華東師范大學(xué)地理科學(xué)學(xué)院邀請(qǐng)康蕾(Emily Lei Kang)教授作了一場(chǎng)題為“Statistical Models for Large Spatial and Spatio-Temporal Datasets(大空間和時(shí)空數(shù)據(jù)集的統(tǒng)計(jì)模型)”的講座。地理科學(xué)學(xué)院是我國(guó)最早具有地理學(xué)一級(jí)學(xué)科博士點(diǎn)授予權(quán)的單位之一，是我國(guó)首批博士后流動(dòng)站建站單位之一，也是我國(guó)最早2個(gè)具有自然地理學(xué)重點(diǎn)學(xué)科的單位之一。講座的主要內(nèi)容是：

　　隨著現(xiàn)代技術(shù)，如地理信息系統(tǒng)(GIS)和全球定位系統(tǒng)(GPS)常規(guī)識(shí)別在當(dāng)今各種學(xué)科的地理坐標(biāo)，科學(xué)家和研究人員的發(fā)展能夠獲得地理編碼數(shù)據(jù)以前所未有的，而這樣的數(shù)據(jù)越來(lái)越高維在觀察位置的數(shù)量方面(以及隨著時(shí)間的推移)。對(duì)于非常大的和大規(guī)模數(shù)據(jù)集的空間數(shù)據(jù)是具有挑戰(zhàn)性的，因?yàn)閿?shù)據(jù)集的大小導(dǎo)致計(jì)算最佳空間預(yù)測(cè)，如克里格問(wèn)題。此外，當(dāng)將數(shù)據(jù)集收集在大的空間域，感興趣的關(guān)聯(lián)的空間過(guò)程通常表現(xiàn)非平穩(wěn)行為超過(guò)該域，和非平穩(wěn)空間相關(guān)結(jié)構(gòu)的柔性家族優(yōu)選在統(tǒng)計(jì)模型。我先介紹一下統(tǒng)計(jì)挑戰(zhàn)及其在分析大型或巨型空間和時(shí)空數(shù)據(jù)的發(fā)展，然后談?wù)勔恍┪乙呀?jīng)在這個(gè)領(lǐng)域做了近期工作。具體來(lái)說(shuō)，我將討論(1)預(yù)測(cè)和降尺度統(tǒng)計(jì)方法; (2)進(jìn)行數(shù)據(jù)融合的統(tǒng)計(jì)方法。這些方法的應(yīng)用也將被討論。

　　原文：With the development of modern technologies such as Geographical Information Systems (GIS) and Global Positioning Systems (GPS) routinely identifying geographical coordinates, scientists and researchers in a variety of disciplines today have access to geocoded data as never before, and such data become increasingly high-dimensional in terms of the number of observed locations (and over time). Spatial statistics for very large and massive datasets is challenging, since the size of the dataset causes problems in computing optimal spatial predictors, such as kriging. In addition, when a dataset is collected on a large spatial domain, the associated spatial process of interest typically exhibits nonstationary behavior over that domain, and a flexible family of nonstationary spatial dependence structure is preferred in statistical models. I will first introduce the statistical challenges and their developments in analyzing large or massive spatial and spatio-temporal data, then talk about some recent work I have done in this field. Specifically, I will discuss (1) statistical methods for prediction and downscaling; (2) statistical methods for data fusion. Applications of these methods will also be discussed.