Fig. 2. Clockwise from upper-left: Visible (VIS), shortwave infrared (SIR), thermal infrared (TIR), and 1.38 micrometer cirrus-band (CIR) for the same date/time/domain as shown in Fig. 1. Each of these channels contains unique information that can be exploited in complement to discriminate between cloud and snow cover.
Knowledge of snow cover is important for numerous applications, including search and rescue (e.g., assisting pilots over mountainous terrain), water supply resource monitoring and management, short-term forecasting (e.g., of radiation fog or monitoring the progress of a major winter storm), and recreation. Satellites offer a distinct perspective on snow cover, particularly in areas where there are few ground observing stations. The purpose of this product is to assist satellite imagery analysts in distinguishing snow vs. cloud cover during the daytime hours, and is particularly relevant during the winter months when snowfall occurs across many parts of the United States. This product takes an additional step beyond the Cloud / Snow Discriminator product (which depicts all clouds as yellow) by differentiating between low and high cloud cover.
The Cloud Layers & Snow Cover Discriminator product demonstrates the kind of imagery that will be possible in the GOES-R era at significantly higher temporal resolution. The GOES-R series of satellites will feature the Advanced Baseline Imager (ABI) sensor, which includes channels that are currently unavailable on GOES satellites. The new channels will enable new and/or improved capabilities which can be demonstrated now only via proxy or simulated datasets.
The Cloud Layers & Snow Cover Discriminator product demonstrates the kind of imagery that will be possible in the GOES-R era at significantly higher temporal resolution. The GOES-R series of satellites will feature the Advanced Baseline Imager (ABI) sensor, which includes channels that are currently unavailable on GOES satellites. The new channels will enable new and/or improved capabilities which can be demonstrated now only via proxy or simulated datasets.
Discriminating between cloud and snow cover over complex terrain using conventional visible and infrared satellite imagery can be extremely challenging, even to experts in field of satellite imagery interpretation. At visible-light wavelengths, snow cover and clouds are both highly reflective and therefore difficult to distinguish. At thermal-infrared wavelengths, the temperature of snow cover can be similar to clouds as well. Stronger absorption by snow/ice at shortwave infrared wavelengths helps to distinguish these features from liquid-phase clouds, but some ambiguity remains between snow on the ground and ice-phase clouds. Even when adding the shortwave infrared channels (e.g., 1.6 and 3.9 micron, which provide improved discrimination of snow from clouds due to preferential absorption by snow cover), analyses are not entirely free of cloud/snow ambiguity due to the spectral behavior of certain varieties of cirrus. Additional channels available to the Moderate-resolution Imaging Spectroradiometer (MODIS), and particularly the 1.38 micron “cirrus detection” channel, improve the handling of this problem. Fig. 2 demonstrates how some of these discriminators appear in single-band imagery for a complex scene over Colorado. Combining information from several channels allows us to isolate the snow cover, but doing so via toggling back and forth between the individual channels can be a tedious and inaccurate process. The Cloud Layers & Snow Cover Discriminator product (see Fig. 1; right panel) attempts to eliminate this legwork by combining the complementary information from multiple channels into a single visual aide.
Fig. 3. Example of color gun usage in the Cloud Layers & Snow Cover Discriminator product, showing how the red (R), green (G), and blue (B) guns combine to provide various tonalities to the features of interest (Land, Snow, Low-Level Liquid Water Cloud, Mid-to-High Mixed-Phase/Ice Cloud).
Channels offering unique information about the features of interest are combined to form a single product. The illustration above shows how this information is fed into the R/G/B “color guns” with varying strength of response (denoted by the height of the blue bar, with a full bar implying high strength) for land, snow, and different levels of cloud. In the current configuration of the product, all three of the color guns respond strongly to snow cover. As shown in the composite key to the right of the table, strong contributions from all color guns produces a white tonality (the intersection of the R, G, and B circles in the diagram) for snow. For low clouds, depression of the blue color gun information while maintaining strong red & green contributions produces a yellow tonality (the intersection of only the R and G circles). In this way, the composite technique reduces ambiguities associated with any single piece of information to yield an improved delineation of cloud and snow.
Channels offering unique information about the features of interest are combined to form a single product. The illustration above shows how this information is fed into the R/G/B “color guns” with varying strength of response (denoted by the height of the blue bar, with a full bar implying high strength) for land, snow, and different levels of cloud. In the current configuration of the product, all three of the color guns respond strongly to snow cover. As shown in the composite key to the right of the table, strong contributions from all color guns produces a white tonality (the intersection of the R, G, and B circles in the diagram) for snow. For low clouds, depression of the blue color gun information while maintaining strong red & green contributions produces a yellow tonality (the intersection of only the R and G circles). In this way, the composite technique reduces ambiguities associated with any single piece of information to yield an improved delineation of cloud and snow.
Fig. 4. Examples of True ColorImagery (left) and the Cloud Layers & Snow Cover Discriminator (right) product for a scene over north/central United States containing snow and multi-layered clouds. Clear-sky surfaces are depicted in dark green.
As mentioned above, the Cloud Layers & Snow Cover Discriminator product is actually an extension to the daytime Cloud / Snow Discriminator GOES-R Proving Ground product. It provides additional information on the cloud levels of the scene-in terms of roughly where the clouds reside in the atmospheric column and their liquid/ice/mixed composition. The basic guidelines for interpretation of the colors in this enhancement are as follows: green = land (clear sky, devoid of snow cover), white/bluish-white = snow cover, yellow = low-level liquid-phase clouds, and orange/magenta = mid/high level ice phase clouds.
Fig. 5. Examples of Visible Imagery (left) and the Cloud Layers & Snow Cover Discriminator (right) product for a scene over Colorado containing snow over southern Wyoming. Clear-sky surfaces are depicted in dark green.
The left image of Fig. 4 shows a true color image spanning portions of Montana, Wyoming, North and South Dakota. It is difficult to tell in this image what is cloud and what is snow cover based on the visible imagery alone, as both features appear white. The Cloud Layers & Snow Cover Discriminator image on the right removes much of the ambiguity, and even provides some information on where thin cloud layers overlap snow fields.
Fig. 5 presents another example of the technique for a complex cloud/snow scene over the Rocky Mountain West. The visible image in the left panel of this example provides little insight on the snow distribution, and particularly any differences between the bright regions over southern Wyoming and western New Mexico. In the right panel, the snow/cloud enhancement reveals an extensive snowfield in Wyoming and a low cloud deck in New Mexico. Also noteworthy is a low cloud/fog layer over the southern Wyoming snowfield that is almost completely obscured in the conventional visible imagery. The 1.38-micron near infrared vapor channel on MODIS enables discrimination between cirrus clouds (pink/magenta) and lower tropospheric clouds, giving some indication of regions where multi-layered clouds exist.
A key advantage of the Cloud Layers & Snow Discriminator imagery is the ability to rapidly discriminate between these features without the need to consult several co-located single-channel images. A general advantage of imagery-based snow-cover identification in contrast to digital masks of snow cover is that the latter typically are provided at much coarser spatial resolution and remove the meteorological/terrain context in which the snow cover resides.