WEBVTT NOTE This file was exported by MacCaption version 7.0.13 to comply with the WebVTT specification dated March 27, 2017. 00:00:11.278 --> 00:00:13.046 align:center line:-1 position:50% size:33% Okay, folks, we're gonna go ahead and get started. 00:00:13.046 --> 00:00:14.281 align:center line:-1 position:50% size:34% My name is Matt Thomas, 00:00:14.281 --> 00:00:15.148 align:center line:-1 position:50% size:35% and thank you for tuning in 00:00:15.148 --> 00:00:17.651 align:center line:-1 position:50% size:37% to the USGS Landslide Hazards Seminar. 00:00:17.651 --> 00:00:20.354 align:center line:-1 position:50% size:42% This meeting is hosted by the Landslide Hazards Program, 00:00:20.354 --> 00:00:21.755 align:center line:-1 position:50% size:23% and co-organized with contributions 00:00:21.755 --> 00:00:24.258 align:center line:-1 position:50% size:31% from Stephen Slaughter and Jaime Kostelnik. 00:00:24.258 --> 00:00:26.293 align:center line:-1 position:50% size:37% For those of you that are new to this meeting, 00:00:26.293 --> 00:00:27.828 align:center line:-1 position:50% size:37% you're gonna have the ability to submit questions 00:00:27.828 --> 00:00:30.297 align:center line:-1 position:50% size:35% via the chat window or use the raise your hand feature 00:00:30.297 --> 00:00:31.431 align:center line:-1 position:50% size:28% in combination with your microphone 00:00:31.431 --> 00:00:32.566 align:center line:-1 position:50% size:24% and video camera. 00:00:32.566 --> 00:00:34.434 align:center line:-1 position:50% size:40% But we're gonna wait until the end of today's presentation 00:00:34.434 --> 00:00:35.335 align:center line:-1 position:50% size:23% to take questions. 00:00:35.335 --> 00:00:36.403 align:center line:-1 position:50% size:26% So in the meantime, 00:00:36.403 --> 00:00:37.738 align:center line:-1 position:50% size:28% if you're not intending to speak, 00:00:37.738 --> 00:00:39.239 align:center line:-1 position:50% size:35% please do the best you can to make sure 00:00:39.239 --> 00:00:42.442 align:center line:-1 position:50% size:38% your microphone is muted and your video camera is off. 00:00:42.442 --> 00:00:44.845 align:center line:-1 position:50% size:30% It's my pleasure today to introduce Ben Mirus. 00:00:44.845 --> 00:00:47.047 align:center line:-1 position:50% size:41% Ben is a supervisor research geologist 00:00:47.047 --> 00:00:49.583 align:center line:-1 position:50% size:37% in the USGS Landslide Hazards Program. 00:00:49.583 --> 00:00:52.386 align:center line:-1 position:50% size:43% He earned his bachelor's degree in geology from Pomona College, 00:00:52.386 --> 00:00:53.787 align:center line:-1 position:50% size:28% and started his career with the USGS 00:00:53.787 --> 00:00:56.757 align:center line:-1 position:50% size:36% as a graduate student intern in Menlo Park, California, 00:00:56.757 --> 00:01:00.127 align:center line:-1 position:50% size:37% while pursuing a PhD in hydrogeology at Stanford. 00:01:00.127 --> 00:01:02.763 align:center line:-1 position:50% size:38% He then served on the faculty at UNC Chapel Hill, 00:01:02.763 --> 00:01:03.864 align:center line:-1 position:50% size:41% teaching and advising students, 00:01:03.864 --> 00:01:05.999 align:center line:-1 position:50% size:39% before returning to the USGS, 00:01:05.999 --> 00:01:09.336 align:center line:-1 position:50% size:38% this time in Colorado in 2015. 00:01:09.336 --> 00:01:10.570 align:center line:-1 position:50% size:24% In his current role, 00:01:10.570 --> 00:01:13.040 align:center line:-1 position:50% size:34% Ben leads a growing team of scientists and postdocs 00:01:13.040 --> 00:01:14.741 align:center line:-1 position:50% size:36% and applies his background in geology, 00:01:14.741 --> 00:01:17.010 align:center line:-1 position:50% size:31% hillslope hydrology, and numerical modeling 00:01:17.010 --> 00:01:19.846 align:center line:-1 position:50% size:42% towards developing new tools for landslide hazard assessment 00:01:19.846 --> 00:01:21.181 align:center line:-1 position:50% size:25% and loss reduction. 00:01:21.181 --> 00:01:22.316 align:center line:-1 position:50% size:25% Ben, thanks for joining us today, 00:01:22.316 --> 00:01:24.685 align:center line:-1 position:50% size:32% and I'm gonna go ahead and turn it over to you. 00:01:24.685 --> 00:01:26.586 align:center line:-1 position:50% size:25% Thank you, Matt, for the introduction. 00:01:26.586 --> 00:01:28.855 align:center line:-1 position:50% size:29% And thanks, everyone, for your attention. 00:01:30.157 --> 00:01:33.160 align:center line:-1 position:50% size:41% It's a very exciting opportunity to share this new work with you. 00:01:33.160 --> 00:01:37.197 align:center line:-1 position:50% size:42% We've developed a new national landslide susceptibility map, 00:01:37.197 --> 00:01:40.100 align:center line:-1 position:50% size:30% and the corresponding journal article title here, 00:01:40.100 --> 00:01:41.535 align:center line:-1 position:50% size:32% "Parsimonious approach 00:01:41.535 --> 00:01:44.871 align:center line:-1 position:50% size:38% to continental-scale landslide susceptibility modeling." 00:01:44.871 --> 00:01:48.442 align:center line:-1 position:50% size:41% I want to take a quick moment to acknowledge my co-authors, 00:01:48.442 --> 00:01:50.077 align:center line:-1 position:50% size:29% Gina Belair and Sabrina Martinez, 00:01:50.077 --> 00:01:51.645 align:center line:-1 position:50% size:38% here at the Geologic Hazards Science Center, 00:01:51.645 --> 00:01:53.180 align:center line:-1 position:50% size:24% and Nate Wood and Jeanne Jones 00:01:53.180 --> 00:01:55.115 align:center line:-1 position:50% size:35% in the Western Geographic Science Center. 00:01:55.115 --> 00:01:56.750 align:center line:-1 position:50% size:36% All of them had pivotal roles in this work, 00:01:56.750 --> 00:01:58.118 align:center line:-1 position:50% size:30% and I really am grateful 00:01:58.118 --> 00:02:00.253 align:center line:-1 position:50% size:28% for such a great team to work with. 00:02:02.089 --> 00:02:03.256 align:center line:-1 position:50% size:24% So just to start off, 00:02:03.256 --> 00:02:05.525 align:center line:-1 position:50% size:39% what are landslide susceptibility maps? 00:02:05.525 --> 00:02:07.828 align:center line:-1 position:50% size:37% Probably everyone's familiar with the concept, 00:02:07.828 --> 00:02:10.163 align:center line:-1 position:50% size:20% but just to state all this up front, 00:02:10.163 --> 00:02:12.432 align:center line:-1 position:50% size:37% landslide susceptibility maps show where landslides 00:02:12.432 --> 00:02:15.635 align:center line:-1 position:50% size:32% are more and less likely throughout a landscape. 00:02:15.635 --> 00:02:17.137 align:center line:-1 position:50% size:29% Usually they use some sort of a terrain, 00:02:17.137 --> 00:02:19.606 align:center line:-1 position:50% size:31% environmental variables as predictors. 00:02:19.606 --> 00:02:21.875 align:center line:-1 position:50% size:35% And there are a wide range of approaches to this now, 00:02:21.875 --> 00:02:23.243 align:center line:-1 position:50% size:23% statistical, physically based, 00:02:23.243 --> 00:02:25.379 align:center line:-1 position:50% size:32% and increasing emphasis on machine learning 00:02:25.379 --> 00:02:27.681 align:center line:-1 position:50% size:37% and data-driven approaches. 00:02:29.316 --> 00:02:32.586 align:center line:-1 position:50% size:36% All of them rely very heavily on landslide position data, 00:02:32.586 --> 00:02:35.555 align:center line:-1 position:50% size:34% which introduces its own sources of biases. 00:02:36.556 --> 00:02:39.326 align:center line:-1 position:50% size:35% But typically they're applied at local scales. 00:02:39.326 --> 00:02:40.260 align:center line:-1 position:50% size:26% There are, however, 00:02:40.260 --> 00:02:42.195 align:center line:-1 position:50% size:30% a few regional and global applications 00:02:42.195 --> 00:02:44.698 align:center line:-1 position:50% size:21% that I'll mention. 00:02:44.698 --> 00:02:47.134 align:center line:-1 position:50% size:30% It is also very important to talk about 00:02:47.134 --> 00:02:49.636 align:center line:-1 position:50% size:35% what a susceptibility model is not. 00:02:49.636 --> 00:02:51.772 align:center line:-1 position:50% size:33% It does not account for frequency, magnitude, 00:02:51.772 --> 00:02:54.074 align:center line:-1 position:50% size:15% or impacts. 00:02:54.074 --> 00:02:55.042 align:center line:-1 position:50% size:23% At the local scale, 00:02:55.042 --> 00:02:56.610 align:center line:-1 position:50% size:38% you might be able to assume some equivalence 00:02:56.610 --> 00:02:58.578 align:center line:-1 position:50% size:29% between susceptibility and hazard, 00:02:58.578 --> 00:03:01.014 align:center line:-1 position:50% size:36% such as shown in these two examples here. 00:03:01.014 --> 00:03:05.218 align:center line:-1 position:50% size:43% There's work by Stephen Hughes and Bill Schulz for Puerto Rico, 00:03:05.218 --> 00:03:07.587 align:center line:-1 position:50% size:42% and Jacob Woodard and others, 00:03:07.587 --> 00:03:10.457 align:center line:-1 position:50% size:35% using slope units for a catchment in Oregon. 00:03:10.457 --> 00:03:11.458 align:center line:-1 position:50% size:29% And at smaller scales, 00:03:11.458 --> 00:03:13.827 align:center line:-1 position:50% size:33% you can sort of assume some of that equivalence, 00:03:13.827 --> 00:03:15.562 align:center line:-1 position:50% size:31% but as you get to bigger and bigger scales, 00:03:15.562 --> 00:03:19.766 align:center line:-1 position:50% size:35% we really do not account for those hazard elements. 00:03:19.766 --> 00:03:23.437 align:center line:-1 position:50% size:29% And so the constraints on timing recurrence 00:03:23.437 --> 00:03:25.672 align:center line:-1 position:50% size:26% are really common across broad areas, 00:03:25.672 --> 00:03:27.507 align:center line:-1 position:50% size:35% and so susceptibility maps, 00:03:27.507 --> 00:03:29.142 align:center line:-1 position:50% size:24% especially at continental scales, 00:03:29.142 --> 00:03:31.745 align:center line:-1 position:50% size:31% are really just where are landslides' potential 00:03:31.745 --> 00:03:33.180 align:center line:-1 position:50% size:35% throughout that landscape. 00:03:35.048 --> 00:03:38.785 align:center line:-1 position:50% size:34% The big question: Why a new national map? 00:03:39.753 --> 00:03:40.554 align:center line:-1 position:50% size:20% Well, first of all, 00:03:40.554 --> 00:03:44.724 align:center line:-1 position:50% size:39% landslides occur in every state and most U.S. territories. 00:03:44.724 --> 00:03:48.995 align:center line:-1 position:50% size:38% But only a few states have a concentrated program 00:03:48.995 --> 00:03:51.131 align:center line:-1 position:50% size:24% on assessing landslide hazards. 00:03:54.167 --> 00:03:57.137 align:center line:-1 position:50% size:31% So there's a lot of areas where that's needed. 00:03:57.137 --> 00:03:59.706 align:center line:-1 position:50% size:33% And then there isn't really a great national picture, 00:03:59.706 --> 00:04:01.208 align:center line:-1 position:50% size:14% as of now. 00:04:01.208 --> 00:04:03.410 align:center line:-1 position:50% size:35% In particular, there's been some recent improvements 00:04:03.410 --> 00:04:05.345 align:center line:-1 position:50% size:21% in USGS data and capabilities, 00:04:05.345 --> 00:04:10.016 align:center line:-1 position:50% size:40% including high-res 3DEP digital elevation models, 00:04:10.016 --> 00:04:11.985 align:center line:-1 position:50% size:38% improved high-performance computing. 00:04:11.985 --> 00:04:12.853 align:center line:-1 position:50% size:21% We use the Yeti, 00:04:12.853 --> 00:04:16.056 align:center line:-1 position:50% size:37% but now there's even a new HBC resource for the USGS. 00:04:16.756 --> 00:04:19.226 align:center line:-1 position:50% size:33% And also what I'll talk about shortly 00:04:19.226 --> 00:04:21.628 align:center line:-1 position:50% size:30% are a compilation of landslide inventories 00:04:21.628 --> 00:04:23.196 align:center line:-1 position:50% size:25% across the country. 00:04:25.465 --> 00:04:27.300 align:center line:-1 position:50% size:27% So there were some existing maps, 00:04:27.300 --> 00:04:31.505 align:center line:-1 position:50% size:31% and those are based on older, limited data. 00:04:32.506 --> 00:04:34.708 align:center line:-1 position:50% size:38% Then, they tended to underrepresent landsliding 00:04:34.708 --> 00:04:36.710 align:center line:-1 position:50% size:26% in moderate terrain. 00:04:36.710 --> 00:04:40.380 align:center line:-1 position:50% size:34% And also many suffered from incomplete coverage 00:04:40.380 --> 00:04:42.449 align:center line:-1 position:50% size:26% over the U.S., especially in Alaska. 00:04:42.449 --> 00:04:44.818 align:center line:-1 position:50% size:31% So, in addition to the coarse resolution, 00:04:44.818 --> 00:04:46.586 align:center line:-1 position:50% size:25% one kilometer is really insufficient 00:04:46.586 --> 00:04:47.621 align:center line:-1 position:50% size:29% for many applications. 00:04:47.621 --> 00:04:50.290 align:center line:-1 position:50% size:41% There is a real need to leverage all of these improvements 00:04:50.290 --> 00:04:52.392 align:center line:-1 position:50% size:28% in data and resources to do better. 00:04:53.493 --> 00:04:55.195 align:center line:-1 position:50% size:23% And that is also within the context 00:04:55.195 --> 00:04:57.164 align:center line:-1 position:50% size:32% of the National Landslide Preparedness Act, 00:04:57.164 --> 00:04:59.666 align:center line:-1 position:50% size:21% passed in 2021. 00:05:00.734 --> 00:05:02.435 align:center line:-1 position:50% size:36% In that there's many regions of the country 00:05:02.435 --> 00:05:04.104 align:center line:-1 position:50% size:27% which are referred to with basically no 00:05:04.104 --> 00:05:06.439 align:center line:-1 position:50% size:36% or minimal landslide reporting and assessments, 00:05:06.439 --> 00:05:09.576 align:center line:-1 position:50% size:40% and a real need for a nationwide understanding 00:05:09.576 --> 00:05:11.478 align:center line:-1 position:50% size:39% of potential exposure and risk. 00:05:11.478 --> 00:05:14.314 align:center line:-1 position:50% size:36% So, this susceptibility model 00:05:14.314 --> 00:05:16.850 align:center line:-1 position:50% size:33% is kind of one of the steps along that path 00:05:16.850 --> 00:05:19.286 align:center line:-1 position:50% size:38% to an improved national-scale understanding. 00:05:19.286 --> 00:05:22.189 align:center line:-1 position:50% size:32% And our objective here was to promptly develop 00:05:22.189 --> 00:05:25.625 align:center line:-1 position:50% size:38% an improved, uniform, high-resolution national map. 00:05:28.395 --> 00:05:30.197 align:center line:-1 position:50% size:40% There are a few existing maps. 00:05:30.197 --> 00:05:32.699 align:center line:-1 position:50% size:37% These were just for CONUS. 00:05:32.699 --> 00:05:36.803 align:center line:-1 position:50% size:36% The first on the left is a landslide overview map 00:05:36.803 --> 00:05:38.705 align:center line:-1 position:50% size:26% of the conterminous United States, 00:05:38.705 --> 00:05:41.274 align:center line:-1 position:50% size:36% published by Radbruch-Hall et al. in 1982. 