Multivariate methods for space-time environmental processes

总的目标是制定多变量方法，在现有数据包括监测测量和确定性（物理）模型输出的情况下模拟环境过程。虽然这两个来源的数据都试图复制现实，但它们具有非常不同的特征。监测数据试图复制微观环境，而物理模型代表宏观环境（例如，MAQSIP：6 km x 6 km网格单元）。所考虑的环境领域可能包括多种污染物（即。多变量响应），并且通常具有空间和时间相关结构以及非平稳和非高斯。 将研究各种各样的研究问题，以便将这些特征纳入环境时空过程的建模中，研究的意义将是新的环境过程一般统计理论，这些理论可以有各种各样的应用。这些措施包括：为健康影响研究改进暴露评估;改进网络监测，以最好地保护人口;通过例如预测干旱和洪水以及飓风等极端气候条件，加强环境风险管理。还可以为确定性模型输出提供不同尺度的分辨率，从而避免计算负担。

The overall objective is to develop multivariate methods for modelling environmental processes where available data include both monitoring measurements and output from deterministic (physical) models. Although data from both sources attempt to replicate reality, they have very different characteristics. The monitoring data attempt to replicate the micro environment whereas the physical model represents the macro environment (eg. MAQSIP: 6km x 6km grid cells). The environmental fields under consideration may include multiple pollutants (ie. Multivariate responses), and generally have both spatial and temporal correlation structures as well as being non-stationary and non-Gaussian.  A variety of research problems will be studied to incorporate these features in the modelling of environmental space-time processes.The significance of the research would be new general statistical theories for environmental processes which can have a wide variety of applications. These include improved exposure assessment for health impact studies; improved network monitoring to best protect human population; and enhanced environmental risk management through, for examples, forecasting drought and flood, as well as extreme climate conditions such as hurricanes. It is also possible to provide different scales of resolution for the deterministic model output, thus avoiding the computational burden.