Assessment of Snow-Cover Mapping Accuracy in a Variety of Vegetation-Cover Densities in Central Alaska

doi:10.1016/S0034-4257(98)00051-0

Remote Sensing of Environment

Volume 66, Issue 2, November 1998, Pages 129-137

https://doi.org/10.1016/S0034-4257(98)00051-0 Get rights and content

Abstract

Field and aircraft measurements were acquired in April 1995 in central Alaska to map snow cover with MODIS Airborne Simulator (MAS) data, acquired from high-altitude aircraft. The Earth Observing System (EOS) Moderate Resolution Imaging Spectroradiometer (MODIS) is a 36-channel system that will be launched on the EOS-AM-1 platform in 1999. A vegetation-density map derived from integrated reflectances (R_i), from MAS data, is compared with an independently-produced vegetation type and density map derived from Thematic Mapper (TM) and ancillary data. The maps agreed to within 13%, thus corroborating the effectiveness of using the reflectance technique for mapping vegetation density. Snow cover was mapped on a 13 April 1995 MAS image, using the original MODIS prototype algorithm and an enhanced MODIS prototype algorithm. Field measurements revealed that the area was completely snow covered. With the original algorithm, snow was mapped in 96% of the pixels having <50% vegetation-cover density according to the R_i map, while in the areas having vegetation-cover densities ⩾50%, snow was mapped in only 71% of the pixels. When the enhanced MODIS snow-mapping algorithm was employed, 99% of the pixels having <50% vegetation-cover density were mapped, and 98% of the pixels with ⩾50% vegetation-cover density were mapped as snow covered. These results demonstrate that the enhanced algorithm represents a significant improvement over the original MODIS prototype algorithm especially in the mapping of snow in dense vegetation. The enhanced algorithm will thus be adopted as the MODIS at-launch snow-cover algorithm. Using this simple method for estimating vegetation density from pixel reflectance, it will be possible to analyze the accuracy of the MODIS snow-cover algorithm in a range of vegetation-cover in places where information on vegetation-cover density is not available from ground measurements.

Section snippets

Introduction and background

During April 1995, a field and aircraft experiment was conducted in the boreal forest of central Alaska in support of the Earth Observing System (EOS) Moderate Resolution Imaging Spectroradiometer (MODIS) snow-cover mapping project. MODIS is scheduled for launch on the EOS AM-1 polar-orbiting platform in 1999. The MODIS Airborne Simulator (MAS), a 50-channel spectroradiometer, was flown onboard a NASA ER-2 aircraft. Field measurements were also acquired. An objective of the mission was to

Results and discussion

Vegetation-cover density map derived from MAS-derived reflectances. Vegetation-cover density was estimated from integrated reflectance (R_i) values in the 13 April 1995 MAS scene acquired over central Alaska (Fig. 3 top) in the following manner. The integrated reflectance is the reflectance integrated over a portion of the electromagnetic spectrum (Hall et al., 1989). The highest R_i values correspond to snow-covered bare areas or areas covered with very low vegetation, snow-covered ice, or

Conclusions

At least in central Alaska, the technique of using the wintertime reflectances as a surrogate measure of vegetation density Robinson and Kukla 1985, Foster et al. 1994 appears to be valid. In this article, we compared a vegetation-density map derived from integrated reflectances (from Channels 1–9 of MAS data), with an independently produced vegetation type and density map (derived from TM and ancillary data). The TM-derived map is assumed to be an accurate depiction of the actual vegetation

Acknowledgements

The authors would like to thank Janet Chien/GSC, Laurel, Maryland, for her expertise in image processing, and Ken Brown of NASA/Goddard Space Flight Center, for discussions about the MODIS Airborne Simulator.

References (24)

J. Dozier
Spectral signature of alpine snow cover from the Landsat thematic mapper
Remote Sens. Environ.
(1989)
D.K. Hall et al.
Comparison of in situ and Landsat derived reflectance of Alaska glaciers
Remote Sens. Environ.
(1989)
D.K. Hall et al.
Development of methods for mapping global snow cover using Moderate Resolution Imaging Spectroradiometer (MODIS) data
Remote Sens. Environ.
(1995)
Bunting, J. T., and d’Entremont, R.P. (1982), Improved cloud detection utilizing defense meteorological satellite...
Chang, A. T. C., Foster, J. L., Hall, D. K., et al. (in press), Snow parameters derived from microwave measurements...
J.L. Foster et al.
Derivation of snow water equivalent in boreal forests using microwave radiometry
Arctic
(1991)
Foster, J. L., Chang, A. T. C., and Hall, D. K. (1994), Snow mass in boreal forests derived from a modified passive...
Goodison, B. E., and Walker, A. E. (1994), Canadian development and use of snow cover information from passive...
Gumley, L., Hubanks, P., and Masuoka, E. (1994), MODIS Airborne Simulator Level 1B Data User’s Guide,...
D.K. Hall et al.
Determination of snow-covered area in different land covers in central Alaska from aircraft data—April 1995
Ann. Glaciol.
(1998)

M.D. King et al.

