Multitemporal Land Use and Land Cover Classification from Time-Series Landsat Datasets Using Harmonic Analysis with a Minimum Spectral Distance Algorithm

Abstract

An understanding of historical and present land use and land cover (LULC) information and its changes, such as urbanization and urban growth, is critical for city planners, land managers and resource managers in any rapidly changing landscape. To deal with this situation, the development of a new supervised classification method for multitemporal LULC mapping with long-term reliable information is necessary. The ultimate goal of this study was to develop a new classification method using harmonic analysis with a minimum spectral distance algorithm for multitemporal LULC mapping. Here, the Jiangning District of Nanjing City, Jiangsu Province, China was chosen as the study area. The research methodology consisted of two main components: (1) Landsat data selection and time-series spectral reflectance reconstruction and (2) multitemporal LULC classification using HA with a minimum spectral distance algorithm. The results revealed that the overall accuracy and Kappa hat coefficients of the four LULC maps in 2000, 2006, 2011, and 2017 were 97.03%, 90.25%, 91.19%, 86.32% and 95.35%, 84.48%, 86.74%, 80.24%, respectively. Further, the average producer accuracy and user accuracy of the urban and built-up land, agricultural land, forest land, and water bodies from the four LULC maps were 92.30%, 90.98%, 94.80%, 85.65% and 90.28%, 93.17%, 84.40%, 99.50%, respectively. Consequently, it can be concluded that the newly developed supervised classification method using harmonic analysis with a minimum spectral distance algorithm can efficiently classify multitemporal LULC maps.

Impact of Multitemporal Land Use and Land Cover Change on Land Surface Temperature Due to Urbanization in Hefei City, China

“…Multitemporal LULC datasets, which included urban and built-up land (U), agricultural land (A), forest land (F), and water bodies (W), were classified using the method developed by Sun and Ongsomwang [42]. In practice, the clearly observed and identified contaminated pixels (with value 0 or 1) from all selected scenes were firstly applied to construct a spatiotemporal cube.…”

“…The average probabilities of an unclassified pixel being a specific LULC type (U, A, F, and W) from among the six spectral bands (blue, green, red, NIR, SWIR1, and SWIR2) were calculated and compared to identify the highest value; the corresponding LULC type that provided the highest probability was then assigned to an unclassified pixel at a specific time point. For further details, see Sun and Ongsomwang [42]. In addition, producer's accuracy (PA), users' accuracy (UA), overall accuracy (OA), and Kappa hat coefficient were assessed based on the error matrix between classified multitemporal LULC data and ground reference information from Landsat data themselves [57].…”

“…Multitemporal LULC classification and change detection were successfully conducted using the algorithm proposed by Sun and Ongsomwang [42] in this research. This approach can be used to quickly classify and map LULC data and their changes at any time and period.…”

“…Using spatial and decomposition analyses, the derived time-series LST data will be further applied to examine the impact of LULC change on LST in Hefei City due to rapid urbanization between 2001 and 2020. In the study, all available Landsat 5/7/8 scenes over the city in this period were downloaded to classify LULC dataset using harmonic analysis with a minimum spectral distance algorithm [42], estimate time-series LST using the single-channel method, reconstruct time-series LST using a harmonic analysis model, and examine the impact of LULC change on LST.…”

An evaluation of urbanisation processes in suburban zones using land-cover data and fuzzy set theory

“…The steady rise in the global population is contributing to the development of various types of human settlements (Albino et al, 2015). In consequence, spatial processes such as migration and spatial diffusion are increasingly leading to important qualitative changes (Sun & Ongsomwang, 2020). The agrarian structure of settlements is being gradually replaced by industrial structures (Bruegmann, 2015).…”