00:05:41.274 --> 00:05:42.442 align:center line:-1 position:50% size:27% This was based on... 00:05:45.045 --> 00:05:48.181 align:center line:-1 position:50% size:35% a kind of national overview and synthesis 00:05:48.181 --> 00:05:50.417 align:center line:-1 position:50% size:27% of geologic mapping available at the time. 00:05:50.417 --> 00:05:52.986 align:center line:-1 position:50% size:27% It went to the 750 million scale. 00:05:52.986 --> 00:05:55.055 align:center line:-1 position:50% size:23% And they included six susceptibility 00:05:55.055 --> 00:05:56.590 align:center line:-1 position:50% size:30% and incidence classes, that you can see here, 00:05:56.590 --> 00:06:02.195 align:center line:-1 position:50% size:40% ranging from the sort of yellow, green, brown, pink, and red. 00:06:02.996 --> 00:06:05.065 align:center line:-1 position:50% size:33% It's a little bit complicated to describe, 00:06:05.065 --> 00:06:05.799 align:center line:-1 position:50% size:26% so I won't get into it. 00:06:05.799 --> 00:06:09.236 align:center line:-1 position:50% size:24% But this map is still getting used 00:06:09.236 --> 00:06:10.770 align:center line:-1 position:50% size:30% and downloaded today. 00:06:11.705 --> 00:06:14.307 align:center line:-1 position:50% size:37% On the right is a topographic slope relief threshold. 00:06:14.307 --> 00:06:15.442 align:center line:-1 position:50% size:31% This is a prototype map 00:06:15.442 --> 00:06:18.478 align:center line:-1 position:50% size:36% that was developed by Godt and others in 2012, 00:06:18.478 --> 00:06:19.346 align:center line:-1 position:50% size:20% and published-- 00:06:19.346 --> 00:06:21.615 align:center line:-1 position:50% size:39% and actually it was being used by Department of Interior 00:06:21.615 --> 00:06:23.383 align:center line:-1 position:50% size:23% until we produced this updated map, 00:06:23.383 --> 00:06:27.153 align:center line:-1 position:50% size:34% and it's in its present form. 00:06:27.153 --> 00:06:29.222 align:center line:-1 position:50% size:42% And I'm gonna describe that one in a little bit more detail 00:06:29.222 --> 00:06:31.491 align:center line:-1 position:50% size:31% because we used a very similar approach. 00:06:32.859 --> 00:06:37.764 align:center line:-1 position:50% size:44% There were a couple global maps, most notably the-- 00:06:37.764 --> 00:06:40.467 align:center line:-1 position:50% size:23% NASA's heuristic fuzzy logic model, 00:06:40.467 --> 00:06:43.737 align:center line:-1 position:50% size:38% which uses slope, forest loss, roads, and geology 00:06:43.737 --> 00:06:46.706 align:center line:-1 position:50% size:36% to develop five global susceptibility classes 00:06:46.706 --> 00:06:49.209 align:center line:-1 position:50% size:28% ranging from very low to very high. 00:06:49.209 --> 00:06:51.344 align:center line:-1 position:50% size:36% They did this based on 70,000 landslides 00:06:51.344 --> 00:06:53.880 align:center line:-1 position:50% size:33% from a handful of high-quality inventories 00:06:53.880 --> 00:06:56.016 align:center line:-1 position:50% size:34% and then some distributed throughout the globe. 00:06:56.016 --> 00:07:00.520 align:center line:-1 position:50% size:37% And they--this is what's used in the LHASA model... 00:07:02.289 --> 00:07:04.224 align:center line:-1 position:50% size:27% the version one of the LHASA model. 00:07:05.859 --> 00:07:07.527 align:center line:-1 position:50% size:22% In the meantime, 00:07:07.527 --> 00:07:10.196 align:center line:-1 position:50% size:30% there's also a global non-susceptibility map. 00:07:10.196 --> 00:07:12.866 align:center line:-1 position:50% size:40% I'll talk a little bit more about how that was developed, 00:07:12.866 --> 00:07:15.602 align:center line:-1 position:50% size:28% but it also uses the slope relief model 00:07:15.602 --> 00:07:18.171 align:center line:-1 position:50% size:32% to define non-susceptibility zones. 00:07:18.171 --> 00:07:21.007 align:center line:-1 position:50% size:44% And so you can see the red areas sort of correspond, actually, 00:07:21.007 --> 00:07:23.410 align:center line:-1 position:50% size:37% reasonably well to the green and blue areas 00:07:23.410 --> 00:07:24.644 align:center line:-1 position:50% size:26% in the NASA model. 00:07:24.644 --> 00:07:27.747 align:center line:-1 position:50% size:40% And they used a modified area under the curve method 00:07:27.747 --> 00:07:29.716 align:center line:-1 position:50% size:39% using about 40,000 landslides 00:07:29.716 --> 00:07:32.085 align:center line:-1 position:50% size:37% from high-quality inventories across the globe. 00:07:33.219 --> 00:07:35.255 align:center line:-1 position:50% size:36% They also looked at kind of landslide type and whatnot, 00:07:35.255 --> 00:07:36.623 align:center line:-1 position:50% size:31% and interesting insights. 00:07:36.623 --> 00:07:38.558 align:center line:-1 position:50% size:28% I recommend looking at both these papers. 00:07:40.460 --> 00:07:42.295 align:center line:-1 position:50% size:32% However, the number of landslides 00:07:42.295 --> 00:07:44.464 align:center line:-1 position:50% size:27% in both those studies is relatively small. 00:07:44.464 --> 00:07:46.599 align:center line:-1 position:50% size:34% And we started compiling, back in 2018, 00:07:46.599 --> 00:07:49.202 align:center line:-1 position:50% size:33% landslide inventories done by all the great folks 00:07:49.202 --> 00:07:51.237 align:center line:-1 position:50% size:22% at state surveys, and researchers, 00:07:51.237 --> 00:07:53.840 align:center line:-1 position:50% size:37% and previous USGS folks mapping all over the country, 00:07:53.840 --> 00:07:56.042 align:center line:-1 position:50% size:29% and put those into one 00:07:56.042 --> 00:07:58.244 align:center line:-1 position:50% size:24% readily accessible, uniform database. 00:07:58.244 --> 00:07:59.446 align:center line:-1 position:50% size:16% And that is-- 00:07:59.446 --> 00:08:02.215 align:center line:-1 position:50% size:28% the initial compilation, published in 2019, 00:08:02.215 --> 00:08:05.118 align:center line:-1 position:50% size:25% included just over 300,000 landslides. 00:08:05.118 --> 00:08:08.054 align:center line:-1 position:50% size:28% Version two, in 2022 was updated, 00:08:08.054 --> 00:08:09.856 align:center line:-1 position:50% size:38% roughly doubled that number. 00:08:09.856 --> 00:08:11.224 align:center line:-1 position:50% size:34% And now we're working on version three, 00:08:11.224 --> 00:08:12.692 align:center line:-1 position:50% size:41% which should be released soon, 00:08:12.692 --> 00:08:14.794 align:center line:-1 position:50% size:30% and we're approaching one million landslides. 00:08:14.794 --> 00:08:15.995 align:center line:-1 position:50% size:20% Really exciting. 00:08:17.330 --> 00:08:18.732 align:center line:-1 position:50% size:25% But we do still have a long way to go. 00:08:18.732 --> 00:08:19.999 align:center line:-1 position:50% size:19% For reference, 00:08:19.999 --> 00:08:21.668 align:center line:-1 position:50% size:28% Italy's national inventory compilation 00:08:21.668 --> 00:08:24.003 align:center line:-1 position:50% size:39% has about 600,000 landslides. 00:08:24.003 --> 00:08:25.672 align:center line:-1 position:50% size:39% So, for a much bigger country, 00:08:25.672 --> 00:08:29.576 align:center line:-1 position:50% size:40% we expect a much, much greater number. 00:08:29.576 --> 00:08:31.444 align:center line:-1 position:50% size:29% There's different types of trigger landslides. 00:08:31.444 --> 00:08:34.347 align:center line:-1 position:50% size:35% There's also highly variable data quality 00:08:34.347 --> 00:08:36.383 align:center line:-1 position:50% size:26% for points, polygons, media reports, 00:08:36.383 --> 00:08:39.219 align:center line:-1 position:50% size:36% and that's what's expressed in these colors here, 00:08:39.219 --> 00:08:42.455 align:center line:-1 position:50% size:32% ranging from light yellow to red. 00:08:42.455 --> 00:08:44.657 align:center line:-1 position:50% size:30% But there's fairly limited information on timing 00:08:44.657 --> 00:08:46.159 align:center line:-1 position:50% size:35% or a triggering mechanism. 00:08:46.159 --> 00:08:49.596 align:center line:-1 position:50% size:37% And if you're interested in downloading that data set, 00:08:49.596 --> 00:08:52.265 align:center line:-1 position:50% size:35% you can visit this DOI here. 00:08:53.166 --> 00:08:54.434 align:center line:-1 position:50% size:32% I'm also gonna just show a quick zoom-in 00:08:54.434 --> 00:08:57.070 align:center line:-1 position:50% size:27% to illustrate that the-- 00:08:57.070 --> 00:08:58.471 align:center line:-1 position:50% size:24% one of the artifacts of ARC online, 00:08:58.471 --> 00:09:00.340 align:center line:-1 position:50% size:37% where this service is hosted, 00:09:00.340 --> 00:09:01.141 align:center line:-1 position:50% size:21% to zoom around, 00:09:01.141 --> 00:09:02.375 align:center line:-1 position:50% size:23% you can look at it, 00:09:02.375 --> 00:09:03.543 align:center line:-1 position:50% size:36% you can see when you start zooming in, 00:09:03.543 --> 00:09:06.646 align:center line:-1 position:50% size:35% you see much of the higher quality polygon mapping, 00:09:06.646 --> 00:09:08.615 align:center line:-1 position:50% size:30% whereas the points just stay the same size 00:09:08.615 --> 00:09:09.716 align:center line:-1 position:50% size:35% when you zoom in and out. 00:09:11.284 --> 00:09:12.886 align:center line:-1 position:50% size:26% So then we have this great database, 00:09:12.886 --> 00:09:14.287 align:center line:-1 position:50% size:38% and I mentioned that the data 00:09:14.287 --> 00:09:16.756 align:center line:-1 position:50% size:33% is a really important part of susceptibility modeling. 00:09:17.657 --> 00:09:21.261 align:center line:-1 position:50% size:39% So now we have the data with which to start moving forward 00:09:21.261 --> 00:09:22.529 align:center line:-1 position:50% size:31% and improving the map. 00:09:22.529 --> 00:09:25.198 align:center line:-1 position:50% size:25% But what method are we gonna use? 00:09:25.198 --> 00:09:27.500 align:center line:-1 position:50% size:32% Now this shows sort of two endmembers 00:09:27.500 --> 00:09:30.670 align:center line:-1 position:50% size:38% in terms of model complexity. 00:09:30.670 --> 00:09:32.005 align:center line:-1 position:50% size:35% And I would strongly argue 00:09:32.005 --> 00:09:34.474 align:center line:-1 position:50% size:26% that more predictors and more variables 00:09:34.474 --> 00:09:36.109 align:center line:-1 position:50% size:33% that you throw into a susceptibility model 00:09:36.109 --> 00:09:38.678 align:center line:-1 position:50% size:30% at the continental scale is not better. 00:09:38.678 --> 00:09:41.114 align:center line:-1 position:50% size:35% So on the left we've got a simple slope index model 00:09:41.114 --> 00:09:44.617 align:center line:-1 position:50% size:40% developed by Earl Brabb and USGS colleagues in 1999. 00:09:44.617 --> 00:09:47.720 align:center line:-1 position:50% size:32% And this is just a simple, "Where across the U.S. 00:09:47.720 --> 00:09:49.689 align:center line:-1 position:50% size:24% are slopes greater than 25 degrees?" 00:09:49.689 --> 00:09:51.191 align:center line:-1 position:50% size:36% This has attempted to focus 00:09:51.191 --> 00:09:52.992 align:center line:-1 position:50% size:28% on shallow landslides and debris flows, 00:09:52.992 --> 00:09:54.294 align:center line:-1 position:50% size:39% which wasn't really considered 00:09:54.294 --> 00:09:56.329 align:center line:-1 position:50% size:41% in the Radbruch-Hall et al. map. 00:09:56.329 --> 00:09:57.530 align:center line:-1 position:50% size:24% And you can see the data they used 00:09:57.530 --> 00:09:59.165 align:center line:-1 position:50% size:32% is sort of the more vivid bright red, 00:09:59.165 --> 00:10:01.301 align:center line:-1 position:50% size:32% and then their model of threshold exceedance 00:10:01.301 --> 00:10:05.371 align:center line:-1 position:50% size:23% is the lighter pink. 00:10:06.172 --> 00:10:09.375 align:center line:-1 position:50% size:25% Now on the right is, in quotes, 00:10:09.375 --> 00:10:12.045 align:center line:-1 position:50% size:23% "state-of-the-art" machine learning. 00:10:12.045 --> 00:10:14.747 align:center line:-1 position:50% size:35% And this uses 32 predictors 00:10:14.747 --> 00:10:19.886 align:center line:-1 position:50% size:36% with a number of automated naive data-driven models. 00:10:19.886 --> 00:10:22.222 align:center line:-1 position:50% size:31% And what this group did was took-- 00:10:22.222 --> 00:10:24.157 align:center line:-1 position:50% size:22% they took 65,000 of our landslides 00:10:24.157 --> 00:10:25.558 align:center line:-1 position:50% size:38% from the national compilation. 00:10:25.558 --> 00:10:26.826 align:center line:-1 position:50% size:17% I'm not sure-- 00:10:26.826 --> 00:10:29.496 align:center line:-1 position:50% size:38% it doesn't explain in the paper how they selected those. 00:10:29.496 --> 00:10:32.999 align:center line:-1 position:50% size:31% But they are shown in the upper figure here. 00:10:32.999 --> 00:10:35.635 align:center line:-1 position:50% size:41% And then the susceptibility map is shown in the lower figure. 00:10:35.635 --> 00:10:39.606 align:center line:-1 position:50% size:35% And what you can see is that it's basically a model 00:10:39.606 --> 00:10:42.108 align:center line:-1 position:50% size:28% of where landslides have been observed, 00:10:42.108 --> 00:10:44.344 align:center line:-1 position:50% size:28% but it's not very useful for predicting 00:10:44.344 --> 00:10:47.514 align:center line:-1 position:50% size:23% where landslides may be occurring. 00:10:47.514 --> 00:10:52.151 align:center line:-1 position:50% size:40% And we know that from looking at our database here, 00:10:52.151 --> 00:10:53.520 align:center line:-1 position:50% size:16% for example. 00:10:53.520 --> 00:10:55.154 align:center line:-1 position:50% size:32% Landslides in Minnesota, 00:10:55.154 --> 00:10:57.657 align:center line:-1 position:50% size:22% parts of Nevada, 00:10:57.657 --> 00:10:59.425 align:center line:-1 position:50% size:30% that just aren't even considered. 00:11:00.593 --> 00:11:03.463 align:center line:-1 position:50% size:39% So this is too many predictors, perhaps. 00:11:03.463 --> 00:11:06.232 align:center line:-1 position:50% size:38% You're over-calibrating the model to just fit your data, 00:11:06.232 --> 00:11:08.268 align:center line:-1 position:50% size:29% such that you develop a model of your data 00:11:08.268 --> 00:11:11.504 align:center line:-1 position:50% size:33% rather than a model of landslide susceptibility. 00:11:11.504 --> 00:11:13.339 align:center line:-1 position:50% size:33% So we wanted something in between, 00:11:13.339 --> 00:11:15.608 align:center line:-1 position:50% size:30% something that's better than just slope, 00:11:15.608 --> 00:11:18.344 align:center line:-1 position:50% size:26% but also is gonna be useful everywhere 00:11:18.344 --> 00:11:21.214 align:center line:-1 position:50% size:32% and not overfit to the data that we have. 00:11:21.214 --> 00:11:22.382 align:center line:-1 position:50% size:30% And one of the reasons for that 00:11:22.382 --> 00:11:24.684 align:center line:-1 position:50% size:31% is that landslide controls are very complicated. 