Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
In recent decades, researchers have increasingly focused attention on how climatic change will affect ecosystem functioning. Much of this research has centred on the effects of rising atmospheric CO2 concentrations and increasing temperatures (e.g. Ainsworth & Long, 2005; Wu et al., 2011). The consequences of altered precipitation patterns, by contrast, have received far less attention. The results of precipitation manipulation experiments were only recently synthesized for the first time (Wu et al., 2011), and the conclusions that can be drawn regarding global patterns remain preliminary. In general, reduced water inputs slow ecosystem processes, while increased rainfall enhances plant productivity (Wu et al., 2011), but it remains unclear how this response differs among ecosystems.
When synthesizing water manipulation experiments, firm conclusions are not only precluded by the insufficient number of data points (Wu et al., 2011), but also by the difficulty of defining the magnitude of the perturbation experienced by the biota (further coined the ‘actual’ treatment). Unlike in the case of, for example, elevated CO2, the magnitude of the imposed treatment (e.g. −20% precipitation vs +300 ppm CO2) does not clearly indicate the actual treatment. Many factors influence the way an ecosystem experiences a change in rainfall; of greatest importance to the ecosystem is not the amount of incoming precipitation, but rather the amount of water that plants have access to. This ‘plant available water’ strongly depends on factors such as soil texture and rooting depth (Tolk, 2003), and the latter can show substantial seasonal and interannual variation (Knapp et al., 2008) and may differ between treatments. Additionally, runoff water and stem flow can complicate estimations of the magnitude of an imposed manipulation (Cotrufo et al., 2011), and plants that access groundwater supplies can complicate interpretations of treatment effects.
If we want to understand why plant responses to altered precipitation differ among ecosystems, comparisons of effects of precipitation manipulations must use a more ecologically meaningful metric to describe the actual treatment than merely the change in precipitation. Without such a relevant ‘common denominator’, observed differences in ecosystem responses to altered precipitation may reflect differences not only in ecosystem properties, but also in the actual treatment.
Potential metrics for quantification of the actual treatment
A common denominator that properly characterizes the actual treatment should reflect the drought stress as experienced by the biota. For plants, the most relevant indicators for drought stress are probably plant water potential and its derived indices (Myers, 1988; Rambal et al., 2003). Measurements of plant water potential are, however, labour-intensive and destructive and are therefore often not suitable for the quantification of the stress intensity integrated over an entire growing season and for multispecies systems. More feasible measurements, applicable for the ecosystem as a whole, are soil water content (SWC) and soil water potential (SWP). If measured over the entire rooting zone, SWP reflects the water that plants can extract from the soil (Tolk, 2003), which provides a useful measure for intersite comparisons. Currently, however, SWP is measured in very few experiments (Table 1), and we can therefore only focus on indices derived from the more commonly measured SWC.
Table 1. Number of experimental sites that provide the indicated data
No. of experimental sites
SWC, soil water content; pF, information about the soil water retention curve, with at least an estimate of SWC at field capacity and SWC at wilting point.
Soil water potential
SWC measured continuously
SWC measured intermittently
pF over entire rooting depth
SWC over entire rooting depth
pF and SWC over entire rooting depth
pF and continuous SWC over entire rooting depth
A first index that can be calculated if SWC and soil water retention curves (pF curves) are available for the entire rooting zone, is total extractable soil water (TEW):
where SWC1 is the soil water content for soil layer 1 and SWC1wp is the soil water content at wilting point for soil layer 1, etc. The entire rooting zone should be included in this way. The wilting point is typically considered to be −1.5 MPa (Tolk, 2003), but is in fact species-specific (Larcher, 2003).
Clay soils have much higher water-storing capacity than sandy soils, but plants can extract a given amount of water more easily from a sandy soil than from a clay soil (White, 2006). Comparison of TEW could therefore indicate that plants in sandy soils are more severely stressed (lower TEW) than plants in clay soils, when in reality the opposite could be true. Moreover, a modelling study indicated that soil texture is a key factor in explaining differences in ecosystem responses among experiments (Weng & Luo, 2008). A useful index that accounts for variability in water-storing capacity is relative extractable water (REW), calculated as:
where TEWmax is the maximum extractable water over the entire rooting zone and SWC1fc is the soil water content at field capacity for layer 1, etc. The entire rooting zone should be included in this way.
