蒸散发饱和现象加剧陆地水产出对气候变化的敏感性

《Nature Communications》：Evapotranspiration saturation amplifies climate sensitivity of terrestrial water yield

本研究针对气候变化下水资源预测的挑战，基于全球FLUXNET通量观测数据，发现生态系统蒸散发(ET)存在约480±210 mm yr-1的饱和上限，显著低于Budyko模型的理论预测。这一饱和现象导致水产出(WY=P-ET)对降水量变化的敏感性被放大，意味着干旱区生态系统的可持续性阈值将更快被触及，而湿润区洪水风险增加。该研究为评估气候变化对水资源及生态系统风险提供了关键指标。

　　

在全球气候变化背景下，水资源的可持续性正面临严峻挑战。降水格局的改变——一些地区日趋干旱，而另一些地区却洪涝加剧——直接影响着人类社会的生存与发展。陆地生态系统中，降水是水分的唯一来源，其中超过60%通过蒸散发（Evapotranspiration, ET，包括植物蒸腾和土壤蒸发）过程返回大气，剩余部分则形成至关重要的水产出（Water Yield, WY），即径流、地下水补给和可供人类使用的水资源。因此，准确理解蒸散发如何响应气候变化，是预测未来水资源可用性的核心。

传统上，水文科学家依赖Budyko框架来预测蒸散发。该理论认为，在干旱地区，蒸散发受水分（降水）供应限制；而在湿润地区，则受能量（如太阳辐射）限制。这意味着只要有充足的水分或能量，蒸散发就可以持续增加。然而，这一基于物理过程的模型是否完全反映了真实生态系统的行为？越来越多的证据表明，植物自身的生理调节在其中扮演着关键角色。植物叶片上的气孔（Stomata）在吸收二氧化碳进行光合作用的同时，会不可避免地损失水分。为了在碳获取和水分消耗之间取得最佳平衡，植物会动态调节气孔开度。这种复杂的生物调控机制，使得实际蒸散发可能并不像理论预测那样“随心所欲”地增长。那么，真实的生态系统蒸散发是否存在一个上限？如果存在，这一现象又将对水资源的未来产生怎样的影响？

为了回答这些问题，由Eyal Rotenberg和Dan Yakir领导的研究团队在《Nature Communications》上发表了他们的最新研究。他们利用覆盖全球的FLUXNET涡度协方差通量观测网络数据，对185个站点、1041个观测年度的数据进行了深入分析。涡度协方差（Eddy Covariance, EC）技术能够直接、连续地测量生态系统与大气之间的二氧化碳、水汽和能量交换，为了解生态系统尺度的蒸散发提供了最直接的证据。

研究人员首先将实际观测到的蒸散发（ET_obs）与基于Budyko方程预测的蒸散发（ET_pred）进行了比较。结果发现，当蒸散发超过约500毫米/年时，观测值便显著低于预测值，并且似乎触及了一个“天花板”。进一步的分析揭示，全球生态系统的年蒸散发存在一个明显的饱和点，平均值为480 ± 210毫米/年。即使年降水量变化范围高达近3500毫米，蒸散发的年际变化也相对保守。只有在少数热带地区高净辐射的森林和稀树草原生态系统中，蒸散发才偶尔超过1000毫米/年。这一饱和现象跨越了不同的气候带和生物群落类型，表明其可能是一个普遍规律。

那么，为什么蒸散发会饱和？研究团队从植物生理学角度给出了解释。核心机制在于蒸散发与光合作用（Gross Primary Production, GPP）的紧密耦合。光合作用本身存在生化饱和点，当光照达到一定强度后，由于羧化、电子传递等过程的限制，光合速率不再随光强增加而线性增加。因此，为了优化碳获取，植物的气孔开度也不会无限增大，从而限制了蒸散发的进一步提升。多余的能量必须通过其他途径耗散，主要是显热通量（Sensible Heat Flux, H）。全球数据中观测到的鲍文比（Bowen ratio, β=H/LE）平均值为1.1 ± 0.6，蒸发分数（Evaporative Fraction, EF=LE/(H+LE)）平均为0.47，证实了显热通量在能量平衡中的重要作用。此外，全球水分利用效率（Water Use Efficiency, WUE=GPP/ET）相对稳定并趋向于生物群落特异的最优值，也支持了碳-水耦合是导致蒸散发饱和的关键。因为植物维持着相对稳定的碳收益与水损失之比，所以蒸散发无法独立地追踪水分或能量可用性的增加。

