et al. [
14] revealed drought vulnerability in a cocoa-based system in the forest–savanna transition zones of West Africa, despite the availability of subsoil water below a depth of 75 cm. This observation suggests that drought vulnerability may be related to trees’ ability to develop sufficiently deep root systems [
21]. Hartl-Meier et al. [
22] found that among three tree species growing under similar temperate climate conditions, those with deep root systems benefited from the access to deep soil water, as reflected in their higher photosynthetic activity and continued biomass production during severe drought conditions. Therefore, the consideration of root systems in the analysis of climate–growth relationships may give new insights on species’ responses to extreme climate events. Another plant strategy to avoid mortality under drought conditions is to shed leaves and reduce transpiration losses; such drought avoidance is likely to reduce the rate of photosynthesis and thus mean growth rates, but also reduce vulnerability [
23].
Assessing the sensitivity and/or response of (early) tree growth to climate change requires the identification of the climate variables that most strongly influence tree growth or have the best predictive power in evaluating the early growth risks, e.g., Teets et al. [
24]. However, experiments carried out in any given year to quantify climate–growth relationships of young trees may exhibit a sample of bad luck, lucky exceptions or be representative of ‘normal’ conditions at the test location, because experiments capture one (or a few) of the many possible weather (e.g., rainfall and temperature) sequences for this location. Early growth sensitivity of trees to predicted climate change cannot thus be directly derived from existing weather statistics because it is uncertain how tree transpiration, soil evaporation and temporal rainfall patterns interact, e.g., Raz-Yaseef et al. [
25]. In this regard, we argue that well-calibrated tree–soil interaction models, simulating sapling growth responses for at least 30 years of weather data can help in (a) identifying the best predictors among a wide set of metrics that can be objectively derived from existing weather records; (b) evaluating where the years with experimental data are positioned within current climate variability according to the metrics with the highest predictive power identified in (a); and (c) using the results of (a) and (b) to assess the probability of success for given species at a given location under various location-specific climate change scenarios. Our study may be the first to test explicitly this generic approach to assess climate change risks for newly established dryland afforestation sites. The datasets stem from a site in a semi-arid zone of West Africa, where tree-based land restoration is high on the agenda and existing information is not sufficient to assess the robustness of afforestation option for a range of tree species [
18,
20,
26].
Dendrochronology has been successfully used to analyze temporal and spatial climate–growth relationships for tree and shrub species in tropical and temperate forests [
9,
23,
27,
28]. However, dendrochronological applications may be inappropriate for young trees and compromised by problematic anatomical features such as missing, vague, discontinuous and false rings common in tropical species [
29,
30], thereby also restricting the use of biomass increments derived from tree-ring series [
31,
32]. An alternative approach to assessing the climate–growth relationships of saplings may be the use of process-based models in combination with field experiments. When supported by reliable soil–plant–climate databases and empirically derived relationships between environmental and plant eco-physiological parameters [
33,
34], process-based models can be used to investigate forest responses to climatic change and silvicultural management [
35,
36], as well as predict tree growth and productivity at both the plant and stand level. There have been few models capable of accurately simulating processes in tree-based agroecosystems.
The process-based Water, Nutrient and Light Capture in Agroforestry Systems (WaNuLCAS) model [
35] has had various applications to improve the understanding of complex ecological processes in tree-based farming systems [
35,
37,
38]. It has also been tested for its performance to simulate crop (e.g.,
Pennisetum glaucum L. and
Sorghum bicolor L.) and tree (e.g.,
Parkia biglobosa Jacq. and
Vitellaria paradoxa C.F. Gaertn) growth in agroforestry parklands in West Africa [
39,
40,
41]. The aim of our research was to assess the climate sensitivity of two afforestation species—
Jatropha curcas L. and
Moringa oleifera Lam.—in the early stages of their development (i.e., the first two years of growth), decisive for the future growth [
42] and also when plants are most sensitive to changes in water availability [
43,
44,
45]. The two species have been subjects of silvicultural trials on degraded cropland in northern Benin [
18,
20,
26]. The specific objectives were to (i) compare different indicators calculated from climate records to identify those with the greatest predictive power for the first two years of tree growth; (ii) quantify the climate–growth relationships of tree growth over the first two years, and (iii) test possible adaptation strategies to reduce the sensitivity of saplings to extreme climate events. Furthermore, we assessed the relevance of the empirical climate data vis-à-vis the long-term climatic variability in the study area based on the most influential climate indicators. We hypothesized that (i) there will be significant differences between the species’ growth responses to climate indicators, particularly to drought-related indicators; (ii) drought sensitivity of sapling growth decreases over time, depending on the establishment of roots; and (iii) deepening of the rooting system may aid the adaptation of saplings to extreme drought conditions.
