Search

Cloud cover and snow misclassifications are the two major limitations for the hydrological application of MODIS snow data. Ground-based meteorological data have the inherent potential to provide the means to reconstruct snow cover for regions in which MODIS snow maps are obstructed by clouds, and to reduce misclassified snow observations. In this study, a multistep method is developed to generate cloud-free MODIS daily snow cover products. The accuracy of the updated MODIS snow products is evaluated for a region in southwestern Germany in which winter cloud coverage and snow variability are typically high. First, we applied Aqua/Terra combination, temporal combination and spatial combination to reduce the cloud coverage and to retrieve the omitted snow. This procedure was not effective since cloud blockage occurs frequently during the snow season. A conditional probability interpolation was then employed to reclassify the remaining cloud cover on MODIS snow maps based on in situ snow depth observations. Finally, we implemented a set of meteorological filters to minimize the misclassified snow in MODIS snow products. The improved cloud-free MODIS daily snow maps showed an overall accuracy of about 92% during the snow season with significantly reduced snow overestimation errors and a slight increase in snow omission errors, compared to the overall accuracy of 87% and 94% for original MODIS Aqua and Terra data, respectively. This study suggests that the fusion of ground-based and satellite-based snow observations is an effective approach for generating cloud-free remote sensing snow data. [ABSTRACT FROM AUTHOR]

Summary MODIS snow data are appropriate for a wide range of eco-hydrological studies and applications in the fields of snow-related hazards, early warning systems and water resources management. However, the high spatio-temporal resolution of the remotely sensed data is often biased by snow misclassifications, and cloud cover frequently limits the availability of the MODIS-based snow cover information. In this study, we applied a four-step methodology that aims to optimize the accuracy of MODIS snow data. To reduce the cloud fraction, 8-day MODIS data from both the Aqua and Terra satellites were combined. Neighborhood analysis was applied as well for this purpose, and it also contributed to the retrieval of some omitted snow. Two meteorological filters were then applied to combine information from station-based measurements of minimum ground temperature, precipitation and air temperature. This procedure helped to reduce the overestimation of snow cover. To test this technique, the methodology was applied to the Rhineland-Palatinate region in southwestern Germany (approximately 20,000 km 2 ), where cloud cover is especially high during winter and surface heterogeneity is complex. The results show that mean annual cloud coverage (reference period 2002–2013) of the 8-day MODIS snow maps could be reduced using this methodology from approximately 14% to 4.5%. During the snow season, obstruction by clouds could be reduced by even a higher degree, but still remains at about 11%. Further, the overall snow overestimation declined from 11.0–11.9% (using the original Aqua–Terra data) to 1.0–1.5%. The method is able to improve the overall accuracy of the 8-day MODIS snow product from originally 78% to 89% and even to 93% during cloud free periods. [ABSTRACT FROM AUTHOR]

The accurate spatial information of snow cover is useful for understanding the impact of global warming, and it is of high significance for hydrological disaster prediction, water resources management, and climate change research. The Normalized Difference Snow Index (NDSI) based approach has been used extensively around the world for mapping snow, and they displayed high accuracy in open areas. However, capturing snow cover in forests remains problematic due to the obstruction effects of the forest canopy, which causes the snow cover area to be seriously underestimated. In this paper, we present a new algorithm based on machine learning (ML) technology to improve the accuracy of binary snow cover (BSC) mapping in forests, using the remotely sensed surface reflectance and ground truth data. A time-lapse photography network with a two-hour resolution was established in the eastern Qilian Mountains in northwestern China to obtain the ground truth data both in forests and open areas. We trained Random Forests (RF) with the 500-m Moderate Resolution Imaging Spectroradiometer (MODIS) reflectance data from bands 1–7 to generate BSC results (RF-BSC). Then we evaluated RF-BSC and the NDSI-derived BSC maps with three different NDSI thresholds (i.e., 0.10, 0.29, and 0.40) against ground-truth data. The results indicate that the proposed algorithm has a high performance in forest BSC mapping in this area, compared to the NDSI-threshold approach. The RF-BSC can retrieve 67% of all real forest snow pixels, while the NDSI-based BSC can only detect 8–14%. We also find that the performance of the algorithm seems to be sensitive to changes in solar illumination conditions and forest coverage. This study suggests that machine learning with the fusion of optical remote sensing and ground-based observations is an effective approach for improving the accuracy of forest snow cover mapping at regional scales. • Forest snow dynamics were obtained by time-lapse photography. • On this basis, high-accuracy forest snow mapping algorithms were developed. • ML with satellite and ground data outperforms NDSI approach in forest snow mapping. • MODIS NDSI using band7/band4 improves forest snow detection than band6/band4. • Influences of cloud, topography and forest coverage on algorithms were analyzed. [ABSTRACT FROM AUTHOR]