New results from Caxiuana

We're gradually going through the data from the Brazil trip, learning as we go. But the results so far are definitely very positive - all the hard work in laying out the targets systematically seems to have paid off in that the registration is better than we could have hoped for really. Andy's put in a great effort to get the drought plot done, and from what we can see there are very few areas where the registration of one scan to the next is anything other than excellent. Good news! The first fly through is done - again, props Andy:

https://www.youtube.com/watch?v=C4-xBCQ291M&feature=youtu.be

A 100 x 5 m slice (40 M points) through the drought plot showing the beautiful detail of the canopy profile.

Here's a zoom in of the slice.

Here's a second slice through the plot - again, coloured by height (note the slice through the runnels on the bottom).

Finally, here's a zoom in of the above slice, with the rainfall runnels more visible.

Kim's paper out in MEE

Kim's paper on testing the TLS-derived biomass estimates against destructive harvests was published today in Methods in Ecology and Evolution. For a local PDF go here. Wageningen University and UCL did press releases on it. It's a good paper - Kim should be proud of it. It's the real hard evidence that the TLS plus QSM really can do volume ok, particularly once you consider the uncertainty in the QSM reconstruction methods more systematically, which is what we've been able to contribute. A key result in my mind is that it really shows how the uncertainty in the TLS-derived values are independent of tree size, unlike the allometry-derived values which tend to grow with tree size, mainly due to the smaller and smaller numbers of larger trees. Now to do the same across the tropical trees we have, and we'll hopefully start seeing how this affects things across the tropics!

TLS point cloud and reconstructed model, from Calders et al. (2014).

And here's a really nice animation of the Brisbane field plot from WU.

Trees in their eyes

Here's a shot of our Wytham tree, scanned, reconstructed and looped, projected onto a screen and then reflected in my eye from close-up. Got it? Lovely.

Shipping out

So we've reached our last day in Brazil, for now. We took the overnight boat from Brevez to Belem, which was an adventure in itself. 14 hours up river, with decks of colourful hammocks slung from every available piece of metal. We had small cabins - comfortable enough despite the tiny bunks and freezing a/c. Better than a hammock though!

Hammocks on the boat from Brevez to Belem.

I watched the sun rise over the river and we made our way off the docks to the hotel, with all our cargo still in tact. Minus the target poles, which we left behind for Alex to use to mark the plots in the forest. We managed to lose/break only about 5 or 6 poles out of 70-odd, which is pretty good going and a testament to Lucy's organisation with the pole layout. That's been a very valuable experience in terms of the registration - we now know how to do this much more effectively than we did before, and how more poles isn't necessarily better. Systematic is the way to go.

Sunrise heading towards to Belem.

We've had a couple of days in Belem, doing some sightseeing - the old colonial parts of the city remain, slightly run down, but with some spectacular architecture including the old cathedral, the Theatre of Peace and the fort. We lucked out with the Theatre - after taking a tour we were given free tickets for a Richard Strauss concert that night. It was really beautiful, and a very strange contrast after getting off the boat from the remote jungle that morning, to sitting in a 19th C concert hall listening to Viennese waltzes.


Fresco on the ceiling of the Theatre de Paz, Belem.

Cathedral in Belem.

Our last thing to do was go to the Institute of Geoscience at UFPA, the University Federal do Para, to see Prof. Lola Da Costa, the driving force behind the Caxiuana experiment. Lola has been working at Caxiuana for more than 20 years, and is the reason the experiment exists, and has survived various funding threats over the years. We showed Lola some of early results from the scanning, and he was excited about the prospects of what we can do with data to elucidate links between structure and physiology. We're all extremely grateful for Lola's help in organising our trip and I'm sure we'll be back.

Pro. Da Costa with Lucy, outside the best fried fish place in town.

Lola took us for dinner on our last night at a small fish place - only thing on the menu was fried fish and prawns, and it was amazing. Some great local Para musicians were playing, (many) beers were drunk and it was a fitting finale for our trip to Brazil. Andy and I chatted about the tasks ahead for processing the lidar data - we worked out that we think we can do the registration much more efficiently, which would be hugely time-saving. Our idea needs needs testing but I don't know why we didn't think of it before - better late than never though eh?

Scanning done

So we finished the two towers today, although I don't know which one was Orthanc. I'm quite happy not to have to lug the scanner up, down, across or under anything for a while, other than on and off the odd boat or plane. The view from the towers is pretty special though, even if it is a little harder to appreciate when you're pouring with sweat and cursing gravity.

