Eyes on Earth Episode 126 – Annual NLCD

[00:04] Hello everyone, and welcome to another episode of Eyes on Earth, a podcast produced at the USGS Eros Center. Our podcast focuses on our ever-changing planet and on the people here at Eros and across the globe who use remote sensing to monitor and study the health of Earth. My name is Tom Adamson. [00:30] nlcd map [00:31] It's a striking large map of the, and people enjoy pointing out where they are from or where they're heading on their road trip. [00:40] And we like to brag about how it's an Eros product.

[00:44] The National Land Cover Database has been around for a while, and we have a few Eyes on Earth episodes about it already. But there is now a new release, and it's being called the Next Generation of USGS Land Cover Mapping. [01:00] It's called annual NLCD. [01:03] And we're going to talk about what's new. [01:05] and what people will do with it. [01:06] So today we're talking with John DeWitts and Jess Brown, who are leading the effort on annual NLCD. [01:14] John, let's start with you. Can you start with what is land cover?

[01:18] Land cover is something that describes what's going on on the land surface. Most people's experience with maps, especially an older generation, starts with a map you used to figure out where you were going. [01:32] and it may seem that a land cover map and a [01:36] state map are the same thing, but they're very different [01:39] Those maps that give you direction and show roads are just a small part of what we do on [01:45] this land cover product. [01:47] Land cover maps themselves are really used to [01:50] categorize and to allow a computer to analyze [01:55] where things are [01:57] how much of them are there, [01:59] and what's happening on the landscape.

[02:01] That really didn't start happening until [02:04] 30 years ago. So when people say maps [02:08] Many people still think of a map as a very basic thing used for navigation. [02:13] the big difference between [02:15] Old paper map that some of us have seen in an LCD is that it's a digital product. [02:20] Humans are very good at looking at a picture and understanding visually what's going on, but we're very poor. [02:26] at doing large number calculations in our head. [02:31] NLCD is digital. It allows us to very quickly use a computer [02:35] to understand how much forest is in an area, [02:38] How much developed is in an area?

[02:40] Visually we can see that. [02:42] but quantifying it and being able to put that into actionable things for policy around the nation is really why we make the map. That brings us to NLCD, the National Land Covered Database. Jess, give us the introduction to NLCD. So the National Land Covered Database is a representation of the Earth's surface into something that can be used [03:07] further in science. [03:09] We try to understand a changing planet, and we need this information to feed into Lali's other purposes. But it is much more than just a map.

[03:19] In the National Land Cover Database, we express land cover in 16 different categories. These 16 categories [03:27] that have been traditionally used. [03:30] divide the landscape up or categorize the landscape into water. [03:35] Ice, snow, there are four classes characterizing developed, the developed surface. [03:41] which is really like our urban landscapes [03:44] three types of forest, [03:46] They're in land. [03:48] shrub, herbaceous, [03:50] to agricultural land cover types, pasture hay and cultivated crops. [03:56] and then two wetlands types. [03:58] woody wetlands and herbaceous wetlands. [04:00] Okay. [04:01] And clearly the landscape is...

[04:04] complicated. The landscape is more complicated than these 16 classes, so this is a simplification [04:10] Yeah, that's a good way to put it, Tom. It's definitely a way to simplify the landscape. So let's talk about these fractional products. [04:19] even though these land cover classes are [04:22] something of a simplification of describing land cover. There is this fractional product where you can get a little bit more detail about something like impervious surface. That's one of them, right? How does that one work? [04:34] When we started NLCD 2001, we saw there was a need for

[04:38] additional discrimination for some of the land cover classes. There are many things in the landscape [04:45] that may only make up a small fraction of the [04:47] pixel. [04:48] but they have an outsized effect on everything surrounding it. [04:52] developed as one of those. [04:55] We started this as a database concept where developed and forest canopy [05:01] inform the land cover. So [05:03] When I initially wrote up the land cover definitions, we worked to figure out how we would split those developed classes from this fractional impervious surface. [05:13] impervious surface has a very outsized effect on the landscape.

It affects drainage, it affects [05:21] air quality. It really shows you where people are and as we know the more people there are in a location the more effects we have on the local vegetation. [05:31] So as we went through, we split those classes into four develop categories, but those four classes [05:38] are derived directly from that percentage product, that 0 to 100%. [05:42] in that pixel if it's only 15% [05:47] impervious surface, it's great to know that because there can also be [05:52] 50% canopy, 20% grass, and those are things that we can understand with a simple [05:59] it's this type of developed in the land cover.

