地理学视角下土地利用/覆被分类发展探讨

随着对地观测和互联网技术的发展，地理大数据时代正在到来，其多尺度、长时序、多模态等海量“超”覆盖数据为土地利用/覆被（Land Use/Land Cover, 简称LULC）分类及变化检测带来巨大的机遇，支撑着新时代人、地两大系统相互作用关系的认知和实践。然而，多数地理学者认为地理学基本原理与核心思想并未因为大数据的到来而发生本质性变化。所以，从地理学基本原理角度理解LULC分类的发展，尤其在地理大数据时代的发展方向，不失为一条可行的途径。为此，本文从区域、尺度、综合三方面的地理学基本原理视角将LULC分类技术的发展划分为地球观测数据匮乏阶段、人类行为数据融合阶段以及地理大数据“超”覆盖阶段分别探讨分析，以期主动把握LULC分类技术及应用的未来发展趋势。研究结果显示：在地球观测数据匮乏阶段，LULC分类多以类型还不丰富的遥感数据源，在空间分辨率较低的像元尺度上，进行以地表覆被状态为主的分类；发展到人类行为数据融合阶段，LULC分类在城市区域率先出现了对地观测数据和人类行为数据相融合，在街区尺度上进行以空间功能异质性划分、识别为主导的城市功能区分类；在地理大数据“超”覆盖阶段，LULC分类将实现多尺度协同、面向全空间的功能异质性划分，并在主体功能的基础上融合“社会-经济-自然”多维定量属性，本文称之为“空间场景”。希望本文的探讨能够为地理大数据时代LULC分类的新技术发展和新产品应用提供有益启示。

Land use/land cover classification development from a geographical perspective

Land Use/Cover Change (LUCC) is a classical geographical method to study the interaction of human system and the earth system. An important basis of LUCC is the Land Use/Land Cover (LULC) classification of the earth surface in spatio-temporal dimension which could recover the interaction patterns, intensity, drivers, tendency, impacts, etc. With the technology development of earth observation and Internet connection, we are entering the era of geographic big data. The multi-scale, long time series, multi-modal and other massive "extra" coverage data greatly enrich the LULC classification and its change detection, and support the cognition and application of the interaction between the two systems of human and earth in the new era. However, most geographers believe that the basic principles and core ideas of geography have not undergone essential changes when entering the big data era. Therefore, it has become a feasible way to understand the development of LULC classification techniques from the perspective of the basic principles of geography, especially the development direction in the big data era. To this end, we describe the technology and application of LULC classification from the three geographic thoughts of regions, scaling and integration by dividing the development stage of LUCC detection into the earth observation data scarce stage, the human behavior data fusing stage, and the geographic big data "extra" coverage stage. In the earth observation data scarce stage, LULC classification uses limited remote sensing data sources to perform land cover categories at the pixel scale with low spatial resolutions; in the human behavior data fusing stage, LULC classification can realize the urban function zoning by the fusion of earth observation data and human behavior data at the block scale; and in the geographic big data "extra" coverage stage, LULC classification will realize multi-scale collaboration and functional heterogeneity division for the whole space, and integrate multi-source quantitative social-economic-natural attributes on the basis of the main spatial function, which we called "spatial scene". Thus, the spatial scene can reveal the spatio-temporal structure of the natural system and also of the human system. We hope the discussion can provide useful enlightenment for the development of new technologies and new applications for LULC classification in the era of geographic big data.