Looking down on the drought plot. Andy top centre, next to the runnel, looking up slightly nervously before making his way up (in harness of course) to help me bring stuff down.

Looking out into the treetops, seeing what the scanner is seeing.

The scanner is picking up the upper part of the canopy beautifully from on the tower platforms. If (and it's a bit of an unknown right now) we can hand-register the scans together from bottom to top, we'll have an amazing profile of wood and foliage all the way up. Another first I reckon.

A first view of a single scan, from the tower (which is the little red dot in the centre). Colours are height (0 to 43m in red).

Another view of the same scan.

These images are just single scans, and of course the platform and scaffolding obscure significant chunks of the view. But we're still seeing trees from out towards the edges of the plot, as well as the upper crowns, which is really encouraging. These were taken from platform 12 of about 20, so the upper views will be even better.

Meanwhile, the guys working on the panels in the drought plot are slowly but surely putting them back on, repairing some, replacing rotten ones. The scale of the experiment is much more obvious with the panels on - it's really something to see. And feel - the temperature underneath is pretty steamy, unsurprisingly, given it's effectively a 10,000 sq m greenhouse.

I bet you could grow some serious courgettes in here.

The team spent the afternoon trying to ID trees from the census data in the scanner data, to make our lives easier later. So far so good on that too. Tomorrow, we'll finish off a few loose ends, and hopefuylly try out the drone if it's calm. We also want to scan the station so we can give them a nice laminated print to remember us by ;-) We'll then pack up our stuff and load it on the boat before supper, as we're heading out downriver in the pitch dark at 4am, to catch the 8am catamaran from Brevez to Belem. Lucy has had her mind very much on a waterfront ice cream in Belem for about a week. I wouldn't say no either. I will miss this though.

Not many of these in North London.

First cuts of data

So today I was dragging the Riegl from level to level up the tall tower in the control plot. 18 levels, the heat, and it's pretty heavy. I was pretty much sopping after about 2 hours of it, and by the end, practically on my knees. Whose stupid idea was that? Oh.

Meanwhile, Andy's been furiously registering, slicing, dicing and generally working like a demon to pull out stem maps from the data so we can try and match them off the field data before we leave on weds night. He's produced a hugely downsampled pointcloud just to make life manageable until we can get our big machines in the lab on it. The examples below are from roughly 15 million points, which is < 1% of the data, and only from the upright scans. When we include the tilt scans, we'll obviously sample preferentially in the upper part of the canopy, and double the number of points.

An overview of the drought plot, coloured by height.

A zoom in of the largest tree in the plot - towards the centre back above.

 Amazingly, even in this (relatively) extremely sparse point cloud, the tops of some of the largest canopy trees are clearly visible. Also, the registration looks better than we could have hoped for. We've really been able to put what we've learned over the last 12 months into practice here, and I think it's really paid off. I may have even indulged in a very dignified whoop of delight on seeing this, slightly frightening Andy in the process.

A slice through the drought plot from 1.3 to 2.3 m above the ground. Lines running right to left are the wooden railings and runnels carrying the panels that keep the rainfall off the plot.

The slice through the data also shows up the wooden runnels that carry the panels very clearly. It's amazing to see them all laid out like this in such detail (the images above are rather low quality just because of the bandwidth) - and the trenches around the edges, clear of any vegetation. Even Lucy and Alex, who are extremely familiar with the plot, and essentially know every tree in there, are seeing it in a new way. It's funny - you can get buried in the details of measurements and kit, and data and logistics, but when you see the results stand out so clearly, it really brings home why we do these things.

Day of rest

We took a well-earned rest day today, had a lazy breakfast, drank a lot of coffee. Ed managed to buy 24 litres of beer in Brevez "by accident" - he says he definitely asked for 24 330ml cans. Riiiight Ed. But he's trying his best to put it right and hence rolled out of bed a little later than usual this morning. We took the smaller boat to a beach about 30 mins up river - really beautiful little spit of sand. Saw river dolphin, as you do. I must admit I didn't entirely believe Lucy's claim to have spotted one until I saw it, but it was definitely a dolphin. Oh, and it was very, very hot. Equator hot. Nothing that a quality hat can't deal with though.

Mat in a hat. Looking like a

Not too shabby for a morning sunbathe.

The local station staff and workers laugh at us for drinking black coffee with no sugar - particularly after lunch - totally unheard of here. They think it's hilarious and now make me two flasks of coffee in the morning, 1 for lunch and then there's often one waiting for me when we get back in from the field. It's tough but I'm getting through it.

