The beta regression

Sorry folks, but this post is (for once) almost purely scientific, and does not include anything about eating or drinking. The reason for this is that the article with which I worked for most of last spring has now been published, and as I think that most of you won’t probably bother to read it, I decided to write something about it into this blog. If advertising would not work, so much time and money would not be spent on it…

The aim of my PhD project is to develop and compare different methods for estimating forest canopy cover. As accurate field measurements of cover percent are quite slow, one alternative is to use statistical model to predict it from easily measurable forest variables. This issue had been around since my master’s thesis, where my first attempts to model the cover with simple linear regressions were made with a data collected near Metla’s research station in Suonenjoki. The results were relatively good, but linear regression models just don’t function very well in modelling percentages: no scientist wants to see percentages that are negative or larger than 100%. In addition, the old models did not function nearly as well when more data were included, so there was still a lot of work to be done when I started the process in early 2007.

I spent some time studying the different modelling techniques: logistic regression, nonlinear curves and generalized linear models, to see if any of these would function better with this kind of data. Nothing very useful appeared, until I found this lecture material by Juha Heikkinen. There was a hint: Juha suggested using the beta distribution in modelling proportions. Fine, but how? Luckily, nowadays there is always one to ask: Google. After some surfing I discovered that there was an R library called betareg, which was just the thing I was looking for: the beta regression was both simple and suitable for this kind of problem. Another interesting thing was that even though a lot of statistics are used in forest science, I could not find any earlier forest-related studies where this technique would have been used, only something related to psychology. So maybe this paper would have something “new” after all.

What remained then was the tough job of curve fitting. In this case it meant writing numerous R-scripts to handle the data, testing the different model shapes, and fixing the small bugs in the library. I tested many different model shapes with not too good results, as the correlations were heavily nonlinear. In the end I decided to use cubic functions to make the polynomials flexible enough, but fortunately the beta regression’s logistic link function would take care of extrapolation problems. As the analysis was done, what remained was the writing of the manuscript, which was finished in May. The manuscript was then offered to Silva Fennica, and after some revisions it was accepted for publication. And now, here it is, and it is time for new challenges.

The Coffee & Cake Ceremony

Yesterday it was once again time for one of our group's traditions: the coffee and cake ceremony. The ceremony takes place every time one of us publishes a paper: the first author takes all the members of the group (who are in town) as well as the co-authors to a café to enjoy a piece of cake and coffee. The ceremony has been extended to cover also other situations: it's called the coffee&cake&cognac ceremony when we receive scientific awards or funding (or something like that) and it's the dinner&wine ceremony when something Really Big happens.

Yesterday, it was Pola's turn (she published a paper), and she took us (Janne, Matti and me) to Fazer in the afternoon. For me it meant a very short day at work since I gave a lecture in the morning and had to prepare the following day's lecture in basically no time before rushing downtown to the café...

While enjoying the cakes we had our daily chance to catch up on all the more or less scientific gossip. Who's done what, who's heard what...

Chilli-chocolate cake - excellent!

Hurtigruten again...

Lauri forgot to mention our latest experiment (carried out during the Hurtigruten trip): does a PhD degree correlate positively with success in card games?

NORSEN-meeting at Hurtigruten, 1.-3.11.2007

Even forest science is mostly office work - the field work is most often outsourced to students or specially educated forestry professionals. Practically the only variety to daily computer routines comes from different conference and meeting trips. So when I heard that there would be a possibility to attend the NORSEN-meeting arranged in Troms, Northern Norway, and the legendary coastal steamer Hurtigruten, I was of course quite eager to go, expecting to learn a little what all that field work last summer had been about, discuss with others, and just have a good time seeing new places and landscapes. Only the 13 hours travel time per direction from Joensuu was something that I did not look forward to. Especially when the route was through Helsinki and Oslo, which seemed slightly irrational from my perspective.

