Thematic Session II: "Remote Sensing of Cryosphere"
Venue: Malhar Hall, Hotel Soaltee, Kathmandu, Nepal
Date: 4 October 2010 (13:30-17:30)
Chair: Bruce Raup, NSIDC, USA
Co-chair: Arun B. Shrestha, ICIMOD
Rapporteur: Ujol Sherchan, ICIMOD
Coordinator: Deo Raj Gurung, ICIMOD
The thematic session focused on the issues of snow cover mapping and modeling. Much of the papers presented focused on the improvement of accuracy of the predominant MODIS snow cover algorithm. Several factors, such as mountain shadows, (ice) clouds, and the snow albedo effect in particular in higher altitudes, have been reducing the accuracy of conventional algorithms.
Imaging spectroscopy of light-absorbing impurities in snow and ice, Thomas H. Painter, Jet Propulsion Laboratory, USA
Studies of aerosols and their impacts on snow are relatively brand new. Dust particles from the deserts of Western America are getting into the mountain systems of the region. If one puts aerosols in snow, it melts faster. However the full implications of this phenomenon are not known. There is not adequate Earth Observation system in place for light –absorbing impurities in snow and ice. In the IPCC report, the understanding of surface albedo land use is seen as medium to low, hence the need to ramp up research in this poorly understood area. Some of the findings of the research done in Western America:
- Snow melts 26-56 days earlier,
- Five- fold increase in dustloading since the 1800s,
- Run off comes 3 weeks earlier and loss of runoff in the order of 5% every year
Comparable study in the Himalaya shows a 4-fold increase in dustloading since the 1850s, this means substantial quantity of dust particles are getting into the atmosphere, including black carbon, which affects albedo.
Snow properties from MODIS show that radiative forcing in snow is relatively qualitative, the greater the red in the image shown the greater the radiative forcing. And where MODIS doesn’t have a band- there is spectral problem. IPCC has been insisting on quantitative analysis, so there is still a long way to go to meet that expectation.
Development and validation of snow cover monitoring algorithm for Himalayan region, Anil V. Kulkarni, Indian Institute of Sciences, India
An algorithm based on Normalised Difference Snow Index (NDSI) has been developed using data of AWiFS sensor of Indian Remote Sensing Satellite. Is NDSI a unique value? Field validation of NDSI was obtained for different Himalayan features (snow, vegetation, water, etc), not to mention slope angle. Slope angle has very little impact on NDSI. Influence of mixed pixels (mixed vegetation: ¼.1/2, ¾, 0, vegetation full) on spectral reflectance and NDSI was also determined. Snow cover mapping of selected basins was conducted for October – June (2002-2005), using the10-DAILY BASIN-WISE MAXIMUM SNOW COVER. Validation of algorithm: Out of 207 observatories, there were 132 matches, 2 unmatches, 73 excluded due to cloud. If there are clouds, there is problem as they are mostly ice clouds. If iced clouds data set was rejected, if ice-free clouds, used. Validation of algorithm: between 95.8% to 99.6% accuracy for various classes such as dense vegetation and snow/ice.
Some results obtained are as follows:
- Mean snow fall Western Himalaya (Bhaga and Ravi basins): 2004-5: 739 cm; 2005-6: 606 cm; 2006-7: 596 cm
- There was some stabilization in the Bhaga basin, but no stabilization in Ravi basin.
A refinement of MODIS snow cover area algorithm for mountainous Tibetan Plateau region, Bohui Tang, Chinese Academy of Sciences, China
Tang’s presentation highlighted NSIDC’s snow cover algorithm and its drawbacks, lack of corrections for atmospheric effect, topography, not eliminating the effects of illumination and viewing geometry on the values of the used reflectance. He proposed a better algorithm that tries to address all these drawbacks, which was tried in the Hindu-Kush Himalayan region, with the result that it maximized snow cover area, reduced cloud contamination and filled the gap. One finding was that NSIDC’s algorithm mistakes many pixels as snow covers in southern Bhutan which does not exist as snow cover in the true color composited image. Validation of the proposed method was done in relation to in situ data obtained from 95 stations in the HKH region for the following five dates – January 10,22, Apr14,Oct16, Nov 22, Dec12 of the year 2003- and the corrected classification (%) obtained was above 91% for each of those days.
