In the past years

the land cover community has strived to develop and supply more datasets at different spatial scales（e.g.

regional

national

or global） with increasing spatial-temporal resolutions and improved classification accuracy. Although these data products have been widely applied in climate change studies

environmental monitoring

sustainable development

and many other societal benefit areas

the user communities constantly propose new demands

such as additional land cover classes

up-to-date time series

and consistency among different datasets. Therefore

the continuous updating and content refinement of land cover data products have become key objectives of the land cover community. The updating and refinement of land cover data products differ from their original creation. Change detection with remotely sensed imagery is a major approach for updating a large area land cover

and the rapidly increasing crowdsourcing information provides another valuable resource. However

as a technical challenge is that no existing change detection algorithm can be applied to all kinds of imageries and geographic regions because of the extremely complex spectral heterogeneity of land cover classes. Therefore

an efficient change detection approach with the consideration of the existing land cover datasets needs to be developed. One cuttingedge issue is to integrate the imagery-based change intensity measurement with prior knowledge represented by existing land cover datasets.Change detection for time series imagery is moving from the comparison of two neighboring points to global trend analysis. The coupling of SAR and infrared images with multispectral images needs to be explored from several aspects

such as relative radiometric correction

spectral matching

and temporal-spatial data fusion. Another key challenge comes from the rational utilization of crowdsourcing information for updating and refining land cover. Crowdsourcing information may differ in terms of data contents

position accuracy

spatial-temporal resolution

and uncertainty

and hence

previous studies have aimed to develop appropriate methods and techniques for evaluating volunteered data quality

discovering useful information from deep web

extracting incremental changes

and integrating multi-source datasets. The increasing amount of freely accessible remote sensing data has increased the data intensiveness of the generating future land cover data products. Specific tools and systems must be designed and developed to support the updating and refining of large area land cover. One of these tools is an online land cover updating system that allows users to execute web-based land cover change detection and processing in an open web environment. The key issues in using this tool include domain-knowledge-based change detection service modeling and dynamic service composition. Data Cube is another system that has a flexible classification concept

but this tool is still under investigation. Nevertheless

this tool is expected to facilitate the on-demand extraction of land cover classes with deep learning and other data mining algorithms.