研究人员还谨慎地评估了数据不确定性（如能量平衡闭合问题）的影响，但即使进行极端修正，蒸散发饱和阈值也仅从480毫米/年升至约550毫米/年，并不改变基本结论。

与蒸散发的“僵化”形成鲜明对比的是，水产出对降水量的响应则几乎是线性的。 across the full precipitation range at the biome scale(R2> 0.8 in most cases; Fig. 2). Grasslands exhibit lower slopes 160 than forests, likely due to faster adjustments of total leaf area to changes in

161 This linearity reveals a striking sensitivity of WY to changes in P. For example, a~40% decline in 162 precipitation(from~700 mm to~400 mm, within the range projected under climate change scenarios1163 would cause nearly a 100% reduction in WY- essentially eliminating any flow of water beyond ET, in 164 the absence of any ecosystem or land cover adjustments. In other words, relative changes in WY are much 165 larger than the underlying changes in P. This amplification is most acute under dry conditions, where WY 166 is already close to zero, emphasizing the vulnerability of ecosystems to drought and mortality. Vegetation 167 responses to such stress- including tree mortality, reduced stand density, or transitions to grasses- can 168 reduce ET and eventually restore positive WY, but often at the cost of biodiversity, carbon storage, and 169 ecosystem services51.

170 The concept of PWY=0, the precipitation threshold for zero WY(the x-intercept in WY-P relationships),171 provides further insight. Globally, WYp=0 averages 371±88 mm(Fig.2, Table S1), a value consistent with 172 the ET saturation point of 480±210 mm noted above. These thresholds likely reflect local adjustments in 173 biome type, stand density, and leaf area52-54. Importantly, WYp=0 defines the limit of ecosystem 174 sustainability: when precipitation balances ET on average, ecosystems cannot persist indefinitely, since soil 175 moisture carry-over buffers water deficits for only 1-2 years17. Indeed, analyses of interannual means of P 176 and ET at sites with≥3 years of data show only one case where ET> P, and none when≥5 years are 177 considered. While shallow groundwater can locally support ET, its influence is limited to 22-32% of the 178 global land area55 and does not alter the global WY-P relationship. Thus, while WY does not replace P or 179 ET as a metric to assess climate change, it offers a clear advantage in this context: it integrates inputs and 180 losses while magnifying their combined effects.

181 Notably, our observationally-based results demonstrating enhanced sensitivity of water yield to changes in 182 precipitation are also observed in the widely used coupled model intercomparison project(CMIP6) future

183 projections. We obtained mean outputs for annual precipitation and evaporation over land(E) from 27

184 CMIP6 models56(see Methods) to examine the relative changes in the water yield(termed here WY-E)

185 between the end of the 20th and the end of the 21st centuries(1980-1999 compared with 2080-2099, using 186 the SSP5-8.5 scenario)(Fig.3).

187 It is important to note that vegetation responses to rising CO2 and climate change-such as changes in 188 stomatal conductance, transpiration, and leaf area-are already embedded in the CMIP6 Earth System

189 Models used here 57,58. These physiological effects are well documented to modify evapotranspiration and 190 water fluxes, and thus are implicitly included in model-based WY=(P-E) projections.

