Princeton post-doc Adam Wolf and his colleagues have developed the Pulse Pod, a low-cost, low-energy device that can collect data from a wide variety of sensors and rapidly make the data available on the internet, even in areas with low internet connectivity. The pods transmit their data to a server via SMS, which makes the system perfect for use in monitoring environmental and ecological variables in regions like Sub-Saharan Africa, where wireless internet connectivity is scarce. Wolf foresees many applications, including crop yield monitoring; if a drought is coming to an agricultural region, up-to-date data can save lives.
This interview has been edited for conciseness.
Travis Korte: What’s wrong with the current system of environmental monitoring and what do the Pulse Pods do to fix that?
Adam Wolf: There’s two ways data is gotten out of Africa. Right now, somebody has a set of rain gauges in Africa that they’d like to use in their research, but essentially some person has to go to those stations once a month, download the data into a text file or a spreadsheet, go back to the office (and often the data hangs up for some period of time, up to a couple of months,) and then at some point the data might be emailed to the researcher. That’s how a lot of environmental data is collected in Africa. Part of that reflects a constraint about the way measurements are taken—they’re almost always taken by a data-logger made by one of two companies that costs some thousands of dollars. But the companies have been extremely slow in having data-loggers that can talk to the internet or cellular, and it’s an extremely expensive way to get data onto the internet from a remote location.
What we’re showing is that a cellular transceiver costs between $15 and $40. A text message costs a few cents, and once the data is on a text message then all it takes is a phone to catch that message and run a command to put it into a database. Once it’s on a database, it’s almost free to store and make that data available. What we’re trying to get at is the critical missing link of having a low-cost way of conveying low-bandwidth data from a remote location onto the internet.
TK: What applications do you see for the pods going forward?
AW: Application one is improving crop yield monitoring and modeling. Number two is forest productivity and mortality. Number three is high resolution hydrological modeling. Number four is satellite measurement and validation. Just to start with the crop yield modeling, my colleagues at Stanford have done a lot of work showing that changes in temperature have large but uncertain impacts on crop productivity, particularly in Sub-Saharan Africa, and particularly in places that are marginal and vulnerable in terms of climate change. We have a picture of a lot of people growing crops, but whether and by how much their yields will decline is unknown. The only view we have into crop productivity is from satellites, but these measurements are very vague; you can see that there’s something green, but you can’t tell that much else about it, either what’s being grown or the yield. So our pods can measure crop productivity in situ and use that to improve crop yield modeling and therefore crop forecasting for highlighting the need for intervention in the case of droughts or other long term changes in climate. There’s also a major effort to deploy our pods in traditional rain gauges at field sites worldwide, to improve the use of the satellite precipitation products that are the backbone of global weather forecasting.
TK: Talk to me about using ground-based sensors like yours to keep remote sensors like satellite imaging systems honest.
AW: From the very beginning, all remote sensing has been coupled to a ground validation effort. In the 1970s, there was a big effort by NASA where they were trying to understand, “We’ve got a satellite that’s looking at the earth, but what are we even looking at?” People went up to Canada to literally measure how tall are the trees, how many leaves do they have, in order to understand what this greenness that the satellite measures refers to on the ground. Always there’s been an effort to ask “what does this mean on the ground?” If this relationship can be established, you can extrapolate widely.
TK: You used some very new software on this project. Can you talk about what you used and your experience with it?
AW: We were faced with the challenge of making the data we have available publicly so that anybody in our community can have their machine query the database using an API and be able to use the data for data assimilation and forecasting. As a scientist I was never faced with the need to create an API for anything until I started on this project. It might seem like a challenge, but not with the tools that are now available. We used [Python-based web development framework] Flask, stored the data on MongoDB (which, unlike MySQL is really easy to query remotely) and [cloud-based application platform] Heroku as a way of testing the internet-based querying of the API. The existence of these tools has made the creation of an easy to use API exceptionally straightforward. That’s an innovation that’s only really happened in the last several years.
TK: Can you speak a bit about the role government research has played in your work?
AW: Organizations like NASA, NOAA and the NSF have played a major role in creating the remote sensing in the first place, and sort of laid the groundwork that has allowed our observations to take place. Even though I work in a university and a lot of the tools we use have been developed in the private sector, there’s a really strong role for government to play in this.