4. Discussion
Climate change is likely to affect trees in the early growth stage, altering the long-term productivity of forests and plantations [
42,
44,
67]. Although climate–growth relationships have been quantified for mature trees of several tropical and temperate species [
9,
29,
68], much remains unknown regarding the sensitivity of tree growth during the early stages of development in semi-arid areas. In this study, we used a generic approach of climate sensitivity assessment that combines field experiments and modelling as an alternative to the traditional dendrochronology technique. First, we calibrated and validated the process-based model WaNuLCAS based on the empirical tree growth data. Second, we quantified the effects of climate variability on the predicted annual biomass increment, based on the most influential climate indicators. We also assessed the relevance of the empirical climate data to historical simulations based on the frequency distribution of the long-term climate indicator data (
Table 3). Our results revealed drought-related indicators as the best predictors, which enabled to quantify their effects on sapling growth. Our approach has an advantage over the dendrochronological analysis because it uses actual biomass increment rather than biomass increment derived from tree-rings as recently suggested [
31,
32], thereby avoiding challenges associated with problematic anatomical features of tree rings common in tropical trees [
29,
30].
4.1. Model Performance
In accordance with previous studies [
41,
57,
69,
70], the range of the GOF statistics and the high correlation between the simulated and observed growth parameters for both the calibration (
Figure 3) and the validation (
Table 4) are indicative of the ability of the WaNuLCAS model to reproduce the early growth dynamics of the tested afforestation species with an acceptable accuracy and precision. A R² value of 0.5, CD value of 0.5–2, and EF value above 0.5 represent a good predicted–to–observed relationship [
69]. However, the high values of the RMSE for AGB during the model calibration and the reduced GOF during the model validation indicated that not all growth limitations occurring under field conditions were adequately captured through simulations [
39]. Most of the discrepancies between the simulations and empirical evidence arose from the limited capacity of the model to reproduce the monitored drought-induced trunk shrinkage and litterfall during the dry season. Shrinking tree diameters are not represented in the current version of the model, although they are known at both the diurnal and seasonal time scales [
71]. Drought-induced litterfall is included in the model by a water potential threshold and a waiting period before new leaves emerge [
72], but an accurate parametrization requires data beyond what is available for the site. While dry-season leaf shed is characteristic for drought-deciduous tree species, its accounting in our simulations resulted in large reductions of the total height rather than D and AGB (results not shown), implying that tree canopy and height are more influenced by this process than D and AGB in the WaNuLCAS model. The lack of calibrated litterfall data may have caused the poor fit of the predicted D and AGB, albeit only during the dry season [
57]. Despite these deviations between the observed and simulated values during the dry season, the well-reproduced growth patterns and accurate prediction of D, H, and AGB at the end of the growing seasons (
Appendix A,
Figure A2) are a sufficient basis for further analyses of plant growth.
4.2. Climate Sensitivity of Afforestation Species
The predicted AGB growth series showed strong variations over the years (
Figure 3), indicative of the influence of climate on the early growth of tree plantations [
68]. Although both species have a semi-deciduous phenology and are fast growing and drought tolerant [
18], their predicted annual growth responses to climate variability differed (
Figure 3), confirming our hypothesis that the responses are species specific. This is likely due to the differences in stress tolerances [
73] and/or allometry. Species differences in climate–growth responses have been reported for older trees of winter-deciduous broad-leaved species (
Fagus sylvatica L. and
Quercus petraea Matt.) under a temperate climate [
68]. Together, these findings suggest that caution must be taken when parametrizing growth models according to plant functional types [
5].
Water availability emerged as the primary driver of the climate sensitivity of trees in the early stages of growth in semi-arid areas. Similar climate–growth relationships were previously found for temperate tree species [
68,
74] and for shrub species across the tundra biome [
28]. Annual biomass accumulation declined with annual water deficit, the length of the dry season, and the length of the longest dry spell, but increased with the annual total wet-day precipitation (
Table 5). This supports our hypothesis that drought-related indicators negatively affect biomass growth in particular. The negative climate–growth relationships with drought-related indicators suggested that drought-reduced sapling growth occurred not only due to reductions in the total amount of precipitation and subsequent longer dry seasons (e.g., AWD), but also due to variability in the distribution of rainfall (e.g., the LDSP). This outcome is reminiscent of the findings by Elliott et al. [
9], which showed that the distribution of precipitation is more influential on the radial growth of deciduous tree species than the amount.
Increased aridity (AWD) had the strongest negative influence on the growth of saplings, suggesting that the projected increase in aridity in northern Benin [
75] may result in a substantial decrease in growth. For instance, the predicted AGB of
J. curcas after two years of growth was 0.34 and 0.49 kg m
−2 under extreme (highest AWD) and mild (lowest AWD) drought conditions, respectively, which represents a potential 31% loss in AGB growth. Applying the same calculations for
M. oleifera resulted in a 14% loss in growth if the aridity is increased. These estimates are obviously to some extent simplified, as AWD is not the only climate factor that is likely to change in the future; however, they do highlight the relative importance of water limitation for the early growth of tree plantations and hence for the planted afforestation efforts in the region.
