The art of canopy image processing

Since 2004, I’ve spent countless hours in taking and analyzing a special kind of photographs: skyward looking images that display tree crowns. Back in 2004 I managed to get a summer job at Metla, so I spent six weeks at Puolanka photographing tree crowns with a digital camera (back then it was not obvious that all cameras are digital). As a continuation project, I was hired to analyze the data: there were something like 1500 images, and emailed instructions how to analyze them with Paint Shop Pro. I needed to separate the sky and canopy pixels in the image so that the percent canopy closure could be calculated as the proportion of canopy pixels. Basically the job was a continuous torture for the keyboard and mouse (not to mention the analyst!), as most of the images needed some manual tuning before the conversion to binary format succeeded.

In spite of all this, I ended up using similar data in my master’s thesis, which meant more of the same cake. In addition, I manually painted the small gaps inside the crowns crowns black so that proportion of between-crown gaps could also be derived. Having gained some hard-earned expertise in this area, I came to contact with Hannu Ilvesniemi, who was planning a large-scale inventory in which crown photographs should also be taken. Fine, but who would analyze all those images? Luckily that wasn’t me: Jaakko Heikkinen happened to work for Hannu at that time, and he knew how Matlab could be used to repeat automatically everything that I had done manually. I was not familiar with automated image processing back then, so it looked almost like magic how he could obtain the same results than I just by typing the name of program into the Matlab command prompt. The trick is actually quite simple – using just the blue RGB channel helps to get rid of problems with clouds and contrast, and an automatic algorithm can be used to select the threshold value consistently. Even better, Jaakko could also separate the between-crown areas automatically with morphological image processing operations. He implemented this all so that results ended up neatly in excel sheets without any human interference, and an additional check image was printed for every image so that any problems could be spotted just by browsing them.

I started to feel that we had done something worth publishing – we had a nice method of analysis, images, and validation data that showed a reasonable correlation. So just the work of putting it all to paper remained. Review process went through quite smoothly, so now you can read it all from this article, download current version of our code here, and everything should work out of the box. This study is really one of the most satisfactory things that I’ve been part of during my career as a canopy researcher – I know that I’ll never need to work with canopy images using such elementary methods as I did five years ago. And the story is not over –I still use different versions of the same code for different tasks, such as estimation of LAI from hemispherical or even aerial images. As a summary: a programmer is the master of the computer, whereas the end user of the software is more or less its slave. So all the students out there – take your programming classes, someday it’ll pay back.

Field work completed

Last week Eeva and I spent a few days in Hyytiälä. The four-month long field campaign is now over, and we have a good looking seasonal LAI time series to analyze (thanks to Eeva!).

I had already forgotten how difficult it is to make LAI measurements in September (not to mention October). The sky is not uniform and the sun sets and rises 'too fast' - the time for making our measurements is very short each day and it seems that the campaign proceeds at a glacial pace. Anyway, we did get nearly everything we had planned done, and even had the time to eat a few blueberries when rushing from one plot to another.

The trip was memorable for another reason too. The roads at the study site are horrendous, and I ended up driving into a rock and breaking the oil pan of my car. The roads of Hyytiälä now have an oil track for several kilometers... and my car was towed 25 kilometers to the closest car workshop.

Two weeks in Boston

Miina has already written about her visit to Connecticut for the Multitemp conference at the end of July. Actually, she was not alone in the US. We were traveling together from Helsinki to Boston and while Miina was visiting the town called Mystic, I was having a good time working on p and having lunches with prof. Ranga Myneni. A large part of justifying the transatlantic flight was to meet the Climate and Vegetation research group at Boston University.

Heat wave in Boston

Boston was hot and tropical, air conditioning in BU was fierce, and welcome was warm. Thus, it was good two weeks. Eventually, prof. Yuri Knyazikhin also arrived from Spain and during the second week the theme of the visit closed in on the spectral invariant, photon recollision probability, or just p for short.

Discussions with Yuri. The photo is misleading: it is actually me who is at a loss, not Yuri. I manage to hide my confusion well, but this is (as I have heard) a common side effect of talking to Yuri about the eigenvalues

Spectral invariants: one of the hot spots of vegetation remote sensing. Everybody knows what they are, but I guess that only a few understand what they (or it, in case there is just one spectral invariant) mean. I have some publications on the subject, but why the spectral invariant approach works so well has remained a mystery to me. I understand the idea behind using the eigenvalue and why the radiation field should be reasonably well approximated by the first eigenvector. Still, sometimes points fall on the line with magical accuracy, and not just for canopy absorption, but also for reflectance.

The discussions we had with Yuri were fruitful, at least for me. I think I am a bit closer to some sort of an understanding. However, it will take some time before I can write this understanding down with sufficient accuracy.

There some other good things about traveling besides meeting interesting people. It is often about the little things, like having morning coffee and a croissant in Paris during a six-hour change at CDG. Some praise for such an opportunity has to go to the global recession: our original flight from Boston to Paris was canceled and we had to take an earlier one. Thus, what is usually considered a nuisance can be turned into a possibility -- to be in Paris before the tourists wake up, sit on a terrace of a boulangerie in the Latin quarter surrounded by a haze in the head created by the time difference watching Parisians cleaning the streets and opening up their businesses.

