Studying water quality with satellites and public data

(piano music) – [Matt Ross] In this
paper, we built a data set that we call Aqua Set. The goal of Aqua Set is
to make remote sensing of water quality as easy
as it has been to do remote sensing of terrestrial variables. The way we are hoping to do that, is by publishing a data
set where we match up Landsat imagery, which
Landsat’s a satellite that’s been flying around
the earth since the ’70s. So we match imagery from Landsat, which contains color information, with field estimates of water quality. And those field estimates come from a federal water quality portal, which integrates state, municipal, and federal water quality
efforts led by people like the USGS, and the EPA. When you combine those two data sets, what you get is color
information from a satellite, so image information. And you also get the
exact, sort of quantity of a optically relevant metric. So something that gives
color to the water. So that would be something
like sediment or carbon. And when you have those
two things together, you can use that data
to predict water quality using Landsat imagery alone. So the goal of publishing this data set is to get more people to do remote sensing of water quality in more places and at larger scales, both
spatially and temporally because the data goes
from 1984 to the present, and it’s the largest such
data set ever assembled, with over 600,000 match ups. Which we can use to use machine learning and other novel approaches
to predict water quality at the continental scale. And so we hope by
publishing this data set, everyone can work on the same thing and we can really make
progress as a whole field, rather than having individual
little islands of progress that don’t necessarily meet up.

Tags:, ,

Add a Comment

Your email address will not be published. Required fields are marked *