Airborne scanning spectrometer for remote sensing of cloud, aerosol, water vapor, and surface properties

J. Atmos. Ocean. Technol.

(1996)

Klein, A. G., Hall, D. K., and Riggs, G. A. (1997), Improving the MODIS global snow-mapping algorithm. In International...

Cited by (136)

Mapping snow cover in forests using optical remote sensing, machine learning and time-lapse photography
2022, Remote Sensing of Environment
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.
A machine learning-based accuracy enhancement on EUMETSAT H-SAF H35 effective snow-covered area product
2022, Remote Sensing of Environment
Citation Excerpt :
The most ideal scenario for the detection of snow cover with high accuracy from space via optical sensors should include i) clear sky, ii) good illumination conditions, and iii) snow cover on the landscape with depth of several centimeters; and all of these conditions may exhibit rapid and extreme daily changes across the landscape. During the development and validation of fSCA retrieval algorithms for coarse resolution sensors at optical wavelengths, it is a commonly preferred approach to generate the necessary reference snow cover maps from higher resolution sensors (also known as the cross-sensor method) (Hall et al., 1998; Hall et al., 1995; Klein et al., 1998; Salomonson and Appel, 2004, 2006). In this study, the reference fSCA maps are produced from higher resolution Sentinel 2 binary snow maps by using the SNOMAP algorithm (Hall and Riggs, 2011; Hall et al., 1995).
Snow is a major element of the cryosphere with significant impact on the Earth's water cycle and global energy budget. Acquiring consistent and long time series data on the spatial extent of snow cover doubtlessly plays a key role in our understanding and modeling of the current and future environmental dynamics. Remote sensing offers a powerful tool for continuous retrieval of snow cover information by utilizing snow's contrasting reflectance characteristics at optical wavelengths. The pre-operational H35 covers the Northern Hemisphere, and it is the successor of the operational Pan-European H12 daily fractional snow-covered area (fSCA) product at ~1 km. Both products are developed through the Satellite Application Facility on Support to Operational Hydrology and Water Management (H-SAF) project of EUMETSAT by exploiting AVHRR channels. This study is focused on developing an alternative fully data-driven H35 product with improved accuracy using a machine learning (ML)-based approach. Multivariate adaptive regression splines (MARS) algorithm is trained by using AVHRR reflectance data as well as the well-known snow and vegetation indices (i.e., NDSI and NDVI) to generate the new version of H35 fSCA product. The reference fSCA maps required for the training of MARS models are obtained from the higher resolution Sentinel 2 multispectral imagery. The MARS-based fSCA models are validated against an initial test dataset composed of 15 Sentinel 2 scenes over European Alps, Tatra Mountain Range, and Turkey. The final MARS-H35 product is then rigorously assessed over the whole Northern Hemisphere within a temporal domain spanning from Nov 2018 to Nov 2019. The quantitative testing process involves the use of reference data in both continuous and dichotomous scales: i) Sentinel 2 derived reference fSCA maps, ii) ERA5-Land snow depth data, iii) MODIS MOD10A1 NDSI snow cover data, and finally iv) in-situ snow depth data. Additionally, qualitative assessment is also performed by visually comparing MARS-H35/MODIS false-color and MARS-H35/Sentinel 2-derived reference fSCA image pairs over various geographic regions. The overall results indicate that: i) the proposed MARS-H35 fSCA product overperforms the original H35, and ii) it has higher capability in detecting the fine variations in the extent of snow cover, especially across the fringes of the slopes in complex mountainous terrains.
Effects of ice cover on soil water, heat, and solute movement: An experimental study
2021, Geoderma
In cold regions, the surface cover plays a pivotal role in understanding the soil water and energy balance during winter. Although the effects of other types of surface cover on soil freezing-thawing process have been studied for decades, the effects of ice cover derived from irrigation remain unclear. To estimate the effects of ice cover on soil water, heat, and solute movement, ice coverage and field observation data were collected for three winter periods from 2013 to 2016 in Inner Mongolia, China. The collected data was then analyzed using the principal component analysis (PCA) method. The dynamics of the water, heat, and solute transport were also investigated using in-situ plot experiments conducted in four lysimeters, which were controlled by manipulating the thickness of the ice cover. The thickness of the ice cover was controlled by the amount of water irrigation in the treatment plots, and no irrigation was performed in the control plot. The results of the field-scale survey and PCA indicated that the ice