[06:02] Well, if I look out my front window and I see a street, [06:06] that's an impervious surface but there's also trees overhanging that street [06:10] there's also a little grass strip next to the street. Is that the kind of thing you're talking about? That's exactly right. Yeah. We also have another product, percent forest canopy layer. [06:21] That also works hand-in-hand with land cover and with impervious surface to help a user figure out those percentages because developed is the dominant feature on the landscape. Even if a pixel is only 15% developed and 50% tree canopy, developed is what comes in the land cover.

But because it's a database, you can use these other pieces to understand more of what's happening in that specific location. [06:51] are just this or that. Agricultural fields, you know, they're not [06:56] 30% agriculture in a pixel. [06:59] a whole field is agriculture. [07:01] but developed and canopy, those things [07:05] They can be very small pieces of each pixel, and that's what we really wanted our users to be able to tease out and understand. [07:12] There's also places that analytically don't make sense. Atlanta, for instance, can have say 50 to 70% developed surface [07:25] underneath that forest canopy and also have 70% canopy covered.

We know percentages only go to 100 and understanding that relationship [07:34] is really important for urban foresters, local land managers, urban development, things like that. Let's get into the product itself. What does annual NLCD provide? Like, what should users expect when they open the box? [07:50] The annual NLCD is a new product suite. [07:53] that we're releasing... [07:55] covering 39 years of [07:57] of history for this country. [08:00] So this database consists of a product suite [08:04] with six products and we're [08:06] in those products were describing [08:08] many different characteristics [08:11] related to the land surface.

Land cover, land cover change, [08:17] confidence product called Land Cover Confidence [08:21] fractional impervious surface [08:24] an impervious descriptor, [08:26] and spectral change day of year. [08:29] So these six products together [08:31] comprise the database. [08:33] So we're pretty aware that [08:35] the land cover product is the most popular. [08:38] people are comfortable using that. They're very interested in [08:42] in using that in their modeling or studies. But we've also included land cover change, which is an annual change. [08:52] indicator [08:54] between two adjacent years of land cover [08:57] The 1986 product represents change from '85.

[09:01] to 86. [09:02] You know, the users could calculate that themselves from the annual land cover, but we did it for them. [09:08] and basically it provides numerically [09:12] the class that was in the prior year and the class that is in the current year. [09:17] The majority of the landscape is actually very stable. [09:20] So what we see when we look across the whole record is about 80% of the pixels across the country. [09:27] didn't change. [09:29] But there's... [09:30] so many changes that have happened and have occurred during that 39-year period.

And this annual land-covered change product helps [09:38] characterize that more specifically through time. [09:41] LandCover Confidence really comes out of our methodology directly. [09:46] And that's a probability value for the land cover class that was derived in our classification methodology. [09:53] gives people an idea of how confident the [09:56] algorithm was [09:58] that this land cover call is correct. And some of the things that John was talking about earlier [10:03] about how [10:04] the pixels aren't pure. So Landsat pixels do not represent [10:09] one land cover type often. There's often heterogeneity within that [10:13] that 30 by 30 meter pixel.

[10:15] And also there are situations where [10:19] the spectral reflectance that Landsat is [10:22] measuring [10:23] is very similar for two different land cover types. For example, shrub and grass might have very similar [10:31] signals in the spectral reflectance that Landsat collects. That's the basis for our classification. So land cover confidence might tell a user [10:40] how well the algorithm was able to identify this particular land cover type. [10:45] So the final product, the spectral change day of year, comes out of our change detection algorithm, which is the shorthand for that is CCD.

[10:53] That is a change detection algorithm that uses a harmonic modeling approach. [11:00] and we are really basically running all of this Landsat data through time. [11:05] in order to detect when an abrupt change has happened. [11:09] Now not all land cover change is abrupt, some is more gradual, [11:12] So this actually helps us pull out [11:17] the day of year when the Landsat surface reflectance [11:21] showed a change. Basically, an anomaly is detected through that harmonic modeling approach. What makes Landsat uniquely suited to being the main input into NLCD?

So I always like to say that Landsat's superpower is time travel. The beauty of the Landsat observing system is that it started [11:43] in the 1970s. Now we're using data from the 1980s forward. There's some specific reasons based on sensor characteristics why we do that. We start with the Landsat 4 record. [11:56] in 1982 and we start that change detection effort, that change detection algorithm, [12:02] in '82. By the time '85 comes around, which is really the first year of the record of land cover, [12:08] We have a pretty decent idea.

