1 00:00:01,187 --> 00:00:02,045 Hello. 2 00:00:02,045 --> 00:00:05,081 My name is Brandon Puckett, and on behalf of my coauthors, 3 00:00:05,279 --> 00:00:08,282 excited to share with you some of our research 4 00:00:08,414 --> 00:00:11,615 assessing the utility of drones and drone 5 00:00:11,615 --> 00:00:16,202 derived products to estimate key indicators of intertidal 6 00:00:16,202 --> 00:00:19,964 oyster reef resilience, persistence and function. 7 00:00:20,063 --> 00:00:22,076 And this comes at the request of managers 8 00:00:22,076 --> 00:00:25,838 who have identified as a need novel tools 9 00:00:25,904 --> 00:00:29,666 and approaches to improve their ability to monitor 10 00:00:29,699 --> 00:00:34,715 and assess the resource at useful spatial and temporal scales. 11 00:00:34,781 --> 00:00:38,774 Because oysters present some pretty unique 12 00:00:38,840 --> 00:00:41,414 management challenges as 13 00:00:41,414 --> 00:00:43,955 as a foundation species that is its own 14 00:00:43,955 --> 00:00:47,915 habitat, yet is often managed to support 15 00:00:47,981 --> 00:00:53,195 commercial and recreational fisheries. 16 00:00:53,261 --> 00:00:53,657 So to do 17 00:00:53,657 --> 00:00:56,924 so, we conducted a coordinated, regionally coordinated effort 18 00:00:56,924 --> 00:01:01,709 from North Carolina to North Florida, working with 19 00:01:01,775 --> 00:01:05,339 and within five National Estuarine Research reserves 20 00:01:05,339 --> 00:01:09,662 highlighted in the map on the left, their respective logos 21 00:01:09,728 --> 00:01:13,655 and really key members of the project team at each of those reserves 22 00:01:13,721 --> 00:01:17,450 are also highlighted here. 23 00:01:17,549 --> 00:01:20,816 Importantly, we also worked with representatives 24 00:01:20,816 --> 00:01:24,314 from each of the state resource agencies responsible 25 00:01:24,314 --> 00:01:29,759 for managing oysters in their respective states. 26 00:01:29,825 --> 00:01:32,828 And from them we heard the need for 27 00:01:32,894 --> 00:01:35,897 tools, approaches that are affordable, 28 00:01:35,930 --> 00:01:38,207 nondestructive 29 00:01:38,207 --> 00:01:42,398 methods that are capable of rapidly and quantitatively in a standardizable, 30 00:01:42,398 --> 00:01:45,533 fashion of monitoring and assessing 31 00:01:45,533 --> 00:01:49,262 intertidal oyster reefs. 32 00:01:49,328 --> 00:01:53,024 A critical part of that is the ability 33 00:01:53,024 --> 00:01:57,677 to develop tools and workflows to really quantify 34 00:01:57,776 --> 00:02:00,845 key indicators of intertidal reef resilience, 35 00:02:00,845 --> 00:02:03,815 persistence and function. 36 00:02:03,914 --> 00:02:07,445 And so conceptually, on this picture of an intertidal Crassostrea virginica 37 00:02:07,445 --> 00:02:11,537 Oyster Reef, which is the focal species of this work, 38 00:02:11,636 --> 00:02:16,949 I just want to highlight five key indicators that that managers prioritized, 39 00:02:17,015 --> 00:02:19,985 the first of which is Reef height and elevation. 40 00:02:19,985 --> 00:02:26,156 Like most organisms in intertidal settings, their abundance and distribution 41 00:02:26,156 --> 00:02:31,205 is really driven by where they sit in the tidal frame. 42 00:02:31,271 --> 00:02:34,373 Extent of the habitat managers really want to know how much of the resource 43 00:02:34,373 --> 00:02:37,772 is out there and how footprints of reefs may be changing over time. 44 00:02:37,772 --> 00:02:41,765 Whether that's expanding or contracting 45 00:02:41,831 --> 00:02:44,867 structural complexity, the 3D matrix of the reef. 46 00:02:44,867 --> 00:02:48,035 We know those interstitial spaces are super critical habitat, 47 00:02:48,035 --> 00:02:51,863 so this gets a habitat function of the reef 48 00:02:51,962 --> 00:02:52,919 shell budget. 