November 20

Meeting with Dr. Robert Kennedy

Assistant Professor

Geospatial analysis, remote sensing, modeling, landscape ecology, disturbance dynamics, computational methods.


For my senior project I'm working with NASA to basically help researchers-- Earth Researchers-- use remote sensed images to do their research-- how we are doing that we're not 100% sure yet.

You're working with NASA, NASA is a big place-- there must be some person or something right? [laughs]


Yeah so very specifically it's NASA JPL, and it's the Global Imagery Browser Service Team-- so that's the team that takes Earth Data and makes images out of it and provides it for people to look at it a tile format so they can like, zoom in and stuff.

Okay, cool.


So, my question for you is-- I saw that you did research looking at Landsat Imagery and how it's used and so I'm really interested in how researchers specifically use remote sensed images-- what kind of challenges do they face when using it.

Okay.. that's a big question, haha. I mean my whole research career is built on image processing-- so I can answer that but I'm not sure I'm your audience 'cause I already do that right? Like my whole world is using the imagery to do stuff.

So it's all pretty easy for you is that what you are saying, or?

Oh no! It's never easy [laughs] but I mean it sounded like part of your goal was to encourage people to use imagery in their work, and that's all my work is--

… and make it easier, yeah

Well you know-- gosh, where to start. So I have traditionally done all of the processing and you know-- all of the atmospheric cloud screening, radiometric normalization, geometric correction, all that stuff, but increasingly that's done, and so that's-- those are key things to have available to have good atmospheric correction and cloud masking, but you can get those from the USGS for Landsat now so that's not an issue and the other part is is bringing the algorithms to the data-- really the thing that has been the biggest issue for all of us in the community-- I'm on the Landsat Science Team and we talk about how access to Landsat data [should] work.

and so-- I'm actually also on the LP DAAC Advisory Committee, so we talk about-- that's another place that serves up imagery for the community-- they do MODIS stuff primarily but a lots of other things. [Checks for Student waiting to talk to him as this is his office hour]

So anyway, you know the real change is that a lot of the questions people are interested in asking-- and the community is pushed forward to make people's expectations about answering-- require LOTS and lots of imagery. Like, the stuff I do, I have 100 Terabytes of imagery sitting in a server here that I had to pull down, right? So-- that's a lot to manage and it requires a certain level of expertise and interest in doing the computer management and all that stuff which not everyone has, so a real barrier to broader use is-- having to managing all that data yourself if you're not already an expert in doing that.

So a lot of the prior barriers were the processing of the imagery, you know-- the atmospheric correction, the cloud masking, the other things I mentioned-- MODIS really changed a lot of the expectations in that, Landsat has pick that up, Sentinel is going to be having that-- that European Satellite soon. And so that's great, now we have all these images and the next barrier is basically how to help people manage that.

And so the answer that a lot of people see is: insteading of having somebody take terabytes of data from the USGS over there and bring it over here and do something over here is to set up a system where people can actually process it on the archive and so you're not moving terabytes of data back and forth, you're moving the results, but not-- like there are hundreds of copies of the exact same image sitting around the world that people have pulled down for whatever place, right?-- that's redundant.

So that's that's the big change-- so there's several changes, so NASA JPL-- so I don't know actually about what you're describing so that's good for me to learn about, but NASA Ames has NEX, the "NASA Earth Exchange" which is-- they also have the entire Landsat archive there and that is a place where you can plug into their super computers and access all that-- so I don't know how that interacts with the JPL folks.


So you process the images remotely, what are you trying to get back, just the answers to your research questions, or are you looking for a specific space and time or what kind of answers are you looking for for those questions?

Okay, so that's a good question-- so I am interested in terrestrial systems, Land Cover change-- what people and natural processes are doing to the observable surface of the earth, and Landsat-- primarily in the Landsat era and the satellite era.

So that's looking at disturbance, people, or you know, or fires, or landslides-- whatever changing what happens, and then what happens after that, what precedes it sometimes if there are biotic things that precede anthropogenic things, et cetera-- so there are bugs that kill the forests that precede a salvage logger that precedes a fire, is there a fire that precedes a sal… like understanding how those things happen, what the processes are that drive them so that eventually we can model them better and you know-- the whole goal is to better understand how what we're doing affects the planet so we better-- have a brighter, greener future.


So you are saying that it's more like-- possibly interactive, like you're not sure when you're looking at the data-- you might not have a specific-- you might not know why this is happening but you'd kind of know that--

Oh, I see, well I'm just trying to paint a broad picture-- we do have specific questions for project of course, right, but that's for the overarching thing-- so one project I have is looking at Carbon, and so there is specific questions about forest Carbon-- the first one is: just documentation.

Documenting every pixel of forest carbon over time: how much it is, if it go away, and if it goes away, who caused it to go away and where. That hasn't been done, it's a very fundamental-- you know, observation thing but that's the starting point. You have to understand what's happening first, right, so, that project is looking at the-- essentially the fade of carbon in Oregon, Washington, California-- and trying to understand and again just document and through the documentation hopefully come up with some understanding of what's happening with policy drivers or whatever.

Land cover change is another one so, biomass land cover descriptors are like forests, non-forest, you know agg, urban, that kind of stuff and watching how those-- again, documentation. Watching how those change over time is sort of a critical first step for a looot of questions about what's actually happening on the landscape. So, you know, the first thing there is again, characterizing who is doing that and what kind of change is happening, and then that leads into questions about habitat and urbanization and what people do.


There are-- Other things that-- Just to sort of I guess final answer your question so-- I also work with national parks, and national marine fishery service, and again a lot of those projects are trying to test specific hypotheses about how much land cover change matters. So, for example for salmon, what they want eventually is to have information about what is happening near streams-- salmon bearing streams-- on a yearly time step that's just very descriptive about what is change is happening because that affects the inputs to the steam-- that affect the salmon.


So a lot of like temporal type stuff, not necessary spatial? Looking at one place over time or looking at--?

Well, it's spatial and temporal. So the spatial part is because we are looking at large areas and we're looking at-- you can look at anything adjacency or whatever--- like with the steams-- the temporal is yeah-- I mean I use-- the reason I have so much imagery is because I'm looking at hundreds of images at the same place through time and then applying time series algorithms to them to extract interesting information.


Alright cool, well that was the most pressing questions so I'll leave you to, you know help your other students and stuff [Okay!] but thanks a lot for meeting with me [Yup!] I really appreciate it, it's been really helpful.

Ok. Sure, yeah, yeah--

[These notes were taken as I was leaving while the recorder was off, not verbatim]

[Showing prototypes]


My advice to you is don't do something that's already been done.

- For example there's this LP DAAC service that nobody uses because people can already get the data elsewhere

- I use Google Earth for example because I can get the imagery and it provides a service on top of it

- If you provide a service on top of the data that's useful, people will use your service.

- Don't do something that's already been done