The conclusions from the study were as follows:
- NSIDC’s MODIS snow cover products overestimate the snow cover
- The proposed algorithm can estimate the snow cover for the Himalayan region more feasible with accuracy better than 90%
- Estimating of the snow cover over heavy mountain regions needs to do atmospheric and topographic corrections
Large scale monitoring of snow cover area and variability in the Hindu Kush-Himalayan region, Deo Raj Gurung, ICIMOD, Nepal
The lack of data on cryosphere is a problem in the Hindu Kush-Himalayan region, which boasts 10 major river systems. The scope of the study was snow cover distribution pattern with respect to aspect, altitude and slope; intra-annual and inter- annual trend and seasonal trend in the HKH region. It used MODIS 8-day products, SRTM DEM v4 for slope, elevation and aspect, and HKH boundary data from ICIMOD file and basin boundary data were generated from SRTM DEM. Cloud removal methods employed in the study were successful in removing clouds significantly and improving the products. Snow cover area was found to be in agreement with the trend. It was depleting except in Amudarya, Indus and Yangtse basins. Depletion rate was highest in Brahmaputra basin and accumulation was highest in Indus basin. Strong linear regression relation was observed between percent snow cover and elevation in all basins except in Tarim basin (R=69). Jan, Feb, Mar, Oct, Nov, Dec indicate maximum monthly variation in snow cover area. While absolute accuracy based on in-situ data from 168 stations varied from 46.5% to 97.30%, relative accuracy based on high resolution snow product stood at 93%.
However with regard to inter-annual and seasonal trends, no clear pattern or trend was seen.
Monitoring Sikkim Himalayan Cryosphere, Smriti Basnet, Department of Science and Technology, Sikkim, India
The study was carried out in the Tista and Rangit Subbasins of Sikkim using AWiFS data of Oct-June taken in 5-day intervals from 2004-2008. It was the first study of its kind in the Sikkim Himalayas. The NDSI-based algorithm was run on fully automated mode. It generated snow and cloud for 5 daily products and maximum snow extent for composite (10 daily) product.
The following results and conclusions were obtained:
- All four years (2004-2008) showed similar snow cover patterns throughout Sikkim state, including in the two sub-basins.
- Maximum snow extent in February suggests Sikkim receives higher snow precipitation from Western Disturbances than North East Monsoon.
- Continuous snow precipitation even during summer months unlike in the Western Himalayas
The challenge remaining is how to address the problem associated with cloud cover and vegetation. The former may be addressed with development of newer microwave technique, where RISAT data would be of prime importance; the latter, through the development of a better algorithm to monitor snow cover under canopy.
Monitoring snow cover dynamic in the trans-Himalayan region of Nepal using MODIS data, Keshav P. Paudel, University of Bergen, Norway
The study focused on the Mustang District of Nepal had a two-fold objective: to improve and develop method to reduce cloud cover as well as examine the spatial and temporal variability of snow cover. At the heart of the study was a 5-step cloud removal methodology. The results obtained were as follows:
- This cloud removal methodology was found to be a robust technique
- High interannual and intra-seasonal variability
- Peak snow day shifted forward by 6.7 days per year
- Feb/March are main snow months
- Declining trend of snow cover duration in agro-pastoral areas of Mustang
Simulation of the spatial distribution of snow cover in the Himalayan river basin of Nepal and comparison with MODIS satellite product, Maheswor Shrestha, University of Tokyo, Japan
In Nepal, many studies have been conducted using conventional hydrological models, but none gives spatial –distributed snow cover. The model used in the study is Water and Energy Budget based Distributed Hydrological Model (WEB-DHM). The study coverd the period 2 Oct 2002 to 25 June 2003 and simulated the spatial distribution of snow cover in the Dudh Koshi basin of Nepal.
The results obtained were as follows:
- In general, WEB-DHM-S with a three layered energy balance based snowmelt module is found to be able to simulate the spatial distribution of snow cover in theDudhkoshi region of eastern Nepal Himalaya.
- Snow cover simulation is comparable to the MODIS 8 daily snow cover product (MOD10A2). Land surface temperature compared to MODIS 8 daily LSTV5 (MOD11A2) shows that LST is underestimated during the entire post monsoon and winter season.
- The snow line altitude has been simulated well and the elevation zone 4500-5500 is found very sensitive in the overestimation of SCA.
