Evaluating past climate variability and modelling its impact on tree growth: by Guangqi Li.

Past climates provide an opportunity to examine the response of the Earth System to large changes in external forcing. The changes in forcing over the last 21,000 years since the Last Glacial Maximum have been as large as those projected to occur over the 21st century as a result of anthropogenic changes in greenhouse gas concentrations and land-use changes. The fact that there have been equally large changes in forcing in the past as expected in the future, coupled with the availability of climat reconstructions of past climates, provides the motivation for using evaluations of simulations of past climates to evaluate how well the models that are used to project future climate changes perform. Terrestrial vegetation is highly sensitive to changes in climate, and records of past vegetation changes are widely used to reconstruct past climate states. Statistical or model-inversion techniques have been used to reconstruct changes in seasonal temperature and water balance from pollen records from lakes and bogs. Most of these pollen records are at comparatively low resolution, and thus provide reconstructions of the long-term changes in mean climate state. Tree-ring series are the most abundant source of information used to reconstruct changes in short-term (interannual to decadal) climate variability. However, most of the available reconstructions focus on relationships with temperature and these relationships appear to break down in recent decades at many sites. An alternative approach is to use forward modelling to translate simulated climate variability into tree growth, which can then be directly compared to observations of tree-ring series. In the first part of this thesis, I have used a global synthesis of pollen-based palaeoclimate reconstructions to evaluate how well state-of-the-art climate models from the fifth phase of the Coupled Model Intercomparison Project (CMIP5) capture large scale (sub-continental to hemispheric) patterns of climate change. I initially examine the scaling of changes in precipitation with temperature at a hemispheric scale in warm and cold climate states (Chapter 2). I then examine how well models simulate large-scale changes in monsoon precipitation in response to changes in orbital forcing during the mid-Holocene, 6000 years ago, focusing on the northern Africa monsoon (Chapter 3). Both of these analyses, and an evaluation of simulated changes in precipitation and temperature in mid-continental Eurasia in the mid-Holocene, are included in Chapter 4 which provides a summary of all of the evaluations that have been done of the CMIP5 palaeoclimate simulations. All of these papers focus on evaluation of changes in the long-term mean climate, but it is also important to evaluate how well the models simulate short-term (annual to decadal) climate variability.To do this, I developed a model to simulate tree growth driven by climate-driven changes in net primary production (Chapter 5). I tested this model using tree-ring data from the historical period in two contrasting climate settings, specifically in a cool climate in the Changbai Mountains, northeastern China (Paper 4) and in the semiarid environment of the Great Western Woodlands, Western Australia (Chapter 6). The final paper (Chapter 7) uses the model to simulate tree growth during the Last 2Glacial Maximum (ca 21,000 years ago) in California, USA. Changes in the hydrological cycle are expected to scale with temperature changes. However, both the observed changes in precipitation in recent decades and model simulations of precipitation changes during the historic period and the 21st century are smaller than would be predicted from the Clausius-Clapeyron relationship which describes the change in atmospheric water vapour content with temperature (~7%/°C ). It has been argued that this reflects energetic constraints on evaporation.To test this hypothesis, I analyzed the scaling of precipitation with temperature in warm (increased CO₂) and cold (Last Glacial Maximum, LGM) climates using six CMIP5 models that have simulated the response to both. Globally, precipitation increases in warm climates and decreases in cold climates. The estimate of the scaling across all the climate states and all models indicates a 2.06%± 0.09% change per degree temperature change at the global scale. The simulated scaling of precipitation to temperature is controlled by energetic constraints on evaporation rather than the atmospheric water-holding capacity, and is also affected by water availability.These constraints lead to a lower sensitivity of precipitation to temperature change over the land than that over the ocean, and a lower sensitivity over tropical land than over extratropical land. The simulated changes in precipitation per degree temperature change are comparable to the observed changes in both the historical period and the LGM, showing the models correctly predict the constraints on precipitation scaling. Temperature-controlled precipitation change is a global large-scale phenomenon. However, regional precipitation change can also be influenced by changes in the large-scale circulation. Monsoon precipitation is one of the most typical circulation controlled evaluated the spatial expression of seasonal climates of the Mediterranean and northern Africa in pre-industrial (piControl) and mid-Holocene (midHolocene, 6 yr BP) CMIP5 simulations using the observed regional pattern and amount of seasonal precipitation. Most of the piControl simulations reproduce the observed modern precipitation patterns in the Mediterranean and equatorial zone, but they overestimate the area influenced by the monsoon and underestimate the extent of desert. The models also fail to capture the observed amount of precipitation. The models simulated a stronger monsoon in response to orbital changes in seasonal insolation receipts inthe mid-Holocene, including a northward expansion of the monsoon and an increase in summer and autumn rainfall. However, the mid-Holocene simulations underestimate the observed changes in annual precipitation, except in equatorial zone. The underestimation of precipitation in the latitude band from 15–30 °N is at least 50%.The failure to capture the observed monsoon expansion is unrelated to biases in the piControl simulations. The failure to capture the observed changes in rainfall over northern Africa is an example of a long-standing modelling problem: current state-of-the-art models do not produce a better match to observations than previous generations of models. The mid-continent of Eurasia provides another example of a persistent problem in the simulation of regional climates during the mid-Holocene. The CMIP5 mid-Holocene simulations produce conditions drier than today in mid-continental Eurasia, particularly between 45° and 60° N, whereas observations systematically show that this region was wetter than today. In the models, dry conditions reduce evapotranspiration and result in an increase in surface temperature compared to today. However, the observations show that the mid-continent was cooler than today. These three analyses form a major part of the evaluation of the CMIP5 palaeoclimate simulations described in Chapter 4. The main conclusion of this summary paper is that while models are able to reproduce the large-scale features (such as precipitation scaling with temperature) of past climates accurately they are poor at reproducing regional changes such as monsoon expansion or the water-balance of the midcontinental regions. This suggests that while we can have confidence in projections of large-scale features of projected future climates, such as the greater warming at high latitudes than in the tropics or enhanced land-sea contrast, predictions of regional climates are very uncertain. Statistical reconstructions of past climate provide one source of information for model evaluation. An alternative approach is to use climate model outputs to drive simple forward models to predict the actual observations of vegetation changes. I have developed a tree growth model (the T model) that predicts carbon allocation to leaves,stem and roots, and thus can simulate tree-ring series. The tree-growth model is driven by a generic light-use efficiency model (the P model). The P model provides values for gross primary production (GPP) per unit of absorbed photosynthetically active radiation (PAR), which is estimated from leaf area. In the tree-growth model,GPP is allocated to foliage, transport tissue, and fine-root production and respiration in such a way as to satisfy well-understood dimensional and functional relationships.The T model represents both ontogenetic effects (the impact of ageing) and the effects of environmental variations and trends (climate and atmospheric CO₂ concentration [CO₂]) on growth. I have tested the T model under modern climate conditions for three species in three different climate settings, including Pinus koraiensis in the cool and mild climate ofthe Changbai Mountains, northeastern China, Callitris columellaris in the semi-arid climate of the Great Western Woodlands, Western Australia, and Juniperus occidentalis in the montane climate of California, USA. In all three regions, when driven by the local climate and [I have tested the T model under modern climate conditions for three species in threedifferent climate settings, including Pinus koraiensis in the cool and mild climate ofthe Changbai Mountains, northeastern China, Callitris columellaris in the semi-aridclimate of the Great Western Woodlands, Western Australia, and Juniperus occidentalisin the montane climate of California, USA. In all three regions, when driven bythe local climate and [CO₂], the T model produces realistic simulations of th .......[See attached thesis files for full Abstract/Description text].