So in the evening of the first travelling day our group - me, Matti, Miina, Pekka, Pola, Terhikki and Juha - were sitting in the auditorium listening a presentation by a local satellite company that uses radar data e.g. to detect oil spills in the Arctic Ocean. After the unforgettable lutefisk dinner we moved to the ship - MS Trollfjorden, which was a moderate shock for someone not used to luxurious travelling… And as the morning brightened and the snowy peaks of Lofoten became visible, it became clear that the yesterdays journey had already paid off.

The meeting itself was also worth listening to. The NORSEN coalition included a number of different participants sharing an interest of using satellite data in environmental monitoring, and to me it was interesting to hear what is remote sensing outside forest sector. Usually it is seems to be much simpler… But after some hours as tiredness began to take over, you could almost imagine being on any university seminar (at least when the room was not swaying to and fro) - until you stepped out of the auditorium and saw the new surroundings. Wow! Lofoten and Hurtigruten are worth the praise they are given. Or maybe we were just lucky, as there was no rain during the whole trip…

In the Friday evening we left the ship in Svolvaer, local governmental center, and after a few hours waiting boarded another, slightly older ship that was heading back north. Next day was free, so there was plenty of time for enjoying the scenery (even sun appeared!), meals (especially dessert!) and some scientific speculations (how do you measure LAI if the trees are covered by snow?). At sunset we were back in Troms, and I’d guess many of us would like to return to Hurtigruten some day. At least as a well-off pensioner, like most of the passengers.

And then things started to go wrong: darkness, chilling wind under the cold sky, crowded restaurants, a terrible movie (that I skipped at the cost of chilling wind), burnt pizzas and slow service, smoky airport hotel, delayed flights, airport bus taking an unusual route back in Joensuu… Well, nothing disastrous actually, but when I once more began to focus on revising a laser manuscript in my room at the university next morning, the ordinary days of the new week did not seem too bad after all.

Official part


Cruising


Hotel yard, sunday morning

Not too successful measurements

Sometimes measurements fail. Especially if carrying them out successfully requires certain kind of weather. And if the certain kind of weather you are waiting for is a standard overcast sky (and especially no rain!). And if it's September or October at our latitudes.

From this introduction you can already guess that my LAI-2000 measurements in Järvselja turned out non-existent last week. I've been carrying out a series of measurements to catch the seasonal change in LAI (and reflectance): first measurements at the study plots were made already in April before budburst, the next ones in July with fully developed foliage. My purpose was to catch the final situation too: yellow (or red) leaves or green needles in early October. But it rained. And rained. And rained. And I got absolutely no measurements done even though I spent hours with the instruments and my field assistant sitting in the car next to the reference site, waiting for the rain to stop. The good news is that while we sat in the car, we ate a lot of sandwiches, cookies and bananas - at least we didn't go hungry...

I must admit that I was quite frustrated not to get the last set of measurements done. But there is nothing I can do about that - the leaves are now on the ground, turning brown. And I'm already waiting for next year and new measurements - I guess.

Leaving the car for a few minutes to see if the rain had stopped. But no, it hadn't...

Field assistant playing with autumn leaves in the rain.

Järvselja in October.

Physical models in remote sensing –course in Helsinki

Graduate school in Forest sciences organized this course 9.-11.10.2007 in Helsinki. Two members of LAI Detectives participated the event: Pola as course leader and I as a student. Our four-member Joensuu student delegation took a train to Helsinki already in Monday and accommodated in hotel omena. In the hotel we met other long distance participants from Estonia and Norway, so we could share the problem of finding the place of lectures, Gardenia building in Viikki. Fortunately we found the right place without any severe straying.

Apart from the opening words from Pola and Jari Varjo, the first two days consisted of lectures by Tiit Nilson and Andres Kuusk. The main theme was the physical basis of remote sensing: radiative transfer in atmosphere and vegetation, ground and canopy reflectance models, measurement techniques, and practical applications. As we Joensuu students are mostly dealing with forest inventory (=statistics with forest and/or remote sensing data), difficulties in following the physics and equations were inevitable, but I think that all of us got a good general overview of the problems and aims of more physically oriented remote sensing. To me personally the most interesting part was the practical demonstration of the FRT-model, of which I had heard many details but never got a general idea how it actually functions.