[12:10] The harmonic model process then can do a better job of determining a change has happened. And we also have fewer... [12:20] observations at the early part of the record [12:22] Landsat was young, there were other challenges to collecting and saving that data. [12:28] so that it's part of our archive. But USGS has done a phenomenal job of processing the entire [12:35] Landsat Archives so it can be used for this kind of purpose. Even though we have [12:39] different [12:40] instruments, [12:42] throughout that record [12:43] the way that we calculate [12:46] surface reflectance, for example, [12:48] helps us be pretty confident that when we are seeing a change, [12:53] in time it's truly a change on the land surface.

[12:57] rather than an atmospheric anomaly. [13:01] or something to do with geometric registration of the data. [13:06] Thank you. [13:06] So what Landsat [13:08] provides [13:09] us in making the National Vancouver Database is that [13:14] long record of wall-to-wall coverage [13:19] of our country. [13:21] So we can now utilize that record and have to make a time series of land cover. [13:28] That is one of the unique properties of the new annual land cover database. [13:33] this 30 meter record [13:35] It's a one of a kind resource. [13:37] There's nothing else that exists like Landsat for studying our Earth.

[13:41] What is the value in having annual NLCD versus what it was before, something like every two to three years? [13:50] So one of the predecessors to this effort [13:53] the LC map. [13:54] project, land change monitoring assessment and projection, was really the first effort that USGS [14:00] did to create [14:03] an annual periodicity of land cover [14:06] across the, [14:08] That project ended in order for us to merge that capability with [14:13] legacy NLCD, which was more on the two to three year cadence. [14:17] Mm-hmm. [14:18] But LCMAP did not have the detail of land cover classes.

[14:24] really the user community was demanding [14:27] all the detail that they can get. [14:29] and the high periodicity [14:31] And a third component was we'd like the longer record. [14:35] So, [14:36] we're trying to satisfy a user community that seems to have evolving and [14:41] than ever increasing needs for more complex information. [14:45] We have a user community that is very interested in performing carbon sequestration modeling. They want to know what is the carbon source and sinks, [14:56] across the country every year. [14:58] They are thrilled to get annual periodicity.

We have [15:03] users [15:04] in the water modeling community, and John mentioned this to do with impervious surface, so if you're going to utilize [15:12] land cover to make a model of flooding, for example, you want to know [15:18] within your watershed [15:21] what the amount of impervious surfaces, what is the amount of developed land cover versus other types of land cover that would absorb the water. And that does change over time. [15:30] So... [15:31] It just helps some of these models [15:34] achieve more realistic representation of [15:38] of the Earth's surface to improve their model accuracy.

[15:41] How do you know that the data is accurate? [15:44] So at our center, we have a huge long history all the way back to the beginning of NLCD of [15:51] doing the best job we could at assessing the accuracy of our products. This has evolved over time. [15:58] And it is one of the most difficult things, I think, that we do, although we do a lot of difficult things. [16:06] involves collecting 10,000 reference plots across the country that our team [16:12] members of our team. [16:15] interpret [16:16] and these locations, these specific plots, [16:20] are used to do an independent accuracy assessment.

[16:23] of these products. [16:25] We're in the middle of collecting all of those. It takes a large amount of time and effort to do that, and we almost sequester those folks. Like, we can't talk to them because we want those... [16:36] reference data to be truly independent and not to skew [16:39] our maps. So we have over the past nine months that team has collected 5,000 of the 10. And then over the next nine months we'll collect more to better [16:50] cover rare classes and land cover change, which is basically rare.

[16:55] And then we'll put together an accuracy assessment and publish it like we have done for many, many years back to the beginning of NLCD. [17:03] It sounds like you already know how to do this. [17:06] We do, but it's still a pretty big job. Yeah. So we are able to, at this point, take a peek into that with the reference data that we've collected. [17:16] And we're not ready to publish numbers, but we're pretty happy with the results. But then I'll say we're never happy with the results because we always want to do better.

And there's always things that we can improve upon. [17:30] I... [17:31] So... [17:32] exactly agree that we know how to do this. This is the first time we're doing land cover accuracy assessment with this complexity. The first time [17:42] It's been done yearly for 40 years with [17:46] 16 classes of land cover assessing every single year. So one point that's been made is even with the 5,000 points, that's multiplied by 40 years. They're assessing every single date. So that's a lot more than 5,000 points. Historically, NLCD is only... [18:05] done that for two or three dates, which is a lot lower level of effort.

