Wednesday, 13 July 2016

Mondah plots done!

2 weeks, 2 hectares, 484 scans, many billion points, several hundred trees including way more 40+ m beauties than I've seen in one place before, some ants in the pants, some blood & definitely sweat, but fortunately no tears. Got some additional 'bonus' data to collect in the time left over with the ZEB, some high res scans of single trees but we got what we came for. We've acquired an additional member of the team for a few days in Melina, a stagiare with ANPN, who has already done some great work manually verifying which tree is which in the lidar stem maps. We're very happy about that and hope she can carry on with us in Lopé (if she wants to!). An example of one of these maps is here.
An example stem map, from plot MNG-04, showing the NE-most 20 x 20m quadrant, with tree sections from the lidar labelled with their tree IDs from the tags. Good work by Melina.
And we really couldn't have done this without the huge help of Leandre, who has been amazing - guiding, carrying, organising, driving and generally being the oil in the wheels of progress. Thanks!
There's nothing like getting that last scan done, and getting the heck out of Dodge for a shower and a nice cup of tea ;-)

Saturday, 9 July 2016

Mondah plot snaps

Andy has been working through the Mondah data from the first two GEM plots, and here are some examples of what they look like. First up is a transect through the MNG-03 plot, which covers 1 full row of 100m, and is about 20m deep. The plot shows the scanner locations on the 10m grid along the bottom (3 rows of them). The colours represent points coming from different scanning locations, with each colour corresponding to the colour of the scan location 'dot' - it's quite hard to see those for any other than the nearest ones on the black background, but you get the idea. For example, the purple points in the upper part of the tall tree in the centre come from scans to the far left and right in this case, showing how scan locations far from a given tree can often reveal the most detail at the top, due to being able to see through the surrounding crowns more easily.
Transect through MNG-03 showing the points from 3 rows of scan locations. There are 121 locations in each plot in total i.e. 11 rows of 11 locations, starting at 0, 0 and moving up to 100, 100 in 10m spacing. The point cloud here is drastically reduced (by factor of 1000) to allow easier manipulation.
 The image below shows a fill resolution example of a typical large tree with very grand buttressed roots, in the centre of the MNG-04 plot. In this case the colours represent the reflectance (returned energy) of the points, so the reflectance of points from the nearby broad trunk and roots is very high (near white), while the reflectance from the leaves and from points further away is much lower (orange through to dark red and black).
Reflectance image of small section of MNG-04 plot. This shows the full point cloud density.
Meanwhile, some more results from the ZEB-Revo scan are shown below. First, a height map of the whole plot, which shows the ZEB sees around 30m in some cases. So we can see the lower parts of the trees well, up to the crowns of the taller trees, but no detail in the upper part of the canopy. There are ~130M points in the 1ha cloud in total. Note that the heights aren't relative to the ground level but to the total height variation across the cloud. So the highest points (upper right) are still only 20-30m above the ground, but are on a raised area and so 30-40m above the lowest points. So the red parts here are largely a map of underlying topography. 
ZEB-Revo scan of MNG-04 plot, with the colour representing height, from 0 (blue) to around 35-40m (red).
Another view of the ZEB scan of the plot, from above this time.
View from above, looking down on the ZEB-Revo scan of the MNG-04 plot.
And a zoom in on one section in the diagonal gap towards the left of the plot. This gap may not be real - I think it may be a function of my scanning pattern going haywire during my walk of the plot. Staying on the grid is hard when you're struggling to walk, not fall down the slope or large holes, trip over logs and vines and hold the scanner at the same time!
Zoom in of small section of the ZEB scan of the MNG-04 plot.

 

Friday, 8 July 2016

Return to Gabon

So we're back in Gabon, returning after 3 years to scan Yadvinder's GEM plots at Mondah where we previously scanned. And then on to Lopé where we'll revisit 2 previously scanned AfriTRON plots, and then a few more new ones established by Simon for the ESA BIOMASS cal/val campaign. This will be the first time we can look at possible changes over 3 years, against a backdrop of the recent very large El Niño event.

Andy is leading the field campaign, over 6 or 7 weeks, with Matheus, myself, Ewan and Joanna coming in and out over the duration. We've had huge help from ANPN Gabon, via Kath Jeffrey, Lee White, David Lehmann and others in organising permits, kit and logistics. Without them it's safe to say this wouldn't be happening. And they've also organised great local help, Leandre and Hector in particular, who are really helping to make the operation much more efficient.

Day 1 and we had this visitor following us on the road!
No you can't have my lunch.