49 00:02:52,919 --> 00:02:55,394 Shell is the currency of the reef. 50 00:02:55,394 --> 00:02:58,001 There are removal terms 51 00:02:58,001 --> 00:03:01,268 through dissolution of shell, burial of shell 52 00:03:01,268 --> 00:03:05,228 and sediment and removal via recreational commercial harvest. 53 00:03:05,228 --> 00:03:07,307 There are also addition terms 54 00:03:07,307 --> 00:03:11,828 recruitment of new oysters to the reef, growth of oysters, and then addition 55 00:03:11,828 --> 00:03:17,240 of reef through excuse me, addition of of shell through restoration 56 00:03:17,339 --> 00:03:18,857 and lastly, abundance. 57 00:03:18,857 --> 00:03:21,464 How many oysters are on the reef 58 00:03:21,464 --> 00:03:24,203 and how many of those are of legal harvestable size? 59 00:03:24,203 --> 00:03:28,394 Those are questions that managers are often asking. 60 00:03:28,460 --> 00:03:29,054 And so these 61 00:03:29,054 --> 00:03:34,895 metrics indicators have generally been estimated in situ with approaches 62 00:03:34,895 --> 00:03:39,515 like using quadrats or in some cases such as extent using remotely 63 00:03:39,515 --> 00:03:43,706 sensed satellite derived imagery and both of those approaches 64 00:03:43,706 --> 00:03:49,217 have their strengths, albeit some severe weaknesses as well. 65 00:03:49,316 --> 00:03:51,659 There's a clear need, I think, to 66 00:03:51,659 --> 00:03:54,563 sort of bridge that gap between quadrats and satellites. 67 00:03:54,563 --> 00:04:01,625 And drones really offer a platform that could potentially do so. 68 00:04:01,691 --> 00:04:03,968 With that said, we set out as part of this work 69 00:04:03,968 --> 00:04:09,413 to really accomplish two objectives, the first of which is to use 70 00:04:09,512 --> 00:04:10,535 drone derived products. 71 00:04:10,535 --> 00:04:13,835 So point clouds, digital surface models and orthomosaics 72 00:04:13,835 --> 00:04:16,805 to quantify reef extent 73 00:04:17,036 --> 00:04:20,303 Elevation, rugosity, which is going to be our proxy 74 00:04:20,303 --> 00:04:26,045 for structural complexity, shell volume, which is our proxy for shell budget 75 00:04:26,111 --> 00:04:30,566 and oyster density and size structure, which is our proxy for abundance. 76 00:04:30,665 --> 00:04:34,988 And then secondly, to compare those drone derived estimates of those indicators 77 00:04:34,988 --> 00:04:38,486 with the same measurements conducted on the ground. 78 00:04:38,486 --> 00:04:41,489 So in situ measurements to assess 79 00:04:41,522 --> 00:04:46,109 essentially the accuracy of drone derived estimates. 80 00:04:46,175 --> 00:04:46,835 And both 81 00:04:46,835 --> 00:04:49,838 of these objectives are sort of a necessary precursor 82 00:04:49,838 --> 00:04:54,524 before operationalizing drone derived products 83 00:04:54,722 --> 00:04:59,606 in sort of a monitoring and assessment framework. 84 00:04:59,705 --> 00:05:00,365 So briefly, 85 00:05:00,365 --> 00:05:04,226 for methods for in-situ sampling and lab processing. 86 00:05:04,226 --> 00:05:07,955 I'm going to focus today on the work we did in 87 00:05:07,955 --> 00:05:11,717 Central North Carolina, the coast of North Carolina. 88 00:05:11,783 --> 00:05:14,951 But this was a coordinated 89 00:05:14,951 --> 00:05:18,647 experiment done across North Carolina to North Florida. 90 00:05:18,713 --> 00:05:21,716 But in North Carolina, we sampled 13 reef sites 91 00:05:21,947 --> 00:05:26,831 during the summer of 2023 and more or less a two by two factorial design 92 00:05:26,897 --> 00:05:31,549 whereby we had replicate patch reefs that were open to harvest and replicate 93 00:05:31,549 --> 00:05:36,169 patch reefs that were closed to harvest as well as replicate patch reefs or 94 00:05:36,169 --> 00:05:40,195 fringing reefs excuse me, open to harvest and fringing reefs closed to harvest. 