Accuracy assessment of MODIS, NOAA and IRS data in snow cover mapping under Himalayan conditions

Accuracy assessment of MODIS, NOAA and IRS data in snow covermapping under Himalayan conditions

SANJAY K. JAIN*{, AJANTA GOSWAMI{ and A. K. SARAF{{Remote Sensing Division, National Institute of Hydrology, Roorkee 247667 India

{Department of Earth Sciences, Indian Institute of Technology, Roorkee, India

(Received 13 September 2006; in final form 23 October 2007 )

Snow cover information is an essential parameter for a wide variety of scientific

studies and management applications, especially in snowmelt runoff modelling.

Until now NOAA and IRS data were widely and effectively used for snow-

covered area (SCA) estimation in several Himalayan basins. The suit of snow

cover products produced from MODIS data had not previously been used in

SCA estimation and snowmelt runoff modelling in any Himalayan basin. The

present study was conducted with the aim of assessing the accuracy of MODIS,

NOAA and IRS data in snow cover mapping under Himalayan conditions. The

total SCA was estimated using these three datasets for 15 dates spread over 4

years. The results were compared with ground-based estimation of snow cover. A

good agreement was observed between satellite-based estimation and ground-

based estimation. The influence of aspect in SCA estimation was analysed for the

three satellite datasets and it was observed that MODIS produced better results.

Snow mapping accuracy with respect to elevation was tested and it was observed

that at higher elevation MODIS sensed more snow and proved better at mapping

snow under mountain shadow conditions. At lower elevation, IRS proved better

in mapping patchy snow cover due to higher spatial resolution. The temporal

resolution of MODIS and NOAA data is better than IRS data, which means that

the chances of getting cloud-free scenes is higher. In addition, MODIS has an

automated snow-mapping algorithm, which reduces the time and errors

incorporated during processing satellite data manually. Considering all these

factors, it was concluded that MODIS data could be effectively used for SCA

estimation under Himalayan conditions, which is a vital parameter for snowmelt

runoff estimation.

1. Introduction

Snow is of great importance as a key environmental parameter. It not only

influences the Earth’s radiation balance but also plays a significant role in river

discharge. A major source of runoff and ground water recharge in middle and higher

latitudes are from snow melt from seasonal snow-covered areas of the Earth’s

mountain region. The central Asian mountains contain about 50% of the total

glaciated area in the world and a large portion of this area drains into the landmass

of the Indian sub-continent. The Himalayan mountain system is the source of one of

the world’s largest supplies of freshwater. All the major south Asian rivers originate

in the Himalayas and their upper catchments are covered with snow and glaciers.

*Corresponding author. Email: [email protected]

International Journal of Remote Sensing

Vol. 29, No. 20, 20 October 2008, 5863–5878

International Journal of Remote SensingISSN 0143-1161 print/ISSN 1366-5901 online # 2008 Taylor & Francis

http://www.tandf.co.uk/journalsDOI: 10.1080/01431160801908129

Downloaded By: [Indian Institute of Technology] At: 14:18 24 November 2008

The Indus, Ganga and Brahmaputra river systems, originating from the Himalayan

region, receive substantial amounts of meltwater and are considered as the lifeline of

the Indian sub-continent. The perennial nature of Himalayan rivers and the

appropriate topographic setting of the region provide a substantial exploitable

hydropower potential in this area (Singh and Jain 2002). Hence, near real-time

estimation of snow cover is of utmost importance for effective management of water

resource and can serve as a guideline for reservoir operation. In addition, the

planning of new hydroelectric projects on the Himalayan rivers emphasizes the need

for reliable estimation of snow and glacier runoff.

Snow-covered area (SCA) is a vital input to many of the hydrological models. An

accurate estimation of SCA in near real time is an essential parameter for Snow

Runoff Models (Ramamoorthi 1987). Conventional methods of SCA estimation,

such as snow surveys, provide detailed information on snow pack properties, but

under Himalayan conditions such surveys are not feasible. The Himalayan region is

topographically highly rugged and climatically very harsh because of which a poor

snow gauge network exists in the high altitude region of the Himalayas, particularly

where heavy snowfall is experienced. Therefore, ground-based estimation of SCA

over the whole basin becomes very difficult under such conditions.

Snow cover mapping in mountainous areas is demanding due to the interfering

topography and the heterogeneous ground properties. According to Rango (1996),

the only efficient way to monitor the dynamically changing seasonal snow cover on

a sufficiently large scale is by satellite remote sensing. But, one must judiciously

select the proper sensor to use for a particular analysis taking into consideration

factors such as wavelengths, resolution or frequency and timing of ground coverage

(Hall and Martinec 1985). A high temporal resolution is important; particularly for

monitoring changes in snow extent due to melt or accumulation. Although snow

cover can be detected and monitored with a variety of remote sensing devices, the

greatest application has been found in the visible (VIS) and the near-infrared (NIR)

region of the electromagnetic spectrum (Hall et al. 2002). For operational snow

cover monitoring, satellite sensors with moderate spatial resolution but with high

repetition rate are important for the advantage of obtaining a cloud-free image.

Although the cloud problem is removed with the use of microwave data (either

passive or active), interpretation of the images is much more difficult with respect to

optical remote sensing data because they are highly affected by surface and

subsurface properties.

