Do satellites sometimes see ice where there isn’t any?

Readers often ask us, “Why does your sea ice map show sea ice where there is none?” Sometimes our Daily Sea Ice Extent images show sea ice in a particular area, but when readers who live in those areas look out their windows, they see open water—or they even may see ice where our maps show open water. This occurs most frequently along rivers or near coasts. Why does this happen?

Ups and downs of passive microwave

Photograph of Qaanaaq, a small town on the Greenland coast

The rugged coast near Qaanaaq, Greenland, illustrates the challenge to satellite sensors, which must distinguish between land and ocean signals within the same image. Credit: Andy Mahoney

These discrepancies are most often caused by the resolution of the satellite sensor. NSIDC relies on passive microwave sensors to compile daily sea ice maps. These sensors have the advantage of being able to see through the Arctic’s cloudy weather and capture surface data even during long, dark winters, making them ideal for tracking sea ice. The disadvantage, however, is that passive microwave sensors often have low spatial resolution. The sensors collect data in “footprints” that are up to 50 to 70 kilometers (31 to 44 miles) in diameter.

Out on the open ocean, for instance, most of these large footprints include only water; inland, most include only land. But along the coasts, footprints often contain both land and water. “The main issue tends to be a land-spillover issue,” said NSIDC scientist Walt Meier. “The land ends at a definitive location, and in our daily maps we mark an area as land or not-land. So there is a discrete boundary. However, the sensor measures areas that include some land and some ocean.”

Is it water? Is it land?

It would be very time-consuming, if not impossible for scientists to pore over the multitude of individual coastal images to determine whether they are mostly land or mostly water. So, they develop algorithms to help classify when the sensor sees land or water, and then further distinguish between water and sea ice. Meier said, “We measure concentration, which detects the fraction of sea ice, with the rest as water. So the mix between sea ice and ocean is accounted for in the data.”

Data graph illustrating the discrepancy in how satellite sensors might see land and ocean

Land ends at a definitive location, forming a discrete boundary between land and ocean, seen as the “step” from land to ocean in the image. Problems arise when the sensor measures an area that includes some land and some ocean, seen as the smooth dashed curve in the image. Although scientists can apply algorithms and masks that help distinguish land and ocean in the data, some pixels may still show sea ice where there is none. Credit: NSIDC

That is not entirely foolproof, however. Meier said, “Unless the boundary between land and ocean falls almost right along the boundary of the those footprints, the sensor will see a mixture of ocean and land. That sometimes fools the algorithm into thinking there is sea ice present.” If the senor detects 15 percent ice in an area, it is considered to be ice covered. This is why readers who live along rivers or near coasts occasionally see open water in places where our maps show sea ice.

Scientists also alleviate these discrepancies by applying masks to the data, which help define land edges so that sea ice does not appear inland, and also help screen out sea ice that is far from the ice edges. NSIDC updates and applies masks on a monthly basis. Meier said, “For example, the ice that was previously seen along the coast of the Gulf of St. Lawrence is now gone in the July images, because we switched to a new mask.”

Meier and other scientists can also use high-resolution sensors when they need to zoom in and verify where sea ice is—or is not. One of these options is the Multisensor Analyzed Sea Ice Extent (MASIE) project, which includes data with resolution as fine as 4 square kilometers (1.5 square miles). “For knowing where there is ice now in specific locations, it is a more useful resource than passive microwave.” However, because MASIE includes a mix of sources and is subject to human analysis, it does not provide the consistent record that passive microwave sensors do.

Even though small amounts of sea ice sometimes still appear in unrealistic places in the passive microwave data that NSIDC uses to compile the Daily Sea Ice Extent images, those errors are small and they do not significantly affect the overall estimates of Arctic sea ice extent. In addition, the passive microwave data record remains valuable because it has generated a consistent time series since 1979, providing scientists with a long-term view of sea ice that higher-resolution satellites cannot yet offer.

References

Arctic Sea Ice News & Analysis

Multisensor Analyzed Sea Ice Extent (MASIE)

Comments are closed.