How do you structure a parametric using a satellite?

“Of course, it’s important when talking about satellite data to acknowledge that satellite data doesn’t do everything perfectly,” said the head of commercial for San Francisco headquartered Floodbase. “In fact, in many cases, it does things terribly.”

However, to get around that issue, Lacovara said a critical initial step is to conduct a validation or testing process. This process, he said, ensures that the data being used to structure a parametric is actually responsive to historical events in the coverage area.

Lacovara’s firm integrates satellite-based observations with hydrological and meteorological data and other ground observations to generate “near real-time flood maps.” According to its website, Floodbase regards itself as a pioneer in the field of satellite flood tracking.

“We furnish carriers and brokers with the data that’s necessary to both price parametric contracts and to trigger them,” said Lacovara, whose previous experience includes head of parametric in North America for the global broker Marsh.

A new satellite based flood parametric

At the WRMA event, Lacovara said during February, Floodbase rolled out a satellite based parametric offering in the US and is looking to do the same in Australia.

“What we’re doing in the US – and we’ll be building for Australia in the coming months – is something that is using a machine learning algorithm that is that is interpolating between direct observations,” he said.

The firm also provides its flood intelligence to governments in the developing world, NGOs, and the United Nations (UN).

Like many parametric offerings, much of the data used, including the satellite images, is publicly available. Lacovara showed flood images taken by NASA’s MODIS satellite. MODIS is a purpose build flood mapping satellite:

“The Flood Product is a daily, near-global, ~250 m resolution product showing flood and surface water detected from the twice-daily overpass of the MODIS optical sensors,” says NASA’s Earth Data website.

The low resolution challenge

One downside of this satellite, said Lacovara, is the relatively low resolution of its flood images.

“The advantage that MODIS brings is twice daily overpasses, so we can see flooding essentially anywhere on the planet and we can make those observations twice daily,” he said. “The trade that we make, and there’s always a trade with satellite data as there is with any data, is that the resolution is somewhat lower.”

Lacovara said a pixel of 250 meters is not in itself a fine enough resolution to build a useful parametric.

“There are some advantages and disadvantages to structuring around this,” he said.

Guinea pig example: Flooding in Echuca

Lacovara showed MODIS satellite images taken during the January flooding around the Victorian town of Echuca. One image clearly showed clusters of dark green pixels which Lacovara said indicated considerable inundations.

“So how do we turn that into something that’s actually useful to build a parametric around?” he said.

Lacovara showed a rainfall data chart for the area taken from NASA’s Global Precipitation Measurement Mission (GPM), that uses satellites to measure global rain and snowfall.

“What we’re looking for here is whether rainfall is an adequate proxy for known historical flood events,” he said.

Lacovara chose three known historical flood events in the area, including the 2011 flood. The GPM data didn’t spike for at least one of these events.

“You can see that in spite of knowing that there was a very severe flood in 2011, there’s actually not a consistent spike in the rainfall data, which tells us that in this particular case, rainfall was actually not a good proxy for the flooding that was observed,” he said.

However, another NASA satellite data source, or index, showing fractional flooded area proved to be more reliable for the Echuca area. This index, said Lacovara, described the amount of water that’s visible in each image pixel.

“You can see all of a sudden that you have an enormous signal that’s quite visible, so a very large spike in the amount of fractional flooded area for that extreme event in 2011,” he said. “We also see meaningful spikes [for at least three recent major flood events], which tells us that actually, this works pretty well for building an index for flood in this particular area.”

During this validation process, Floodbase also uses its algorithms to process older, historical satellite images as far back as 20 years – when NASA started collecting the data.

“This allows us to generate these flood maps that show what we call recurrence or basically return period of flooding in different areas,” said Lacovara. “So we can go back and look at what actual events looked like in the satellite record.”

He said stream gauges can also be very good at providing local information that is accurate and are often used. However, Lacovara said one advantage of using satellites to construct a parametric program, rather than stream gauges, is that gauges tend to fail “as do many other ground based sensors.”

“So using satellites in this particular context gives a lot of confidence and we can actually capture those events and we can build an index that will respond to them,” he said.

During last week’s conference, Floodbase announced that it was selected by the US Federal Emergency Management Agency (FEMA), to provide a “near-real time flood intelligence system.”

What do you think of parametric flood offerings? Please tell us below.