Snow was observed in the first image obtained from the Television and Infrared

Observation Satellite (TIROS-1) following its April 1960 launch (Singer and

Popham 1963). In the mid-1960s, snow was successfully mapped from space on a

weekly basis following the launch of the Environmental Science Service

Administration (ESSA-3) satellite, which carried the Advanced Vidicon Camera

System (AVCS) that operated in the spectral range of 0.5–0.75 mm with a spatial

resolution at nadir of 3.7 km. The National Oceanographic and Atmospheric

Administration (NOAA) has measured snow cover on weekly basis in the Northern

Hemisphere since 1966 using a variety of sensors, including the Scanning

Radiometer (SR), Very High Resolution Radiometer (VHRR) and the Advanced

Very High Resolution Radiometer (AVHRR) (Matson et al. 1986). Now there are

many sensors providing images with great spectral, spatial and temporal resolution,

which can be used, based on the need of the study. With the availability of large

number of satellite sensor data, it is now left up to the choice and requirement of the

5864 S. K. Jain et al.


user community which data is used. The revolutionary role of remote sensing in

snow study is discussed in detail by Rango (1996), Singh and Jain (2003) and Hall

et al. (1995). Nevertheless, field measurements are still required to validate the

satellite sensor data (Saraf et al. 1999).

Satellite images have been extensively used for snow study in Himalayan

conditions by several researchers. Saraf et al. (1999) has used passive microwave

(scanning multichannel microwave radiometer (SMMR)) data for snow-depth and

snow-extent estimation in a part of Satluj basin in the Himalayan Mountains.

Rango (1996) has analysed and reported that NOAA-AVHRR and Aqua/Terra-

Moderate Resolution Imaging Spectroradiometer (MODIS) data are useful for

snow study with basin area above 200 km2. Dey and Goswami (1983) used NOAA-

VHRR data to predict seasonal snowmelt runoff in the Indus basin. Gupta et al.

(2005) have used Indian Remote Sensing (IRS) satellite LISS (Linear Imaging Self

Scanning) III multispectral data to map dry/wet snow cover in Himalaya. In

addition, many studies have been conducted on snow cover study in Himalaya using

satellite sensor data (Gupta et al. 1982, Agarwal et al. 1983, Dey et al. 1988,

Upadhyay et al. 1991, Thapa 1993, Jain 2001).

With the availability of satellite data in a wide range of spectral, spatial and

temporal resolution, the choice is left up to the requirement of the user community.

Many of the data are made freely available to the user community through World

Wide Web, and some can be procured through other agencies against a nominal

charge. The NASA’s Earth Observing System (EOS) Data and Information System

(EOSDIS) and Satellite Active Archive site (NESDIS 2006) are among the few of the

communities which are involved in processing, archiving, and distributing different

satellite data, thereby promoting the inter-disciplinary study and understanding of

the integrated Earth system. The present study was undertaken to estimate SCA

using IRS WiFS images, NOAA-AVHRR images and MODIS SCA data product

for the upper reaches of Satluj river basin. The aim of the present study was to

analyse the snow cover mapping potential of these three datasets under different

conditions, accuracy achieved in snow cover mapping and selection of the most

suitable datasets for SCA estimation for the present study area. Optical remote

sensing data were used in this study due to its easy interpretability. The data have

the added advantage of easy availability, and it is also relatively easy to distinguish

snow from snow-free areas. Besides, the visible satellite sensor data is available in a

wide range of spatial and temporal resolutions. The SCA was estimated from these

three datasets for selected cloud-free dates for the years 2000–2004. The temperature

lapse rate method was used to estimate SCA from air temperature data. The SCA

estimated from both remote sensing and field-based methods were compared. The

findings were considered as an aid to select datasets for the Satluj river basin in snow

melt runoff estimation.

2. The study area

The study area consists of the Satluj River Basin up to the Bhakra dam located in

the Himalayan mountain range (figure 1). The Satluj, also known as Sutlej, is the

longest of the five rivers of Punjab state that flows through Northern India. The

river rises in the lakes of Mansarovar and Rakastal in the Tibetan Plateau at an

elevation of about 4572 m and is one of the main tributaries of the Indus River. It

receives the Beas River in the state of Punjab, India and continues into Pakistan to

join the Chenab river to form the Panjnad river which further joins the Indus river at

Accuracy assessment of MODIS, NOAA and IRS data 5865


Mithankot (Wikipedia 2006). The total catchment area of Satluj River up to the

Bhakra dam is about 56 500 km2, out of which about 22 305 km2 lies in India. The

basinal area considered for the study is about 25 000 km2. The elevation of

the catchment varies widely from 500 to 7000 m; although only a very small area

exists above 6000 m. Mean elevation of the basin is about 3600 m. The gradient of

the river is very steep at its source and it gradually reduces as it moves downstream.

Owing to large differences in seasonal temperatures and a great range of elevation in

the catchment, the snowline is highly variable, descending to an elevation of about

2000 m during summer. The Satluj River receives cool, snowmelt water from the

upper catchment during the spring and summer months and from monsoon

precipitation during July–September in its lower catchment.