Paris in the morning

The jet lag, of course, had its revenge once we were back in Helsinki.

Cajanus tube conquers Norway

I returned last Friday from Norway, where I participated my first ever field measurement campaign outside Finland. Svein Solberg from Skog & Landskap contacted us to ask if some Cajanus tube experiments in Norway could be arranged. As Miina was busy with her travels, I volunteered for a substitute. And so after the usual flight and train travels I found myself in a farmhouse guest cabin, sitting nicely above a river valley near Svarstad in Southern Norway. And the river Lågen seemed to swarm with fisherman and salmon, wish I had known to bring my fishing tackle…

River landscape

The plan for next days was to make some Cajanus tube measurements and compare results to terrestrial laser scans obtained from the same plots. So I headed to forest with the TLS team, and had some time to follow their work before Svein and others arrived to discuss how the measurements should be arranged. We decided to do very typical dot count sampling but record also gaps inside the foliage, which is more difficult than usual between-crown gap measurement.

Terrestrial laser

Following days were typical field work with small differences to Finnish conditions, like climbing up to 30 minutes before arriving to the plot, cow dung and sound of the bells in the forest, and a Norwegian Cajanus tube that was considerably sturdier than the Finnish ones. Norway spruce forests were quite similar to Southern Finland, except usually more sloping. Also the rain that hindered our attempts was a very familiar experience… All in all, even though the final plot result remained once again smaller than planned, the campaign was definitely a success: also Norway now belongs to the international Cajanus tube family.

Two Cajanus tubes and an American implementation, GRS densitometer.

Towards hypertemporal remote sensing

I just attended a conference on multitemporal remote sensing images in Mystic, Connecticut. The conference was already the fifth in a series of workshops focusing on the analysis of time series imagery - for me it was the first one. The topics covered in the meeting were quite diverse, and included also radar images. Various methods for change detection and data mining were presented. There was discussion on cross-calibration of data from different sensors, accuracy assessment and detection of land cover changes and vegetation dynamics (the most interesting part for me). From many of the talks it was evident that we hope to be going towards hypertemporal (and not only hyperspectral) remote sensing - especially if we want to monitor environmental processes continuously and near real-time.

Tiit recommended this conference to me already last winter. I was interested to attend, and so we decided to prepare a presentation on our seasonal reflectance time series of birch stands. We used a radiative transfer model to identify the key factors which influence the seasonal pattern of stand reflectance in medium resolution satellite images (Landsat, SPOT), and briefly compared our results to 'landscape level' MODIS LAI and phenology products.

Signposting

Muutama viikko sitten löytyi meille uusi koeala. Nimi on "Parasmetsa" ja sijainti Saarenmaalla. Voisiko parempaa ollakaan PARAS-mallimme testaamiseen?

ALS, ALS, ALS

This blog has been too long without a decent Airborne Laser Scanning post. After a few months of laser data processing and one week on an intensive ALS course at Mekrijärvi research station, it is a good time for me to correct this deficiency.

In case that somebody not yet familiar with ALS happens to read this, the idea is that an aircraft equipped with laser scanner and inertia measurement unit flies above a forest area, measuring XYZ coordinates of laser return echoes. This way the 3D-structure of underlying forest can be measured very accurately. Here at University of Joensuu's forest inventory research group, ALS has been the main topic of research since 2004. First Finnish experiment with originally Norwegian area-based laser inventory method started then with the collection of the famous Matalansalo dataset. The idea in this method is that the height distribution of pulses hitting a pre-measured plot can be used to predict predict practically anything, from stem volume and LAI (easy cases) to biodiversity indicators (very difficult). Alternative approach is to detect individual trees in a stand and use allometric models to predict their dimensions, which requires higher pulse density.

Now five years have gone, and forest inventory is going through greatest revolution in its history. Numerous ALS papers have been published during this period, and Joensuu has become one of the leading knowledge centers in this field. Airborne forest inventories have entered commercial stage, first in company and state-owned forests, and next year also first private forest-owners can order a forestry plan that is based on laser inventory. National land survey of Finland has already started gathering data for laser-based elevation model, which will eventually cover entire country. This means that huge amounts of data will be available for practical applications of ALS research.

Personally I've been in great place to see this all happen, but my own start with laser data (which I've had since 2006) has delayed and delayed as other more urgent projects have moved forward. Now I actually feel that I'm a bit late - so many scientific papers concerned with laser measurements of canopy structure have already been published that the remaining job is to figure out how all those studies could be done a little better. Luckily I think I have some new ideas to try... So the last winter months before the yearly travelling season in may-june were quite intensive laser data processing, which will now continue for a few weeks before the start of the holiday season.

Last week's Nova course was certainly one of the highlights of Univ. of Joensuu's ALS history. PhD students and teachers, mostly from the Nordic countries, came together to study basics and more advanced topics of ALS in forestry, and also to have a good time in the peaceful countryside. For me the most interesting part was Ilkka Korpela's photogrammetric/physical approach to ALS, and especially his work with LiDAR-vegetation interactions. I totally agree with his view that also ALS research should next turn to a more physically-based direction, which will however require plenty of work and technological developments. But that's what science is all about - mapping the unknown.