PalaeoQUMP:using palaeodata to reduce uncertainties in climate prediction

数值模型是我们研究人类活动对气候的影响的唯一工具，例如大气中温室气体浓度的增加或土地利用的广泛变化。根据全世界所有主要模拟小组为政府间气候变化专门委员会（气专委）上一份评估报告进行的模拟，对大气中二氧化碳浓度加倍造成的全球气温变化的估计为1.5至4.5摄氏度。这种对二氧化碳加倍的“气候敏感性”的广泛范围的值使得政府很难定义和获得广泛接受的政策，以防止“危险的”气候变化。许多重要的过程在气候模型中以简化的方式描述，因为否则模型将过于复杂而无法运行。据认为，分配给关键过程的参数值的不确定性是一个主要来源的不确定性的模型预测的反应加倍二氧化碳。DEFRA-funded项目QUMP（量化模型预测中的不确定性）通过运行一系列现代气候模拟来研究这一点，其中参数值在基于当前知识的合理范围内系统地改变。然后将这些模拟结果与近期气候观测结果进行比较。从模拟中生成概率分布函数，根据再现观察结果的真实性对各个运行进行加权。根据这一比较，QUMP估计，大气CO2浓度加倍导致全球温度变化的5-95%概率范围（即气候敏感性）为2.4至5.4摄氏度。这意味着最近的气候观测所提供的限制不足以减少预测的不确定性。改进我们对气候敏感性的估计的一种方法是使用其他信息来评估模型预测，例如，过去地质学中关于气候与今天非常不同的时期的信息。冰盖覆盖了北方的大部分地区，海平面在末次盛冰期（LGM，21,000年前）时较低。海洋要冷得多，海冰更广泛，大气中温室气体的浓度比现在低。这些冰盖实际上已经消失了6000年前（中全新世，MH），但整个北方半球的入射太阳能的分布增加，在夏季和冬季减少相比，今天，导致季节性温度对比度和季风的强度增加。在这些非常不同的气候下测试气候模型应该对气候敏感性施加更大的限制。我们将运行一个经过充分测试的气候模型，使用已知的太阳辐射变化，冰盖分布，以及LGM和MH的温室气体浓度，并运行与QUMP为现代气候所做的相同系列的模拟（关键过程的不同值）。我们将使用LGM和MH气候的重建来评估这些模拟。有大量的证据表明，在湖泊和泥炭沼泽沉积物中的花粉在冬季温度的变化所造成的植被格局的变化，生长季节的长度和水的可用性在末次冰期和MH。通过强迫植被模型重现这些植被模式，我们可以得到气候这三个方面的定量估计。降水变化引起的湖泊面积变化有丰富的古岸线和湖泊沉积物证据。再一次，通过强迫一个湖泊模型再现这些面积变化，我们可以得到定量的降水估计。这些重建与基于同位素或地球化学数据的气候重建相结合，将成为我们模拟的目标。我们的项目旨在为IPCC第五次评估报告提供更好的气候敏感性估计。

Numerical models are the only tool we have to examine the consequences of human activities, such as increasing greenhouse gas concentrations in the atmosphere or widespread changes in land use, on climate. Estimates of the change in global temperature caused by doubling the atmospheric CO2 concentration, based on simulations run by all the major modeling groups worldwide for the last Assessment Report of the Intergovernmental Panel on Climate Change (IPCC), ranged from 1.5 to 4.5 degrees C. This wide range of values for the 'climate sensitivity' to a doubling of CO2 makes it difficult for governments to define and gain wide acceptance for policies to prevent 'dangerous' climate change. Many important processes are described in a simplified way in climate models, because the models would otherwise be too complex to run. It is thought that uncertainties about parameter values assigned to key processes are a major source of uncertainty in model predictions of the response to doubling CO2. The DEFRA-funded project QUMP (Quantifying Uncertainties in Model Prediction) has investigated this by running a series of simulations of the modern climate in which parameter values are systematically altered within plausible ranges based on current knowledge. These simulations are then compared with observations of the recent climate. A probability distribution function is generated from the simulations, with individual runs weighted according to their realism in reproducing the observations. From this comparison, QUMP estimated that the 5-95% probability range for the global temperature change caused by doubling the atmospheric CO2 concentration (i.e. the climate sensitivity) is 2.4 to 5.4 degrees C. This means the constraints supplied by recent observations of climate are insufficient to reduce uncertainties in prediction. A way of improving our estimate of climate sensitivity is to use other information to evaluate the model predictions, e.g. information from the geological past about times when climate was very different from today. Ice sheets covered much of the northern hemisphere and sea level was lower at the last glacial maximum (LGM, 21,000 years ago). The oceans were much colder, sea ice was more extensive and the atmospheric concentration of greenhouse gases was lower than present. These ice sheets had virtually disappeared by 6000 years ago (mid-Holocene, MH), but the distribution of incoming solar energy across the northern hemisphere was increased in summer and decreased in winter compared to today, resulting in increases in seasonal temperature contrast and the strength of the monsoons. Testing climate models under these very different climates should put stronger limits on climate sensitivity. We will run a well-tested climate model, using known changes in solar radiation, ice-sheet distribution, and greenhouse gas concentrations for the LGM and MH, and run the same series of simulations (with different values for key processes) as QUMP has done for modern climate. We will evaluate these simulations using reconstructions of LGM and MH climate. There is abundant evidence from pollen in lake and peat-bog sediments for changes in vegetation patterns caused by changes in winter temperature, the length of the growing season and water availability at the LGM and MH. By forcing a vegetation model to reproduce these vegetation patterns we can derive quantitative estimates of these three aspects of climate. There is abundant evidence in the form of old shorelines and lake sediments for changes in lake area caused by changes in precipitation. Again, by forcing a lake model to reproduce these changes in area we can derive quantitative precipitation estimates. These reconstructions, combined with climate reconstructions based on isotopic or geochemical data, will form targets for our simulations. Our project is intended to provide a better estimate of the climate sensitivity to doubling CO2 in time for the IPCC Fifth Assessment Report.