The Indus water Treaty, 1960 entitled India to use of the water of rivers Ravi,

Beas and Satluj. However, it was provided that full/unrestricted use of the waters of

Rivers Satluj, Beas and Ravi will be made by India only after 31 March 1970. With a

view to fully utilize the waters of these three rivers, the Government of India has

approved various schemes from time to time (BBMB 2006). Several dams were

erected over the Satluj River to create a reservoir. The Bhakra–Nangal dam is a

huge multipurpose dam on the river. The Bhakra Beas Management Board

(BBMB), constituted by the Government of India in 1966, is responsible for the

management and distribution of water and hydro-power generated out of the river

water. Therefore, daily runoff estimation is essential for the management of the

hydro-power houses and efficient distribution of the river water among the

beneficiaries. Hence, SCA estimation for the Satluj river basin is essential.

Figure 1. USGS DEM of Satluj basin (Indian part) up to Bhakra Reservoir with location ofmeteorological stations (marked with stars).



3. Data used

In this study, data from three different satellites have been used and the details of

datasets used in the study are as follows.

3.1 IRS 1C/1D WiFS data

A total of 15 IRS 1C/1D Wide Field Sensor (WiFS, path/row 29/45) cloud-free data

were procured from the National Remote Sensing Agency (NRSA), Hyderabad, India

for the years 2000–2004. The sensor acquired the images around 11am. The WiFS

instrument is a spaceborne optical sensor that was designed for observation of

vegetation and land surfaces. The specifications of the WiFS sensor are given in table 1.

With a swath of 814 km, WiFS provides temporal resolution of five days at 188 mspatial resolution. Twenty-two scenes cover the entire of India. Unlike scanner sensors

(e.g. AVHRR), the WiFS instrument uses a linear array sensor and thus produces high

quality imagery at regional/national scale for vegetation mapping.

3.2 NOAA-AVHRR Images

The cloud-free NOAA-16 and -17 images used in this study were acquired at around

11am Indian Standard Time (IST). The NOAA images were acquired for the datesand years corresponding to the available IRS 1C/1D WiFS data. The NOAA series

of Sun-synchronous satellites circle the Earth in a near-polar orbit. The specification

of the NOAA-16/17 data used in the study is summarized in table 2. The raw

AVHRR High Resolution Picture Transmission (HRPT) data used in the present

study have been acquired by the Indian Institute of Technology-Roorkee Satellite

Earth Station (SES), which was established on 24 October 2002 and has been

operational ever since (Saraf and Choudhury 2006). This NOAA-HRPT satellite

Earth station receives about 12 images per day and covers an area with a radius ofaround 3000 km. The NOAA images of the study area, before October 2002 were

acquired from Satellite Active Archive site (NESDIS 2006).

3.3 MODIS daily snow cover maps (MOD10A1)

MODIS is an environmental satellite operating in visible, near- and short-wave

infrared and thermal portions of the electromagnetic spectrum and acquires imagesin 36 spectral bands. It has spatial resolution of 250, 500 and 1000 m depending on

the spectral band and has a swath width of 2330 km, enabling the entire surface of

the Earth to be viewed every 2 days. MODIS images have the potential to provide

quantitative measures of numerous geophysical parameters, including snow cover

(Justice et al. 1998).

Table 1. Key parameters of the IRS 1C/1D WiFS sensors.

1. Radiometric resolution: 7 bits2. Local time for equatorial crossing: 10:30 a.m.3. Temporal resolution: 5 days with the off-nadir cross-track steering facility

Bands Spectral coverage (mm) Spatial resolution (m) Swath (km)

B3 0.62–0.68 188 810B4 0.77–0.86 188 810

Source: Pandya et al. (2002), Euromap (2006), NRSA (2006).



The MODIS snow-mapping algorithm utilizes the seven MODIS bands designed

especially to image the land surface. MOD10A1 consists of 1200 km by 1200 km tiles

of 500 m resolution data gridded in a sinusoidal map projection. MODIS/Terra

V004 data extend from 24 February 2000 to the present. The product uses MODIS

snow-mapping algorithm, which is based on a long lineage of snow detection

algorithms and was developed using Landsat Thematic Mapper (TM) data (Hall

et al. 1995, Klein et al. 1998, Hall et al. 2002, Klein and Barnett 2003). The MODIS

snow-mapping algorithm is a fully automated and computationally frugal approach

to snow detection. It is based on a series of tests for snow detection including

Normalized Difference Snow Index (NDSI) and other criteria tests. Currently, a

suite of snow cover products is being produced from MODIS data and is distributed

by the Distributed Active Archive Center (DAAC) located at the National Snow

and Ice Data Center (NSIDC). The inputs for MODIS snow-mapping algorithm

are: calibrated radiances (MOD02HKM), the MODIS cloud mask (MOD35) and

the solar and satellite viewing geometries, as well as other ancillary information

from the MOD03 geolocation product. The resulting snow product is the MODIS/

Terra Snow Cover 5-Min L2 Swath 500 m (MOD10_L2) product. It is so named

because it represents snow cover acquired in approximately 5 min of MODIS scans.

It has a nominal spatial resolution of 500 m and a swath width of 1354 km across

track and 2030 km along track length. MOD10A1 data product is the spatial

composite of individual snow observations from all daily MOD10_L2 products

covering a particular area (Klein et al. 1998). A set of five MODIS snow cover

products is currently being distributed to the user community through the NSIDC

DAAC. This suite of snow products represents a progression of products ranging

from snow cover maps produced from individual MODIS swaths (scenes) to spatial

and temporal composites covering 8-day periods at spatial resolutions of

0.05u60.05u. The MODIS Snow Products Users’ Guide contains detailed

information pertaining to the structure of and information contained in all

MODIS snow products (Riggs et al. 2003). The summary of MOD10A1 data is

listed in table 3 (adopted from Riggs et al. 2003).

