189
MODIS-Based Snow Cover Products, Validation, and Hydrologic Applications
snow in vegetated areas was disabled for Aqua imagery, because the use of band 7
resulted in too much false snow detection (Hall et al. 2003).
9.3 MODIS SNOW COVER VALIDATION
The resolution and accuracy of the actual snow cover to which remote sensing products
can represent are critically important for both climate and hydrologic studies, as they
are the main determinants of the products’ usefulness. Numerous studies have been
conducted to evaluate the accuracy of MODIS snow products, either based on comparisons with other satellite-derived products, reanalyses of regional climate simulations, or comparisons with point ground-based (in situ) SD measurements (Table 9.2).
Table 9.2 contains summary information about study region, MODIS product tested,
validation data set, and the overall accuracy (OA) for clear sky conditions when given.
Early validation studies assessing the relative accuracy of MODIS against other
satellite snow products often used a limited number of MODIS images. Hall et al.
(2002b) and Klein and Barnett (2003), for example, found that MODIS tended to
map more snow than the National Operational Hydrologic Remote Sensing Center
(NOHRSC) and Special Sensor Microwave Imager (SSM/I), especially at the beginning of the snow season when the more frequent temporal coverage of MODIS
permitted mapping of shallow snow deposits from fleeting storms. In similar comparisons, Bitner et al. (2002) found that MODIS mapped more discontinuous snow
cover under the forest canopy than the NOHRSC product, particularly when large
areas of discontinuous snow cover occurred in the forested areas of the mountains.
Maurer et al. (2003) reported that, on the average, the MODIS images classified
fewer pixels as cloud and misclassified fewer pixels than did the NOHRSC product.
Numerous later validation studies examined the accuracy of MODIS against in situ
station data. The number of in situ observations applied in the validation varied between
4 (e.g., Tong et al. 2009b) and more than 2000 (Simic et al. 2004; Ault et al. 2006) stations. As is indicated in Table 9.2, nine validation studies were performed in Northern
America (including continental United States, Canada, and Alaska), five in China,
and two in Austria and Turkey. Overall, most of them reported 85%–99% OA during
clear-sky conditions. The MODIS product summary page (MODIS 2010) states that
“the maximum expected errors are 15 percent for forests, 10 percent for mixed agriculture and forest, and 5 percent for other land covers. The maximum monthly errors are
expected to range from 5 percent to 9 percent for North America, and from 5 percent to
10 percent for Eurasia. The maximum aggregated Northern Hemisphere snow mapping
error is estimated to be 7.5 percent. The error is highest, around 9 percent to 10 percent,
when snow covers the Boreal Forest roughly between November and April.” If one
interprets this error as 100-OA used in Table 9.2, these figures are consistent with those
in Table 9.2. As is documented, for example, in Parajka and Blöschl (2006, 2008a), the
OA of MOD10A1 and MYD10A1 over Austria was about 95%. An example of seasonal
evaluation of mapping accuracy is presented in Figure 9.1. The left and right panels
show MODIS (MOD10A1) accuracy to map snow and land, respectively.
Figure 9.1 shows that the mapping accuracy of MODIS over Austria has a clear
seasonal pattern related to the overall percentage of snow coverage with smaller errors
in summer and larger errors in winter. However, for a given percentage of snow cover,
MODIS-Based Snow Cover Products, Validation, and Hydrologic Applications
snow in vegetated areas was disabled for Aqua imagery, because the use of band 7
resulted in too much false snow detection (Hall et al. 2003).
9.3 MODIS SNOW COVER VALIDATION
The resolution and accuracy of the actual snow cover to which remote sensing products
can represent are critically important for both climate and hydrologic studies, as they
are the main determinants of the products’ usefulness. Numerous studies have been
conducted to evaluate the accuracy of MODIS snow products, either based on comparisons with other satellite-derived products, reanalyses of regional climate simulations, or comparisons with point ground-based (in situ) SD measurements (Table 9.2).
Table 9.2 contains summary information about study region, MODIS product tested,
validation data set, and the overall accuracy (OA) for clear sky conditions when given.
Early validation studies assessing the relative accuracy of MODIS against other
satellite snow products often used a limited number of MODIS images. Hall et al.
(2002b) and Klein and Barnett (2003), for example, found that MODIS tended to
map more snow than the National Operational Hydrologic Remote Sensing Center
(NOHRSC) and Special Sensor Microwave Imager (SSM/I), especially at the beginning of the snow season when the more frequent temporal coverage of MODIS
permitted mapping of shallow snow deposits from fleeting storms. In similar comparisons, Bitner et al. (2002) found that MODIS mapped more discontinuous snow
cover under the forest canopy than the NOHRSC product, particularly when large
areas of discontinuous snow cover occurred in the forested areas of the mountains.
Maurer et al. (2003) reported that, on the average, the MODIS images classified
fewer pixels as cloud and misclassified fewer pixels than did the NOHRSC product.
Numerous later validation studies examined the accuracy of MODIS against in situ
station data. The number of in situ observations applied in the validation varied between
4 (e.g., Tong et al. 2009b) and more than 2000 (Simic et al. 2004; Ault et al. 2006) stations. As is indicated in Table 9.2, nine validation studies were performed in Northern
America (including continental United States, Canada, and Alaska), five in China,
and two in Austria and Turkey. Overall, most of them reported 85%–99% OA during
clear-sky conditions. The MODIS product summary page (MODIS 2010) states that
“the maximum expected errors are 15 percent for forests, 10 percent for mixed agriculture and forest, and 5 percent for other land covers. The maximum monthly errors are
expected to range from 5 percent to 9 percent for North America, and from 5 percent to
10 percent for Eurasia. The maximum aggregated Northern Hemisphere snow mapping
error is estimated to be 7.5 percent. The error is highest, around 9 percent to 10 percent,
when snow covers the Boreal Forest roughly between November and April.” If one
interprets this error as 100-OA used in Table 9.2, these figures are consistent with those
in Table 9.2. As is documented, for example, in Parajka and Blöschl (2006, 2008a), the
OA of MOD10A1 and MYD10A1 over Austria was about 95%. An example of seasonal
evaluation of mapping accuracy is presented in Figure 9.1. The left and right panels
show MODIS (MOD10A1) accuracy to map snow and land, respectively.
Figure 9.1 shows that the mapping accuracy of MODIS over Austria has a clear
seasonal pattern related to the overall percentage of snow coverage with smaller errors
in summer and larger errors in winter. However, for a given percentage of snow cover,