So we're transferring the knowledge from doing that 16 class accuracy assessment for just a couple years across 40 years on the landscape, [18:23] where we don't have high res imagery for all of the dates like we've had in the past. We're going back into the 80s where there's only the Landsat record and having to rely on assessors. [18:36] experience understanding what Landsat looks like and what the class should be respectively. Let me see if I understand what one of these plots is like. What is one of those 10,000 plots? [18:48] It's a Landsat pixel.

[18:49] A plot is a 30 meter pixel. [18:51] Yes. So are you just looking at the imagery? You're kind of doing this remotely? Yes. [18:57] so you're looking at the imagery the imagery gives you perspective on the larger landscape [19:03] it's a pixel in the middle of a crop field. That's [19:07] easier to understand with the context. [19:10] the same as if it's a pixel in the middle of Lake Erie, [19:13] The context lets you know you're likely going to get that water pixel right. [19:17] that water pixel [19:19] is say the little stock pond on the way to work here may only be one or two pixels [19:26] that's a lot harder to get the context for what's going on in the landscape.

[19:29] especially if you don't have high res imagery. So you're looking at [19:33] the Landsat image, but you're also looking for some other source imagery to compare it to? Exactly. If it exists. Sometimes there are mid-90s digital ortho quarter quads that we can use. Those are black and white. [19:50] It goes from one in the mid-90s to usually one in the 2000s. [19:56] in the mid 2000s and then 2010 you start getting [20:01] more frequent updates. [20:02] NAIP, National Agriculture Imagery Program comes to mind? Is that something that's used to?

[20:08] we used [20:09] Everything that we can do. Whatever's out there. Okay. It's easy for me to say, well, why don't you just look at Google Earth and compare it and figure it out. [20:17] We use Google Earth. Okay. We use Google Earth extensively for... [20:22] accuracy assessment. [20:24] But again, Google Earth can only [20:26] provide imagery that was there in history. So... [20:30] You can look in Google Earth 1985 [20:32] what you'll find is a Landsat pixel instead of high-res imagery. Yeah, sure thing. [20:37] Maybe one of the important things is how this is all evolution, like science [20:43] our capabilities are evolving, [20:47] our [20:47] Imagery is evolving.

Our tools are evolving. [20:52] all of this evolves over time and we're trying to [20:54] in our position stay as [20:57] close to the cutting edge as we can, but at the same time, John has said many times, [21:02] we can't be on the cutting edge because we need to be repeatable, [21:06] We need to be transparent. We need to have systems that can be run again and again to produce the data [21:13] reliably year after year. [21:15] So we have to strike this balance between [21:19] trying to [21:19] to do the best we can and use new methods.

[21:24] updated methods. [21:25] but yet be consistent. I would add to that, I would say we're touching the cutting edge, but we can't be bleeding edge. Yeah. [21:34] It's very delicate, it sounds like. We might get cut. You still might. You still might. [21:39] Can you name a few examples of benefits of annual NLCD? What areas is it especially useful for? [21:46] for developed areas, for instance. [21:49] understanding where urban population [21:52] growth has taken place, especially [21:56] and large urban centers. [21:59] urban sprawl, the spread of infrastructure, all of those things have a profound effect on the local landscape.

[22:06] and [22:07] that has not been able to be quantified in a way where you could [22:11] find change over time. This is the first... [22:15] land cover at this scale to go from 85 to current and have [22:21] comparable change through that time frame. [22:23] one of the projects in [22:25] also in this building has been delving into [22:29] quantifying and modeling urban heat island effects. [22:33] And so this data is [22:36] revolutionary for them. Not only do we have a 39-year record [22:40] we have this detail of [22:42] land cover and impervious [22:44] surface, fractional impervious surface over these cities that will help them [22:49] be able to model [22:51] and map urban heat island effects through that 39 year record as well.

[22:57] We just had a noon seminar today on [23:01] projection scenarios of future and past land cover [23:05] and land use. [23:07] done through a modeling approach called 4C. [23:12] and [23:12] annual land cover feeds into their scenarios so that they can take the land cover record back. [23:20] before Landsat and go back even into the 1600s and then go forward in time to understand what future land cover might look like. [23:29] The elegant part of that too is bringing in, they can bring in information. [23:35] related to [23:37] what we expect out of future climate and how those land cover [23:41] types might be affected by [23:43] by climate.

[23:44] Okay, so I just got an email from a person who said he was an environmental economist. [23:50] I don't even know what an environmental economist really does, but he's interested in using the new products [23:56] the annual NLCD, to help him with modeling flood impacts [24:02] on underserved communities. [24:05] Thank you. [24:06] One of the recent questions that came into the help desk was trying to understand how our data was used to create this map of [24:16] impaired air quality [24:18] and aligning that with the Head Start program to show where children may experience incidences of increased asthma attacks, things like that.