00:11:24.684 --> 00:11:25.552 align:center line:-1 position:50% size:28% There's a lot of things 00:11:25.552 --> 00:11:27.754 align:center line:-1 position:50% size:24% that influence landslide initiation. 00:11:27.754 --> 00:11:30.557 align:center line:-1 position:50% size:33% On the left here are just five pictures of landslides 00:11:30.557 --> 00:11:32.392 align:center line:-1 position:50% size:26% I've visited throughout the U.S. 00:11:32.392 --> 00:11:34.627 align:center line:-1 position:50% size:29% And the first one here, Boulder County. 00:11:34.627 --> 00:11:36.396 align:center line:-1 position:50% size:32% Aspect is a really important control 00:11:36.396 --> 00:11:37.263 align:center line:-1 position:50% size:31% in landslide occurrence, 00:11:37.263 --> 00:11:40.700 align:center line:-1 position:50% size:42% north-facing versus south-facing slopes and vegetation. 00:11:40.700 --> 00:11:43.102 align:center line:-1 position:50% size:30% Here in North Carolina, 00:11:43.102 --> 00:11:46.039 align:center line:-1 position:50% size:39% actually, this is Macon County in the study here, 00:11:46.039 --> 00:11:47.006 align:center line:-1 position:50% size:26% on the right I'm showing a figure 00:11:47.006 --> 00:11:49.876 align:center line:-1 position:50% size:29% from Jacob Woodard's JGR paper in '23. 00:11:49.876 --> 00:11:54.781 align:center line:-1 position:50% size:39% That slope is not necessarily a great predictor of landslides. 00:11:54.781 --> 00:11:55.848 align:center line:-1 position:50% size:32% In Macon County, in fact, 00:11:55.848 --> 00:11:57.016 align:center line:-1 position:50% size:29% we see an inverse relationship 00:11:57.016 --> 00:11:58.451 align:center line:-1 position:50% size:23% between slope and susceptibility. 00:11:58.451 --> 00:12:00.219 align:center line:-1 position:50% size:34% And it's because there's some of these big, 00:12:00.219 --> 00:12:02.622 align:center line:-1 position:50% size:35% slow-moving landslides occurring at gentler slopes, 00:12:02.622 --> 00:12:05.091 align:center line:-1 position:50% size:23% and there's some geologic controls. 00:12:05.091 --> 00:12:06.092 align:center line:-1 position:50% size:31% Here in Custer County-- 00:12:06.092 --> 00:12:07.827 align:center line:-1 position:50% size:19% Custer Conic-- 00:12:07.827 --> 00:12:10.296 align:center line:-1 position:50% size:42% Contra Costa County, California, 00:12:10.296 --> 00:12:14.867 align:center line:-1 position:50% size:31% is some--a big landslide on a very planar slope, 00:12:14.867 --> 00:12:17.203 align:center line:-1 position:50% size:32% so not occurring in those typical hillslope hollows 00:12:17.203 --> 00:12:18.404 align:center line:-1 position:50% size:21% that we think of. 00:12:18.404 --> 00:12:20.173 align:center line:-1 position:50% size:40% Whereas here in Sitka, Alaska, 00:12:20.173 --> 00:12:23.676 align:center line:-1 position:50% size:35% it is on the very steepest conversion hillslope hollow. 00:12:23.676 --> 00:12:24.544 align:center line:-1 position:50% size:10% Woops. 00:12:25.612 --> 00:12:27.246 align:center line:-1 position:50% size:35% And then here on the right, 00:12:27.246 --> 00:12:29.449 align:center line:-1 position:50% size:43% in the Fremont County, Wyoming, 00:12:29.449 --> 00:12:30.683 align:center line:-1 position:50% size:34% it's not the steepest slope. 00:12:30.683 --> 00:12:32.151 align:center line:-1 position:50% size:31% Again, where landslides are initiating, 00:12:32.151 --> 00:12:34.554 align:center line:-1 position:50% size:35% they're initiating right below these bedrock cliffs, 00:12:34.554 --> 00:12:37.023 align:center line:-1 position:50% size:41% and so there's probably some other processes going on 00:12:37.023 --> 00:12:38.257 align:center line:-1 position:50% size:32% with the hydrology there. 00:12:39.692 --> 00:12:41.694 align:center line:-1 position:50% size:36% And essentially what Jacob's paper showed 00:12:41.694 --> 00:12:43.129 align:center line:-1 position:50% size:30% is that you can develop 00:12:43.129 --> 00:12:44.764 align:center line:-1 position:50% size:26% a very robust susceptibility model 00:12:44.764 --> 00:12:46.733 align:center line:-1 position:50% size:35% for a location where you have good data. 00:12:46.733 --> 00:12:49.736 align:center line:-1 position:50% size:42% And you can take data from across a number of regions 00:12:49.736 --> 00:12:51.304 align:center line:-1 position:50% size:28% and develop a susceptibility model 00:12:51.304 --> 00:12:53.406 align:center line:-1 position:50% size:27% that works really well in all those regions. 00:12:53.406 --> 00:12:55.842 align:center line:-1 position:50% size:35% But if you take the data out for one region, 00:12:55.842 --> 00:12:58.645 align:center line:-1 position:50% size:37% and then calibrate the model for all those other regions, 00:12:58.645 --> 00:13:00.046 align:center line:-1 position:50% size:27% and then try to use it for the location 00:13:00.046 --> 00:13:01.280 align:center line:-1 position:50% size:39% where you pulled the data out, 00:13:01.280 --> 00:13:02.515 align:center line:-1 position:50% size:30% it performs quite poorly. 00:13:02.515 --> 00:13:03.316 align:center line:-1 position:50% size:23% And that's, again, 00:13:03.316 --> 00:13:04.884 align:center line:-1 position:50% size:29% because there's these very complex controls 00:13:04.884 --> 00:13:06.285 align:center line:-1 position:50% size:27% on landslide initiation 00:13:06.285 --> 00:13:08.554 align:center line:-1 position:50% size:36% that we don't want to overparametrize a model 00:13:08.554 --> 00:13:09.789 align:center line:-1 position:50% size:28% so that it is not useful 00:13:09.789 --> 00:13:12.225 align:center line:-1 position:50% size:34% in places where we don't have data. 00:13:12.225 --> 00:13:14.227 align:center line:-1 position:50% size:35% And that's exactly the point of the susceptibility model 00:13:14.227 --> 00:13:15.161 align:center line:-1 position:50% size:13% here now. 00:13:16.262 --> 00:13:20.066 align:center line:-1 position:50% size:36% So to do a little bit better than just topographic slope, 00:13:20.066 --> 00:13:22.802 align:center line:-1 position:50% size:30% we looked to a really excellent study 00:13:22.802 --> 00:13:24.704 align:center line:-1 position:50% size:30% by Kevin Schmidt and Dave Montgomery, 00:13:24.704 --> 00:13:26.506 align:center line:-1 position:50% size:24% published in 1995, 00:13:26.506 --> 00:13:27.774 align:center line:-1 position:50% size:34% looking at slope and relief. 00:13:27.774 --> 00:13:30.376 align:center line:-1 position:50% size:36% And they used 17th century 00:13:30.376 --> 00:13:34.047 align:center line:-1 position:50% size:29% engineering geologist Culmann's 1875 paper 00:13:34.047 --> 00:13:37.550 align:center line:-1 position:50% size:32% to show that there's this rock strength control 00:13:37.550 --> 00:13:39.152 align:center line:-1 position:50% size:25% on slope and relief, 00:13:39.152 --> 00:13:40.687 align:center line:-1 position:50% size:29% and that in highest relief areas, 00:13:40.687 --> 00:13:42.789 align:center line:-1 position:50% size:32% you've got stronger earth materials exposed, 00:13:42.789 --> 00:13:45.358 align:center line:-1 position:50% size:36% such that those can support steeper slopes. 00:13:45.358 --> 00:13:46.693 align:center line:-1 position:50% size:40% And you've got a steeper slope for landslide-- 00:13:46.693 --> 00:13:48.194 align:center line:-1 position:50% size:20% threshold slope for landsliding. 00:13:48.194 --> 00:13:51.264 align:center line:-1 position:50% size:29% And lower relief is much gentler slopes 00:13:51.264 --> 00:13:54.434 align:center line:-1 position:50% size:34% where it might be unstable and fail. 00:13:54.434 --> 00:13:56.335 align:center line:-1 position:50% size:40% And I've just got some nice pictures from Nepal 00:13:56.335 --> 00:13:58.171 align:center line:-1 position:50% size:22% to really illustrate what a beautiful 00:13:58.171 --> 00:14:02.308 align:center line:-1 position:50% size:34% high relief landslide terrain can look like. 00:14:02.308 --> 00:14:04.677 align:center line:-1 position:50% size:36% So this is sort of the theoretical underpinning 00:14:04.677 --> 00:14:06.179 align:center line:-1 position:50% size:37% for the slope relief threshold. 00:14:07.046 --> 00:14:12.018 align:center line:-1 position:50% size:42% And here's how Godt and others implemented it back in 2012. 00:14:12.018 --> 00:14:13.586 align:center line:-1 position:50% size:25% They plotted slope, 00:14:13.586 --> 00:14:14.987 align:center line:-1 position:50% size:28% cumulative frequency distribution of slope, 00:14:14.987 --> 00:14:16.389 align:center line:-1 position:50% size:35% for five different inventories 00:14:16.389 --> 00:14:19.025 align:center line:-1 position:50% size:37% from New Jersey to North Carolina to Oregon, 00:14:19.025 --> 00:14:20.860 align:center line:-1 position:50% size:20% New Mexico, San Francisco. 00:14:20.860 --> 00:14:23.696 align:center line:-1 position:50% size:38% And then also the same CDF of relief. 00:14:23.696 --> 00:14:25.364 align:center line:-1 position:50% size:25% And then they took the 10th percentile 00:14:25.364 --> 00:14:27.133 align:center line:-1 position:50% size:32% for each of those five inventories, 00:14:27.133 --> 00:14:29.135 align:center line:-1 position:50% size:41% plotted them against each other for slope and relief, 00:14:29.135 --> 00:14:32.238 align:center line:-1 position:50% size:39% and developed this really nice stepwise linear function. 00:14:32.238 --> 00:14:34.140 align:center line:-1 position:50% size:25% And you can see it works very nicely. 00:14:34.140 --> 00:14:36.275 align:center line:-1 position:50% size:38% Above that threshold, there's some potential for landslides. 00:14:36.275 --> 00:14:37.677 align:center line:-1 position:50% size:30% Below it, there is none. 00:14:39.312 --> 00:14:41.681 align:center line:-1 position:50% size:30% And that was based on 30,000 landslides 00:14:41.681 --> 00:14:44.817 align:center line:-1 position:50% size:31% from five high-quality inventories. 00:14:44.817 --> 00:14:48.187 align:center line:-1 position:50% size:37% Marchesini and others explored the same approach 00:14:48.187 --> 00:14:51.023 align:center line:-1 position:50% size:38% in the Mediterranean in 2014, 00:14:51.023 --> 00:14:53.826 align:center line:-1 position:50% size:42% and found that they could slightly improve the performance 00:14:53.826 --> 00:14:55.695 align:center line:-1 position:50% size:27% by developing this nonlinear model. 00:14:56.662 --> 00:14:57.764 align:center line:-1 position:50% size:32% Let's see what that looks like here. 00:14:57.764 --> 00:14:59.866 align:center line:-1 position:50% size:38% And then they collaborated with some Chinese scientists 00:14:59.866 --> 00:15:02.568 align:center line:-1 position:50% size:24% to extrapolate that across the globe, 00:15:02.568 --> 00:15:06.339 align:center line:-1 position:50% size:29% and that's that 2021 non-susceptibility map 00:15:06.339 --> 00:15:08.307 align:center line:-1 position:50% size:32% that I showed you earlier in the talk. 00:15:10.676 --> 00:15:13.412 align:center line:-1 position:50% size:31% So they also used about 40,000 landslides 00:15:13.412 --> 00:15:15.148 align:center line:-1 position:50% size:22% in their inventory. 00:15:15.148 --> 00:15:15.848 align:center line:-1 position:50% size:27% And so now we have 00:15:15.848 --> 00:15:17.617 align:center line:-1 position:50% size:38% this much greater landslide inventory approach, 00:15:17.617 --> 00:15:20.987 align:center line:-1 position:50% size:39% but we're inventory (inaudible) use this similar approach 00:15:20.987 --> 00:15:23.256 align:center line:-1 position:50% size:24% of slope and relief. 00:15:23.256 --> 00:15:24.690 align:center line:-1 position:50% size:26% So here's just an overall summary 00:15:24.690 --> 00:15:27.693 align:center line:-1 position:50% size:28% of how we improved or updated the model 00:15:27.693 --> 00:15:30.396 align:center line:-1 position:50% size:31% relative to the prototype developed in 2012. 00:15:31.764 --> 00:15:33.766 align:center line:-1 position:50% size:11% We've... 00:15:33.766 --> 00:15:36.202 align:center line:-1 position:50% size:29% basically an order of magnitude increase 00:15:36.202 --> 00:15:39.305 align:center line:-1 position:50% size:36% in terms of the number of landslides and resolution. 00:15:39.305 --> 00:15:44.477 align:center line:-1 position:50% size:31% And also, our analysis is at 10-meter elevation, 00:15:44.477 --> 00:15:46.212 align:center line:-1 position:50% size:41% so we're using 10-meter slopes. 00:15:46.212 --> 00:15:47.446 align:center line:-1 position:50% size:28% And I didn't have that on the slide here, 00:15:47.446 --> 00:15:49.482 align:center line:-1 position:50% size:28% but to calculate relief, 00:15:49.482 --> 00:15:51.284 align:center line:-1 position:50% size:37% we used a 100-meter moving window 00:15:51.284 --> 00:15:54.153 align:center line:-1 position:50% size:42% and then just differenced the highest and lowest elevation 00:15:54.153 --> 00:15:55.621 align:center line:-1 position:50% size:34% within that 100-meter moving window 00:15:55.621 --> 00:15:59.058 align:center line:-1 position:50% size:34% for each 10-meter grid-cell in our model. 00:15:59.058 --> 00:15:59.992 align:center line:-1 position:50% size:36% Then we downsampled that 00:15:59.992 --> 00:16:02.695 align:center line:-1 position:50% size:29% to produce a 90-meter output resolution map, 00:16:02.695 --> 00:16:03.696 align:center line:-1 position:50% size:18% and that just-- 00:16:03.696 --> 00:16:05.998 align:center line:-1 position:50% size:34% 90-meter coincides with some other products, 00:16:05.998 --> 00:16:07.567 align:center line:-1 position:50% size:32% such as population data, et cetera, 00:16:07.567 --> 00:16:08.801 align:center line:-1 position:50% size:36% that we thought was useful, 00:16:08.801 --> 00:16:10.336 align:center line:-1 position:50% size:34% and felt also that 10-meter 00:16:10.336 --> 00:16:12.171 align:center line:-1 position:50% size:24% was just a little too high-resolution 00:16:12.171 --> 00:16:13.806 align:center line:-1 position:50% size:28% for a national product. 00:16:15.174 --> 00:16:17.577 align:center line:-1 position:50% size:34% It covers CONUS, Alaska, Hawaii, and Puerto Rico. 00:16:17.577 --> 00:16:18.344 align:center line:-1 position:50% size:23% And I should note 00:16:18.344 --> 00:16:20.346 align:center line:-1 position:50% size:31% that we could do this for other U.S. territories, 00:16:20.346 --> 00:16:22.315 align:center line:-1 position:50% size:28% but there just isn't enough landslide 00:16:22.315 --> 00:16:23.616 align:center line:-1 position:50% size:29% and high-res topo data 00:16:23.616 --> 00:16:28.087 align:center line:-1 position:50% size:32% to do it for all the other U.S. territories. 