National Scale Land Cover Classification Using the Semiautomatic High-Quality Reference Sample Generation (HRSG) Method and an Adaptive Supervised Classification Scheme

Abstract

The advent of new high-performance cloud computing platforms (e.g., Google Earth Engine (GEE)) and freely available satellite data provides a great opportunity for land cover (LC) mapping over large-scale areas. However, the shortage of reliable and sufficient reference samples still hinders large-scale LC classification. Here, selecting Turkey as the case study, we presented a Semi-automatic High-quality Reference Sample Generation (HRSG) method using the publicly available scientific LC products and the linear spectral unmixing analysis to generate high-quality ground samples for the years 1995 and 2020 within the GEE platform. Furthermore, we developed an adaptive random forest classification scheme based on Köppen-Geiger climate zone classification system. Our rationale was related to the fact that large-scale study areas often contain multiple climate zones where the spectral signature of the same LC class may vary within different climate zones that can lead to a poor LC classification accuracy. To have a robust assessment, the generated LC maps were evaluated against independent test datasets. In regard to the proposed sample generation method, it was observed that HRSG can generate high-quality samples independent of the characteristics of scientific LC products. The high overall accuracy of 92% for 2020 and 90% for 1995 and satisfactory results for producer's accuracy (ranging between 83.4% and 99.3%) and user's accuracy (ranging between 86.1% and 99.7%) of nine LC classes demonstrated the effectiveness of the proposed framework. The presented methodologies can be incorporated into future studies related to large scale LC mapping and LC change monitoring studies.

Land cover dataset of the China Central-Asia West-Asia Economic Corridor from 1993 to 2018

“…On the other hand, large-scale studies require an immense size of input features, making processing of the entire region often unfeasible 22 . Our recently published research 19 has shown that adopting an adaptive classification approach that divides the large-scale region into several sub-areas can effectively address these issues. As such, we divided the Corridor region into 25 sub-areas identified by Köppen-Geiger climate classification system 23 (Fig.…”

“…We used three different techniques to collect high-quality reference samples for calibration and testing. The first method used was the Semi-automatic High-quality Reference Sample Generation method (HRSG) developed by our team 19 . This method acknowledges that no scientific LC product/map can be entirely free of errors, and a limited number of training samples are capable of representing intra-class diversity.…”

“…Similarly, Naboureh et al . 19 , claimed that Globland30 had a relatively low OA of 62.3 percent for Turkey. Thus, it is impractical to study historical LC changes using the available LC products.…”

“…Addressing this deficiency, the Landsat archive’s free sharing policy in 2008 2 , 20 and the advent of Earth science data cloud computing and analysis tools, such as Google Earth Engine (GEE), have facilitated the production of various types of LC products 19 . As a result, several valuable LC datasets for various regions of the world have been generated.…”

MSDMCCG: Design of an efficient Multimodal Satellite Data Processing Model for Component-level analysis of Contextual Geographic entities

“…Every model has advantages and disadvantages, and the selection of a model depends on the needs of the specific application [30,31,32]. Recent research has demonstrated that deep learning models such as CNNs and DBNs outperform more conventional machine learning models such as SVMs and DTs in satellite image classification tasks [33,34,35]. To examine the full potential of ensemble models for the classification of satellite images, additional research is necessary for a variety of scenarios.…”