This is the lower Mangoky river, southwest Madagascar, as seen with the EVI2 vegetation index as derived from data collected by NASA’s MODIS sensors in 2011 and 2012. To the south of the river’s mouth, you can see Lake Ihotry shrink in 2012. It’s been on a downward trend for a few years, and at the end of 2012’s dry season it almost disappeared for a few weeks. North of the delta, some dark areas appear – deforestation. The largest of them (just south of the second largest river in our view, Maintapaka) is about 6.5 km (4 mi) on a side, or roughly half the area of Manhattan.

If you look closely, you may also see the course of the Mangoky river shift very slightly around an island in its delta. The Mangoky is full of soil that’s washed off deforested upstream hillsides, and it’s constantly filling its streambed with silt. In fact, in maps only a few decades old, you can find that it used to flow due northwest out of its delta, many miles north of today’s course.

We are trying to make it easier to quickly communicate complex environmental issues. This post is an early look at how we’re thinking about investing in our analytical stack at MapBox.

Developing an analytical stack

Producing clear satellite imagery analysis is traditionally very complicated. First there’s the trouble of accessing large data sets, and then there’s the difficulty of working with tools (both software and hardware) with the power to crank through big data. The power of a clear, data-driven story on a map is huge, but for a long time it simply too hard. We’re turning our cloud expertise into tools to help make analysis easier. Some of this comes naturally out of our cloudless imagery work, one of the most common patterns we notice is – unfortunately – deforestation. Sometimes it’s very clear, as in the Amazon rainforest in the state of Pará, Brazil:

Pará Cloudless Atlas with OpenStreetMap borders and labels.

Small roads branching to the north and south off the Trans-Amazonian Highway around Uruará are each surrounded by clearings. This zipper or herringbone pattern is clear in much of the Amazon basin, and in the Congo basin of central Africa. Luckily, space-based analysis by organizations like NASA, the World Resources Institute, Brazil’s National Institute for Space Research (INPE), and InfoAmazonia (who we’ve been proud to work with!) has supported regulatory changes that are slowing this type of deforestation in most areas.

As I worked on some of our tools and processes, I looked for regions where deforestation would be harder to spot with the unaided eye. One of the most interesting is the island of Madagascar, which has unusual vegetation and an especially serious environmental crisis. Here’s an overview of Madagascar and neighboring islands in the visible spectrum:

Madagascar and environs

Cloudless Atlas with OpenStreetMap borders and labels.

As you might guess from the mix of colors, there’s a lot of environmental variation. The World Wildlife Fund lists four major ecosystems on the island: dry deciduous forests along the west and north, lowland forests on the east coast, succulent woodlands in the southwest, and subhumid forests in the central uplands. It can be hard to tell what you’re looking at in the complicated patterns of topography, settlement, and land use – yellow might be healthy in a scrubland, for example, but a sign of serious problems in a forest.

To get a more quantitative view of the situation, I used EVI2 again, the vegetation index I used last week to look at the effects of forest fires. Within limits, it’s a powerful indicator of relative plant thickness. Here’s Madagascar’s EVI2 in every quarter from 2009 through 2012:


Madagascar’s rainy season is November through April, which you can see reflected in higher growth in the first half of the each year, when plants have more water. Reading down the columns, there’s year-on-year variation from weather and climate. These can interact dangerously with a complicated political situation: after a crisis in 2009, itself driven partly by conflicts over land use, foreign aid suspensions combined with a drought led to hunger. A period of lax environmental enforcement allowed a boom in illegal logging of rosewood and other trees in high demand for the international luxury trade. Such big events are often visible at the regional level with techniques like difference overlays, but for me it’s sometimes the details that are most interesting. That animation again:


You’ll see more of this kind of work on Twitter (@MapBox) in the coming weeks. Feel free to talk to me (@vruba) or Chris (@hrwgc) on the satellite team to get into more specifics.