cover had positive effect on soil salinization control and moisture preservation after autumn irrigation (AI). An unique air space created during the ice formation might enhance the heat insulation effect of ice cover. Furthermore, the results of the in-situ plot experiment revealed that the freezing of the irrigated soil could be divided into the unstable freezing period (UFP) and stable freezing period (SFP) when considering the ice cover. During the UFP, the upward movement of the solute that was leached by the irrigation water was reduced as the development of the freezing front was hindered by the irrigation water. During the SFP, the penetration of the freezing front was further reduced by the combination of the ‘inherited’ thermal regime from the UFP and the input solar energy reduced by the remaining ice cover and air space, which decreased the water and solute transport. The study findings aid in achieving an improved understanding of ice cover on agricultural fields in cold regions, and have important implications for water conservation and salinization control.
Changes of glacier facies on Hornsund glaciers (Svalbard) during the decade 2007–2017
2020, Remote Sensing of Environment
Citation Excerpt :
MLC, on the other hand, is a supervised classification method which – based on means and covariance matrixes provided by unsupervised ISODATA clustering – assigns a given data point to a class with the highest probability of membership for that point. The combination of ISODATA and MLC has been applied in various studies with good results (e.g. Sulebak et al., 1997; Hall et al., 1998; Townsend et al., 2009; Jones et al., 2014; Ganju et al., 2017). The combination of ISODATA and MLC (ISODATA+MLC) classification of single- and dual-polarization SAR images was limited to the areas of the analysed glaciers derived from Błaszczyk et al. (2013).
Changes in glacier facies (glacier zones), such as firn or superimposed ice (SI), are good indicators of glacier response to climate change. They are especially important for fast-warming Svalbard, where only a few glaciers are under glaciological mass balance monitoring. This paper presents a first study of changes of glacier facies extent for three tidewater glaciers located in southern Spitsbergen (Svalbard) and it is based on both satellite remote sensing and terrestrial data analysis, covering two time spans: 2007–2017 for Hansbreen and 2012–2017 for Storbreen and Hornbreen. Satellite remote sensing analysis include unsupervised classification of Synthetic Aperture Radar (SAR) data from both decommissioned (ENVISAT ASAR) and modern satellite missions (RADARSAT-2, Sentinel-1). The results of the SAR classification are compared to the information on glacier zones retrieved from terrestrial data, i.e. shallow cores and visual interpretation of 800 MHz Ground Penetrating Radar (GPR) profiles. In addition, a novel application of the Internal Reflection Power (IRP) coefficient as an objective method of distinguishing glacier zones based on GPR data is discussed. Changes in glacier facies areas over time are analysed, as well as their correlation to Hansbreen's mass balance. The main finding of the study is that firn and SI of Hansbreen, Storbreen and Hornbreen significantly decreased over the study period. For example, due to continuous negative mass balance between 2010 and 2017, the contribution of firn area to Hansbreen's total area decreased ca. 14% (cumulative firn area loss during that time: ~45%) whereas since 2012 SI has not been distinguished as a vast area on this glacier. In addition, an east–west gradient of firn area loss was observed as a result of differences in local climate conditions. Therefore, for the common time span (i.e. 2012–2017) Hansbreen recorded a ca. 12% loss of firn contribution to glacier area whereas Hornbreen recorded ca. 9%. Finally, application of the IRP coefficient as an objective method of glacier zones discrimination by GPR data gave very good results, so the method is recommended for future analysis of glacier zones instead, or as a support, to popular visual interpretation of the GPR profiles.
Reconstructing the Chongbaxia Tsho glacial lake outburst flood in the Eastern Himalaya: Evolution, process and impacts
2020, Geomorphology
Glacial lake outburst floods (GLOF) are one of the most destructive natural disasters. Understanding GLOF evolution, and their impacts, plays a fundamental role in GLOF hazard assessment and risk management. Reconstructing historical GLOFs is an important exercise because detailed case studies of such glacial hazards are rare, which hinders the capacity of glacial hazard practitioners to learn from these events. In this study, we reconstruct a historical GLOF from moraine-dammed Chongbaxia Tsho (89.745°E, 28.211°N) in the Eastern Himalaya, which is a unique case study because the outburst flood cascaded into two further lakes downstream. We employ a combination of i) multi-source and multi-temporal satellite imagery, ii) field investigation (including an unmanned aerial