Plots 1 and 2 complete

So after 7 days continuous work, with 10 hour days of scanning, we finished the main plots we came for - the 1ha drought and control plots. The weather has been nigh on perfect - calm, almost no rain. We'll all be grateful for a day off tomorrow, and maybe some beach action, cleaning clothes, and really just relief at not lugging stuff around the forest. We're all seeing reflectance targets wherever we look and I'm hearing the Riegl's bleeping scan noise in my sleep. Here's what we came for.

Literally the first cut - a tiny slice through only 6 upright scans, representing about 40m along and 20m across.

The scanner data look really, really good. The systematic target layout seems to have worked better than we could have hoped, with a team of 3 marshalled by Lucy keeping track of 60+ poles in 2 duplicate sets of 6 colours - a triumph of organisation! Andy cracks a big grin when the scans match up to within a few mm; when he frowns I know it means more work. But this is the first time anyone has seen this forest like this, and there is so much to come.

Done and done. Happy team after the 144th and final scan of the two plots.

On the way out of the plot, we saw this little fella.

Climbing snake - no ID as yet.

And so what's next after the rest day? Well, right on cue the heavens opened this evening, so looks like we got our core work done just in time. But I've thought that scanning up through the canopy would be great *if* we could do it - the two 45m towers give us a really good shot, and the detailed profiles would be very interesting - and never been done. We discussed how we might go about it and we've got a plan. Monday, we put it into action.

An old jedi mind trick.

School by the river

We had a great morning visiting the local (only) school, St Sebastian. Lucy has been building links between the school and her local school in Edinburgh, to help both sets of kids appreciate what life is like for each other. I'm not sure which is more difficult: for Scottish children to imagine what life is like living beside a huge river in the middle of a humid tropical jungle, where the temperature never drops below mid-20s (if that), snow might as well be from the moon and where the school run involves an intermittently running boat; or for the Brazilian children to imagine having to put on a coat to go out, for days to get dark at 3pm (or earlier!) and for the biggest threat to their lives to be traffic and crossing the road.

Everyone standing still waiting to be scanned, whilst being bitten by ants.

Lucy brought school uniforms from Edinburgh, loom bands (of course!) as friendship gifts, and a met station so the children can record weather data to compare with Scotland. They showed us their work and their school garden - their teacher Cleyson, is showing them how to grow manioc, banana, sugar cane and pineapple without needing to burn and clear forest first. Fire is the primary land clearance method at the moment, with all the attendant risk and long-term damage that it brings.

Snow? What's snow?

We got everyone out, including the school dog, and scanned them. Once again, everyone loves seeing the view the lidar provides! I'd love to come back and see how they're getting on & I really admire what both Cleyson and Lucy are doing here.

Everyone looks more real in lidar world.

The plots thicken

So we're now a week in to the trip, with 5 full days of scanning we've done all of the drought plot and half the control plot. The second plot is much harder to work in so we've had to improvise, using hand-made wooden tripods for some of the targets - maybe 3m tall. And just moving the scanner and tripod through the understory takes some doing at times. I find myself swearing to no one in particular a fair amount as the tripod snags on another liana, or I bark my knuckles on something for the twelfth time, all the while literally dripping with sweat. I've never sweated this profusely except in a sauna. But with 3 people on the target shake and bake, and me on the lidar, we're moving much faster than we thought: 13 locations today - 26 scans in all - which is our all-time record I think.

I went swimming before dawn yesterday, to do some filming of the sunrise for the ESA online course I'm contributing to. I took the camera they leant me down to the water and saw a herd of these on my way.



Little red rodent

A dawn swim is a thing of beauty in a place like this. This is a screengrab from a 30 minute sequence where I left the camera running as I swam, while the sun came up.

Sun rise over the river at Caxiuana.

Tomorrow, Lucy is taking us to a local school, where she is working with the teacher who organises the kids, the curriculum, the travel, everything. She's paired the school with a school local to her in Edinburgh, so the children in each can get an idea of what life is like across the world. What a great thing to do. I'm going to go along and scan the kids with the lidar, and send them a laminated print of themselves. Hopefully we can do the same for the kids in Edinburgh too.

Amazon scanning II

So far so good. We've been scanning for four days and have managed to capture the whole of the drought plot, plus 1/6th of the control plot. This is a major achievement I'd say - 72 scans, from 36 locations in the drought plot. 130GB of data, comprising around 700 million points, and 360 photos. This is an unprecedented dataset from this sort of forest, and we're all really excited to see what it brings in the way of science.