In the final day lectures were given by Jouni Pullianen and Sanna Kaasalainen, whose lectures covered the physical basis of radar and lidar remote sensing. These were more familiar to our Joensuu group, as Jouni had lectured about microwave remote sensing in Joensuu earlier this year, and laser scanning is currently the most important research topic in Joensuu's forest inventory group. It was still good to hear a laser lecture from a different perspective, and also the microwave part was considerably easier to follow with some existing background information. All in all this was a nice, well arranged course, but for some reason in these study trips the things happening outside the lecture hall tend to be even more interesting than the actual lessons...

Sodankylä field campaign 2007

There’s nothing very exciting going on at the moment here in Joensuu, just the daily business of working with future research articles. But as I should somehow contribute to this new-born blog, I thought I could write something about last summers NORSEN field measurement campaign in Sodankylä. Altogether I spent two four weeks periods up there in Lapland, starting in the middle of July and ending well in September. Many people would probably get really frustrated about spending such a long time 130 km away from the closest supermarket together with reindeers and mosquitoes, but I took it more as a vacation. I had long before looked for some nice hiking and fishing places to visit in the weekends, but unfortunately I had also set a goal of coming back with at least 70 sample plots data set, which turned out to require more work than expected...

The center of the campaign was Finnish meteorological institutes arctic research centre in Tähtelä, 7 km south of the center of Sodankylä, by the side of “river” Kitinen. Tähtelä is also an important remote sensing research center, so it was useful to make measurements in an area where a lot of data were already available. Typical days included searching for new, different plot locations and measurements of forest stand variables, canopy cover and LAI. Originally I had planned to measure three plots per day (if weather was fine), but soon it became clear that two was a more realistic aim, and finally I had to content myself with 68 plots. Huge thanks for Pekka, who came there for two weeks to help me with the measurements, otherwise the plot quantity would have been even smaller. Even though the days in the forest became often fairly long and the plentiful mosquito population was a real nuisance especially during the first weeks, I still liked to be there. It was easy remember the long winter full of office work, so it was better to enjoy the summer while it lasted!

To end with I’ll tell one special incident that happened somewhere out there, in an unusually distant place tens of kilometers away from the closest inhabited place. I was doing some standard canopy measurements when I heard something rustling somewhere behind me. I turned to look expecting to see a reindeer, but it was - a berry picker. Well, normally that would have been nothing special, but this one was from Thailand! That was a small surprise, as I knew that the Finnish berry refiners bring labour from Far East, but I from what I had heard they like to move around in big groups. However, this one was clearly on his own, and I had seen no signs of other people when coming there, so I was really amazed how he had got there. We had no common language so we just waved to each other, and continued with our own businesses, one picking lingonberries and the other measuring trees with some funny instruments, tape measures set all around. I think that he was even more amazed of me than I was of him! After a while I heard a van approaching, honking as it came closer, clearly giving signals to the lost one. And so the puzzle of a mysterious berry picker had its solution - the van was filled with others belonging to the same group, so the one I met only had his own ideas about the best berry sites.

PHYSENSE workshop in Tartu

To get the blog properly started I'll write about the latest news from Tartu. Last week we hosted the first seminar of the PHYSENSE (Physically-based remote sensing of forests) network here in Tartu. It's meant for the younger generation of Nordic and Baltic researchers working in our field. Everything went smoothly: the presentations were interesting and covered both measurements and modeling, and there was plenty of food to eat.

Soon it's time to start making plans for the next workshop of the network. I've had quite positive feedback from participants by email - may be there is a point in organizing the next event! (And in trying to gather a bigger crowd for audience.)

LAI Detectives blog started!

We decided to start a LAI Detectives blog, Lauri came up with this good idea just last week!