Table 2. Key parameters of the NOAA-16/17 AVHRR data.

1. Radiometric resolution: 10 bits2. Swath width: 2800 km3. Temporal resolution: four images per satellite per day4. Spatial resolution: 1.1 km

Name of band ChannelSpectral

coverage (mm) Typical use

Visible 1 0.58–0.68 Daytime cloud, ice and snow, vegetationNear-infrared 2 0.73–1.10 Daytime cloud, vegetationInfrared 3(a) 1.58–1.64 Soil humidity, ice/snow discrimination

(operates only in daytime)Near-infrared 3(b) 3.55–3.93 Heat source, night-time cloud (operates

only at night)Thermal infrared 4 10.3–11.3 SST, LST, day/night cloudThermal infrared 5 11.5–12.5 SST, LST, day/night cloud

Source: Saraf et al. (2006).SST, sea surface temperature; LST, land surface temperature.



The automated snow-mapping algorithm uses at-satellite reflectance in MODIS

band 4 and band 6 to calculate the Normalized Difference Snow Index (NDSI) (Hall

et al. 1995):

NDSI~B4{B6

B4zB6ð1Þ

where B4 stands for reflectance in 0.545–0.565 mm and B6 for reflectance in 1.628–

1.652 mm.

A pixel in a non-densely forested region will be mapped as snow if the NDSI is

>0.4 and reflectance in MODIS band 2 (0.841–0.876 mm) is .11%. However, if the

MODIS band 4 reflectance is 10%, then the pixel will not be mapped as snow even if

the other criteria are met. This prevents pixels containing very dark targets from

being mapped as snow. This is required because very low reflectance causes the

denominator in the NDSI to be quite small, and only a small increase in the visible

wavelengths are required to make the NDSI value high enough to classify a pixel

erroneously as snow (Hall et al. 2002).

3.4 SRTM 90 m digital elevation data

On 19 February 2000, the space shuttle carried onboard, for the first time, a space-

borne, single-pass interferometer. The mission was referred to as the Shuttle Radar

Topographic Mission (SRTM). SRTM successfully mapped the topographic

features of the Earth’s landmasses using radar interferometry (Leblanc et al.

2006). A radar interferogram is produced by measuring the radar phase difference

between two spatially separated antennas, A1 and A2 (Zebker et al. 1994). Two

antennas can be mounted on a single platform or can use repeat passes over the

same site with a single antenna. A more detailed description of the methodology for

extracting interferograms from radar phase measurements is given by Zebker et al.

(1994). The SRTM mission employed two antennae that were separated by a

baseline of 60 m. The prime antenna transmitted and received a radar signal, while

the second only received a signal. Two antennae operating in C- and X-bands

simultaneously illuminated and recorded radar signals over the entire landmass

between 60uN and 57u S (Blumbarg 2006). The C- band radar with a wavelength of

5.6 cm has an electronically steerable antenna and is therefore able to operate in

ScanSAR mode with a 225 km swath. However, the X-band radar system operated

at 3 cm wavelength with its passive primary antenna is historically limited to a 45 km

wide swath. The near global coverage DEM was produced from the C-band data

and processed by NASA’s Jet Propulsion Laboratory (JPL) and the X-band data

provided slightly higher resolution and were processed by the German space

agency’s aerospace centre (DLR). This method requires no ground control, and

hence is very useful for inaccessible regions. The overall absolute horizontal and

vertical accuracy of these 1 arcsecond data is estimated to be significantly better

Table 3. Summary of the MOD10A2 product.

Earth sciencedata type (ESDT)

Productlevel

Nominal data arraydimensions

Spatialresolution

Temporalresolution Map projection

MOD10A1 L3 1200 km by 1200 km 500 m daily GCTP sinusoidal

Source: Riggs et al. 2003.GCTP, General Cartographic Transformation Package.



than the original mission requirements of 20 and 16 m, respectively (Rosen et al.

2001, Sun et al. 2003). The spatial resolution of the SRTM DEM is 3 arcseconds

(approximately 90 m). A 1 arcsecond data product was also produced, but is not

available for all countries. Vertical reference of the SRTM DEMs is the WGS84

EGM96 geoid. These data are currently distributed free of charge by the US

Geological Survey (USGS) and is available for download from the National Map

Seamless Data Distribution System, or the USGS ftp site. The DEM files have been

mosaiced into a seamless global coverage, and are available for download as 5u65utiles, in geographic coordinate system – WGS84 datum. These files are available for

download in both Arc-Info ASCII format, and as GeoTiff, for easy use in most GIS

and remote sensing software applications.

3.5 Air temperature data

Air temperature data utilized in this study were collected by Bhakra Beas

Management Board (BBMB) from five ground stations, namely, Kaza, Raksham,

Namigia, Kalpa and Nangal, located within the Satluj river basin (figure 1). The daily

average air temperature data for the years 2000–2004 were utilized in the study.