Relative extractable water can be useful when comparing, for example, stomatal conductance measurements across experiments. For ecosystem measures such as annual plant productivity that reflect differences in water availability over longer periods, however, intersite comparisons require an indicator of the stress that plants experienced during the entire period relevant to that process. Such an indicator should take into account the fact that soil moisture becomes a limiting factor only when it falls below a certain threshold. Such indicators can describe the duration of the stress period and the intensity of the stress. The duration of a stress period can be calculated as the number of days with REW lower than a threshold value. Stress intensity (Is) can be computed as in Granier et al. (2007):
where TH is the threshold (e.g. REW = 0.4; Granier et al., 2007) and REWt is the relative extractable water on day t.
Both stress duration and stress intensity reflect the actual treatment as experienced by the plants and can therefore be used as a common metric when comparing responses to altered rainfall in different experiments. Note that other indices (e.g. the indices used in the TECO model; Weng & Luo, 2008) can also be calculated with the indicated data.
Data needed from manipulation experiments
In the best case, SWP for the entire rooting zone is used for quantification of water availability and intersite comparison. Indices such as stress intensity can then be computed, without the need for pF curves to estimate water availability. When SWP is not available, SWC can be used instead, if SWC integrated over the entire rooting zone and SWC at field capacity and wilting point, applicable for the entire rooting zone, are available. Ideally, SWC (or SWP) would be measured continuously, although interpolation can offer a solution when SWC is measured less frequently. SWC at field capacity and wilting point are best derived from pF curves of the soil under study, but, if not available, pF curves may be estimated from soil texture (e.g. Van Genuchten, 1980). Information on soil texture is therefore needed as well.
Root distribution also needs to be assessed. For many ecosystems, maximum rooting depth and root depth distribution can be estimated via soil cores or minirhizotron analysis (e.g. Hendricks et al., 2006; this measurement is admittedly much more difficult to make in ecosystems with deeply rooted species). Once the rooting depth is determined, several options exist for nondestructively measuring SWC for the entire rooting zone (at high temporal frequency), including neutron moisture meters, time domain reflectometry systems and frequency domain reflectometry (Hignett & Evett, 2008). In some cases, however, plant roots penetrate the bedrock to extract considerable amounts of water (Schwinning, 2010), thus complicating the estimation of plant available water. Soil drying may also disrupt the contact between soil and sensor, resulting in unreliable measurements. We therefore recommend that researchers also measure plant water potential and/or sap flow to verify the estimated stress intensities based on SWC. Plant water potential, in particular, is of great interest because it is also useful for simulation modelling (Hoff & Rambal, 2003).
Drought quantification in reality
In order to test how different ecosystems respond to altered precipitation, we contacted site coordinators of all the precipitation manipulation experiments we were aware of (via networks such as Carbo-Extreme, CLIMMANI and INTERFACE), asking for the data needed to quantify the actual treatment, as described earlier, as well as for data related to ecosystem functioning. Data from 62 experimental sites were gathered (see Supporting Information, Table S1). Table 1 shows the data related to soil moisture that were available for the different experiments. Only four of the 62 experiments had collected the data needed to quantify the actual treatment as experienced by the plants, limiting our ability to compare the effects of altered precipitation on ecosystem processes across sites.
Using above-ground plant productivity (ANPP) as an example, we illustrate the need for measurements to quantify actual precipitation treatments (Fig. 1). For four experiments that had sufficient data, we computed the relative effect of precipitation manipulation on ANPP and plotted this against different indices: (1) relative difference in annual precipitation between treatment and control plots; (2) relative difference in growing season precipitation; (3) difference in stress intensity for topsoil only; and (4) difference in stress intensity for multiple soil depths. This analysis demonstrates that both within- and between-site comparisons vary with each index that is used; the relative positions of the different symbols change depending on the index. For example, according to the calculations based on data for topsoil only, the difference in stress intensity between treatment and control was larger in Soyface (both experimental years) than in Chamau (2006). Similar calculations including deeper soil layers, however, indicate the opposite. From the topsoil-only calculations, one would thus conclude that ANPP in Soyface was less sensitive to decreases in water availability than ANPP in Chamau (2006). The calculations including multiple soil layers, however, suggest that the difference in stress intensity in Soyface was minor compared with Chamau (2006). Note that Fig. 1 is intended merely to illustrate the problem, and is not for testing responses of ANPP to precipitation manipulation, because ‘all depths’ indicates only that SWC was measured at multiple depths. These depths did not cover the entire rooting zone.