The scanner in action next to one of the runnels in the drought plot.

We've got things working fairly smoothly, operating two teams - one person manning the scanner, and then 3 people moving and placing the reflecting targets in a grid pattern. This can get complicated, and is the key to making sure we can register all the scans together. Lucy has taken charge of this aspect as it needs one person to have an overview of the process - it can get so confusing, particularly late in the afternoon when everyone's overheating and really tired. It's easy to make mistakes which can cost time and good data. Ed has arrived (after an epic 3 day journey!) and is now working with Lucy and Andy to place the targets. This is a much harder job in the control plot, where the understory is so much denser - unsurprisingly, having not been dried out for 10 years. But we seem to have got a system fairly smoothly. We were very lucky to be able to get the workmen to knock up some makeshift tripods for us to raise the targets to 3 or 4m, making them much more visible above the understory. We scanned 6 locations in 3 hours this afternoon. A couple more good days like that and barring any technical mishaps and we'll have cracked it. Oh, and Andy's laptops have both recovered, so he's a very relieved man.

Panorama of the drought plot - panels removed, but runnels still in place.

The scale of the operation to remove and replace the panels is pretty impressive. The local workmen are now replacing the panels since we've finished scanning the drought plot, and now we move on to the control plot. Days are long, hot and very very humid - every pore is dripping after a couple of hours, and it becomes an effort to drag yourself up and down after a while. But the thought of this is pretty enticing after a long day .....

Not a bad way to end a day. Or start one for that matter.

Up the Amazon

Amazing sunset from the boat as we make our way to the field station at Caxiuana.

We have been very fortunate to be invited to participate in a long-running drought experiment at Caxiuana, Para State, Brazil. The experiment is to establish what happens to the trees in the Amazon when they experience a severe drought. This has happened twice recently,with so-called 1 in a 100 year drought events occurring in 2005 and 2010. These resulted in large-scale mortality or trees and loss of carbon. But many questions remain about resilience, rates of loss and recovery and these are hugely important to understand what will happen to the Amazon under predicted increases in drought frequency and severity.

A view of some large Amazonian trees in the drought plot.

A major difficulty is quantifying the relatively subtle changes in structure that are hypothesised to occur in the upper canopy, not just the loss of large trees (or even small ones). So we are collaborating with Patrick Meir, Ed Mitchard and Lucy Rowland from the University of Edinburgh, along with their Brazilian colleagues at the Museo Goeldi in Brevez, to scan the drought plot at Caxiuana. The 1 hectare plot has had 50% rainfall exclusion for over a decade, making it one of the most droughted pieces of forest in the world. The neighbouring 1 ha plot has been set up as a control, to allow a direct comparison and assessment of the impact of drought.

We arrived this week to scan the two plots, aiming to explore the impact of drought on the canopy in terms of its strcutre, biomass and then differences in the upper canopy. The journey in and the field station and site are spectactular - 500 km by boar up-river to a beatiful, incredibly remote location. We have spent 3 days so far and have scanned nearly all of the drought plot. A few hiccups along the way, including two laptops that have expired due to the humid air, we think. We've managed to revive them with judicious drying and cooling, but we've been warned - the tropics kill electronic stuff. No kidding!

Scanning in suburbia

We spent a couple of days in August scanning in suburban Luton. Not necessarily the most glamorous, or even obvious location for scanning vegetation. But this work was for a colleague, Dr. Steve Hancock from Exeter University, who we've worked with for several years - since he graduated as one of our PhD students in fact. Steve's research interests include using terrestrial and airborne lidar to measure canopy properties, and using this information in radiation models, and ecological models more generally. Steve's currently working a 6-year NERC-funded research programme called BESS (Biodiversity and Ecosystem Service Sustainability). One of the themes of BESS is F3UES (Fragments, Functions, Flows and Urban Ecosystem Services). Steve is looking at ways to measure the structure of vegetation in urban areas in a fine-grained way, in order to assess its utility as an ecosystem. Related work is going on mapping bird and insect distributions in the same areas in order to assess ecosystem value. Small, fragmented urban areas of vegetation (gardens, parks etc) are generally overlooked when spatial estimates of habitat are made at scales of 10s of m or greater - they are often too small or irregular, or variable, to be included.