4. Methodology

Snow maps derived from satellite sensor data are a pixel-based representation of a

snow-covered area. With spatial resolution of a few hundred metres up to 1 km, a

pixel, either classified as ‘snow’ or ‘no-snow’, often consists of snow-covered and

snow-free parts. In theory, the snow line defines the line separating snow-covered

from snow-free areas. However, because of the patchiness of the edge of the snow

cover, no distinct line can be drawn. Instead, a more or less narrow belt has to be

defined as the snow line, which represents a zone of approximately 50% snow

coverage (Seidel and Martinec 2004).

Snow cover mapping is a process that involves distinguishing snow pixels from

non-snow pixels. For snow cover mapping, three kinds of methods, such as training

sites supervised classification (SC), reflectance statistics and Normalized Difference

Snow Index (NDSI) and unsupervised classification have generally been used to

map snow cover distribution and then calculate SCA. As stated above, the satellite

remote sensing datasets from IRS-WiFS, MODIS and NOAA were utilized for

mapping snow cover. Before using all these three datasets for snow cover area

mapping, they were brought to a common projection system.

Using the Indian tile of the global SRTM-DEM, the DEM of the study area was

extracted. It was in geographic coordinate system and WGS 84 datum plane. All

the 15 IRS 1C/1D WiFS images, corresponding NOAA-AVHRR images and the

MOD10A1 data products were georeferenced (figure 2) by taking the DEM as the

master image and the rest of the images as slave images. More than 30 ground

control points were selected for each image in such a way that they were well spread

through out the study area to achieve higher accuracy in georeferencing. The

second-order transformation and nearest-neighbour re-sampling technique was

adopted. The root mean square error was within a pixel size.

One set of SRTM DEM was resampled to 188 m pixel size to match it with the

spatial resolution of the IRS WiFS data. Another set of SRTM DEM was resampled to

500 m pixel size to match with the MODIS data. The NOAA-AVHRR images were

also resampled to 500 m pixel size for better comparison with MODIS data products.



4.1 Preparation of snow cover map

The most sensitive part of the whole exercise was to estimate SCA from satellite

images especially under Himalayan conditions. The first problem faced was due to

cloud, which has the same reflectance as snow in the visible part of the

electromagnetic spectrum, which leads to overestimation of SCA (e.g. Hall and

Martinec 1985, Jian and Wenjun 2003, Gupta et al. 2005). The second problem

faced was due to mountain shadow. Himalayan terrains are very rugged. The

sensors acquired the images used in this study at around 11am, and the prevailing

lower angle of solar illumination leads to shadow. Hence it was very difficult todiscriminate snow from snow-free areas.

Snow exhibits high reflectance in the visible band and strong absorption in the

SWIR band. Cloud, on the other hand, shows uniform reflectance due to non-

selective scattering. A spectral band ratio can enhance features, if there are

differences in spectral slopes (e.g. Gupta et al. 2005).

The NDSI uses the above spectral characteristics of snow and is based on the

concept of Normalized Difference Vegetation Index (NDVI) used in vegetation

mapping from remote sensing data (Dozier 1989, Hall et al. 1995, Gupta et al. 2005).

The NDVI is defined as the difference of reflectance observed in a visible band and

the short-wave infrared band divided by the sum of the two reflectance (Gupta et al.

2005). The equation is given below:

NDSI~Visible band{SWIR band

Visible bandzSWIR band: ð2Þ

Figure 2. Schematic representation of the methodology adopted to compare IRS, NOAAand MODIS images for snow cover mapping.



NDSI maps were prepared from NOAA-AVHRR images. The NDSI map wasfurther classified into two classes: (a) snow and (b) snow-free areas based on a

threshold value of 0.4 (Dozier 1989). This type of classification provided an

advantage of SCA estimation under mountain shadow condition and discrimination

between snow and cloud. For IRS WiFS data, unsupervised classification using the

Iterative Self-Organizing Data Analysis Techniques (ISODATA) has been applied

in ERDAS Imagine. The ISODATA uses the minimum spectral distance to assign a

cluster for each candidate pixel. The classified output has been categorized into two

classes, i.e. snow and non-snow.The MODIS snow cover product is a classified image. The images were further

classified by combining snow and lake ice into the snow category and the rest of the

classes into the non-snow category. Thus, all the images, namely, MODIS, NOAA

and WiFS images were classified into snow and non-snow categories.

Using the classified snow maps, the total percentage of snow cover in the study

area was estimated for different dates. A comparison was made between the

percentages of SCA estimated using different images (figure 3).

4.2 Preparation of elevation and aspect map

The SRTM-DEM was used to prepare an elevation map of the study area with two

different spatial resolutions namely, 188 and 500 m to match the data with IRSWiFS images and MODIS data products. The aspect map of the study area was

generated using the two sets of DEM. It was classified into eight categories, namely,

east, south-east, south, south-west, west, north-west, north and north-east. Further,

the distribution of snow cover in different aspect classes was determined for

different dates using the three sets of images and aspect map.

Figure 3. Graphical representation of snow cover area estimated for Satluj basin usingMODIS, IRS and NOAA images.



5. SCA estimation using ground data

The accuracy of satellite-based estimation of SCA was compared with ground data.