95 00:05:40,294 --> 00:05:44,056 And that was just so we'd get a pretty wide span 96 00:05:44,056 --> 00:05:49,006 of sort of reef types as well as management regimes. 97 00:05:49,105 --> 00:05:52,174 At each of these reefs, we took 4 to 6 rugosity measurements, 98 00:05:52,174 --> 00:05:54,682 as shown in the second picture from the left, 99 00:05:54,682 --> 00:05:58,114 using the traditional sort of chain and tape approach 100 00:05:58,147 --> 00:05:59,731 where you can see the chain where we tried 101 00:05:59,731 --> 00:06:03,955 to lay that throughout the interstitial spacing on the reef, 102 00:06:04,054 --> 00:06:06,034 we excavated six quadrats 103 00:06:06,034 --> 00:06:09,565 at each reef and then took those samples 104 00:06:09,565 --> 00:06:13,327 back to the lab and measured and counted oysters to estimate density. 105 00:06:13,459 --> 00:06:16,198 And then estimated volume of material. 106 00:06:16,198 --> 00:06:21,346 Shell excavated via water displacement. 107 00:06:21,445 --> 00:06:24,019 So for drone based sampling, our methods were as such. 108 00:06:24,019 --> 00:06:27,022 We used the DJI Phantom four Pro 109 00:06:27,088 --> 00:06:29,728 with a 20 megapixel red green blue sensor. 110 00:06:29,728 --> 00:06:33,391 So pretty consumer grade level airframe and sensor. 111 00:06:33,457 --> 00:06:36,955 We flew at low tide when when reefs were exposed 112 00:06:37,054 --> 00:06:42,004 at approximately solar noon to minimize shadowing on the reef, 113 00:06:42,070 --> 00:06:45,007 we set front and side overlap at 75%. 114 00:06:45,007 --> 00:06:49,033 And generally speaking, the reefs that we were flying 115 00:06:49,198 --> 00:06:55,072 for the study were on the scale of 1 to 5 hectares in size. 116 00:06:55,171 --> 00:06:55,963 Our ground sampling 117 00:06:55,963 --> 00:07:00,022 distance was one centimeter per pixel, albeit at a subset of sites. 118 00:07:00,022 --> 00:07:03,553 We did fly at lower altitudes to 119 00:07:03,553 --> 00:07:06,556 achieve a ground sampling distance of half a centimeter per pixel. 120 00:07:06,754 --> 00:07:11,176 To look at the sensitivity of some of the metrics to ground sampling distance. 121 00:07:11,176 --> 00:07:12,661 And I'll highlight that on the next slide. 122 00:07:12,661 --> 00:07:15,202 When talking about rugosity, 123 00:07:15,202 --> 00:07:18,667 at each site we deployed and surveyed with RTK GPS, 124 00:07:18,733 --> 00:07:22,825 12 ground control points as well as six checkpoints to 125 00:07:22,891 --> 00:07:26,851 to assess the accuracy of our our elevation models. 126 00:07:26,917 --> 00:07:30,217 We flew before and after our in-situ quadrat excavation. 127 00:07:30,217 --> 00:07:34,408 So flights before and after were staggered by about an hour 128 00:07:34,507 --> 00:07:37,081 and used structure for motion photogrammetry software. 129 00:07:37,081 --> 00:07:40,777 In this case Agisoft Metashpe to generate orthomosaics 130 00:07:40,843 --> 00:07:45,793 point clouds and digital elevation models and analyzed those products in ArcGIS Pro. 131 00:07:45,793 --> 00:07:49,192 So again using consumer grade airframes and sensors 132 00:07:49,192 --> 00:07:53,218 and software that is very readily available to the management community, 133 00:07:53,317 --> 00:07:55,132 The picture here on the bottom just shows 134 00:07:55,132 --> 00:07:59,488 sort of a typical field set up with our ground control points in yellow 135 00:07:59,587 --> 00:08:04,108 quadrat locations in red, rugosity measurements in blue dots 136 00:08:04,108 --> 00:08:08,827 and then elevation along axial profiles in purple. 