A Multi-Stage Approach Combining Very High-Resolution Satellite Image, GIS Database and Post-Classification Modification Rules for Habitat Mapping in Hong Kong

Ivan H Y Kwong,Eric K Y Liu,Tung Fung,Roger H Lee,Terence P T Ng,Frankie K K Wong

doi:10.3390/rs14010067

Ivan H Y Kwong, Eric K Y Liu + Show 4 more

Open Access

https://doi.org/10.3390/rs14010067

Copy DOI

Abstract

Identification and mapping of various habitats with sufficient spatial details are essential to support environmental planning and management. Considering the complexity of diverse habitat types in a heterogeneous landscape, a context-dependent mapping framework is expected to be superior to traditional classification techniques. With the aim to produce a territory-wide habitat map in Hong Kong, a three-stage mapping procedure was developed to identify 21 habitats by combining very-high-resolution satellite images, geographic information system (GIS) layers and knowledge-based modification rules. In stage 1, several classification methods were tested to produce initial results with 11 classes from a WorldView-2/3 image mosaic using a combination of spectral, textural, topographic and geometric variables. In stage 2, modification rules were applied to refine the classification results based on contextual properties and ancillary data layers. Evaluation of the classified maps showed that the highest overall accuracy was obtained from pixel-based random forest classification (84.0%) and the implementation of modification rules led to an average 8.8% increase in the accuracy. In stage 3, the classification scheme was expanded to all 21 habitats through the adoption of additional rules. The resulting habitat map achieved >80% accuracy for most of the evaluated classes and >70% accuracy for the mixed habitats when validated using field-collected points. The proposed mapping framework was able to utilize different information sources in a systematic and controllable workflow. While transitional mixed habitats were mapped using class membership probabilities and a soft classification method, the identification of other habitats benefited from the hybrid use of remote-sensing classification and ancillary data. Adaptive implementation of classification procedures, development of appropriate rules and combination with spatial data are recommended when producing an integrated and accurate map.

Highlights

The highest overall accuracies (OA) was obtained from pixel-based random forest (RF) classification (84.0% and 0.82 Kappa)
Before the application of modification rules, the accuracies were between 67.7% and 73.1%, which suggested that this set of modification rules were able to rectify the misclassified pixels and resulted in an average increase of 8.8% in OA
This study has achieved the following: (i) it translated the habitat mapping process into a straightforward and controllable workflow with enhanced accuracies; (ii) it revealed varying performances on specific habitats using different classification methods which could be context-dependent; (iii) it exploited manyto-many relationships between habitat categories through geographical data and contextual knowledge in the post-classification phase; (iv) it presented a method to identify mixed habitats by combining the soft probability outputs from the RF classification model; and (v) it demonstrated the benefits of integrating remote-sensing classification and geographic information system (GIS) data to extract particular habitats

Summary

Introduction

Identification and mapping of natural and artificial habitats can serve as the basis for assessments of biodiversity and ecosystem services, supporting environmental planning and management [1,2]. By reflecting changing ecological patterns at different spatial and temporal scales, habitat mapping provides baseline data to understand potential anthropogenic pressures and establish conservation policies [3,4]. Compared to traditional field surveys, remote sensing offers a cost-effective, rapid and repeatable option for habitat mapping, as it provides a synoptic view of phenomena on the ground continuously and consistently from a wide range of sensors with various spatial and spectral resolutions [5,6]. Medium-resolution imageries, such as those from Landsat satellites at a spatial resolution of Remote Sens.

Methods

Results

Discussion

Conclusion