00:16:30.089 --> 00:16:32.058 align:center line:-1 position:50% size:30% Godt et al. just sort of explored this approach 00:16:32.058 --> 00:16:33.826 align:center line:-1 position:50% size:29% and demonstrated that it's very useful, 00:16:33.826 --> 00:16:35.561 align:center line:-1 position:50% size:38% and we wanted to really have a robust model 00:16:35.561 --> 00:16:37.396 align:center line:-1 position:50% size:32% that we were confident was useful for prediction, 00:16:37.396 --> 00:16:41.133 align:center line:-1 position:50% size:36% and so we did quite a bit of calibration and validation. 00:16:41.133 --> 00:16:43.202 align:center line:-1 position:50% size:31% We evaluated first four alternative models, 00:16:43.202 --> 00:16:44.303 align:center line:-1 position:50% size:27% linear and nonlinear, 00:16:44.303 --> 00:16:46.005 align:center line:-1 position:50% size:37% I'll explain those in a second. 00:16:46.005 --> 00:16:46.839 align:center line:-1 position:50% size:34% For each of those models, 00:16:46.839 --> 00:16:48.374 align:center line:-1 position:50% size:36% we did 50 independent runs 00:16:48.374 --> 00:16:51.110 align:center line:-1 position:50% size:31% using a 70/30 split sample calibration. 00:16:51.110 --> 00:16:53.346 align:center line:-1 position:50% size:22% So, really robust. 00:16:53.346 --> 00:16:55.548 align:center line:-1 position:50% size:40% And then used the mean value of those 50 iterations 00:16:55.548 --> 00:16:57.049 align:center line:-1 position:50% size:37% to evaluate different models. 00:16:57.817 --> 00:17:00.720 align:center line:-1 position:50% size:26% And then, after that, 00:17:00.720 --> 00:17:02.655 align:center line:-1 position:50% size:26% we started this work a few years ago, 00:17:02.655 --> 00:17:05.157 align:center line:-1 position:50% size:31% where we had the 2022 inventory compilation, 00:17:05.157 --> 00:17:06.058 align:center line:-1 position:50% size:26% but in the meantime 00:17:06.058 --> 00:17:08.828 align:center line:-1 position:50% size:33% identified some additional statewide inventories. 00:17:08.828 --> 00:17:09.829 align:center line:-1 position:50% size:20% And in the end, 00:17:09.829 --> 00:17:12.765 align:center line:-1 position:50% size:25% had an additional 195,000 landslides 00:17:12.765 --> 00:17:14.100 align:center line:-1 position:50% size:30% that we could use for independent testing 00:17:14.100 --> 00:17:14.700 align:center line:-1 position:50% size:17% of our model. 00:17:14.700 --> 00:17:17.303 align:center line:-1 position:50% size:20% So really a nice extra validation 00:17:17.303 --> 00:17:18.304 align:center line:-1 position:50% size:22% of our approach. 00:17:18.304 --> 00:17:20.873 align:center line:-1 position:50% size:35% And I'll talk about all of that here shortly. 00:17:22.508 --> 00:17:24.877 align:center line:-1 position:50% size:32% There was a lot of data pre-processing involved, 00:17:24.877 --> 00:17:25.645 align:center line:-1 position:50% size:23% though, first of all, 00:17:25.645 --> 00:17:27.613 align:center line:-1 position:50% size:20% to just extract slope and relief 00:17:27.613 --> 00:17:29.248 align:center line:-1 position:50% size:33% across the entire country, 00:17:29.248 --> 00:17:32.518 align:center line:-1 position:50% size:29% and then assign that to all of the landslides. 00:17:32.518 --> 00:17:34.854 align:center line:-1 position:50% size:26% We did have to turn every landslide 00:17:34.854 --> 00:17:38.758 align:center line:-1 position:50% size:38% into a 10-meter point or grid-cell across the nation. 00:17:38.758 --> 00:17:41.127 align:center line:-1 position:50% size:27% And so for polygons, 00:17:41.127 --> 00:17:44.630 align:center line:-1 position:50% size:35% we decided to use the highest elevation point. 00:17:44.630 --> 00:17:46.832 align:center line:-1 position:50% size:36% Now some of the polygons in the inventory compilation 00:17:46.832 --> 00:17:50.069 align:center line:-1 position:50% size:32% are alluvial fans or debris flows deposits. 00:17:50.069 --> 00:17:51.704 align:center line:-1 position:50% size:28% Some of them are the entire polygon 00:17:51.704 --> 00:17:54.674 align:center line:-1 position:50% size:25% of a landslide from source to toe. 00:17:54.674 --> 00:17:56.575 align:center line:-1 position:50% size:36% Some of them are just the initiation zones. 00:17:58.344 --> 00:17:59.779 align:center line:-1 position:50% size:27% Some of them are property outlines 00:17:59.779 --> 00:18:01.580 align:center line:-1 position:50% size:26% that were impacted, for example. 00:18:01.580 --> 00:18:04.050 align:center line:-1 position:50% size:38% So there's a lot of heterogeneity in our data set, 00:18:04.050 --> 00:18:05.851 align:center line:-1 position:50% size:41% and there's many different ways we could have approached this. 00:18:05.851 --> 00:18:08.287 align:center line:-1 position:50% size:35% In the end, we thought that the highest elevation point 00:18:08.287 --> 00:18:10.823 align:center line:-1 position:50% size:42% was the most robust assumption we could make. 00:18:12.658 --> 00:18:15.661 align:center line:-1 position:50% size:30% So, there's lots and lots of landslide data, 00:18:15.661 --> 00:18:16.796 align:center line:-1 position:50% size:29% and I think there's a lot of things-- 00:18:16.796 --> 00:18:19.332 align:center line:-1 position:50% size:24% you can dig into just the data alone 00:18:19.332 --> 00:18:21.600 align:center line:-1 position:50% size:33% and have some interesting insights. 00:18:21.600 --> 00:18:22.835 align:center line:-1 position:50% size:27% But there are several 00:18:22.835 --> 00:18:26.105 align:center line:-1 position:50% size:35% of the U.S. North American ecoregions 00:18:26.105 --> 00:18:27.106 align:center line:-1 position:50% size:27% that really dominate, 00:18:27.106 --> 00:18:28.674 align:center line:-1 position:50% size:28% especially in the Pacific Northwest, 00:18:28.674 --> 00:18:32.178 align:center line:-1 position:50% size:32% West Coast, Appalachia, and Puerto Rico. 00:18:33.112 --> 00:18:34.547 align:center line:-1 position:50% size:41% So we looked at the percentage of landslides, 00:18:34.547 --> 00:18:35.815 align:center line:-1 position:50% size:41% but then also the concentration. 00:18:35.815 --> 00:18:37.683 align:center line:-1 position:50% size:38% You can see that Puerto Rico is just off the map 00:18:37.683 --> 00:18:39.251 align:center line:-1 position:50% size:39% in terms of their concentration of landslides, 00:18:39.251 --> 00:18:42.054 align:center line:-1 position:50% size:29% and that was--reflects the devastating effects 00:18:42.054 --> 00:18:44.390 align:center line:-1 position:50% size:25% of Hurricane Maria. 00:18:47.259 --> 00:18:48.794 align:center line:-1 position:50% size:33% Looking at the cumulative frequency distributions, 00:18:48.794 --> 00:18:50.463 align:center line:-1 position:50% size:24% you can see, wow, 00:18:50.463 --> 00:18:51.697 align:center line:-1 position:50% size:35% there's a lot of landslides that are actually happening 00:18:51.697 --> 00:18:55.001 align:center line:-1 position:50% size:23% at very low slopes and relief values. 00:18:57.169 --> 00:18:58.704 align:center line:-1 position:50% size:35% But then there's also some 00:18:58.704 --> 00:19:00.473 align:center line:-1 position:50% size:30% at very high slope and relief values. 00:19:00.473 --> 00:19:02.875 align:center line:-1 position:50% size:25% So, this was sort of the starting point 00:19:02.875 --> 00:19:05.177 align:center line:-1 position:50% size:35% just of the data for looking at linear models 00:19:05.177 --> 00:19:07.380 align:center line:-1 position:50% size:34% in the same way that Godt and others did in 2012, 00:19:07.380 --> 00:19:10.449 align:center line:-1 position:50% size:34% and Marchesini et al. did for the Mediterranean. 00:19:10.449 --> 00:19:13.285 align:center line:-1 position:50% size:40% And we used the ecoregions instead of the state inventories 00:19:13.285 --> 00:19:16.722 align:center line:-1 position:50% size:36% to constrain those linear threshold calibrations. 00:19:16.722 --> 00:19:19.558 align:center line:-1 position:50% size:38% We also did look at weighting by ecoregion density. 00:19:19.558 --> 00:19:20.693 align:center line:-1 position:50% size:24% But for that model, 00:19:20.693 --> 00:19:23.229 align:center line:-1 position:50% size:35% we did remove Puerto Rico 00:19:23.229 --> 00:19:24.296 align:center line:-1 position:50% size:30% just because otherwise 00:19:24.296 --> 00:19:26.465 align:center line:-1 position:50% size:27% it turns into a Puerto Rico model. 00:19:27.800 --> 00:19:31.237 align:center line:-1 position:50% size:37% So here are the four different linear and nonlinear models 00:19:31.237 --> 00:19:32.571 align:center line:-1 position:50% size:26% that we considered. 00:19:32.571 --> 00:19:35.674 align:center line:-1 position:50% size:26% There's a slope linear function relief, 00:19:35.674 --> 00:19:38.577 align:center line:-1 position:50% size:38% and then a nonlinear function of slope and relief. 00:19:38.577 --> 00:19:40.513 align:center line:-1 position:50% size:40% And we looked at the weighted and unweighted. 00:19:40.513 --> 00:19:43.249 align:center line:-1 position:50% size:38% So the unweighted is just like what Godt and others did. 00:19:43.249 --> 00:19:45.051 align:center line:-1 position:50% size:20% The weighted, as I mentioned, 00:19:45.051 --> 00:19:48.287 align:center line:-1 position:50% size:41% based on ecoregion density of landslides in each ecoregion, 00:19:48.287 --> 00:19:49.555 align:center line:-1 position:50% size:25% minus Puerto Rico. 00:19:49.555 --> 00:19:51.390 align:center line:-1 position:50% size:41% And then the nonlinear models, 00:19:51.390 --> 00:19:53.626 align:center line:-1 position:50% size:35% we looked at the fifth percentile quantile, 00:19:53.626 --> 00:19:55.061 align:center line:-1 position:50% size:40% just like Marchesini and others, 00:19:55.061 --> 00:19:56.762 align:center line:-1 position:50% size:35% and then we looked at a slightly less conservative 00:19:56.762 --> 00:19:58.764 align:center line:-1 position:50% size:21% tenth percentile, 00:19:58.764 --> 00:20:00.533 align:center line:-1 position:50% size:36% kind of more similar to what Godt and others did 00:20:00.533 --> 00:20:01.667 align:center line:-1 position:50% size:26% for the linear model. 00:20:02.668 --> 00:20:03.869 align:center line:-1 position:50% size:33% On the figure on the right, 00:20:03.869 --> 00:20:07.807 align:center line:-1 position:50% size:38% you can see the 50 trial runs, or iterations, 00:20:07.807 --> 00:20:11.377 align:center line:-1 position:50% size:28% of that model, 70/30 split calibration. 00:20:11.377 --> 00:20:13.379 align:center line:-1 position:50% size:40% And the linear weighted model, 00:20:13.379 --> 00:20:14.480 align:center line:-1 position:50% size:21% the orange one, 00:20:14.480 --> 00:20:16.449 align:center line:-1 position:50% size:37% looks really well-constrained. 00:20:16.449 --> 00:20:19.485 align:center line:-1 position:50% size:37% The linear unweighted didn't really have as much success 00:20:19.485 --> 00:20:23.556 align:center line:-1 position:50% size:35% in converging on a uniform parameter value. 00:20:23.556 --> 00:20:24.623 align:center line:-1 position:50% size:25% And for that and other reasons, 00:20:24.623 --> 00:20:27.159 align:center line:-1 position:50% size:29% we decided not to use that model. 00:20:27.159 --> 00:20:27.860 align:center line:-1 position:50% size:29% The nonlinear models, 00:20:27.860 --> 00:20:29.662 align:center line:-1 position:50% size:21% also really tight calibration here, 00:20:29.662 --> 00:20:30.763 align:center line:-1 position:50% size:17% you can see. 00:20:30.763 --> 00:20:33.432 align:center line:-1 position:50% size:25% And so, in the end, 00:20:33.432 --> 00:20:34.700 align:center line:-1 position:50% size:34% moving into the next stage 00:20:34.700 --> 00:20:36.535 align:center line:-1 position:50% size:27% of evaluation and model selection, 00:20:36.535 --> 00:20:39.305 align:center line:-1 position:50% size:38% we eliminated that linear unweighted model. 00:20:40.639 --> 00:20:43.042 align:center line:-1 position:50% size:33% So here's what all four of those look like, though. 00:20:43.042 --> 00:20:45.344 align:center line:-1 position:50% size:24% And this considers a downsampling 00:20:45.344 --> 00:20:48.814 align:center line:-1 position:50% size:39% from 10-meter analysis to our 90-meter model output. 00:20:50.249 --> 00:20:52.251 align:center line:-1 position:50% size:32% You can see plotted here slope and relief 00:20:52.251 --> 00:20:54.487 align:center line:-1 position:50% size:25% of all the landslides in our database. 00:20:54.487 --> 00:20:55.254 align:center line:-1 position:50% size:28% Number of landslides 00:20:55.254 --> 00:20:57.156 align:center line:-1 position:50% size:29% on each sort of little binned pixel here. 00:20:58.390 --> 00:21:00.059 align:center line:-1 position:50% size:28% The linear weighted-- 00:21:00.059 --> 00:21:01.794 align:center line:-1 position:50% size:32% linear unweighted model is the highest, 00:21:01.794 --> 00:21:03.462 align:center line:-1 position:50% size:27% and this is a lot of landslides as well. 00:21:03.462 --> 00:21:05.397 align:center line:-1 position:50% size:34% So not only was it kind of poorly constrained, 00:21:05.397 --> 00:21:08.701 align:center line:-1 position:50% size:29% it also doesn't capture a lot of the landslides. 00:21:08.701 --> 00:21:10.703 align:center line:-1 position:50% size:35% Linear weighted does a really nice job here. 00:21:10.703 --> 00:21:13.172 align:center line:-1 position:50% size:31% This is a little orange dotted line you can see. 00:21:13.172 --> 00:21:16.375 align:center line:-1 position:50% size:41% And through the range of sort of normal slope and relief values, 00:21:16.375 --> 00:21:20.346 align:center line:-1 position:50% size:34% it's very similar to the N10, or the nonlinear 10 model. 00:21:21.480 --> 00:21:23.482 align:center line:-1 position:50% size:30% And then the N5 model is the lowest. 00:21:23.482 --> 00:21:25.818 align:center line:-1 position:50% size:38% You can see, captures the most landslides. 00:21:25.818 --> 00:21:26.986 align:center line:-1 position:50% size:33% So that's really attractive. 00:21:26.986 --> 00:21:28.521 align:center line:-1 position:50% size:30% And overall we wanted to develop a model 00:21:28.521 --> 00:21:30.589 align:center line:-1 position:50% size:26% that captured the most landslides, 00:21:30.589 --> 00:21:32.825 align:center line:-1 position:50% size:32% but we also want to take some false positives 00:21:32.825 --> 00:21:34.627 align:center line:-1 position:50% size:24% into consideration. 