137 00:08:08,893 --> 00:08:11,797 So the first indicator I want to speak about is structural complexity, 138 00:08:11,797 --> 00:08:16,417 which we're going to rugosity as a proxy of that. 139 00:08:16,417 --> 00:08:19,717 And so I've shown an elevation profile extracted 140 00:08:19,717 --> 00:08:23,017 from a linear measurement from our surface model of one of our reefs. 141 00:08:23,017 --> 00:08:27,934 So you have distance and meters on the x axis and an elevation on the Y axis. 142 00:08:28,000 --> 00:08:31,168 And the way rugosity works is essentially it is the distance 143 00:08:31,168 --> 00:08:36,052 it would take to traverse all of those 3D structure intricacies, 144 00:08:36,052 --> 00:08:39,253 interstitial spaces of the reef as a function of linear distance. 145 00:08:39,253 --> 00:08:43,840 So a flat table has a rugosity of 1 to 1 because those values are the same 146 00:08:43,906 --> 00:08:46,678 as your 3D structure of your reef increases. 147 00:08:46,678 --> 00:08:49,417 So does your rugosity. 148 00:08:49,483 --> 00:08:51,496 And so how do our estimates 149 00:08:51,496 --> 00:08:55,852 of rugosity measured from our surface models compare to what we measure in situ? 150 00:08:55,852 --> 00:09:00,736 So here that is plotted so in situ rugosity estimates on the x axis 151 00:09:00,835 --> 00:09:05,587 and rugosity estimated from digital surface models on the Y axis, the little 152 00:09:05,719 --> 00:09:09,877 dashed line there is the 1 to 1 line, which would be perfect agreement. 153 00:09:09,943 --> 00:09:11,560 And so a couple of take homes here. 154 00:09:11,560 --> 00:09:16,081 So first of all, we are underestimating rugosity pretty considerably. 155 00:09:16,081 --> 00:09:21,658 So all of those observations you see there are below that 1 to 1 line 156 00:09:21,658 --> 00:09:24,760 and that underestimation tends to get worse 157 00:09:24,760 --> 00:09:28,027 as our ground sampling distance gets larger. 158 00:09:28,093 --> 00:09:31,822 So the light green dots are one centimeter per pixel, whereas the dark 159 00:09:31,822 --> 00:09:36,442 blue dots are half a centimeter per pixel. 160 00:09:36,541 --> 00:09:39,412 The third point here, though, is that the functions are linear, 161 00:09:39,412 --> 00:09:43,306 and I think we need to have discussion with our management group to see if 162 00:09:43,405 --> 00:09:46,243 it's important that the relationship here is 1 to 1 or if 163 00:09:46,243 --> 00:09:47,893 if this linear relationship, albeit 164 00:09:47,893 --> 00:09:51,787 an underestimate, is substantial for their purposes in terms of estimating 165 00:09:51,919 --> 00:09:54,658 and thinking about the 3D structural complexity of a reef. 166 00:09:54,658 --> 00:09:55,846 But just to highlight 167 00:09:55,846 --> 00:09:58,849 sort of the differences of what those profiles look like at those two 168 00:09:58,981 --> 00:10:02,281 ground sampling distances, I've shown the profiles here 169 00:10:02,512 --> 00:10:05,515 and what you see is that at some of the 170 00:10:05,647 --> 00:10:10,531 some of the larger peaks like here, you lose about a centimeter in elevation 171 00:10:10,531 --> 00:10:13,534 as you increase your ground sampling distance and you fail to pick up 172 00:10:13,534 --> 00:10:16,933 some of the the minor details and smaller peaks. 173 00:10:16,933 --> 00:10:20,035 So those peaks that are here aren't as readily 174 00:10:20,035 --> 00:10:24,952 picked up as our ground sampling distance increases. 175 00:10:25,051 --> 00:10:25,546 The second 176 00:10:25,546 --> 00:10:29,506 indicator I want to speak about is shell budget. 