Since field data of SCA was not available for the present study, an indirect approach

was used to estimate snow cover using temperature lapse rate method. Air

temperature data was collected from BBMB for five ground stations as mentioned in

previous paragraphs. Using the maximum and minimum air temperature data

average air temperature was calculated. To make the data smooth, air temperature

data of two pre- and post-date were added with the data of that same date for which

satellite images were available and an average air temperature was calculated. Using

this air temperature data and temperature lapse rate method, the altitude

corresponding to 0uC was computed. The area above this altitude was considered

to be covered with snow (Gupta et al. 2005). The temperature lapse rate value of

0.65uC/100 m (Singh and Jain 2002) was used for estimating SCA. The SCA was

estimated for all the dates for which satellite images were available. The result shows

(table 4) that ground-based estimation of SCA are in very close agreement with the

satellite-based estimation.

6. Results and discussion

It was observed that the total SCA estimated using three different sets of data were

in good agreement. Overall, MODIS mapped slightly higher snow than WiFS

images and NOAA-AVHRR images mapped slightly less compared to the other two

(figure 3). The standard deviation of the three results was calculated and it was

observed that the minimum standard deviation is 0.57 for 21 April 2001 and

maximum standard deviation is 1.51 for 27 March 2001. This indicates that the three

datasets are mutually in good agreement and have nearly equal ability to map total

SCA.

To better examine the datasets, distribution of snow with respect to different

aspect classes were examined. It was observed that in all the images, snow cover is

Table 4. Comparison of SCA estimated from NOAA, MODIS, IRS data and fieldtemperature lapse rate method.

Dates MODIS (%) WiFS (%) NOAA (%)Temperature lapserate method (%)

Standarddeviation

11 April 2000 45.91 44.70 44.21 45.60 0.792 May 2000 38.12 36.76 37.36 37.70 0.5827 March 2001 51.96 51.03 49.01 51.70 1.3321 April 2001 42.78 42.47 41.68 42.31 0.4618 September

20016.09 5.58 5.02 5.56 0.44

29 March 2002 59.89 60.10 58.08 59.60 0.9122 April 2002 46.67 45.41 45.02 46.40 0.7916 May 2002 38.59 37.10 36.68 38.20 0.9028 September

200220.15 18.91 17.56 19.80 1.15

22 March 2003 54.42 53.33 52.12 54.20 1.0411 May 2003 45.56 44.77 45.98 45.44 0.506 March 2004 47.66 47.83 45.46 47.40 1.1031 March 2004 37.23 36.55 34.12 37.10 1.4525 April 2004 30.96 30.25 29.68 31.10 0.6628 August 2004 4.56 7.43 6.68 5.70 1.24



minimum in south and south-west directions and is maximum in north and north-east directions. The estimation of SCA in between north to east is equal in all the

datasets. But the snow cover estimation in case of west and north-west directions

varies in the three datasets. Along west and north-west directions MODIS mapped

more snow compared to NOAA and WiFS images (figure 4). The reason is

mountain shadow. Since all the images were acquired at around 11am when solar

illumination is in east and south-east direction, shadows appear in the opposite

quadrant, i.e. in west and north-west. NDSI classification, although capable of

mapping snow cover under mountain shadow condition, is not effective enough tomap it completely. MODIS, on the other hand, utilizes a set of grouped decision

Figure 4. Relation between aspect classes and snow cover area as estimated from MODIS,WiFS and NOAA images.



tests for snow detection, which makes it more effective in snow cover mapping under

shadow condition.

To spatially estimate the distribution of snow cover in all the images, a

comparison was made between snow cover and elevation. An interesting result was

observed for all the images. In general, MODIS data product mapped slightly moresnow in higher elevation (figure 5). But in the lower part of the study area, WiFS

images mapped slightly more snow. A possible reason for this disagreement is the

difference in spatial resolution of the three datasets. Along the valley area, patchy

snow cover exists which may be of 10–100 m in extent. Since the resolution of

NOAA images and MODIS data products are 1 km and 500 m, respectively, these

sensors could not sense snow cover of a few tens of metres. WiFS, on the other

hand, has a spatial resolution of 188 m, which is more effective in sensing snow cover

of such a small area.

The comparison of ground-based estimation and satellite-based estimation of

snow cover shows that both the approaches are mutually in good agreement

(table 4). Hence, satellite data can be effectively used for SCA estimation under

Himalayan conditions.

7. Conclusions

With the progress in space technology, a wide variety of satellite sensors and data

products with different characteristics is available to the scientific community. But,

the availability of large numbers of satellite images presented a new challenge ofchoosing the most suitable among them all for a particular requirement. Estimation

of SCA is such a field where remote sensing and GIS has attained operational status

Figure 5. Comparison of SCA estimated in Satluj basin using MODIS, NOAA and IRSimages. (Not to scale).



in many parts of the world. The present study was conducted for Satluj basin to

compare and select the suitable image among MODIS, IRS WiFS and NOAA for

snow cover mapping. All the three images estimated almost equal amounts of snow

in between the north-east to west direction of slope. But in the west, north-west and

north direction, MODIS is more effective at estimating snow cover.

When elevation was considered as a criterion, MODIS proved more effective in

estimating snow cover in higher elevation. But in lower elevation where snow cover

is patchy and thin, IRS images gave better results because of its better spatial

resolution.

Due to high temporal resolution, the chance of getting cloud-free images is more

likely in the cases of MODIS and NOAA images compared to IRS. Considering all

these criteria it was observed that MODIS images are more suitable for use in snow

cover mapping.