191 Consistent with our observational results, the model projections demonstrate that in drying regions, a 21%

192 decrease in P(e.g., from~500 to~400 mm, as projected for the Mediterranean) is associated with an~87%

193 decline in WYp-E(Fig. 3)- a four-fold amplification. Conversely, a 21% increase in P(e.g., from~800 to 194~970 mm) yields a~44% increase in WYp-E, a two-fold amplification. Across drying regions, the mean

195(negative) enhancement ratio( ΔWYP?E/ΔP) is 155%, compared with 112% in wetter regions. This

196 asymmetric amplification resembles that observed in productivity responses59,54. While uncertainties remain

197 in the magnitude of these responses across models, their incorporation reinforces our conclusion that WY 198 is a sensitive, and amplifying, indicator of future climate projections.

Taken together, these results highlight that all water-budget terms matter, but WY has a particular advantage

200 for climate change impact assessment. By combining integration across terms with amplified sensitivity,

201 WY provides a powerful lens for identifying ecosystem thresholds, sustainability limits, and risks to water

202 resources60.

203 Insights from site-scale evapotranspiration and water yield: The 56 sites that reported at least six years of

204 evapotranspiration data with a sufficiently large range in precipitation(average Pmax/Pmin~ 2.5) permitted

evapotranspiration data with a sufficiently large range in precipitation(average Pmax/Pmin~ 2.5) permitted 205 the analysis of water yield versus precipitation at the single site level. In some cases, paired sites with

206 different vegetation types could also be used(Fig. 4, Table S2). Such higher resolution analysis revealed

207 patterns similar to those observed at the larger biome scale, with large inter-annual variations in

208 precipitation(e.g., a P range of~600 mm; Fig.4A), but a conservative range of evapotranspiration values 209(ET range of~250 mm; Fig. 4A red data points), and a highly linear WY to P correlations in all sites 210(R2>0.8). Examining paired forest and non-forest ecosystems located in close proximity to each other 211 demonstrates the impact of vegetation type on the sustainability limit(i.e., on their WYp=0 value obtained 212 from the equations in Fig. 4), which is as high as 551 mm in the forests but only 276 mm in the adjacent 213 non-forest grassland site(Fig.4B). These site-scale results highlight both the high sensitivity of water yield 214 to decreases in precipitation and the significance of land cover type for sustainability. Such results also 215 indicate the ecosystem‘safety margin’, that is, the difference between the annual mean precipitation and

216 the WYp=0 values. Such a safety margin demonstrates, first, the extent to which average evapotranspiration

217 can be maintained as average precipitation declines while still permitting residual water yield(P>ET). For

218 example, the results in Fig. 4B(and the equations therein) indicate a safety margin of 195 mm for forest

219 biomes(being the difference between mean P= 746 and the local WYp=0= 551 mm y-1). Second, it can be 220 greatly extended to 470 mm in the grassland at the same location. Alternatively, ecosystem adjustments

221 through tree mortality and reduced stand density can occur and will also improve water availability for 222 societal needs.

In conclusion, we show that, for a given land cover and land use, ecosystem evapotranspiration values are

224 surprisingly conservative. Evapotranspiration exhibits apparent saturation at a value below the energy

225 limiting values and largely independent of climate and biome type. This inflexibility amplifies the

227 disproportionately large impacts on WY. This sensitivity is evident across observations and model

228 projections with implications for increased flood risk in wetter regions and ecological collapse in drier

229 zones. In such cases, land cover and water management must change2,7. Given its capacity to integrate both

230 climatic forcing and biological regulation, WY should be adopted as an important indicator of hydrological

231 sensitivity and ecosystem sustainability under climate change.