177 00:10:29,506 --> 00:10:34,786 And so again, we're going to use volumetric change as a proxy for that. 178 00:10:34,852 --> 00:10:41,254 And so what I've shown here on the left is essentially a difference map. 179 00:10:41,254 --> 00:10:45,346 So we subtract our surface model before and after excavation. 180 00:10:45,346 --> 00:10:48,514 So again, flights before and after staggered by about an hour. 181 00:10:48,646 --> 00:10:50,230 And what you're seeing is a lot of purple. 182 00:10:50,230 --> 00:10:53,167 And if you look at the color bar there, that means no difference. 183 00:10:53,167 --> 00:10:53,695 Right? 184 00:10:53,695 --> 00:10:56,764 Shouldn't be very much difference in this reef surface over the course of an hour, 185 00:10:56,962 --> 00:11:00,822 except where we went out and excavated material. 186 00:11:00,888 --> 00:11:05,739 So zoomed in here are our quadrat plots, the corners there are the dots there, 187 00:11:05,739 --> 00:11:10,722 the corners of the quadrats and the teal color indicates a loss in elevation. 188 00:11:10,722 --> 00:11:12,999 So where we excavated material. 189 00:11:12,999 --> 00:11:18,081 So if we take and we can sum the volume, the total volume of reef material 190 00:11:18,081 --> 00:11:22,404 that we removed, we can then look at how that sum total compares 191 00:11:22,404 --> 00:11:23,856 with what we measured via water 192 00:11:23,856 --> 00:11:26,859 displacement in the lab versus what we predict from our surface models. 193 00:11:27,024 --> 00:11:32,205 And so I've shown that here with in situ volumetric change on the x axis 194 00:11:32,205 --> 00:11:37,353 and surface volume, volumetric change on the Y axis and a couple of points. 195 00:11:37,353 --> 00:11:42,336 So the dots there are 13 reefs, many of which fall along that dashed 196 00:11:42,336 --> 00:11:46,362 1 to 1 line, albeit the best fitting linear regression suggests 197 00:11:46,362 --> 00:11:50,718 that it's not completely 1 to 1, albeit the slopes there are close to one 198 00:11:50,718 --> 00:11:52,038 and the Y intercepts are very small. 199 00:11:52,038 --> 00:11:56,163 So very good agreement between observed and predicted 200 00:11:56,295 --> 00:11:59,562 volumetric change at the reef scale. 201 00:11:59,628 --> 00:12:01,839 And just for context here, we're predicting 202 00:12:01,839 --> 00:12:04,314 we're picking up pretty small changes in volume. 203 00:12:04,314 --> 00:12:07,977 The largest volumetric change that we observed 204 00:12:08,076 --> 00:12:10,419 down here is about half a bushel. 205 00:12:10,419 --> 00:12:13,587 So if we think about removal from recreational commercial fisheries, 206 00:12:13,587 --> 00:12:18,042 that's a pretty minute amount. 207 00:12:18,108 --> 00:12:21,309 In the third and final indicator I want to 208 00:12:21,309 --> 00:12:25,137 discuss is abundance, and density 209 00:12:25,170 --> 00:12:27,216 and size structure will be the proxy we'll use for that. 210 00:12:27,216 --> 00:12:31,572 So here I've shown observed density on the x 211 00:12:31,572 --> 00:12:35,796 axis versus predicted density on the Y axis. And the way we predicted 212 00:12:35,796 --> 00:12:40,383 total density shown here is via 213 00:12:40,416 --> 00:12:44,475 sort of multiple linear regression modeling whereby we use kind of structural 214 00:12:44,475 --> 00:12:49,326 characteristics of the quadrats and spectral characteristics of the quadrats. 215 00:12:49,326 --> 00:12:52,956 So things like the number of of light and dark pixels tend to be important 216 00:12:52,956 --> 00:12:53,814 predictors, as 217 00:12:53,814 --> 00:12:59,094 does the average elevation of the quadrat and deviation around that average. 