00:21:34.627 --> 00:21:37.530 align:center line:-1 position:50% size:40% And also think about where are the false negatives occurring, 00:21:37.530 --> 00:21:40.332 align:center line:-1 position:50% size:41% and that's what I'm gonna show in the next slide here. 00:21:43.102 --> 00:21:44.436 align:center line:-1 position:50% size:30% When you're looking at continental scales 00:21:44.436 --> 00:21:47.706 align:center line:-1 position:50% size:28% in a very clearly incomplete inventory, 00:21:47.706 --> 00:21:48.574 align:center line:-1 position:50% size:7% we... 00:21:51.277 --> 00:21:55.481 align:center line:-1 position:50% size:28% can't really, truly have confidence 00:21:55.481 --> 00:21:57.016 align:center line:-1 position:50% size:22% that a landslide hasn't happened 00:21:57.016 --> 00:21:59.218 align:center line:-1 position:50% size:31% just because we haven't observed it yet. 00:21:59.218 --> 00:22:02.188 align:center line:-1 position:50% size:32% And so while we can say 00:22:02.188 --> 00:22:03.222 align:center line:-1 position:50% size:23% with some degree of confidence 00:22:03.222 --> 00:22:04.089 align:center line:-1 position:50% size:20% that landslides have happened 00:22:04.089 --> 00:22:05.524 align:center line:-1 position:50% size:23% where we haven't observed them, 00:22:05.524 --> 00:22:07.860 align:center line:-1 position:50% size:34% we can't use all the non-landslide pixels 00:22:07.860 --> 00:22:13.165 align:center line:-1 position:50% size:37% within our model as negative or non-landslides. 00:22:13.165 --> 00:22:16.468 align:center line:-1 position:50% size:28% And so instead of actually calculating 00:22:16.468 --> 00:22:19.238 align:center line:-1 position:50% size:28% these false negatives or false positives, 00:22:19.238 --> 00:22:21.173 align:center line:-1 position:50% size:37% we use the true positive rate 00:22:21.173 --> 00:22:25.244 align:center line:-1 position:50% size:37% and then the area of landslide-susceptible terrain. 00:22:25.244 --> 00:22:27.880 align:center line:-1 position:50% size:38% And that's what's shown here in the upper left-hand plot. 00:22:27.880 --> 00:22:30.683 align:center line:-1 position:50% size:38% We've got CONUS landslides captured by the N5, N10, 00:22:30.683 --> 00:22:32.518 align:center line:-1 position:50% size:36% and linear weighted models, 00:22:32.518 --> 00:22:34.253 align:center line:-1 position:50% size:33% and then the percent area covered 00:22:34.253 --> 00:22:36.422 align:center line:-1 position:50% size:40% by those respective thresholds. 00:22:36.422 --> 00:22:38.490 align:center line:-1 position:50% size:28% And what we can see is that we only lose 00:22:38.490 --> 00:22:41.093 align:center line:-1 position:50% size:28% not even a percent of the total landslides 00:22:41.093 --> 00:22:42.194 align:center line:-1 position:50% size:21% in the inventory, 00:22:42.194 --> 00:22:43.996 align:center line:-1 position:50% size:35% but are able to reduce the area covered 00:22:43.996 --> 00:22:45.731 align:center line:-1 position:50% size:22% by eight percent. 00:22:45.731 --> 00:22:47.166 align:center line:-1 position:50% size:22% And that really greatly increases 00:22:47.166 --> 00:22:49.468 align:center line:-1 position:50% size:34% the frequency ratio, which are also calculated. 00:22:49.468 --> 00:22:51.437 align:center line:-1 position:50% size:27% So, in the end, we sort of tossed out 00:22:51.437 --> 00:22:53.105 align:center line:-1 position:50% size:20% this N5 model for that reason, 00:22:53.105 --> 00:22:54.740 align:center line:-1 position:50% size:31% and there's just too much area covered, 00:22:54.740 --> 00:22:57.509 align:center line:-1 position:50% size:31% even though it captures slightly more landslides, 00:22:57.509 --> 00:23:00.646 align:center line:-1 position:50% size:22% but imperceptibly slightly more. 00:23:00.646 --> 00:23:04.416 align:center line:-1 position:50% size:38% We then compared the N10 and LW models here 00:23:04.416 --> 00:23:07.086 align:center line:-1 position:50% size:37% to the four statewide inventories that I mentioned. 00:23:07.086 --> 00:23:09.555 align:center line:-1 position:50% size:27% We have inventories from Idaho, Maine, 00:23:09.555 --> 00:23:10.856 align:center line:-1 position:50% size:23% North Dakota, and West Virginia 00:23:10.856 --> 00:23:12.825 align:center line:-1 position:50% size:33% that were not in our calibration data set 00:23:12.825 --> 00:23:15.094 align:center line:-1 position:50% size:31% but we used for independent testing, 00:23:15.094 --> 00:23:17.563 align:center line:-1 position:50% size:38% and what you can see in here is that in each case, 00:23:17.563 --> 00:23:19.865 align:center line:-1 position:50% size:26% the N10 model does slightly better 00:23:19.865 --> 00:23:22.668 align:center line:-1 position:50% size:27% at capturing all those additional landslides. 00:23:22.668 --> 00:23:24.069 align:center line:-1 position:50% size:35% And then we look at Maine, 00:23:24.069 --> 00:23:26.805 align:center line:-1 position:50% size:30% just the 405 landslides mapped across Maine. 00:23:26.805 --> 00:23:30.376 align:center line:-1 position:50% size:32% Our N10 model captures all the modern ones 00:23:30.376 --> 00:23:32.378 align:center line:-1 position:50% size:38% and quite a few of the historic and prehistoric landslides 00:23:32.378 --> 00:23:37.149 align:center line:-1 position:50% size:32% that occurred post-glaciation in Maine. 00:23:37.149 --> 00:23:42.354 align:center line:-1 position:50% size:33% So that sort of addresses the true-positive 00:23:42.354 --> 00:23:44.857 align:center line:-1 position:50% size:31% and false-positive part of evaluating our model, 00:23:44.857 --> 00:23:48.827 align:center line:-1 position:50% size:38% and it's looking like the N10 is the one to choose. 00:23:48.827 --> 00:23:50.195 align:center line:-1 position:50% size:34% But we also wanted to see 00:23:50.195 --> 00:23:52.998 align:center line:-1 position:50% size:25% where were we missing landslides, 00:23:52.998 --> 00:23:54.800 align:center line:-1 position:50% size:34% and so, we did that to some extent 00:23:54.800 --> 00:23:56.502 align:center line:-1 position:50% size:35% with the proximity analysis. 00:23:56.502 --> 00:23:59.538 align:center line:-1 position:50% size:40% So this is just the 1% of landslides in our compilation 00:23:59.538 --> 00:24:01.874 align:center line:-1 position:50% size:24% that we missed with those models, 00:24:01.874 --> 00:24:06.378 align:center line:-1 position:50% size:29% and what you can see, essentially, from this 00:24:06.378 --> 00:24:08.881 align:center line:-1 position:50% size:32% is that almost all the missed landslides 00:24:08.881 --> 00:24:11.583 align:center line:-1 position:50% size:29% are within one to three grid cells, 00:24:11.583 --> 00:24:13.319 align:center line:-1 position:50% size:24% 90-meter grid cells in our model, 00:24:13.319 --> 00:24:17.222 align:center line:-1 position:50% size:39% so less than 300 meters away from some susceptible terrain. 00:24:17.222 --> 00:24:18.557 align:center line:-1 position:50% size:31% So that comes into play 00:24:18.557 --> 00:24:21.126 align:center line:-1 position:50% size:28% when you're applying these models 00:24:21.126 --> 00:24:22.728 align:center line:-1 position:50% size:25% across the country. 00:24:24.396 --> 00:24:26.365 align:center line:-1 position:50% size:20% So, this shows, 00:24:26.365 --> 00:24:27.666 align:center line:-1 position:50% size:29% this gray shaded area, 00:24:27.666 --> 00:24:29.034 align:center line:-1 position:50% size:24% shows everywhere 00:24:29.034 --> 00:24:31.003 align:center line:-1 position:50% size:26% in the United States and Puerto Rico 00:24:31.003 --> 00:24:33.505 align:center line:-1 position:50% size:25% that exceeds that N10 threshold, 00:24:33.505 --> 00:24:36.542 align:center line:-1 position:50% size:35% with the white being the negligible susceptibility. 00:24:36.542 --> 00:24:38.877 align:center line:-1 position:50% size:31% So that's about 42% of the land surface area 00:24:38.877 --> 00:24:41.613 align:center line:-1 position:50% size:28% across the study area 00:24:41.613 --> 00:24:43.882 align:center line:-1 position:50% size:34% and 99% of the landslides. 00:24:43.882 --> 00:24:45.250 align:center line:-1 position:50% size:25% So, this is fantastic, 00:24:45.250 --> 00:24:47.853 align:center line:-1 position:50% size:35% this is a great improvement in a number of ways 00:24:47.853 --> 00:24:52.091 align:center line:-1 position:50% size:23% over the previous prototype model. 00:24:52.091 --> 00:24:53.359 align:center line:-1 position:50% size:26% But we also realized 00:24:53.359 --> 00:24:54.727 align:center line:-1 position:50% size:24% that there was a lot of information 00:24:54.727 --> 00:24:57.796 align:center line:-1 position:50% size:40% that we were throwing out in our down-sampling analysis. 00:24:57.796 --> 00:24:59.365 align:center line:-1 position:50% size:35% So I think I mentioned this, 00:24:59.365 --> 00:25:02.167 align:center line:-1 position:50% size:41% that we used 10-meter analysis, 00:25:02.167 --> 00:25:03.836 align:center line:-1 position:50% size:36% and then we down sampled to 90 meters 00:25:03.836 --> 00:25:06.071 align:center line:-1 position:50% size:32% such that anywhere within the 90-meter grid, 00:25:06.071 --> 00:25:10.142 align:center line:-1 position:50% size:39% if there was one or more cells, 10-meter cells, 00:25:10.142 --> 00:25:12.344 align:center line:-1 position:50% size:37% that exceeded the threshold, 00:25:12.344 --> 00:25:15.247 align:center line:-1 position:50% size:25% then we'd consider that entire cell 00:25:15.247 --> 00:25:17.449 align:center line:-1 position:50% size:23% to be susceptible. 00:25:17.449 --> 00:25:18.684 align:center line:-1 position:50% size:39% But there's a lot of information 00:25:18.684 --> 00:25:21.086 align:center line:-1 position:50% size:28% that we were sort of throwing away in that, 00:25:21.086 --> 00:25:22.821 align:center line:-1 position:50% size:31% and that you could have a 90-meter cell 00:25:22.821 --> 00:25:24.623 align:center line:-1 position:50% size:25% with only one pixel, 00:25:24.623 --> 00:25:26.759 align:center line:-1 position:50% size:24% one 10-meter pixel exceeding that, 00:25:26.759 --> 00:25:29.261 align:center line:-1 position:50% size:37% or you could have all 81 of those 10x10 meters 00:25:29.261 --> 00:25:30.729 align:center line:-1 position:50% size:32% exceeding the threshold. 00:25:30.729 --> 00:25:32.631 align:center line:-1 position:50% size:34% So we were curious whether this might give us 00:25:32.631 --> 00:25:35.501 align:center line:-1 position:50% size:33% more insight into variable susceptibility 00:25:35.501 --> 00:25:37.569 align:center line:-1 position:50% size:25% across the country. 00:25:37.569 --> 00:25:41.273 align:center line:-1 position:50% size:33% So we started looking at this range from 0 to 81 00:25:41.273 --> 00:25:43.876 align:center line:-1 position:50% size:41% in terms of does it actually add some value 00:25:43.876 --> 00:25:45.110 align:center line:-1 position:50% size:28% in landslide prediction 00:25:45.110 --> 00:25:47.379 align:center line:-1 position:50% size:31% and kind of constraining relative susceptibility? 00:25:47.379 --> 00:25:49.782 align:center line:-1 position:50% size:39% So, here, we have that shown. 00:25:49.782 --> 00:25:53.352 align:center line:-1 position:50% size:36% The graph on the left shows cumulative percent 00:25:53.352 --> 00:25:55.087 align:center line:-1 position:50% size:24% and the cell count. 00:25:55.087 --> 00:25:57.689 align:center line:-1 position:50% size:32% And so, the blue line is the cumulative percent 00:25:57.689 --> 00:25:59.091 align:center line:-1 position:50% size:29% of landslides captured, 00:25:59.091 --> 00:26:01.360 align:center line:-1 position:50% size:40% and the green line is the land surface area covered, 00:26:01.360 --> 00:26:03.796 align:center line:-1 position:50% size:39% and so, when you have 0 or 1, 00:26:03.796 --> 00:26:06.598 align:center line:-1 position:50% size:35% a threshold of 1 cell count-- 00:26:06.598 --> 00:26:08.534 align:center line:-1 position:50% size:37% this is essentially our threshold model in gray-- 00:26:08.534 --> 00:26:10.436 align:center line:-1 position:50% size:32% it's 42% of the land surface area, 00:26:10.436 --> 00:26:13.338 align:center line:-1 position:50% size:31% 1 minus the value here, 00:26:13.338 --> 00:26:15.073 align:center line:-1 position:50% size:41% and then 99% of the landslides. 00:26:15.073 --> 00:26:16.241 align:center line:-1 position:50% size:37% But what you can see is that 00:26:16.241 --> 00:26:17.743 align:center line:-1 position:50% size:24% the vast majority of these landslides 00:26:17.743 --> 00:26:20.512 align:center line:-1 position:50% size:39% are all happening in this much higher cell count, 00:26:20.512 --> 00:26:24.283 align:center line:-1 position:50% size:35% and that that's actually not really a very large area 00:26:24.283 --> 00:26:26.652 align:center line:-1 position:50% size:26% of the United States that's covered. 00:26:26.652 --> 00:26:29.121 align:center line:-1 position:50% size:35% We're going to drill into that a little bit more. 00:26:29.121 --> 00:26:32.057 align:center line:-1 position:50% size:39% I want to stress that the model that we're releasing 00:26:32.057 --> 00:26:35.461 align:center line:-1 position:50% size:41% and the analysis is a continuum from 0 to 81, 00:26:35.461 --> 00:26:36.762 align:center line:-1 position:50% size:32% but for the colored maps, 00:26:36.762 --> 00:26:38.997 align:center line:-1 position:50% size:30% I'm just gonna show equal intervals 1 to 15, 00:26:38.997 --> 00:26:44.336 align:center line:-1 position:50% size:29% 16 to 31, 32 to 47, and 48 to 63, 64 to 81 00:26:44.336 --> 00:26:46.705 align:center line:-1 position:50% size:33% being these sort of five equal interval breaks. 00:26:46.705 --> 00:26:50.175 align:center line:-1 position:50% size:44% That's shown for the Puget Sound region on the right here. 00:26:50.175 --> 00:26:51.643 align:center line:-1 position:50% size:25% You can nicely see, for example, 00:26:51.643 --> 00:26:57.816 align:center line:-1 position:50% size:34% the coastal bluffs around Edmonds, Everett, 00:26:57.816 --> 00:27:02.354 align:center line:-1 position:50% size:39% and along West Seattle are all very high susceptibility, 00:27:02.354 --> 00:27:03.989 align:center line:-1 position:50% size:28% relative to the interior. 00:27:03.989 --> 00:27:05.557 align:center line:-1 position:50% size:34% And then, as you move up into the Cascades 00:27:05.557 --> 00:27:06.859 align:center line:-1 position:50% size:30% towards Mount Rainier, 00:27:06.859 --> 00:27:09.561 align:center line:-1 position:50% size:32% we again have these very high concentrations 00:27:09.561 --> 00:27:11.497 align:center line:-1 position:50% size:26% of susceptible cells. 00:27:13.665 --> 00:27:17.035 align:center line:-1 position:50% size:31% So, now, we have this national map that is 00:27:17.035 --> 00:27:18.237 align:center line:-1 position:50% size:30% a little bit more colorful, 00:27:18.237 --> 00:27:19.638 align:center line:-1 position:50% size:31% a little bit more exciting, 00:27:19.638 --> 00:27:23.242 align:center line:-1 position:50% size:37% and also sort of nicely shows what intuitively we know 00:27:23.242 --> 00:27:25.511 align:center line:-1 position:50% size:35% about landslide occurrence and susceptibility 00:27:25.511 --> 00:27:26.545 align:center line:-1 position:50% size:20% across the U.S. 00:27:26.545 --> 00:27:29.248 align:center line:-1 position:50% size:29% It really highlights that, for example, 00:27:29.248 --> 00:27:31.817 align:center line:-1 position:50% size:28% the Appalachians, the Rocky Mountains, 00:27:31.817 --> 00:27:33.585 align:center line:-1 position:50% size:31% the Marine West Coast, et cetera, 00:27:33.585 --> 00:27:37.222 align:center line:-1 position:50% size:39% are much higher susceptibility relative to... 00:27:37.222 --> 00:27:38.490 align:center line:-1 position:50% size:32% you know, there is still some hazard 00:27:38.490 --> 00:27:42.027 align:center line:-1 position:50% size:36% in Minnesota, North Dakota, central United States, 00:27:42.027 --> 00:27:44.796 align:center line:-1 position:50% size:36% but it's just quite as dark red as some of the areas 00:27:44.796 --> 00:27:47.266 align:center line:-1 position:50% size:33% where we know there's lots of landsliding. 