Oh, wow. So by intersecting NLCD, where developed is and how much developed is in an area, you can get an estimate of potentially poor air quality from cities. Intersect that with where Head Start reports locations, and all of a sudden now you have a population estimate for children [24:48] may be affected over the long term just by their location on the map. Wow. They can mitigate that by... [24:55] putting money into [24:57] green space in those areas. [24:59] far-reaching examples and benefits anyway. [25:03] This product uses a lot of data.

[25:06] How do we deal with so much data? [25:08] To make that 39-year record of land cover, we're actually touching... [25:14] Okay, so this is the statistics that... [25:20] basically 1.4 [25:22] some million Landsat observations [25:26] through history, through time. [25:29] we are looking at [25:32] over 3,000 average observations over a spatial area. [25:37] called a tile we process in something called tiles. [25:42] Eight Landsat Bands. [25:44] per observation. [25:46] in those tiles. [25:48] are the different parts of the electromagnetic spectrum that Landsat collects in. [25:53] surface reflectance over, collects and processes surface reflectance [25:57] And we estimated that we processed or touched over 295 trillion pixels.

[26:04] to make this land cover data set. [26:07] Is that a lot? That seems like a lot. [26:10] We think it's a lot. So we did all of this processing process [26:14] in the commercial cloud. I don't believe this would have been possible, at least not in the time that we processed. [26:21] if we weren't using cloud compute capabilities. We have a stunningly talented group of scientists and engineers who have set this system up for us. [26:32] It is designed to be rerun. [26:35] It's not exactly like [26:37] snap your fingers and we're going to do it again, but we are planning on rerunning [26:42] the system...

[26:43] next year when we can add another year's worth of Landsat data and produce an updated [26:49] data set that will cover [26:51] through time through 2024. And we want to do that every year. [26:55] into the future. [26:56] It doesn't sound like this is something we could have done. [26:59] A decade ago? No. [27:01] Absolutely not. We needed some of these advancements to happen recently, cloud computing, to make this work. [27:08] cloud computing and [27:09] artificial intelligence [27:11] We're using several deep learning [27:14] models. [27:16] chained together in our [27:18] classification architecture.

And what's the output? How many pixels are being released in annual NLCD? [27:25] Over $2 trillion. [27:27] That's pretty cool too. [27:29] Where can people view the map? Is there like an online viewer that people can work with? [27:34] And how about researchers? Can they download the data? [27:37] The data will continue to be available, as it has in the past, on gov. [27:42] We're also adding additional venues. It will be available for direct processing on AWS. [27:50] It will also be available on Earth Explorer, and we're currently just finalizing a new website to showcase some of the science and linkages that go into making the National Land Cover Database.

It'll function very similar to the previous slide. [28:06] MRLC viewer but of course with 40 years of data [28:10] it changes a bit to be more usable. [28:12] Our objective is to offer people as many different ways to get their hands on these data as they want. [28:18] Some people are not going to want to download. [28:21] 2 trillion pixels. They're just going to want their area of interest for the years they're interested in. The viewer will give them that. If you are a super user and you're already in the cloud, [28:32] working, you don't have to move the data to your own desktop.

You can do processing in the cloud yourself. [28:39] That's why we ran in the cloud. We did not want to move the Landsat archive to another system to process it, because that's yet more time. [28:47] What's next? I assume calling it [28:49] annual NLCD, [28:51] means a new release maybe around this time next year. [28:55] Yeah, our hope is actually to do it quicker. [28:58] next year [28:59] and our plan is to try and release data in the [29:04] late spring or early summer. [29:06] that will be an update [29:08] for the contermis US for the lower 48 states.

[29:12] adding the year 2024. So then it truly will be a 40-year record of annual land cover. [29:19] Our plans are to keep doing that into the future. [29:22] on an annual basis. [29:24] to try and make improvements to the system, [29:27] and then in 2026 at Alaska and Hawaii Land Cover for annual time periods. The new website is gov slash annual NLCD. [29:43] Thank you, Jess Brown and John DeWitts, for talking with us about annual NLCD. [29:50] And thank you, listeners. Check out our social media accounts to watch for our future episodes.

[29:57] us on Apple and YouTube podcasts. This podcast is a product of the S. Geological Survey, Department of Interior. [30:21] you [30:21] you