NOAA and IRS data have previously been used successfully for SCA estimation

under Himalayan conditions, which was used for snowmelt runoff estimation.

Previous to this study, there was no record of using MODIS snow cover product for

SCA estimation. The present study clearly shows that MODIS snow cover product

can provide better results under certain conditions. Hence it can be effectively used

for SCA estimation and snowmelt runoff generation under Himalayan conditions.

Acknowledgements

Ajanta Goswami was supported by the Council of Scientific and Industrial

Research, India under the CSIR-JRF Fellowship Grant No 9/143 (495)/04-EMR-I.

The authors are thankful to Bhakra Beas Management Board (BBMB), Nangal,

India for providing the air temperature data used in this study. Thanks are due to

Santosh Panda, research scholar, IIT Roorkee for providing timely help and

information related to this work.

ReferencesAGARWAL, K.C., KUMAR, V. and DAS, T., 1983, Snowmelt runoff for a catchment of Beas

basin. In Proceedings of the First National Symposium on Seasonal Snowcover, 28–30

April 1983, Snow & Avalanche Study Establishment (SASE), Manali, Vol II, pp.

43–63.

BBMB, 2006, Formation of BBMB. Available online at: http://bhakra.nic.in/english/

menu1.asp (accessed on 18 August 2006).

BLUMBERG, D.G., 2006, Analysis of large aeolian (wind-blown) bedforms using the Shuttle

Radar Topography Mission (SRTM) digital elevation data. Remote Sensing of

Environment, 100, pp. 179–189.

DEY, B. and GOSWAMI, D.C., 1983, Application of Remote Sensing for seasonal runoff

prediction in the Indus basin, Pakistan. In Hydrological Applications of Remote

Sensing and Remote Data Transmission (Proceedings of the Hamburg Symposium,

August 1983), IAHS Publication No. 145, pp. 637–645.

DEY, B., SHARMA, V.K., GOSWAMI, D.C. and SUBBA RAO, P., 1988, Snow cover, snowmelt

and runoff in the Himalayan river basins. Final technical report, National Aeronotics

and Space Administration, NASA-CR-182434.

DOZIER, J., 1989, Spectral signature of alpine snow cover from the Landsat Thematic

Mapper. Remote Sensing of Environment, 28, pp. 9–22.

EUROMAP, 2006, IRS-1C, IRS-1D. Available online at: http://www.euromap.de/site/index.

html (accessed on 18 August 2006).



GUPTA, R.P., HARITASHYA, U.K. and SINGH, P., 2005, Mapping dry/wet snow cover in the

Indian Himalayas using IRS multispectral images. Remote Sensing of Environment,

97, pp. 458–469.

GUPTA, R.P., DUGGAL, A.J., RAO, S.N., SHANKAR, G. and SINGHAL, B.B.S., 1982, Snowcover

area in snowmelt runoff relation and its dependence on geomorphology – a study

from Beas catchment (Himalaya, India). Journal of Hydrology, 58, pp. 325–339.

HALL, D.K. and MARTINEC, J., 1985, Remote Sensing of Ice and Snow, p. 189 (London:

Chapman and Hall).

HALL, D.K., RIGGS, G.A. and SALOMONSON, V.V., 1995, Development of methods for

mapping global snow cover using moderate resolution imaging spectroradiometer

data. Remote Sensing of Environment, 54, pp. 127–140.

HALL, D.K., RIGGS, G.A., SALOMONSON, V.V., DIGIROLAMO, N.E. and BAYR, K.J., 2002,

MODIS snow cover products. Remote Sensing of Environment, 83, pp. 181–194.

JAIN, S.K., 2001, Snowmelt runoff modeling and sedimentation studies in Satluj basin using

remote sensing and GIS. PhD thesis, University of Roorkee.

JIAN, W. and WENJUN, L., 2003, Comparison of methods of snow cover mapping by

analyzing solar spectrum of satellite remote sensing data in China. International

Journal of Remote Sensing, 24, pp. 4129–4136.

JUSTICE, C., VERMOTE, E., TOWNSHEND, J.R.G., DEFRIES, R., ROY, D.P., HALL, D.K.,

SALOMONSON, V.V., PRIVETTE, J., RIGGS, G., STRAHLER, A., LUCHT, W., MYNENI, R.,

KNJAZIHHIN, Y., RUNNING, S., NEMANI, R., WAN, Z., HUETE, A., VAN LEEUWEN, W.,

WOLFE, R., GIGLIO, L., MULLER, J.-P., LEWIS, P. and BARNSLEY, M., 1998, The

Moderate Resolution Imaging Spectroradiometer (MODIS): land remote sensing

for global change research. IEEE Transactions on Geoscience and Remote Sensing, 36,

pp. 1228–1249.

KLEIN, A.G. and BARNETT, A.G., 2003, Validation of daily MODIS snow cover maps of the

Upper Rio Grande River Basin for the 2000–2001 snow year. Remote Sensing of

Environment, 86, pp. 162–176.

KLEIN, A.G., HALL, D.K. and RIGGS, G.A., 1998, Improving snow-cover mapping in

forests through the use of a canopy reflectance model. Hydrological Processes, 12,

pp. 1723–1744.