Without a common denominator that characterizes the actual treatment, it is not possible to accurately compare the responses of ecosystem processes to changes in water availability across different experiments. Differences in ecosystem responses between experiments imposing the same precipitation manipulation remain attributable to differences in the actual treatment as well as to differences in biological factors. Conversely, similar responses across sites could indicate either similar process responses and similar changes in water availability, or offsetting differences in ecosystem process responses and the actual treatment (see also Box 1). Despite these problems, available data currently dictate that cross-site analyses focus on precipitation differences. Analyses based on precipitation data are useful in that they are easily compared with model results. However, because of the reasons highlighted earlier, it is important to recognize that extrapolations are highly uncertain. Once the necessary site-level data are available, synthesis studies can further improve our mechanistic understanding of ecosystem responses to changes in water availability. Questions such as ‘Are ecosystems in dryer regions better adapted to drought?’ can then be answered without speculation on potential confounding factors, such as rooting depth and soil texture.
Box 1 Two hypothetical examples, illustrating why the change in rainfall may not represent the difference in actual treatment experienced by the plants and is therefore not a reliable index for intersite comparison of effects of precipitation manipulation.
Observation: A similar rainfall reduction reduces plant production in grasslands more than in shrublands.
1. Shrubs have deeper roots and therefore experience less water stress than grasses.
2. Shrubs are better adapted to drought than are grasses.
3. The grasslands used in this hypothetical study were located on sandier soils (with low water-holding capacity), while the shrublands under study were located on loamy soils (with higher water-holding capacity). Hence, the grasslands experienced more stress than the shrublands.
4. The grasslands studied were on a steeper slope than the shrublands, causing a larger fraction of precipitation to run off. In the drier treatment, the fraction of precipitation lost as runoff increased more in the grasslands than in the shrublands.
5. All possible combinations of 1, 2, 3, and 4.
Observation: Plant production in shrublands and forests responds similarly to a similar reduction in rainfall.
1. The advantage given to forests by having deeper roots is offset by shrubs being better adapted to drought.
2. The difference in plant available water resulting from a certain reduction in rainfall varied to such a high degree that differences in the effect on plant productivity remained undetected.
We thank Ivan Janssens for organization of the (amusing) workshop during which this paper was discussed and for comments on previous versions of this manuscript. We acknowledge support from the European Community’s Seventh Framework Programme (FP7/2007–2013) under grant agreement no. 226701 (CARBO-Extreme) and from the European Science Foundation through the Research Networking Programme CLIMMANI. S.V. is a postdoctoral research associate of the Fund for Scientific Research – Flanders. L.v.d.L. was financially supported by the Villum Kann Rasmussen Foundation. M.E., R.O. and J.S. were supported by Spanish government projects CGL2010-17172/BOS and Consolider-Ingenio Montes CSD2008-00040. M.Z.O. was supported by Ministry of Science, Education and Sports project 024-0242049-2106 and Croatian Forests Ltd project HS-UE17. M.S. was supported by FWF Project P22214-B017. S.B.G. was supported by the DOE Global Change Education Program Graduate Research on the Environment Fellowship and A.D.B.L. was supported by a Beckman Fellowship from the Center for Advanced Studies, UIUC, and a Fellowship from the Environmental Change Institute, UIUC. The SoyFACE facility was supported by the US Department of Agriculture Agricultural Research Service; the US DOE through the Office of Science (BER) Midwestern Regional Center of the National Institute for Climatic Change Research at Michigan Technological University, under award number DE-FC02-06ER64158; the National Research Initiative or Agriculture and Food Research Initiative Competitive Grants Program, grant no. 2010-65114-20343 from the USDA National Institute of Food and Agriculture. The Boston-Area Climate Experiment was supported by the NSF and by the US Department of Energy’s Office of Science (BER), through the Northeastern Regional Center of the National Institute for Climatic Change Research.