00:27:47.266 --> 00:27:50.068 align:center line:-1 position:50% size:32% Southeast Alaska, again, for example. 00:27:50.068 --> 00:27:51.770 align:center line:-1 position:50% size:16% Puerto Rico. 00:27:51.770 --> 00:27:54.673 align:center line:-1 position:50% size:37% So this is now our new map, and it's really exciting. 00:27:54.673 --> 00:27:57.576 align:center line:-1 position:50% size:35% I'm very happy this is going to be released soon. 00:27:57.576 --> 00:28:00.379 align:center line:-1 position:50% size:40% But we wanted to compare this to existing models. 00:28:00.379 --> 00:28:02.614 align:center line:-1 position:50% size:35% So this is just for the four models I talked about here, 00:28:02.614 --> 00:28:05.050 align:center line:-1 position:50% size:32% the Radbruch-Hall, et al. landslide overview map, 00:28:05.050 --> 00:28:09.388 align:center line:-1 position:50% size:33% the Godt, et al., prototype nonsusceptibility map, 00:28:09.388 --> 00:28:13.158 align:center line:-1 position:50% size:23% NASA's heuristic fuzzy logic model, 00:28:13.158 --> 00:28:17.496 align:center line:-1 position:50% size:28% and the Jia, et al., nonsusceptibility map 00:28:17.496 --> 00:28:20.465 align:center line:-1 position:50% size:34% relative to our N10 model. 00:28:20.465 --> 00:28:22.234 align:center line:-1 position:50% size:33% What we're showing here on the left column 00:28:22.234 --> 00:28:24.836 align:center line:-1 position:50% size:29% is the landslides by susceptibility class, 00:28:24.836 --> 00:28:26.371 align:center line:-1 position:50% size:21% and on the right, 00:28:26.371 --> 00:28:27.873 align:center line:-1 position:50% size:31% the land area covered by those corresponding 00:28:27.873 --> 00:28:29.541 align:center line:-1 position:50% size:28% susceptibility classes. 00:28:29.541 --> 00:28:31.176 align:center line:-1 position:50% size:26% All the existing ones have this sort of 00:28:31.176 --> 00:28:34.046 align:center line:-1 position:50% size:28% qualitative description that, to me, 00:28:34.046 --> 00:28:35.480 align:center line:-1 position:50% size:23% is in some ways very unsatisfying, 00:28:35.480 --> 00:28:36.848 align:center line:-1 position:50% size:32% but a reviewer asked me 00:28:36.848 --> 00:28:39.718 align:center line:-1 position:50% size:23% to specifically give some values 00:28:39.718 --> 00:28:40.819 align:center line:-1 position:50% size:31% to these equal intervals, 00:28:40.819 --> 00:28:43.388 align:center line:-1 position:50% size:42% so you could really put whatever qualitative descriptors 00:28:43.388 --> 00:28:45.257 align:center line:-1 position:50% size:26% make sense to you, 00:28:45.257 --> 00:28:46.625 align:center line:-1 position:50% size:21% but essentially, what I really like 00:28:46.625 --> 00:28:48.994 align:center line:-1 position:50% size:37% is that we have these nice, continuous numerical values 00:28:48.994 --> 00:28:49.828 align:center line:-1 position:50% size:27% for our susceptibility, 00:28:49.828 --> 00:28:54.132 align:center line:-1 position:50% size:34% rather than the qualitative descriptors. 00:28:54.132 --> 00:28:55.667 align:center line:-1 position:50% size:21% The big thing that jumps out-- 00:28:55.667 --> 00:28:58.470 align:center line:-1 position:50% size:36% there's lots of different ways you can look at this, 00:28:58.470 --> 00:29:01.239 align:center line:-1 position:50% size:31% but I'm gonna point out two really critical things. 00:29:01.239 --> 00:29:04.209 align:center line:-1 position:50% size:41% One, the best thing, in my view, about our model, 00:29:04.209 --> 00:29:05.811 align:center line:-1 position:50% size:21% is that the most, 00:29:05.811 --> 00:29:07.579 align:center line:-1 position:50% size:33% or the highest percentage of landslides, 00:29:07.579 --> 00:29:11.249 align:center line:-1 position:50% size:35% are observed in the highest susceptibility class. 00:29:11.249 --> 00:29:13.185 align:center line:-1 position:50% size:31% And that is not the case for the other two, 00:29:13.185 --> 00:29:14.620 align:center line:-1 position:50% size:38% Radbruch-Hall, et al., and Stanley and Kirschbaum, 00:29:14.620 --> 00:29:16.722 align:center line:-1 position:50% size:28% that have multiple susceptibility classes. 00:29:16.722 --> 00:29:18.223 align:center line:-1 position:50% size:23% Instead, in those, we actually see 00:29:18.223 --> 00:29:20.025 align:center line:-1 position:50% size:32% the greatest percentage of landslides 00:29:20.025 --> 00:29:23.495 align:center line:-1 position:50% size:37% in sort of moderate- and low- susceptibility terrains, 00:29:23.495 --> 00:29:25.464 align:center line:-1 position:50% size:29% so that's potentially problematic, and, yes, 00:29:25.464 --> 00:29:27.299 align:center line:-1 position:50% size:36% there is some reporting bias 00:29:27.299 --> 00:29:31.436 align:center line:-1 position:50% size:39% in that more landslides are reported and documented 00:29:31.436 --> 00:29:34.106 align:center line:-1 position:50% size:34% in potentially lower-susceptibility terrain 00:29:34.106 --> 00:29:36.375 align:center line:-1 position:50% size:39% because that's where humans are more likely to interface 00:29:36.375 --> 00:29:38.644 align:center line:-1 position:50% size:20% with landslides. 00:29:38.644 --> 00:29:41.113 align:center line:-1 position:50% size:29% However, we're still really happy 00:29:41.113 --> 00:29:45.784 align:center line:-1 position:50% size:37% with this kind of general finding in the model. 00:29:45.784 --> 00:29:48.720 align:center line:-1 position:50% size:41% And then, relative to the "some" and "none" classes, 00:29:48.720 --> 00:29:50.722 align:center line:-1 position:50% size:38% it's a little harder to compare, but again, 00:29:50.722 --> 00:29:52.424 align:center line:-1 position:50% size:28% we also think that we're doing great 00:29:52.424 --> 00:29:54.526 align:center line:-1 position:50% size:29% in terms of percentage of landslides captured. 00:29:54.526 --> 00:29:56.328 align:center line:-1 position:50% size:27% You can pick any cutoff you'd like, 00:29:56.328 --> 00:30:00.132 align:center line:-1 position:50% size:41% compared to the ones captured by those other two models 00:30:00.132 --> 00:30:04.636 align:center line:-1 position:50% size:35% for corresponding surface area of the country. 00:30:04.636 --> 00:30:06.405 align:center line:-1 position:50% size:35% So, whereas some models, for example, 00:30:06.405 --> 00:30:08.240 align:center line:-1 position:50% size:28% the NASA model was, specifically, 00:30:08.240 --> 00:30:10.609 align:center line:-1 position:50% size:34% the classes were selected to be decreasing in area 00:30:10.609 --> 00:30:12.277 align:center line:-1 position:50% size:39% as you increase susceptibility, 00:30:12.277 --> 00:30:14.046 align:center line:-1 position:50% size:24% ours is developed, 00:30:14.046 --> 00:30:15.681 align:center line:-1 position:50% size:25% or the result of ours is that we have 00:30:15.681 --> 00:30:18.417 align:center line:-1 position:50% size:35% an increasing likelihood or percentage of landslides 00:30:18.417 --> 00:30:20.185 align:center line:-1 position:50% size:25% as you increase susceptibility class, 00:30:20.185 --> 00:30:21.420 align:center line:-1 position:50% size:28% and so, that's, I think, 00:30:21.420 --> 00:30:23.555 align:center line:-1 position:50% size:33% a really great contribution here. 00:30:26.358 --> 00:30:27.859 align:center line:-1 position:50% size:26% I just showed you analysis for CONUS 00:30:27.859 --> 00:30:30.495 align:center line:-1 position:50% size:41% so that we could do sort of this unbiased cross comparison 00:30:30.495 --> 00:30:32.497 align:center line:-1 position:50% size:37% across all those five models, 00:30:32.497 --> 00:30:35.033 align:center line:-1 position:50% size:31% but now, I'm looking at-- 00:30:35.033 --> 00:30:36.301 align:center line:-1 position:50% size:24% five different ways of looking at it 00:30:36.301 --> 00:30:39.004 align:center line:-1 position:50% size:29% for the entire country, including Puerto Rico, 00:30:39.004 --> 00:30:40.338 align:center line:-1 position:50% size:31% just Alaska, just Hawaii, 00:30:40.338 --> 00:30:41.707 align:center line:-1 position:50% size:27% just CONUS, and just Puerto Rico, 00:30:41.707 --> 00:30:43.141 align:center line:-1 position:50% size:22% and you can see there's quite a bit 00:30:43.141 --> 00:30:44.576 align:center line:-1 position:50% size:28% of spatial variability in model performance 00:30:44.576 --> 00:30:47.312 align:center line:-1 position:50% size:39% and in susceptible terrain across those different regions, 00:30:47.312 --> 00:30:49.347 align:center line:-1 position:50% size:27% especially looking at Puerto Rico, wow. 00:30:49.347 --> 00:30:51.583 align:center line:-1 position:50% size:39% The northern half of the island 00:30:51.583 --> 00:30:54.453 align:center line:-1 position:50% size:36% is covered by this highest equal interval class, 00:30:54.453 --> 00:30:57.756 align:center line:-1 position:50% size:34% and 92% of the landslides are within that class. 00:30:57.756 --> 00:31:00.358 align:center line:-1 position:50% size:7% Um... 00:31:00.358 --> 00:31:02.527 align:center line:-1 position:50% size:37% So, I think what this points to is that there's-- 00:31:02.527 --> 00:31:04.696 align:center line:-1 position:50% size:33% there's kind of a need to-- 00:31:04.696 --> 00:31:06.765 align:center line:-1 position:50% size:32% although this helps show nationally 00:31:06.765 --> 00:31:08.567 align:center line:-1 position:50% size:30% quite a bit of variability, 00:31:08.567 --> 00:31:09.835 align:center line:-1 position:50% size:26% there's also a need, 00:31:09.835 --> 00:31:11.236 align:center line:-1 position:50% size:34% if you were gonna use this at a local, state, 00:31:11.236 --> 00:31:13.038 align:center line:-1 position:50% size:27% or non-national level, 00:31:13.038 --> 00:31:14.239 align:center line:-1 position:50% size:24% there's a need to kind of consider 00:31:14.239 --> 00:31:15.807 align:center line:-1 position:50% size:27% what are appropriate classes and cutoff 00:31:15.807 --> 00:31:19.144 align:center line:-1 position:50% size:35% for different risk tolerances. 00:31:19.144 --> 00:31:22.347 align:center line:-1 position:50% size:35% Another type of spatial evaluation we did 00:31:22.347 --> 00:31:24.349 align:center line:-1 position:50% size:21% is in some ways a lot simpler, 00:31:24.349 --> 00:31:27.052 align:center line:-1 position:50% size:22% but I think, a very, very cool, 00:31:27.052 --> 00:31:29.221 align:center line:-1 position:50% size:33% and provided some really cool insights. 00:31:29.221 --> 00:31:32.758 align:center line:-1 position:50% size:32% Here, we are just plotting two things... 00:31:32.758 --> 00:31:35.627 align:center line:-1 position:50% size:24% oh, and I guess-- hang on a second. 00:31:35.627 --> 00:31:39.097 align:center line:-1 position:50% size:33% Um, here, this little inset here shows 00:31:39.097 --> 00:31:42.534 align:center line:-1 position:50% size:23% that on the y-axis in the pink colors 00:31:42.534 --> 00:31:44.469 align:center line:-1 position:50% size:27% is the percentage of susceptible terrain 00:31:44.469 --> 00:31:45.704 align:center line:-1 position:50% size:25% within each county. 00:31:45.704 --> 00:31:48.340 align:center line:-1 position:50% size:30% So I should say this is a county-level analysis. 00:31:48.340 --> 00:31:49.541 align:center line:-1 position:50% size:39% And then, the blue colors here 00:31:49.541 --> 00:31:51.543 align:center line:-1 position:50% size:33% show the number of documented landslides 00:31:51.543 --> 00:31:54.346 align:center line:-1 position:50% size:28% per square kilometer of susceptible terrain. 00:31:54.346 --> 00:31:56.314 align:center line:-1 position:50% size:35% So sort of landslide density 00:31:56.314 --> 00:32:00.252 align:center line:-1 position:50% size:28% versus amount of susceptible terrain. 00:32:00.252 --> 00:32:03.488 align:center line:-1 position:50% size:38% Essentially, what this is doing is showing us 00:32:03.488 --> 00:32:05.157 align:center line:-1 position:50% size:37% where our model works well, 00:32:05.157 --> 00:32:06.658 align:center line:-1 position:50% size:23% where our model doesn't work well, 00:32:06.658 --> 00:32:09.427 align:center line:-1 position:50% size:34% and where we may expect to find more landslides, 00:32:09.427 --> 00:32:11.463 align:center line:-1 position:50% size:24% if we went digging. 00:32:11.463 --> 00:32:12.764 align:center line:-1 position:50% size:27% In gray, this is sort of 00:32:12.764 --> 00:32:15.667 align:center line:-1 position:50% size:27% low percentage of susceptible terrain 00:32:15.667 --> 00:32:17.369 align:center line:-1 position:50% size:29% and a low density of mapped landslides. 00:32:17.369 --> 00:32:19.104 align:center line:-1 position:50% size:29% That's good, that's what we expect. 00:32:19.104 --> 00:32:21.206 align:center line:-1 position:50% size:32% In the upper-right corner, this bright violet is 00:32:21.206 --> 00:32:23.241 align:center line:-1 position:50% size:28% a high percentage of susceptible terrain, 00:32:23.241 --> 00:32:25.143 align:center line:-1 position:50% size:33% and again, a high density of mapped landslides. 00:32:25.143 --> 00:32:26.478 align:center line:-1 position:50% size:22% That's also good, 00:32:26.478 --> 00:32:29.181 align:center line:-1 position:50% size:34% and this sort of light purple in between is also great. 00:32:29.181 --> 00:32:31.316 align:center line:-1 position:50% size:30% But then, up here in the upper-left corner, 00:32:31.316 --> 00:32:32.617 align:center line:-1 position:50% size:22% this bright pink is 00:32:32.617 --> 00:32:33.852 align:center line:-1 position:50% size:27% a high percentage of susceptible terrain 00:32:33.852 --> 00:32:35.987 align:center line:-1 position:50% size:29% with a low percentage of mapped landslides, 00:32:35.987 --> 00:32:37.689 align:center line:-1 position:50% size:26% and so, that's areas, for example, 00:32:37.689 --> 00:32:41.226 align:center line:-1 position:50% size:42% if we look across the border from Kentucky to North Carolina, 00:32:41.226 --> 00:32:42.394 align:center line:-1 position:50% size:25% there's nothing magical happening 00:32:42.394 --> 00:32:44.563 align:center line:-1 position:50% size:38% across the Appalachians here on the state borders 00:32:44.563 --> 00:32:47.065 align:center line:-1 position:50% size:34% in Virginia and Tennessee; 00:32:47.065 --> 00:32:50.202 align:center line:-1 position:50% size:29% there's just not reported mapping. 