主要技术方法

本研究主要依赖于全球FLUXNET网络（包括EuroFlux、AmeriFlux、AsiaFlux）的涡度协方差（EC）站点长期观测数据，数据时间步长为半小时。研究团队对来自全球365个站点（涵盖森林、草地、稀树草原和灌丛生态系统）的2745个观测年度的数据进行了严格的质量控制，确保每年有超过300天的有效数据才用于年尺度分析。关键变量包括蒸散发（ET，由潜热通量LE计算得出）、降水量（P）、净辐射（R_n）和总初级生产力（GPP）。此外，研究还利用了第六次国际耦合模式比较计划（CMIP6）中27个地球系统模型对未来（2080-2099年相对于1980-1999年，SSP5-8.5情景下）陆地降水和蒸发（E）的模拟输出，以评估水产出（WY = P - E）的相对变化。

研究结果

蒸散发饱和跨越生态系统

全球观测数据清晰地表明，生态系统蒸散发存在饱和现象。当蒸散发超过约500毫米/年后，其实测值（ET_obs）显著低于基于Budyko方程以Priestley-Taylor方法估算的潜在蒸散发（PET）得到的预测值（ET_pred）。全球1041个数据年的分析显示，蒸散发的饱和上限约为480 ± 210毫米/年，且在不同生物群落和气候带中变化很小，而同期降水量变化范围近3500毫米。这表明蒸散发对水分供应增加的响应存在刚性。

能量限制并非主导因素

理论上，在湿润地区，蒸散发应受净辐射（R_n）等能量因素限制。然而，全球数据表明，潜热通量（LE，与ET成正比）随净辐射增加的斜率仅为0.42，远低于能量限制线（LE = R_n）。这意味着大部分过剩能量并非用于蒸散发，而是通过显热通量（H）等其他途径耗散。即使在斜率较高的生物群落中，也远未达到能量限制。这表明能量可用性并非实际蒸散发饱和的唯一或主要解释。

水分产出对降水量的敏感性

与蒸散发的饱和特性相反，水分产出（WY）对降水量（P）的响应在生物群落尺度上呈现高度线性关系。例如，降水量减少40%（从约700毫米降至400毫米，这在气候变化情景预测范围内）可能导致水分产出近乎100%的减少，即在没有生态系统调整的情况下，几乎完全耗尽其可用水资源。这种放大效应在干旱条件下尤为显著，因为此时水分产出已接近零。全球平均的水分产出为零的降水阈值（WY_P=0）为371 ± 88毫米，与观测到的蒸散发饱和点一致，定义了生态系统的可持续性极限。

模型验证与未来预测

基于CMIP6多模型集合的未来气候预测支持了观测结果。在干旱化区域（如地中海地区），降水量减少21%（例如从500毫米降至400毫米）预计将导致模型估算的水分产出（WY_P-E）减少约87%，放大倍数达四倍。而在降水增加的区域，增幅也被放大。这种不对称的放大效应与生产力响应模式相似，进一步证实了水分产出作为气候变化影响放大器的角色。

站点尺度的验证与启示

对56个具有至少六年观测数据且降水量变化范围较大的站点进行分析，结果与生物群落尺度一致。年际降水量波动很大，但蒸散发变化范围相对保守，而水分产出与降水量保持高度线性相关。对比相邻的森林和非森林（如草地）生态系统发现，植被类型显著影响可持续性极限（WY_P=0）。例如，一处站点森林的WY_P=0高达551毫米，而相邻草地仅为276毫米。这定义了生态系统的“安全边际”，即年平均降水量与WY_P=0的差值，草地的安全边际远大于森林。

结论与意义

本研究通过整合全球通量观测和气候模型模拟，揭示了生态系统蒸散发存在一个相对固定的饱和上限，这主要源于植物碳-水耦合的生理优化机制。这一饱和现象使得水分产出对降水量变化的敏感性被显著放大。这意味着，在未来气候变化下，干旱地区生态系统将比预期更快地逼近其生存阈值，面临更高的退化风险；而湿润地区则可能因水分产出对降水增加的线性响应而面临更大的洪水风险。因此，水分产出（WY）作为一个能够整合气候强迫和生物调控的指标，比单独的降水量或蒸散发更能敏感地反映气候变化对水资源和生态系统可持续性的影响，应被纳入气候影响评估和水资源管理的关键指标体系。