Phenology-based temporal mixture analysis for estimating large-scale impervious surface distributions

Abstract

Along with rapid urbanization, the prevalence of urban impervious surfaces, a major biophysical component of urbanized areas, has increased concurrently. As a key indicator of environmental quality and urbanization intensity, the accurate estimation of impervious surfaces is essential. To address this problem, numerous automated estimation approaches have been developed in the past several decades. Among these approaches, spectral mixture analysis (SMA) is an especially powerful and widely used technique. Although SMA has proved valuable in impervious surface estimation, the issues of seasonal sensitivity and spectral confusion have not been successfully addressed. In particular, impervious surface estimation is likely to be sensitive to seasonal variations, largely due to the shadowing effects of vegetation canopy during summer and confusion between impervious surfaces and soil during winter. In this study, we developed two temporal mixture analysis methods: phenology-based temporal mixture analysis (PTMA) and phenology-based multi-endmember temporal mixture analysis (PMETMA), to quantify impervious surface areal fractions using multi-temporal MODIS NDVI data. Specifically, 1 year-continuous MODIS NDVI series were employed to address seasonal sensitivity and spectral confusion issues. Furthermore, the estimated results were compared to TMAs that applied only to summer and winter data. The results indicate that both PTMA and PMETMA perform well for estimating the percentage of impervious surface areas. Moreover, a comparative analysis indicates that PMETMA performs slightly better than PTMA root mean square error (RMSE) of 7.27%, SE of 3.25%, and MAE of 4.03%) and much better than summer TMA and winter TMA, with a RMSE of 7.54%, an SE of 2.13%, an MAE of 3.36%, and an R² of 0.7623.

Acknowledgements

We would like to acknowledge the anonymous reviewers for their constructive and valuable suggestions on the earlier drafts of this manuscript.

Funding

This research was partially supported by UWM Research Growth Initiative (RGI) grant and the National Natural Science Foundation of China [nos. 41030743 and 41171322].