There has been a lot of attention on the drought in the Sao Paulo region, and the lack of potable water resources available to the metropolitan communities in and around Sao Paulo. (Here is a very brief sample of the type of coverage on the drought: The Guardian, The Economist, BBC, and Circle of Blue).

In my own effort to understand the drought in Sao Paulo, I’ve searched for the first thing I typically search for when there’s a story about water scarcity, namely the surface watershed that provides water to Sao Paulo. Surprisingly, I have found scant information online regarding the watershed, either in the form of simple online maps, or easily accessible GIS shapefiles. Now, I could have continued expending lots of time trying to find this information, but I figured it was “quicker”, more interesting, and more educational to simply generate the watershed myself using GIS software.*

Now, before we go any further…. this blogpost will not present an analysis of Sao Paulo’s vulnerability to drought, etc. However, it will provide a brief overview of the sources of Sao Paulo’s surface water. Likewise, I’m making the GIS shapefiles that I create publicly available so that other hydro-nerds who want to do more detailed analysis can get down with their bad selves 🙂

Where to start?

Circle of Blue has provided fantastic ongoing commentary and analysis on the Sao Paulo drought, and I took a look at their very useful infographic which shows the rivers, canals and cities that that are part of the greater Sao Paulo region.


Created by Codi Yeager-Kozacek. Copyright Circle of Blue. (used here with permission)

This information helped to frame how I would delineate the watershed of Sao Paulo, primarily because it indicates that Sao Paulo’s water comes from four separate catchments, rather than a single watershed. This is evidenced by the Sistema Cantareira, which is a connection of canals that connects the Jaguari river to the Cachoeria river, to the Atibaia river, and finally to the Juqueri river. This canal system brings water that would otherwise flow downstream to other metropolitan areas (such as Campinas), to the faucets of Sao Paulo.

Watershed delineation methods

Now, I have to give credit where credit is due. I learned basic watershed delineation using ArcGIS while I studied for my Masters degree at the University of Washington. Rather than provide my own walkthrough, I’ll just direct you to the very excellent notes provided for the course I took, which were drafted by the estimable Dr. Akira Kato (who at the time was merely Mr. Kato). If for some reason these notes have disappeared since I made this blogpost, leave me a comment, and I can pass along what I have. The basics of the method are:

1. Start with a Digital Elevation Model (DEM) of the region; make sure the DEM is “void-filled”

2. Use the Hydrology toolbox in ArcGIS to generate Flow Direction,

3. Use Flow Direction to generate the Flow Accumulation,

4. Along the Flow Accumulation path of interest select a pour-point, or outflow point of the watershed.

5. Using the Flow Direction, Flow Accumulation and Pour Points, use the watershed delineation tool to backtrack the extent of the watershed.

Note that the only starting data that you need are the DEM and an idea of where the watershed outflow point will be (i.e. Pour Points). In this case there are four outflow points. This was straightforward in the cases of Jaguari, Atibaia and Cachoeira (given that the reservoirs are the functional outflow points), but this was less clear for Juqueri. Since the Sao Paulo metropolitan area encompasses nearly the entirety of the Juqueri river before it enters the Rio Tiete, I chose the confluence of the Rio Tiete and the Rio Juqueri as the outflow point for the fourth catchment.

In this way, the figure below reveals four distinct catchments that supply water to Sao Paulo (Click the image for a bigger version).

The Catchments of Sao Paulo


As you can see in the figure, there are four distinct catchments, with the top three flowing from the Northeast to the Southwest. Likewise, both the Jaguari and Cachoeria catchments have their headwaters in the adjacent state of Minas Gerais, with the Jaguari having much of the catchment area upstream of the reservoir within Minas Gerais (…which could lead to interstate water allocation issues… a topic for another post).

In generating these catchments, one of the things that struck me about the river and stream network in the region is that the rivers flow in completely different directions. There are rivers flowing north to Minas Gerais, rivers flowing west-then-east towards Rio de Janeiro, and many rivers flowing west towards northeastern Sao Paulo state, and then towards Peru. This isn’t particularly out of the ordinary, but it does mean that Sao Paulo’s geographical position places it at the upper reaches of its catchment (i.e. the Rio Juqueri/Tiete) and near the headwaters of its other catchments (Jaguari, Cachoeira, and Atibaia). This reality is part of the reason that Sao Paulo may struggle in the future to find new, unappropriated sources of water for its municipal needs.


A criticism of my method for calculating these catchments is that my methods may lead to over-estimating the amount of water flowing out of the Rio Juqueri and Rio Tiete as part of Sao Paulo’s water supply. This means that some of the water included in Sao Paulo’s water supply may flow downstream to the adjacent metropolitan area of Campinas, as well as many other smaller communities.


My eventual goal with this information is to better understand the coupled land-atmosphere dynamics which are linked to the Sao Paulo drought. In particular, I’m interested in the land surface boundaries of Sao Paulo’s catchments, so that I can now backtrack the precipitationshed, to better understand the moisture recycling dynamics associated with Sao Paulo’s surface watershed. For more on precipitationsheds, checkout this earlier blogpost.


This is not meant to drag on, so here is the shapefile for the watershed of the catchments (and some of the other layers you can see in the above map). NOTE, that the files do not contain everything to make the map above, but they do contain the essentials.

Good luck and let me know if you end up using this for anything interesting!

* In the interest of full disclosure, I rarely get to work with GIS, so I didn’t look that hard online to find the maps… I mostly just wanted to play with ArcGIS.