vehicle survey), iii) numerical dam breach and hydrodynamic modelling and, iv) qualitative and quantitative cryospheric and meteorological analysis, to investigate the evolution of the GLOF hazard, simulate moraine dam failure and GLOF propagation, and explore the role that long- and short-term climate trends played in providing the conditioning factors for the outburst. Chongbaxia Tsho expanded rapidly between 1987 until 2001 in response to glacier recession most likely caused by a regional warming trend of +0.37 °C per decade. Based on satellite image analysis we refine the outburst date to be 6 August 2001, instead of 6 August 2000, as previously reported, and attribute an ice avalanche into the glacial lake originating from the receding parent glacier as the most likely trigger for moraine dam failure. Through DEM differencing and lake level decrease, we estimate that a total water volume of 27.1 ± 1.6 × 10⁶ m³ was released from the lake during the event, and using dam breach modelling we estimate that the peak discharge at the breach was >6600 m³ s⁻¹. The GLOF flowed through downstream Chongbamang Tsho and Chongbayong Tsho, both of which served to attenuate the GLOF and reduce downstream losses; the latter stored an estimated 96% of the flood volume. Precipitation totals in the weeks preceding the GLOF exceeded the historical mean by up to 40%, and may have contributed to instability of the parent glacier, and generation of an ice avalanche with enough impact energy to cause lake water to overtop the moraine dam and initiate breach development. A future GLOF from Chongbaxia Tsho cannot be ruled out, but more field data, including detailed lake bathymetry, and information pertaining to the sedimentological and geotechnical characteristics of the moraine dam, are required for a more robust parameterization of a predictive GLOF model and quantification of the hazard posed by a future GLOF.
Change detection of glaciers and snow cover and temperature using remote sensing and GIS: A case study of the Upper Indus Basin, Pakistan
2020, Remote Sensing Applications: Society and Environment
Citation Excerpt :
Therefore, its anti-noise capability is not satisfactory (Xiao et al., 2019). According to Hall et al. (1998), NDSI threshold value of ≥0.4 is an optimal value to map a snow pixel. Likewise, several investigators (Kulkarni et al., 2010; Xiao et al., 2001) used the threshold value of ≥0.4 while Burns and Nolin (2014) considered threshold value of ≥0.42 to be the best for debris free glacier areas after comparing the result of Landsat images to the IKONOS-2 and QuickBird high resolution imagery.
Glaciers and snow cover in the Himalayas, Karakoram, and Hindu-Kush (HKH) mountain ranges are expected to decrease, resulting in serious consequences in terms of water availability for downstream population. This study is an attempt to detect temporal trend of snow cover in the Upper Indus Basin (UIB), in northern Pakistan, with respect to temperature change, spatial variation in land surface temperature, and to evaluate the accuracy of Landsat imagery for detecting land surface temperature. Landsat images, acquired for the period 1993 to 2019 and the Normalized Difference Snow Index (NDSI), were used to map the snow cover. Maximum snow cover of about 20,675 km² was observed in April 2017 followed by 1996, and 1999, when it was about 18,800 km²; the minimum of about 14,500 km² was recorded in April 1993, 2011 and 2018. The results show that there has been a very slight increase in snow cover area with time. The maximum elevation of snow-covered area, extracted from the Digital Elevation Model of Shuttle Radar Topography Mission (SRTM DEM) in the study area, is 7889 m while the minimum elevation is 991 m. Freezing temperature occurs above 3000 m elevation, indicating that these areas are covered by snow and glaciers for much of the year. Temperatures, of −20 °C to −25 °C, were present only at a few places–at the very high elevation of above 6000 m. We found inconsistencies in the Landsat temperature data, which were corrected using in-situ daily temperature data. Validation of Landsat extracted temperatures before and after using the in-situ daily temperature data improved the accuracy and removed the inconsistencies. We recommend use of in-situ diurnal temperature data to correct Landsat extracted temperature data.

View all citing articles on Scopus

View full text

Assessment of Snow-Cover Mapping Accuracy in a Variety of Vegetation-Cover Densities in Central Alaska

Abstract

Section snippets

Introduction and background

Results and discussion

Conclusions

Acknowledgements

Remote Sens. Environ.

Remote Sens. Environ.

Remote Sens. Environ.

Derivation of snow water equivalent in boreal forests using microwave radiometry

Arctic

Determination of snow-covered area in different land covers in central Alaska from aircraft data—April 1995

Ann. Glaciol.

Airborne scanning spectrometer for remote sensing of cloud, aerosol, water vapor, and surface properties

J. Atmos. Ocean. Technol.