00:32:50.202 --> 00:32:51.603 align:center line:-1 position:50% size:34% So these are areas where, if we went in and mapped, 00:32:51.603 --> 00:32:55.040 align:center line:-1 position:50% size:24% we'd expect to see more landslides. 00:32:55.040 --> 00:32:57.475 align:center line:-1 position:50% size:30% Conversely, the blue is areas where there's 00:32:57.475 --> 00:32:59.878 align:center line:-1 position:50% size:30% a high density--relative, at the national level-- 00:32:59.878 --> 00:33:01.646 align:center line:-1 position:50% size:28% high density of mapped landslides 00:33:01.646 --> 00:33:04.482 align:center line:-1 position:50% size:28% and a low percentage of susceptible terrain, 00:33:04.482 --> 00:33:06.818 align:center line:-1 position:50% size:37% so that's areas where maybe our model didn't do so well-- 00:33:06.818 --> 00:33:10.822 align:center line:-1 position:50% size:42% for example, along the North Dakota-Minnesota border, 00:33:10.822 --> 00:33:14.125 align:center line:-1 position:50% size:32% other parts of Minnesota, 00:33:14.125 --> 00:33:18.463 align:center line:-1 position:50% size:32% and a couple of interval locations as well. 00:33:18.463 --> 00:33:20.165 align:center line:-1 position:50% size:29% And I should point out, 00:33:20.165 --> 00:33:22.400 align:center line:-1 position:50% size:44% Puerto Rico and Southeast Alaska are also showing up, 00:33:22.400 --> 00:33:25.036 align:center line:-1 position:50% size:33% as areas in the Pacific Northwest and California, 00:33:25.036 --> 00:33:26.238 align:center line:-1 position:50% size:27% where there's a lot of landslide mapping 00:33:26.238 --> 00:33:28.306 align:center line:-1 position:50% size:26% here in the Rockies, 00:33:28.306 --> 00:33:31.877 align:center line:-1 position:50% size:32% areas where our model's doing well. 00:33:31.877 --> 00:33:34.045 align:center line:-1 position:50% size:37% And then, really, just all up and down Appalachia, 00:33:34.045 --> 00:33:35.513 align:center line:-1 position:50% size:15% the Ozarks. 00:33:35.513 --> 00:33:37.682 align:center line:-1 position:50% size:38% Parts of the Northern Rockies could handle 00:33:37.682 --> 00:33:39.784 align:center line:-1 position:50% size:32% a little bit more mapping, potentially. 00:33:39.784 --> 00:33:41.586 align:center line:-1 position:50% size:20% So this is great. 00:33:41.586 --> 00:33:43.221 align:center line:-1 position:50% size:26% It shows us at the national level, 00:33:43.221 --> 00:33:44.756 align:center line:-1 position:50% size:26% in the coarsest way, 00:33:44.756 --> 00:33:46.625 align:center line:-1 position:50% size:32% you could look at this in more detail, of course, 00:33:46.625 --> 00:33:48.059 align:center line:-1 position:50% size:39% where is our model doing well, 00:33:48.059 --> 00:33:49.494 align:center line:-1 position:50% size:39% where do we need more data, 00:33:49.494 --> 00:33:53.064 align:center line:-1 position:50% size:35% and that is potentially useful guidance for funding 00:33:53.064 --> 00:33:55.500 align:center line:-1 position:50% size:27% or for more research on susceptibility. 00:33:57.702 --> 00:33:59.271 align:center line:-1 position:50% size:35% And it also really highlights 00:33:59.271 --> 00:34:01.039 align:center line:-1 position:50% size:37% the value of statewide mapping efforts 00:34:01.039 --> 00:34:03.341 align:center line:-1 position:50% size:35% by state geological surveys and other entities, 00:34:03.341 --> 00:34:04.676 align:center line:-1 position:50% size:22% and I really want to stress that, 00:34:04.676 --> 00:34:08.213 align:center line:-1 position:50% size:30% that our compilation of inventories is largely 00:34:08.213 --> 00:34:10.282 align:center line:-1 position:50% size:29% individual inventories mapped by the USGS, 00:34:10.282 --> 00:34:12.751 align:center line:-1 position:50% size:40% and then statewide inventories mapped by state surveys 00:34:12.751 --> 00:34:16.187 align:center line:-1 position:50% size:33% and similar organizations. 00:34:16.187 --> 00:34:19.624 align:center line:-1 position:50% size:39% So, again, this is now showing that same schema 00:34:19.624 --> 00:34:23.495 align:center line:-1 position:50% size:34% with the pink, purple, blue, et cetera, 00:34:23.495 --> 00:34:27.599 align:center line:-1 position:50% size:32% but now, we're looking at sort of before and after 00:34:27.599 --> 00:34:29.367 align:center line:-1 position:50% size:27% for the four statewide inventories 00:34:29.367 --> 00:34:31.636 align:center line:-1 position:50% size:35% that we included in our independent testing. 00:34:31.636 --> 00:34:34.706 align:center line:-1 position:50% size:34% So, Idaho, almost 500,000 mapped landslides, 00:34:34.706 --> 00:34:36.708 align:center line:-1 position:50% size:30% Maine, a little over 400 mapped landslides, 00:34:36.708 --> 00:34:38.476 align:center line:-1 position:50% size:29% North Dakota, 60,000, 00:34:38.476 --> 00:34:41.079 align:center line:-1 position:50% size:24% and West Virginia, over 100,000. 00:34:41.079 --> 00:34:46.284 align:center line:-1 position:50% size:35% So what these show now is sort of this great change 00:34:46.284 --> 00:34:47.619 align:center line:-1 position:50% size:38% before the statewide mapping and after, 00:34:47.619 --> 00:34:49.254 align:center line:-1 position:50% size:22% and in Idaho, a lot of the areas 00:34:49.254 --> 00:34:51.423 align:center line:-1 position:50% size:34% where you'd expect to find the most landslides 00:34:51.423 --> 00:34:55.760 align:center line:-1 position:50% size:32% are in fact shifting from pink to dark purple. 00:34:55.760 --> 00:34:57.562 align:center line:-1 position:50% size:35% There's also a few counties where, oh, 00:34:57.562 --> 00:34:59.497 align:center line:-1 position:50% size:41% there's more landslides mapped than we may have expected, 00:34:59.497 --> 00:35:01.066 align:center line:-1 position:50% size:27% based on the model. 00:35:01.066 --> 00:35:02.667 align:center line:-1 position:50% size:29% And Maine, there's not that many landslides; 00:35:02.667 --> 00:35:04.636 align:center line:-1 position:50% size:34% probably, there's still some mapping 00:35:04.636 --> 00:35:06.171 align:center line:-1 position:50% size:23% that would reveal further landslides, 00:35:06.171 --> 00:35:11.242 align:center line:-1 position:50% size:30% especially on the northeast coast. 00:35:11.242 --> 00:35:13.211 align:center line:-1 position:50% size:23% But North Dakota really stands out, 00:35:13.211 --> 00:35:15.180 align:center line:-1 position:50% size:32% and I should point out that North Dakota's done 00:35:15.180 --> 00:35:17.449 align:center line:-1 position:50% size:21% a really nice job 00:35:17.449 --> 00:35:19.884 align:center line:-1 position:50% size:40% doing what's probably the most complete statewide inventory 00:35:19.884 --> 00:35:21.753 align:center line:-1 position:50% size:20% in the Midwest. 00:35:21.753 --> 00:35:23.722 align:center line:-1 position:50% size:40% Minnesota also has a nice one. 00:35:23.722 --> 00:35:27.792 align:center line:-1 position:50% size:40% And so, there's also some glacial geology involved, 00:35:27.792 --> 00:35:29.127 align:center line:-1 position:50% size:30% and things that maybe, 00:35:29.127 --> 00:35:31.162 align:center line:-1 position:50% size:36% reasons that our simple slope relief threshold 00:35:31.162 --> 00:35:33.164 align:center line:-1 position:50% size:30% is not working that well, 00:35:33.164 --> 00:35:34.499 align:center line:-1 position:50% size:22% so this is an area 00:35:34.499 --> 00:35:36.701 align:center line:-1 position:50% size:30% where further research might be helpful. 00:35:36.701 --> 00:35:38.670 align:center line:-1 position:50% size:34% Whereas in West Virginia, wow, okay, 00:35:38.670 --> 00:35:41.006 align:center line:-1 position:50% size:41% everywhere that was bright pink is now dark purple 00:35:41.006 --> 00:35:45.477 align:center line:-1 position:50% size:33% when you include these 108,000 landslides. 00:35:45.477 --> 00:35:47.645 align:center line:-1 position:50% size:24% So this, again, just really stresses 00:35:47.645 --> 00:35:49.180 align:center line:-1 position:50% size:40% the value of statewide mapping 00:35:49.180 --> 00:35:53.551 align:center line:-1 position:50% size:34% and how this model is also maybe useful everywhere 00:35:53.551 --> 00:35:56.855 align:center line:-1 position:50% size:27% but not really correct everywhere, either. 00:35:58.523 --> 00:36:00.191 align:center line:-1 position:50% size:22% We drilled down a little bit deeper. 00:36:00.191 --> 00:36:02.460 align:center line:-1 position:50% size:28% This chart is a little bit more difficult to read 00:36:02.460 --> 00:36:04.162 align:center line:-1 position:50% size:23% with all the states. 00:36:04.162 --> 00:36:05.497 align:center line:-1 position:50% size:22% We're basically looking at states, 00:36:05.497 --> 00:36:07.699 align:center line:-1 position:50% size:17% all 50 states, 00:36:07.699 --> 00:36:10.301 align:center line:-1 position:50% size:43% and then, the number of counties 00:36:10.301 --> 00:36:13.471 align:center line:-1 position:50% size:37% with land that's susceptible but zero mapped landslides. 00:36:13.471 --> 00:36:15.040 align:center line:-1 position:50% size:28% This is just a number of counties, 00:36:15.040 --> 00:36:16.608 align:center line:-1 position:50% size:31% and we cut it off at 50%, 00:36:16.608 --> 00:36:18.243 align:center line:-1 position:50% size:25% so if more than half of the county 00:36:18.243 --> 00:36:19.677 align:center line:-1 position:50% size:35% is susceptible to landslides 00:36:19.677 --> 00:36:21.246 align:center line:-1 position:50% size:25% and it has zero mapped landslides, 00:36:21.246 --> 00:36:22.514 align:center line:-1 position:50% size:33% then it's showing up here. 00:36:22.514 --> 00:36:23.848 align:center line:-1 position:50% size:35% And really, what stands out 00:36:23.848 --> 00:36:26.584 align:center line:-1 position:50% size:29% are Appalachia and Ozark Mountains. 00:36:26.584 --> 00:36:31.356 align:center line:-1 position:50% size:31% So, Alabama, Arkansas, Mississippi, Georgia, 00:36:31.356 --> 00:36:37.262 align:center line:-1 position:50% size:40% Ohio, Tennessee, and Virginia, among others. 00:36:37.262 --> 00:36:39.697 align:center line:-1 position:50% size:29% So, really, those areas where there is 00:36:39.697 --> 00:36:42.200 align:center line:-1 position:50% size:26% a potentially higher susceptibility 00:36:42.200 --> 00:36:43.501 align:center line:-1 position:50% size:28% throughout the county 00:36:43.501 --> 00:36:45.270 align:center line:-1 position:50% size:33% and not a single landslide that is in our inventory, 00:36:45.270 --> 00:36:47.105 align:center line:-1 position:50% size:37% at least that we've compiled. 00:36:48.606 --> 00:36:52.043 align:center line:-1 position:50% size:29% This new model has a lot of potential value 00:36:52.043 --> 00:36:53.778 align:center line:-1 position:50% size:36% for systematic risk analysis. 00:36:53.778 --> 00:36:57.449 align:center line:-1 position:50% size:32% For example, this is just two locations, 00:36:57.449 --> 00:36:59.684 align:center line:-1 position:50% size:30% southwest Washington, north central Florida, 00:36:59.684 --> 00:37:01.586 align:center line:-1 position:50% size:33% and what's shown in gray 00:37:01.586 --> 00:37:04.589 align:center line:-1 position:50% size:41% is that previous prototype model from Godt and others, 00:37:04.589 --> 00:37:07.158 align:center line:-1 position:50% size:44% and our new model showing these different cell class counts, 00:37:07.158 --> 00:37:09.461 align:center line:-1 position:50% size:40% and again, you could add whatever qualitative descriptor 00:37:09.461 --> 00:37:10.695 align:center line:-1 position:50% size:19% you want here, 00:37:10.695 --> 00:37:13.031 align:center line:-1 position:50% size:23% but, essentially, zero is negligible, 00:37:13.031 --> 00:37:15.567 align:center line:-1 position:50% size:37% and there's some increasing degree of susceptibility 00:37:15.567 --> 00:37:18.770 align:center line:-1 position:50% size:42% as you get from orange up to red in our model. 00:37:18.770 --> 00:37:20.038 align:center line:-1 position:50% size:27% The interesting thing, 00:37:20.038 --> 00:37:21.406 align:center line:-1 position:50% size:29% if you look at before our new model, 00:37:21.406 --> 00:37:24.275 align:center line:-1 position:50% size:36% these two locations essentially look very similar. 00:37:24.275 --> 00:37:27.212 align:center line:-1 position:50% size:22% A lot of gray area that is potentially 00:37:27.212 --> 00:37:28.613 align:center line:-1 position:50% size:35% having some susceptibility. 00:37:28.613 --> 00:37:30.315 align:center line:-1 position:50% size:32% But then, when you start looking at our model, 00:37:30.315 --> 00:37:32.283 align:center line:-1 position:50% size:33% you can see, oh my gosh, in southwest Washington, 00:37:32.283 --> 00:37:35.220 align:center line:-1 position:50% size:40% there's way more areas with much higher susceptibility, 00:37:35.220 --> 00:37:38.123 align:center line:-1 position:50% size:31% whereas in north central Florida, there's less. 00:37:38.123 --> 00:37:40.225 align:center line:-1 position:50% size:36% Additionally, that higher-resolution output 00:37:40.225 --> 00:37:41.659 align:center line:-1 position:50% size:31% allows us to look at some of these areas, 00:37:41.659 --> 00:37:44.696 align:center line:-1 position:50% size:17% for example, north of I-75, 00:37:44.696 --> 00:37:47.665 align:center line:-1 position:50% size:43% where there's clearly some anthropogenic features going on, 00:37:47.665 --> 00:37:49.167 align:center line:-1 position:50% size:18% and actually, if you go look, 00:37:49.167 --> 00:37:52.003 align:center line:-1 position:50% size:39% there's some mining and other industrial operations here 00:37:52.003 --> 00:37:54.205 align:center line:-1 position:50% size:30% that is creating sort of-- 00:37:54.205 --> 00:37:56.374 align:center line:-1 position:50% size:32% even a fairly gentle slope and lower-leaf terrain 00:37:56.374 --> 00:37:58.743 align:center line:-1 position:50% size:34% can exceed that threshold and potentially be 00:37:58.743 --> 00:38:00.278 align:center line:-1 position:50% size:26% a landslide concern, 00:38:00.278 --> 00:38:04.616 align:center line:-1 position:50% size:22% even if it's not a natural feature. 00:38:04.616 --> 00:38:07.852 align:center line:-1 position:50% size:24% So, this right away just helps illustrate 00:38:07.852 --> 00:38:09.554 align:center line:-1 position:50% size:38% what a nice advance we have 00:38:09.554 --> 00:38:11.556 align:center line:-1 position:50% size:33% in terms of the increase in resolution 00:38:11.556 --> 00:38:15.160 align:center line:-1 position:50% size:29% and the increase in information content. 00:38:15.160 --> 00:38:19.297 align:center line:-1 position:50% size:36% There's a really nice LHP seminar on March 20th 00:38:19.297 --> 00:38:20.465 align:center line:-1 position:50% size:27% by Sabrina Martinez. 00:38:20.465 --> 00:38:22.200 align:center line:-1 position:50% size:37% I encourage you to go see it. 00:38:22.200 --> 00:38:25.336 align:center line:-1 position:50% size:27% This is just a teaser of what sort of things 00:38:25.336 --> 00:38:28.006 align:center line:-1 position:50% size:32% this new nationwide map can facilitate. 