LEBLANC, M., FAVREAU, G., MALEY, J., NAZOUMOU, Y., LEDUC, C., STAGNITTI, F.,

OEVELEN, P.J., DELCLAUX, F. and LEMOALLE, J., 2006, Reconstruction of Megalake

Chad using Shuttle Radar Topographic Mission data. Palaeogeography,

Palaeoclimatology, Palaeoecology, 239, pp. 16–27.

MARKHAM, B.L. and BARKER, J.L., 1987, Thematic Mapper band-pass solar exoatmospheric

irradiances. International Journal of Remote Sensing, 8, pp. 517–523.

MATSON, M., ROEPLEWSKI, C.F. and VARNADORE, M.S., 1986, An Atlas of Satellite-derived

Northern Hemisphere Snow Cover Frequency (Washington, DC: National Weather

Service).

NESDIS, 2006, Comprehensive large array-data stewardship system. Available online at:

http://www.class.noaa.gov (accessed on 18 August 2006).

NRSA, 2006, Indian remote sensing satellites. Available online at: www.nrsa.gov.in (accessed

on 18 August 2006).

PANDYA, M.R., SINGH, R.P., MURALI, K.R., BABU, P.N., KIRAN KUMAR, A.S. and

DADHWAL, V.K., 2002, Bandpass solar exo-atmospheric irradiance and Rayleigh

optical thickness of sensors onboard Indian Remote Sensing Satellites-1B, 1C, 1D,

and P4. IEEE Transactions on Geoscience and Remote Sensing, 40, pp. 714–718.

RAMAMOORTHI, A.S., 1987, Snow cover area (SCA) is the main factor in forecasting snowmelt

runoff from major river basins. In Large Scale Effects of Seasonal Snow Cover

(Proceedings of the Vancouver Symposium, August, 1987), IAHS Publication No. 166,

pp. 187–198.

RANGO, A., 1996, Spaceborne remote sensing for snow hydrology applications. Hydrological

Sciences – Journal des Sciences Hydrologiques, 41, pp. 477–494.



RIGGS, G.A., HALL, D.K. and SALOMONSON, V.V., 2003, MODIS snow products users’ guide.

Available online at: http://www.modis-snow-ice.gsfc.nasa.gov/sug.pdf (accessed on 18

August 2006).

ROSEN, P.A., HENSLEY, S., GURROLA, E., ROGEZ, F., CHAN, S., MARTIN, J. and

RODRIGUEZ, E., 2001, SRTM C-band topographic data: quality assessments and

calibration activities. In Geoscience and Remote Sensing Symposium, July 2001.

International Geoscience and Remote Sensing Symposium (IGARSS’01). IEEE 2001

International, Sydney, Australia, pp. 739–741.

SARAF, A.K., FOSTER, J.L., SINGH, P. and TARAFDAR, S., 1999, Passive microwave data for

snow-depth and snow-extent estimations in the Himalayan Mountains. International

Journal of Remote Sensing, 20, pp. 83–95.

SARAF, A.K. and CHOUDHURY, S., 2006, NOAA-HRPT and FY-CHRPT Satellite Earth

Station at IITR. GIS Development, 9, No. 12, pp. 44–46.

SEIDEL, K. and MARTINEC, J., 2004, Remote Sensing in Snow Hydrology; Runoff Modeling,

Effect of Climate Change (Berlin: Springer).

SINGER, F.S. and POPHAM, R.W., 1963, Non-meteorological observations from satellites.

Astronautics and Aerospace Engineering, 1, pp. 89–92.

SINGH, P. and JAIN, S.K., 2002, Snow and glacier melt in the Satluj River at Bhakra Dam in

the Western Himalayan region. Hydrological Sciences Journal, 47, pp. 93–109.

SINGH, P. and JAIN, S.K., 2003, Modelling of stream flow and its components for a large

Himalayan basin with predominant snowmelt yields. Journal of Hydrological

Sciences, 48, pp. 257–276.

SUN, G., RANSON, K.J., KHARUK, V.I. and KOVACS, K., 2003, Validation of surface height

from shuttle radar topography mission using shuttle laser altimeter. Remote Sensing

of Environment, 88, pp. 401–411.

THAPA, K.B., 1993, Estimation of snowmelt runoff in Himalayan catchments incorporating

remote sensing data. In Snow and Glacier Hydrology, International Symposium,

Kathmandu, Nepal, 16–21 November 1992, Proceedings International Association of

Hydrological Sciences, G.J. Young (Ed). IAHS/AISH Publication, no. 218,

pp. 69–74.

UPADHYAY, D.S., MISHRA, D.K., JOHRI, A.P., MISHRA, D.K. and SRIVASTAVA, A.K., 1991,

Use of satellite based information in snowmelt run-off studies. Mausam, 42,

pp. 187–194.

WIKIPEDIA, 2006, Chenab River. Available online at: http://en.wikipedia.org/wiki/

Chenab_River (accessed on 18 August 2006).

ZEBKER, H.A., FARR, T.G., SALAZAR, R.P. and DIXON, T.H., 1994, Mapping the world’s

topography using RADAR interferometry – the Topsat Mission. Proceedings of the

IEEE, 82, pp. 1774–1786.

5878 Accuracy assessment of MODIS, NOAA and IRS data


Accuracy assessment of MODIS, NOAA and IRS data in snow cover mapping under Himalayan conditions

Documents