- The seasonal snowmelt contribution to the discharge at Rabuwabazar is found insignificant however multiyear simulation is necessary to draw a solid conclusion. Nevertheless, this study provides a basis for the application of WEB-DHM-S in cold and high elevated regions to simulate spatial distribution of seasonal snow coverage.
Remote sensing inputs for snowmelt runoff modeling and glacial lakes mapping
K Abdul Hakeem, NRSC, India
The study focused on the snowmelt runoff modelling in the Sutlej Basin of India, especially with reference to seasonal forecasting in April and revised forecasting in May. Although there are lots of observation points in the lower part of the basin, the upper part actually lies outside India (and in China). The timely forecast is important for planning and adaptive management of multipurpose projects (drinking water, electricity, etc) in the region. It was found that the highflow season (April, May, June) contributed to 27% of annual flows, out of that, the month of June alone contributed about 57%, while the rest was made up by April and May. Temperature trend closely followed the discharge trend in the Bagra reservoir. Also efforts were underway to develop an inventory of glacial lakes and water bodies in Tawang basin and other basins of the India Himalayas.
The research of snow cover monitoring and analysis with multi data source over Tibetan Plateau, Liu Yujie, China Meteorological Administration, China
The research objective was to figure out the main characteristics of snow cover over the Tibetan Plateau using satellite data useful in climate analysis. The data showed the main climatic characteristics of snow cover in duration, variation and distribution over the Tibetan Plateau and helped better understand the relationship between winter snow over TP and the next summer precipitation in the middle and lower reaches of the Yangtze River.
Based on the 13-year data, in both winter half year (Nov-Apr) and summer half year (May-Oct), there were two rich snow cover regions seen in the western and eastern parts of the Tibetan Plateau. This finding calls for detailed sub-regional analysis to better understand these two rich snow cover regions. The snow coverage duration varied from year to year. Snow cover decreased from 1980 to 1990, and increased until 1998, and then decreased again until recently, but snow cover above 3000 m has been decreasing more slowly. The snow cover over TP has 18-20 years large period variation and 3-5 years short period variation The snowline over TP has been upward of 3000 m height in summers and autumns in recent years. Moreover, winter snow cover in Tibetan Plateau has positive correlation with spring and summer precipitation along the Yangtze River, and Inverse correlation with spring and summer precipitation in Hunan and Heibei regions. This analysis is useful as it can potentially help identify factors contributing to “flooding in South China and drought in North China in spring and summer time.”
QUESTIONS AND ANSWERS & DISCUSSIONS
A few of the salient points that emerged from the question and answer session and ensuing discussions are summarized below:
- There was a case study presented on the application of snow melt run off forecast in the Sutlej basin of India, first in April and then the revised forecast in June based on the availability of newer data. As the timely forecast is very useful for planning and adaptive management of multi-purpose projects downstream, this good practice has a potential for replication and upscaling in the other major river basins of the region.
- With regard to snow cover mapping and modelling the issue of time period covered in the study was seen as crucial. The general impression was that the 10-year period was inadequate for snow trend analysis or for drawing any broad generalisations. The suggestion was to go for longer-period data, up to 30-years. A suggestion was put forward to also consider historical data such as AVHRR snow products.
- There was a suggestion from participants to complement data obtained from optical remote sensing with microwave remote sensing since the latter has the capability to penetrate cloud covers. However, there are limitations associated with microwave remote sensing due to mountain topography, not to mention cost. As one participant put it: “Passive microwave may introduce far greater errors than waiting for clouds to go away!”
- As India is developing an inventory of glacial lakes and water bodies in the basins of the Indian Himalayas, perhaps the institute in charge – the Central Water Commission under the Ministry of Water Resources - could inform ICIMOD of the goings-on and may be also collaborate together – as ICIMOD already has some experience in inventorying glacial lakes in the region.
- It was felt that the culture of data sharing has quite a way to go in the Hindu Kush-Himalayan region. It was suggested that access to data should be provided to all in the region and beyond in order to accelerate the progress of science for the benefits of humanity. For instance, ICIMOD makes data available through its GeoPortal. Perhaps the countries of the region should have a clear and more generous policy toward provision of satellite images to aid with disaster preparedness (not just in post disaster situations) as well as climate change adaptation.
- Ø There was a call for communicating the (cryospheric) research findings to the world-at-large in layman's terms, including their implications on the livelihoods of mountain inhabitants and those living downstream, especially in the context of climate change adaptation.