00:38:28.006 --> 00:38:31.609 align:center line:-1 position:50% size:38% What Sabrina did was look at population exposure analysis 00:38:31.609 --> 00:38:33.811 align:center line:-1 position:50% size:40% and just some basic questions: 00:38:33.811 --> 00:38:35.580 align:center line:-1 position:50% size:31% How many people are vulnerable to landslides 00:38:35.580 --> 00:38:37.348 align:center line:-1 position:50% size:33% and when, day and night; 00:38:37.348 --> 00:38:38.883 align:center line:-1 position:50% size:33% where are the most people exposed 00:38:38.883 --> 00:38:40.285 align:center line:-1 position:50% size:24% and who are they? 00:38:40.285 --> 00:38:43.788 align:center line:-1 position:50% size:37% So we leveraged this new national susceptibility model. 00:38:43.788 --> 00:38:45.623 align:center line:-1 position:50% size:37% We looked at the Oak Ridge 00:38:45.623 --> 00:38:48.660 align:center line:-1 position:50% size:24% national LandScan population model, 00:38:48.660 --> 00:38:49.794 align:center line:-1 position:50% size:30% which is also 90 meter, 00:38:49.794 --> 00:38:51.462 align:center line:-1 position:50% size:43% and then, the American Community Survey 00:38:51.462 --> 00:38:52.697 align:center line:-1 position:50% size:32% from the Census Bureau 00:38:52.697 --> 00:38:54.465 align:center line:-1 position:50% size:34% to look at those questions, 00:38:54.465 --> 00:38:56.768 align:center line:-1 position:50% size:28% and the key findings-- and again, 00:38:56.768 --> 00:38:59.070 align:center line:-1 position:50% size:35% I refer you to Sabrina's talk to dig into this, 00:38:59.070 --> 00:39:00.838 align:center line:-1 position:50% size:27% but 75 million people during the day, 00:39:00.838 --> 00:39:02.173 align:center line:-1 position:50% size:23% 85 million at night 00:39:02.173 --> 00:39:03.775 align:center line:-1 position:50% size:27% are in some sort of susceptible terrain 00:39:03.775 --> 00:39:05.276 align:center line:-1 position:50% size:33% across the United States. 00:39:05.276 --> 00:39:08.179 align:center line:-1 position:50% size:38% That's a substantial percentage of the population. 00:39:08.179 --> 00:39:10.248 align:center line:-1 position:50% size:36% If we look at just the highest susceptibility class-- 00:39:10.248 --> 00:39:11.816 align:center line:-1 position:50% size:39% in the previous slide I showed, 00:39:11.816 --> 00:39:15.086 align:center line:-1 position:50% size:26% it's sort of the pinks and reds-- 00:39:15.086 --> 00:39:17.989 align:center line:-1 position:50% size:34% it's 9 million during the day and 16 million at night, 00:39:17.989 --> 00:39:20.658 align:center line:-1 position:50% size:37% so, certainly, people more exposed in their home, 00:39:20.658 --> 00:39:22.227 align:center line:-1 position:50% size:27% more people living in susceptible terrain 00:39:22.227 --> 00:39:24.429 align:center line:-1 position:50% size:20% than in working during the day. 00:39:24.429 --> 00:39:26.598 align:center line:-1 position:50% size:33% And the greatest total population exposure 00:39:26.598 --> 00:39:27.832 align:center line:-1 position:50% size:37% not surprisingly in California; 00:39:27.832 --> 00:39:30.568 align:center line:-1 position:50% size:32% whereas the highest proportion of population, 00:39:30.568 --> 00:39:31.769 align:center line:-1 position:50% size:40% West Virginia and Puerto Rico, 00:39:31.769 --> 00:39:33.638 align:center line:-1 position:50% size:30% and that proportion has a big implication 00:39:33.638 --> 00:39:37.742 align:center line:-1 position:50% size:28% in terms of resilience and that's state level. 00:39:37.742 --> 00:39:41.646 align:center line:-1 position:50% size:37% Then, lastly, of those highest-susceptibility terrain, 00:39:41.646 --> 00:39:44.616 align:center line:-1 position:50% size:26% 5.5 million of those in the most exposed 00:39:44.616 --> 00:39:46.384 align:center line:-1 position:50% size:31% are in counties with a high poverty rate, 00:39:46.384 --> 00:39:48.119 align:center line:-1 position:50% size:27% so that also gives us some information 00:39:48.119 --> 00:39:50.455 align:center line:-1 position:50% size:30% in terms of vulnerability and exposure. 00:39:50.455 --> 00:39:52.824 align:center line:-1 position:50% size:24% So, again, this is a really cool study 00:39:52.824 --> 00:39:54.459 align:center line:-1 position:50% size:27% that we're working on getting published, 00:39:54.459 --> 00:39:56.628 align:center line:-1 position:50% size:34% and Sabrina's got a nice presentation of that 00:39:56.628 --> 00:40:00.098 align:center line:-1 position:50% size:27% you can view online. 00:40:00.098 --> 00:40:01.766 align:center line:-1 position:50% size:28% So, lots of other potential applications, 00:40:01.766 --> 00:40:04.435 align:center line:-1 position:50% size:30% and I'm excited to talk with anyone about that. 00:40:04.435 --> 00:40:06.404 align:center line:-1 position:50% size:26% But I do want to say, first of all, 00:40:06.404 --> 00:40:09.340 align:center line:-1 position:50% size:42% we are going to make both the N10 and linear weighted models 00:40:09.340 --> 00:40:10.675 align:center line:-1 position:50% size:23% publicly available. 00:40:10.675 --> 00:40:13.645 align:center line:-1 position:50% size:23% That data release will be accessible 00:40:13.645 --> 00:40:17.548 align:center line:-1 position:50% size:28% at this DOI very soon. 00:40:17.548 --> 00:40:20.318 align:center line:-1 position:50% size:36% I want to say there are limits to the slope relief concept. 00:40:20.318 --> 00:40:21.786 align:center line:-1 position:50% size:33% This is inherently general, 00:40:21.786 --> 00:40:24.656 align:center line:-1 position:50% size:43% and we wanted a model that worked well enough everywhere, 00:40:24.656 --> 00:40:26.357 align:center line:-1 position:50% size:17% but in reality, 00:40:26.357 --> 00:40:29.527 align:center line:-1 position:50% size:39% you've got to really take some good judgment and caution 00:40:29.527 --> 00:40:32.063 align:center line:-1 position:50% size:23% when using it at the local scale. 00:40:32.063 --> 00:40:33.498 align:center line:-1 position:50% size:34% For example, the highest elevation point 00:40:33.498 --> 00:40:35.099 align:center line:-1 position:50% size:34% is not always the steepest. 00:40:35.099 --> 00:40:36.834 align:center line:-1 position:50% size:46% Some of the assumptions we baked into our model and analysis 00:40:36.834 --> 00:40:38.369 align:center line:-1 position:50% size:28% maybe aren't correct. 00:40:38.369 --> 00:40:40.238 align:center line:-1 position:50% size:40% This is an example of one of our missed landslides 00:40:40.238 --> 00:40:41.606 align:center line:-1 position:50% size:23% from Glacier Bay. 00:40:41.606 --> 00:40:43.074 align:center line:-1 position:50% size:36% You can see it's surrounded 00:40:43.074 --> 00:40:44.642 align:center line:-1 position:50% size:37% by all this very high susceptible terrain. 00:40:44.642 --> 00:40:46.377 align:center line:-1 position:50% size:38% We have a unique landscape with glaciers, 00:40:46.377 --> 00:40:47.879 align:center line:-1 position:50% size:39% some flat stuff, some benches 00:40:47.879 --> 00:40:50.515 align:center line:-1 position:50% size:40% that maybe don't work that well with the slope relief model. 00:40:50.515 --> 00:40:52.817 align:center line:-1 position:50% size:37% I want to remind everyone about our proximity analysis, 00:40:52.817 --> 00:40:55.420 align:center line:-1 position:50% size:25% that if you're within two, three grid cells 00:40:55.420 --> 00:40:57.989 align:center line:-1 position:50% size:32% of a susceptible terrain-- well, 00:40:57.989 --> 00:40:59.390 align:center line:-1 position:50% size:23% we don't explicitly consider runout, 00:40:59.390 --> 00:41:01.025 align:center line:-1 position:50% size:28% we don't consider the size of landslides, 00:41:01.025 --> 00:41:04.295 align:center line:-1 position:50% size:26% so use with caution, 00:41:04.295 --> 00:41:06.364 align:center line:-1 position:50% size:24% and perhaps even further upscaling, 00:41:06.364 --> 00:41:10.201 align:center line:-1 position:50% size:28% if you're looking at some larger areas. 00:41:12.503 --> 00:41:15.039 align:center line:-1 position:50% size:27% Also, there's nothing that is a substitute 00:41:15.039 --> 00:41:18.343 align:center line:-1 position:50% size:36% for looking at the landscape in a detailed site analysis. 00:41:18.343 --> 00:41:20.378 align:center line:-1 position:50% size:30% This is hopefully just sort of a screening tool 00:41:20.378 --> 00:41:23.081 align:center line:-1 position:50% size:34% to use at the national level and/or in places 00:41:23.081 --> 00:41:25.783 align:center line:-1 position:50% size:30% where there is no other local-scale tool, 00:41:25.783 --> 00:41:29.754 align:center line:-1 position:50% size:36% so whether you should send boots on the ground 00:41:29.754 --> 00:41:31.489 align:center line:-1 position:50% size:33% to look at landslide-prone training. 00:41:31.489 --> 00:41:33.391 align:center line:-1 position:50% size:37% This is those same locations from central Washington 00:41:33.391 --> 00:41:35.360 align:center line:-1 position:50% size:27% and northern Florida. 00:41:35.360 --> 00:41:38.062 align:center line:-1 position:50% size:35% You can see that obviously the landscape also tells 00:41:38.062 --> 00:41:40.398 align:center line:-1 position:50% size:36% a whole other story about what's going on here. 00:41:42.166 --> 00:41:44.836 align:center line:-1 position:50% size:31% So I'm just really happy to have had this chance 00:41:44.836 --> 00:41:47.739 align:center line:-1 position:50% size:32% to share the exciting contributions 00:41:47.739 --> 00:41:49.640 align:center line:-1 position:50% size:33% from my co-authors and I 00:41:49.640 --> 00:41:51.809 align:center line:-1 position:50% size:38% on parsimonious approaches to national- 00:41:51.809 --> 00:41:55.012 align:center line:-1 position:50% size:30% and continental-scale susceptibility modeling. 00:41:55.012 --> 00:41:56.314 align:center line:-1 position:50% size:23% We've developed this updated map 00:41:56.314 --> 00:41:57.815 align:center line:-1 position:50% size:29% with uniform coverage over CONUS, 00:41:57.815 --> 00:42:00.752 align:center line:-1 position:50% size:22% Alaska, Hawaii, and Puerto Rico. 00:42:00.752 --> 00:42:02.520 align:center line:-1 position:50% size:32% The improved resolution and performance 00:42:02.520 --> 00:42:03.821 align:center line:-1 position:50% size:28% over existing models, 00:42:03.821 --> 00:42:06.691 align:center line:-1 position:50% size:38% so it's really from 1 kilometer to 90 meters 00:42:06.691 --> 00:42:09.127 align:center line:-1 position:50% size:31% and is performing better at capturing 00:42:09.127 --> 00:42:11.796 align:center line:-1 position:50% size:33% nearly a million landslides that we've done, 00:42:11.796 --> 00:42:14.465 align:center line:-1 position:50% size:26% that we've compiled across the country. 00:42:14.465 --> 00:42:15.867 align:center line:-1 position:50% size:22% And in particular, 00:42:15.867 --> 00:42:18.269 align:center line:-1 position:50% size:33% this down-scaling method is really intriguing, 00:42:18.269 --> 00:42:20.838 align:center line:-1 position:50% size:34% and hopefully, this method will be useful for others 00:42:20.838 --> 00:42:22.640 align:center line:-1 position:50% size:23% in other locations. 00:42:22.640 --> 00:42:24.342 align:center line:-1 position:50% size:31% We like that it's really transparent and flexible 00:42:24.342 --> 00:42:26.310 align:center line:-1 position:50% size:33% in terms of allowing a full continuum of values 00:42:26.310 --> 00:42:27.879 align:center line:-1 position:50% size:29% that users could have, 00:42:27.879 --> 00:42:30.548 align:center line:-1 position:50% size:31% and we think this is in potential applications 00:42:30.548 --> 00:42:33.551 align:center line:-1 position:50% size:30% of federal, state, and for other agencies, 00:42:33.551 --> 00:42:35.420 align:center line:-1 position:50% size:33% and in particular, in areas where there isn't already 00:42:35.420 --> 00:42:37.155 align:center line:-1 position:50% size:29% some local-scale map. 00:42:37.155 --> 00:42:39.824 align:center line:-1 position:50% size:28% But it is not intended to supplant or replace 00:42:39.824 --> 00:42:41.659 align:center line:-1 position:50% size:25% better local-scale susceptibility maps 00:42:41.659 --> 00:42:44.662 align:center line:-1 position:50% size:34% that are developed with similar or better data. 00:42:44.662 --> 00:42:46.197 align:center line:-1 position:50% size:34% For example, Puerto Rico. 00:42:46.197 --> 00:42:47.765 align:center line:-1 position:50% size:32% We included Puerto Rico just to sort of give 00:42:47.765 --> 00:42:51.068 align:center line:-1 position:50% size:33% this national--how it fits in at the national level. 00:42:51.068 --> 00:42:53.304 align:center line:-1 position:50% size:34% But there's a great existing susceptibility map 00:42:53.304 --> 00:42:56.574 align:center line:-1 position:50% size:38% of the USGS for Puerto Rico. 00:42:56.574 --> 00:42:58.042 align:center line:-1 position:50% size:30% So, it is overly general. 00:42:58.042 --> 00:42:59.811 align:center line:-1 position:50% size:24% We don't consider regional variability, 00:42:59.811 --> 00:43:02.046 align:center line:-1 position:50% size:31% and so, if you're looking for local applications, 00:43:02.046 --> 00:43:04.682 align:center line:-1 position:50% size:43% you may want to be careful about selecting appropriate cutoffs 00:43:04.682 --> 00:43:07.285 align:center line:-1 position:50% size:34% within that 0-81 spectrum. 00:43:07.285 --> 00:43:09.554 align:center line:-1 position:50% size:36% It's a gridded map that may be overly detailed, 00:43:09.554 --> 00:43:11.088 align:center line:-1 position:50% size:24% so down sampling, 00:43:11.088 --> 00:43:12.490 align:center line:-1 position:50% size:26% or using slope units, for example, 00:43:12.490 --> 00:43:14.859 align:center line:-1 position:50% size:34% is maybe the next iteration we're going to look at. 00:43:14.859 --> 00:43:16.227 align:center line:-1 position:50% size:26% And just a reminder, 00:43:16.227 --> 00:43:17.862 align:center line:-1 position:50% size:26% this is susceptibility, not hazard, 00:43:17.862 --> 00:43:19.363 align:center line:-1 position:50% size:34% and that's really important 00:43:19.363 --> 00:43:21.065 align:center line:-1 position:50% size:33% when considering frequency and magnitude 00:43:21.065 --> 00:43:24.202 align:center line:-1 position:50% size:39% in impacts of landslides from one location to the other. 00:43:24.202 --> 00:43:25.369 align:center line:-1 position:50% size:39% These are not the same thing; 00:43:25.369 --> 00:43:28.673 align:center line:-1 position:50% size:30% it's just potential for landslides. 00:43:28.673 --> 00:43:30.541 align:center line:-1 position:50% size:26% Thank you so much for your attention, 00:43:30.541 --> 00:43:34.779 align:center line:-1 position:50% size:37% and I hope this work is going to be useful for all of y'all. 00:43:34.779 --> 00:43:35.713 align:center line:-1 position:50% size:27% Thank you so much.