Fanning the flames: past and future links between climate change and fire activity across Siberia.

西伯利亚横跨欧亚大陆北部，面积超过1300万平方公里。它的大部分被森林覆盖，包括少数树种，其分布是不相交的，例如，落叶松（落叶针叶）在东部，云杉和松树（常绿针叶）在西部。这片北方森林是地球上最大的森林，形成了全球重要的碳库，被称为“北半球之肺”：因此，其碳汇能力的任何变化都有可能对未来“全球变暖”的程度产生重大影响。西伯利亚森林是由气候、森林火灾、昆虫爆发和人类活动之间复杂的相互作用形成的。在过去20年里，区域性森林火灾的频率和范围增加了10倍。到2020年8月，约有1000万公顷的森林被毁，今年上半年出现了创纪录的高温，这一现象直接归因于人为造成的“全球变暖”。虽然似乎普遍的“北极放大”（即最近近地表区域温度的变暖是低纬度地区的两倍多）导致了西伯利亚火灾干扰的增加，但显然仍需要更好地了解火-气候相互作用中涉及的过程和驱动因素的时空尺度范围。此外，物种对火灾的特定反应——与火灾生态学和燃料负荷的确定有关——对于确定火灾影响的空间范围和时间频率以及长期生态系统的净碳损失非常重要。该项目旨在阐明这些过程/驱动因素，并使用最先进的机器学习技术，将它们与更广泛的大气变化联系起来。该项目旨在回答以下问题：影响西伯利亚北方针叶林的主要火-气候过程和气候驱动因素是什么？Q2。最近的气候变化在多大程度上导致了西伯利亚火灾干扰的增加？第三季。预计气候变化对西伯利亚地区未来火灾干扰频率的可能影响是什么？这将通过承担以下任务来实现：根据最近的火灾活动水平和树种，从西伯利亚选定地区的观测和遥感数据中获得火灾活动/危险措施。T2。从现有的北极CORDEX区域气候模式运行中获得西伯利亚这些地区的高分辨率（~10公里）气象参数输出。T3。采用最先进的机器学习技术，开发气象变量与火灾活动/危险观测之间的非线性多变量多时相关系。答案是Q1。T4。采用最先进的机器学习技术，在T3获得的影响火灾活动的关键气象变量与相对粗糙（~100公里）gcm集合获得的更大尺度大气环流之间建立区域“指纹”。基于这些指纹，学生将能够有效地将GCM输出缩小到火灾活动的规模。答案是Q1。T5。利用经过再分析数据验证的历史GCM模型运行，以及T4中导出的指纹，估计关键气象变量变化对观测到的火灾活动的可能贡献。答案是Q2。T6。利用一系列选定的gcm和未来气候情景的输出估计21世纪西伯利亚火灾活动的变化。这将包括对基于自然变率、模式不确定性和情景不确定性的预估中的总体不确定性的估计。这将回答问题3。

Siberia extends across northern Eurasia and encompasses more than 13 million square kilometres. Much of it is covered in forest, comprising a small number of tree species, whose distributions are disjunct e.g., larch (deciduous needles) to the east, spruce and pine (evergreen needles) to the west. This boreal forest is the largest on Earth, forming a globally important carbon reservoir and described as the 'lungs of the Northern Hemisphere': therefore, any change in its capacity as a carbon sink has the potential to significantly impact the magnitude of future 'global warming'. Siberian forest is shaped by complex interactions between climate, forest fires, insect outbreaks and human activities. The frequency and extent of regional forest fires has increased tenfold in the last 20 years. In 2020 about 10 million hectares of forest had been destroyed by August and followed record-breaking high temperatures during the first half of the year, a phenomenon that has been directly attributed to anthropogenically-forced 'global warming'.While it appears that general 'Arctic amplification', whereby recent warming in near-surface regional temperatures has been more than twice that at lower latitudes, has contributed to the increase in Siberian fire disturbance, there remains a clear need to better understand the range of spatial and temporal scales of the processes and drivers involved in fire-climate interactions. Furthermore, species-specific responses to fire - in relation to fire ecology and determination of fuel load - are important in determining the spatial extent and temporal frequency of fire impacts and the net carbon loss from the ecosystem over the longer-term.This project aims to elucidate these processes/drivers and, using state-of-the-art machine learning techniques, link them to broader-scale atmospheric variability. The project aims to answer the following questions: Q1.What are the key fire-climate processes and climatological drivers affecting the Siberian boreal forest? Q2.To what extent is recent climate change responsible for observed increases in fire disturbance in Siberia? Q3. What will be the likely impact of projected climate change on the frequency of future fire disturbance in regions of Siberia?This will be achieved by undertaking the following tasks:T1. Obtain measures of fire activity/danger from observations and remotely-sensed data for selected regions of Siberia, based on the level of recent fire activity and tree species. T2. Obtain high-resolution (~10 km) output of meteorological parameters from existing Arctic CORDEX regional climate model runs for these regions of Siberia. T3. Employ state-of-the-art machine learning techniques to develop non-linear multivariate multitemporal relationships between meteorological variables and fire activity/danger observations. This will answer Q1. T4. Employ state-of-the-art machine learning techniques to develop regional 'fingerprints' between the key meteorological variables affecting fire activity, obtained from T3, and the broader-scale atmospheric circulation as derived from an ensemble of relatively coarse (~100 km) GCMs. Based on these fingerprints, the student will be able to efficiently downscale the GCM output to the scale of the fire activity. This will answer Q1. T5. Using the historical GCM model runs, validated against reanalysis data, and the fingerprints derived in T4, estimate the likely contribution that changes in the key meteorological variables have made to observed fire activity. This will answer Q2. T6. Using the output from a range of selected GCMs and future climate scenarios estimate the change in Siberian fire activity during the 21st Century. This will include estimates of the overall uncertainty in the projections based on natural variability, model uncertainty and scenario uncertainty. This will answer Q3.