A method for climate and vegetation reconstruction through the inversion of a dynamic vegetation model
A method for climate and vegetation reconstruction through the inversion of a dynamic vegetation model
复制标题
DOI：
10.1007/s00382-009-0629-1
复制
发表时间：
2010-08-01
期刊：
影响因子：
4.6
通讯作者：
Litt, Thomas
中科院分区：
文献类型：
--
作者：
Garreta, Vincent;Miller, Paul A.;Litt, Thomas
根据对过去气候敏感的数据进行的气候重建提供了对这些气候的估计。将这些重建与气候模型的模拟进行比较，可以验证用于未来气候预测的模型。已经表明，对于化石花粉数据，通过反演植被模型获得估计值可以将过去的二氧化碳值变化包括在内。随着新一代动态植被模型的出现，我们提出了一种新的反演方法LPJ-GUESS。当这种新方法用于高分辨率沉积物时，它使我们能够绕过（1）样本之间的气候和花粉独立性以及（2）植被（以花粉表示）和气候之间的平衡的经典假设。我们的动态反演方法是基于一个统计模型来描述气候，模拟植被和花粉样本之间的联系。由于粒子滤波算法的反演实现。我们进行了30个现代欧洲网站的验证，然后将该方法应用于沉积物芯Meerfelder马尔（德国），其中包括全新世的时间分辨率约为每30年一个样本。我们证明，重建的温度受到约束。重建的降水是不太好的约束，由于考虑的维度（一个降水季节），和低灵敏度的LPJ-GUESS降水变化。
Climate reconstructions from data sensitive to past climates provide estimates of what these climates were like. Comparing these reconstructions with simulations from climate models allows to validate the models used for future climate prediction. It has been shown that for fossil pollen data, gaining estimates by inverting a vegetation model allows inclusion of past changes in carbon dioxide values. As a new generation of dynamic vegetation model is available we have developed an inversion method for one model, LPJ-GUESS. When this novel method is used with high-resolution sediment it allows us to bypass the classic assumptions of (1) climate and pollen independence between samples and (2) equilibrium between the vegetation, represented as pollen, and climate. Our dynamic inversion method is based on a statistical model to describe the links among climate, simulated vegetation and pollen samples. The inversion is realised thanks to a particle filter algorithm. We perform a validation on 30 modern European sites and then apply the method to the sediment core of Meerfelder Maar (Germany), which covers the Holocene at a temporal resolution of approximately one sample per 30 years. We demonstrate that reconstructed temperatures are constrained. The reconstructed precipitation is less well constrained, due to the dimension considered (one precipitation by season), and the low sensitivity of LPJ-GUESS to precipitation changes.