218 00:12:59,193 --> 00:12:59,985 And so what we see 219 00:12:59,985 --> 00:13:03,516 here, again, the dots are the 13 reefs with their 220 00:13:03,516 --> 00:13:07,377 associated errors around average total density. 221 00:13:07,377 --> 00:13:10,413 The dash line again is the 1 to 1 line. 222 00:13:10,611 --> 00:13:15,594 And what you see is that in many cases our estimates predicted values do fall 223 00:13:15,726 --> 00:13:17,706 right along that one, two, one line, albeit 224 00:13:17,706 --> 00:13:20,775 the best fitting linear regression would suggest that at really high 225 00:13:20,775 --> 00:13:25,560 observed oyster densities, we are now underestimating total density 226 00:13:25,560 --> 00:13:29,916 and really low estimates of observed density, 227 00:13:29,916 --> 00:13:32,160 we're probably overestimating density. 228 00:13:32,160 --> 00:13:35,823 The cool thing about the the drone derived products is we can start to project those 229 00:13:35,823 --> 00:13:38,826 at sort of reef wide scales. 230 00:13:38,826 --> 00:13:41,202 And so I've shown that here a map at a small patch reef 231 00:13:41,202 --> 00:13:45,327 where by total density, the highest total density values are indicated 232 00:13:45,327 --> 00:13:51,564 by the yellow and red colors and lowest values by the blue. 233 00:13:51,663 --> 00:13:54,336 And so I think maybe with the exception of rugosity, 234 00:13:54,336 --> 00:13:57,801 we've demonstrated to some extent that that drones are a tool 235 00:13:57,801 --> 00:13:59,781 that can be used to quantify some key 236 00:13:59,781 --> 00:14:03,576 indicators of oyster reef resilience, persistence and function. 237 00:14:03,576 --> 00:14:07,305 And while I didn't have time to to talk about our results for reef extent 238 00:14:07,305 --> 00:14:11,859 in aerial and reef elevation, you'll just suffice it 239 00:14:11,859 --> 00:14:12,750 to say that we were able 240 00:14:12,750 --> 00:14:15,951 to capture those very accurately with drone derived products. 241 00:14:16,050 --> 00:14:20,109 So for next steps we're going to refine the workflows that I've sort of shown here 242 00:14:20,109 --> 00:14:23,112 and apply those throughout the geography of the rest of the study. 243 00:14:23,376 --> 00:14:26,247 South Carolina, Georgia and Florida, 244 00:14:26,247 --> 00:14:29,547 we're going to conduct a coordinated experiment across that geography 245 00:14:29,547 --> 00:14:33,738 whereby we're going to try to mimic harvest by removing shells in sort of a 246 00:14:33,738 --> 00:14:38,127 more manner that more closely resembles fishery removal. 247 00:14:38,292 --> 00:14:42,186 And we'll fly before and after that removal and measure 248 00:14:42,186 --> 00:14:43,935 the same response variables. 249 00:14:43,935 --> 00:14:45,948 And then we'll add shell to 250 00:14:45,948 --> 00:14:49,941 mimic restoration and fly before and after that addition. 251 00:14:49,941 --> 00:14:53,241 And then I think lastly, and something that could be pretty cool 252 00:14:53,274 --> 00:14:56,805 is sort of a first of its kind with our ability to project these 253 00:14:56,904 --> 00:15:00,963 indicators across the reef is to develop a multi metric index of reef resilience 254 00:15:00,963 --> 00:15:05,715 that can really inform, I think, oyster restoration and management. 255 00:15:05,715 --> 00:15:09,873 And with that, I would like to thank you for your time 256 00:15:09,939 --> 00:15:12,843 and I'd like to thank NOAA's National Center for Coastal Ocean Science 257 00:15:12,843 --> 00:15:15,219 and the NERRS Science Collaborative for funding this work. 258 00:15:15,219 --> 00:15:17,397 And my contact information is provided here.