The influence of drought on tree growth and the identity of climate drivers were not uniform between years. In year one, the duration of the LDSP in the rainy season had the best predictive power, while in year two, the annual water balance (here, the difference between precipitation and ET
0) was the best predictor for both species (
Table 5). This supports our hypothesis that the climate sensitivity to ‘immediate’ drought is stronger at the very early stages of tree growth, when root systems are not yet fully established [
43,
44,
45], compared to later stages where the water balance dominates results. Sensitivity to AWD and LDS was greater in the second year compared to the first. This could be attributed to the increased tree water use and consequent reduction in soil available water related to increased canopy interception and water drainage as a function of tree growth in the WaNuLCAS model [
51]. For instance, the water use efficiency of
J. curcas, defined as the water uptake to total rainfall ratio, was 4% in year one and 10% in year two under the most extreme historical drought scenario (1984–1985). This implies that the same AWD or LDS would result in more water stress in year two than in year one. Hence, the climatic factors related to the amount of precipitation showed greater sensitivity in year two, whereas greater sensitivity to LDSP, which describes the distribution of precipitation, was more evident in year one.
Just as empirical data gives insights in the credibility of model predictions, the current analysis of sensitivity of tree growth to a range of rainfall metrics can help to better judge the representativeness of the empirical data regarding climate variability in the study area. The year 2014 had the longest LDS of the available climate data set for the test site, while the AWD and LDSP of both years were in the upper quartiles (
Table 3). The values of the wetness-related indicator (ATWP) were in the lowest quartile for both years. Apparently, the years in which experimental data were collected were already foreboding what climate change predictions point out as a ‘new normal’ for the study area: less predictable rain, with more extremes on both the high and the low end [
12,
13]. Therefore, the tree performance measured during the two years is probably a conservative estimate of what can be expected in years closer to the current average values and is more representative of the expected growth under projected climate conditions.
4.3. Effects of Rooting Depth on Biomass Growth under Extreme Dry Conditions
The simulation results showed that deeper rooting depth was an advantage to sapling growth of both species under extreme dry conditions (
Figure 5). Moreover, the AGB at the end of year two was greater for deep–rooted than shallow–rooted saplings in both extreme dry and wet conditions. The improved performance can be attributed to enhanced tree water use under deep rooting depth conditions as a result of increased water uptake and canopy interception, and reduced water drainage and runoff (
Figure 6). The relative increase in AGB, induced by the deep rooting depth, was greater under severe water stress compared to mild water stress, suggesting that deep-rooted saplings are less sensitive to extreme drought than shallow-rooted saplings. Similarly, Coulibaly et al. [
41] has reported that tree species with deep rooting systems are less vulnerable to water deficits. These results can be attributed to the fact that during extreme drought conditions, deep rooted saplings take advantage of their access to deeper soil horizons, which could result in increased photosynthetic activity and continued biomass production, as reported by Hartl-Meier et al. [
22] for beech (
Fagus sylvatica L.) and larch (
Larix decidua Mill.) under a temperate climate.
4.4. Replicability of the Approach
Where trees of a wide range of ages can be found in a local environment of interest, dendrochronological analysis can, in combination with climate records, be used for assessing the effects of climate change on young forests. In the absence of such, the combination we used of experiments, model calibration and identification of the local climate metric with the best predictive skill can improve the value of experiments and increase their interpretation, at the interface of climate change mitigation and adaptation strategies.
5. Conclusions
The WaNuLCAS model was successfully calibrated and validated to simulate sapling growth of two afforestation species in semi-arid northern Benin. Although the stem diameter and biomass predictions during the dry season were poorly simulated due to the limited ability of the model to reproduce trunk shrinkage and litterfall during this period, the overall outputs of the model calibration and validation were satisfactory. The model application to simulate the early growth of the afforestation species under past climate conditions permitted the quantification of the climate sensitivity of sapling growth.
The aboveground biomass growth was most sensitive to water availability, as evidenced by the negative relationships between growth and drought-related indicators, and the positive growth response to annual total wet-day precipitation. The distribution, rather than the total amount, of precipitation was the main factor limiting sapling growth at the very early stages of growth (year one), when tree root systems are not yet well established. Given the projected increase in variability of precipitation distribution, extended arid conditions, and longer dry spells in West Africa, the current results suggest that increased aridity could play an increasingly important role in limiting future tree establishment and forest growth. Based on the observed species-specific responses to climate variability and the importance of root depth in buffering the negative effects of extreme drought on sapling growth, a multi-species afforestation system with species that are able to develop deep-penetrating root systems may increase the resilience of plantations to climate change.
The current results illustrate that process-based modelling combined with field experiments can be effective in integration of multi-source data to assess the climate–growth relationships of tree species. Further verification of the simulation results under field conditions, through dendrochronological and rhizological studies, would be needed to develop confidence in the application of WaNuLCAS for climate–growth analyses.