Steve asked us to come and scan some of their core sites in Luton that he had identified from airborne lidar as potentially containing interesting pieces of vegetation. We took the TLS into these areas (including a Girl Guide hut garden, suburban garden, street front and the large, mature garden of a residential care home) and collected scans of the dominant vegetation.

RGB (top) and range (bottom) from a suburban garden, flowerbeds and ornamental shrubs.

Scanning in back gardens, sitting on a bench and with the kind offer of a nice cup of tea from the house owner certainly makes a change from struggling through tropical forests, or even standing in English woodlands in the rain.  

RGB (top), intensity (centre) and range (bottom) from an urban carpark with a large willow in the foreground. 

I'm looking forward to working more on these data with Steve over the coming months, relating to the airborne lidar data he has, and trying to explain the properties of the lidar waveform data. Ultimately, Steve aims to provide a better characterisation of the ecosystem structure and value from this type of analysis, which will be both novel and illuminating.  

Housekeeping

My 9 year old son was asking whether we could scan ourselves and the house, and so, having the equipment at home before visiting one of our field sites I thought "why not?". Scanning in a tight, built-up environment like this illustrates some interesting aspects of the scanner geometry and the camera properties.

RGB (top) and distance-coloured scan (bottom) of rear of house, and me and Rudi.

The detail you can see always amazes me, and this is far from the highest resolution the scanner is capable of (100 times finer) and the image resolution is low to get it under the blogger 300k image size limit.

Close up of scan showing rear detail and shadows. 

The lower view shows the scanner shadow behind our figures, with my very characteristic (apparently) ears showing prominently. It can be hard for people who don't know how the scanner works to understand how you can view from any point, even where the scanner can't see due to shadowing/occlusion.

Rear elevation of the house, from high in the virtual lime trees.

View from above, showing the upturned paddling pool and even Rudi's crocs in front of him.

Next time I've got the scanner at home I might try and scan with targets to do a registered point cloud.

A year in the life ....

We visited Wytham again this week, to start a new project: scanning a single, isolated oak tree, as part of a documentary being made for the BBC by Paul Sen and his production company, furnace.tv. The documentary will tell the story of a year in the life of an oak tree, through the changing seasons and faces of the English climate and landscape. Paul had picked Wytham as a possible location, before hearing of our work and getting in touch. He had selected a tree, and wanted to film us scanning it if possible, analysing the dat and showing the final results. So Andy and I visited on a slightly too windy, but very English later summer's day, to scan the tree. Our colleague Eric Casella from Forest Research (see earlier posts) also brought his Leica scanner so we can compare data.

The star of the show. Wytham oak, looking North, with Paul and Eric to lower right.

 You can see why Paul picked the location - a gentle valley leading down to Wytham Abbey (nee House), just west of the banks of the Thames just beyond, the dreaming spires (TM) of Oxford further to the East beyond Wytham Village, and the gently rolling Oxfordshire landscape stretching away in the distance. The tree itself is fairly large. not tall, but broad and squat, separated by 20 m or so from the edge of the main woods, and sitting in a large pasture that seems to be used for cattle grazing. We scanned the tree from 8 locations around the compass, with Paul and his colleagues filming while we worked, asking ys about what we were doing, and how the instruments and process worked. Paul also set up some DSLRs to collect time-lapse shots of us working around the tree (I guess we'll look like cumbersome wasps in the edit of that!). Over the coming months we will revisit the tree to scan it with the leaves off in the winter, and in spring during leaf emergence if possible. Paul and crew will be returning throughout the next year to follow the tree and the changing landscape of its home, as well as exploring the ecology and biology that goes into making the tree what it is. In the longer term, we're interested in scanning individual trees in greater detail to explore the relationship of structure and function, so this is a very useful focus for us and the TLS as well. A fun project for all of us and I'm really looking forward to seeing how the film turns out.

Panorama of the tree, looking South, with Wytham Abbey and Village to the lower left.

A Wytham Tree

A first result from the Wytham scans - a single relatively small sycamore tree, around 20 m high with a dbh of 32 cm. 

The lefthand image shows the point cloud recorded by the TLS, coloured by height (blue to red). The righthand image shows the reconstructed cylinder model, same colour scheme. The reconstruction does a very good job of picking out the main trunk and 1st and 2nd order branching. Less well on the upper parts as usual. This is exacerbated by being only scanned from 5 locations here. As we expand through the plot we will collect more information at the upper part of the trees, and then redo the reconstruction to compare with these results.

May the sun always shine.