A 'typical' view of the second GEM plot.
Leandre carrying the Riegl out of the plot. A job for a younger man than me these days!
One of the giants we're here to scan, in the centre of our plot. Photos just don't do these trees justice - this one is 50+ m tall, with the enormous buttressed root system and a canopy that seems to be in the clouds.
One thing we're doing here is testing out our new Geoslam ZEB-Revo handheld scanner, to see what it can do. In particular we want to see how well it can capture the tree trunks and diameter, given its relatively short range, and how quickly we can do plots. First results are very encouraging. I scanned our 1h plot in about 80 minutes, generating 500G of data, which took ~6 hrs to process. But Andy and I were surprised at the level of detail and how well the tree trunks were captured. The snapshots here don't really do the detail justice, but this looks very good indeed. It opens up the possibility of us covering many ha at this level of detail, for trunk diameter and shape, to augment the more detailed Riegl data.
A very small slice through the 1ha ZEB scan of the plot.
I also scanned a single large tree in the corner of the plot, looping around it 3 times. The results from this are also very encouraging indeed. Here's a slice showing the tree with 2 slightly smaller (but still huge!) trees next to it. The white points are a second horizontal slice through the cloud, shown in plan view below.

Scan of single large tree (well, 3 really).
A slice through the trunk of the large tree, at about 4m above the ground, with the other two smaller trees next to it, and the rest of the point cloud in gray around them. The larger tree diameter is around 1m.
So, good progress and some interesting results so far and I'm really looking forward to seeing what we can get from the ZEB in combination with the Riegl. It's hard work doing the high density scanning in this landscape though! ;-)



Saturday, 25 June 2016

New papers from the group

Three new papers from our group, either from our work directly, or members of the team working with colleagues elsewhere. First up is Phil's paper written when he was in Australia, on characterising the vertical structure of forest canopies using airborne lidar data via a new metric, the Number of Strata (NOS). Phil et al.'s analysis shows that NOS can be a useful descriptor of canopy structure to complement canopy height and cover. They also propose this might be useful as a candidate Ecological Biodiversity Variable (EBV) for characterizing habitat structure.

Next we have work led by Sébastien Bauwens in University of Liège, and involving our own Kim Calders and colleagues from Wageningen University, comparing mobile hand-held and static TLS for forest inventory work, such as estimating canopy DBH rapidly across field plots. Their work shows that low-cost hand-held scanners can be a useful tool to supplement more expensive, longer range TLS instruments for DBH, particularly if speed is a key constraint, but struggle with height due to their much lower range.  

Lastly, we have a paper in PLOS One by former UCL postdoc Aida Cuni-Sanchez, with colleagues from the Agence Nationale des Parcs Nationaux and the Institute de Recherche en Ecologie Tropicale in Gabon and myself, Andy Burt, Kim Calders, Jose Gomez-Dans and Simon Lewis from UCL. Aida's paper presents work quantifying forest-savannah boundary dynamics over 20 years. Our contribution was to use our TLS to characterise the vertical forest structure quantitatively, allowing differentiation of forest types across these boundaries. Aida shows how colonising forests show much slower changes in biodiversity than structure or AGB. Aida also shows that all the forest types studied store substantial quantities of carbon.

Friday, 3 June 2016

Name-checked by Chris Packham and Frank Skinner!

Nic Stacey, who directed the Oak Tree documentary for Furnace TV we were involved in, alerted me to the fact that our work on that got a very nice mention by Chris Packham on Frank Skinner's weekly  On Demand With ... roundup of BBC highlights on the iPlayer. Both of them picked up on the number of leaves and seemed amused by the thought of me counting the leaves by hand. In practice, counting leaves like that is anything but amusing ;-) Hey ho, that's what the lidar is for I guess! Link to the program here, and from about 02:20 in: http://www.bbc.co.uk/iplayer/episode/p03qb629/frank-skinner-on-demand-with-chris-packham. It was also great to see that the documentary won a Royal Television Society award for Best Science and Nature programme. It's good to see Nic, Paul Sen, George McGavin and Furnace getting recognised for this, a really gripping and beautifully-made programme.


Friday, 20 May 2016

Karen's blog post on our recent lidar paper

Karen Anderson has written a really nice blog post on our recent paper "A comparison of LiDAR approaches for vegetation and landscape characterization" (published in the open access Remote Sensing in Ecology and Conservation journal). Karen's blog "Weighing up the benefits and costs of waveform LiDAR", and hosted on the ZSL blog site, outlines some of the benefits of waveform lidar for ecology. Our work shows how waveform lidar can provide habitat measurements that just aren't available any other way. Karen concludes that the next few years are likely to see an explosion in applications of waveform lidar in spatial ecology.
An image from Karen's blog showing our TLS work in suburban Luton (taken from: http://www.zsl.org/sites/default/files/styles/wysiwyg/public/media/2016-05/Figure%203.png?itok=ckeELLou)

Friday, 13 May 2016

First results from Ankasa

Andy's been busy processing the Ghana data. First cut from this is a flythrough - low res at the moment but you can see the tower at the start, and the topography towards the end. This was ..... a challenge shall we say.
Andy's also pulled out the 17 largest trees in the ESA AfriSCAT RADAR footprint - and the poibt cluds, and the resultiung models, are shown below. Our initial estimate is that there's around 150 t C in these 17 trees, probably close to half of the total for the whole plot!
Point clouds of 17 largest trees in the plot, extracted automatically from the co-registered full point cloud. Colours are just to differentiate each tree (same below).

Initial results from fitting cylinder models to point clouds. Mostly looking good, and the volume estimates seem robust from the parameter variation.
More tree models to follow!