We're sorry but this page doesn't work properly without JavaScript enabled. Please enable it to continue.
Feedback

Public sharing of semi-automatically detected dead trees in remote sensing images

00:00

Formal Metadata

Title
Public sharing of semi-automatically detected dead trees in remote sensing images
Title of Series
Number of Parts
156
Author
Contributors
License
CC Attribution 3.0 Unported:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
Identifiers
Publisher
Release Date
Language

Content Metadata

Subject Area
Genre
Abstract
Forestry as well as the efforts to delay climate change, both need that tree stand in forest would be healthy and high growing potential. Even if tree damaging and killing pests are a natural part of forest ecosystem, the extensive pest outbreaks hamper the support of ecosystem services expected to be provided by forest. Therefore, instant and highly detailed awareness about the health status of trees in mature and old-age stands is vital to maintain ecosystem services, to apply timely salvage cuttings rescuing the timber of dead trees supporting local rural economy and to heal gaps in the damaged forest. We tested several automated and semi-automated image analysis methods to pin-point dead trees from high-resolution summer orthophotos combined with ALS (Aerial Laser Scanning) derived nDSM. Both are open data provided by Estonian Land Board. Starting with object-based machine learning we reached the situation where simple map algebra was even more efficient in dead trees detection and computational resources. The methodological testing revealed multiple sources of false-positive observations. We had to apply various cleaning algorithms to reduce the proportion of biased objects. The removal occurred to be the major task. Finally, the large-scale test object layer was produced for one quarter of the country (16600km2), and these results were shared with experts for review. When feedback was collected, additional algorithms and parameters were tested to improve the results. Only then the final version was published. The resulting object-layer is published in open-access GIS platform XGIS provided by Estonian Land Board, which has many stakeholder-oriented thematic maps (CountrysideGIS https://xgis.maaamet.ee/xgis2/page/app/maaeluGIS). The specific thematic map also provides many other open access data. For example, we will show, how the indicated dead tree locations can be assessed for the current state and assess the outbreak using the latest Sentinel-2 images and their derivates. All combined, the pile of open data will improve the forest management, maintenance of ecological quality as well as public awareness on the forest processes as a natural ecosystem in Estonia. The experience, surely, can be transferred to other countries.
Keywords
127
Streaming mediaComputer-generated imageryWordLocal ringNatural numberComputer programmingCartesian coordinate systemMultiplication signLocal area networkLecture/ConferenceComputer animation
Computer-generated imageryForestAerodynamicsMobile appCartesian coordinate systemComputer programmingCausalityNetwork topologyReal numberVideo gameForestEvent horizonComputer animation
Network topologyForestProduct (business)Cubic graphReal numberMetreGoodness of fitAreaForestComputer animation
ForestNetwork topologyNatural numberEvent horizonData managementConservation lawContrast (vision)Network topologyCompilation albumForestFrequencyNatural numberView (database)Computer animation
ForestExecution unitForestMultiplication signOffice suiteNatural numberNetwork topologyPoint (geometry)Computer animation
ForestDrum memoryFluxTwin primeSummierbarkeitNatural numberNetwork topologyPoint (geometry)Uniform resource locatorPosition operatorForestNumberAreaComputer animation
Source codeDigital signalData modelSurfacePixelGreen's functionNumbering schemeType theoryBitFreewareMultiplication signCombinational logicDigital photographyForestMappingCovering spaceGraph coloringNetwork topologyNeuroinformatikVirtual machineOffice suiteComputer animation
ForestMathematical analysisNeighbourhood (graph theory)Random numberData modelShift operatorBildsegmentierungArithmetic meanObject-oriented programmingThresholding (image processing)Level (video gaming)AlgebraCoprocessorMedical imagingComplex (psychology)Office suiteThresholding (image processing)Object-oriented programmingLevel (video gaming)Pattern recognitionMultiplication signDifferent (Kate Ryan album)Network topologyAuditory maskingRandomizationPhysical systemAreaError messageFood energyForestEndliche ModelltheorieElectric generatorNeuroinformatikObject-oriented programmingVirtual machineWhiteboardComputer animation
CoprocessorAlgebraMathematical analysisData managementMultiplicationForestConservation lawMaxima and minimaObject-oriented programmingNeighbourhood (graph theory)Mathematical analysisAreaSound effectCircleNetwork topologyForestTraffic reportingFitness functionConservation lawInformationBitData managementSystem administratorState of matterProcess (computing)Function (mathematics)Different (Kate Ryan album)Natural numberFilm editingComputer animation
WhiteboardWeb pageData managementNetwork topologyForestState observerPrice indexUniform resource locatorPosition operatorPlanningTotal S.A.Sound effectRevision controlElectronic mailing listMultiplication signDifferent (Kate Ryan album)BitTexture mappingOpen setAreaData managementWeb servicePhysical systemMappingSource codeCartesian coordinate systemDatabaseService (economics)Computer animation
Web pageData managementMaxima and minimaLink (knot theory)Sound effectGraph (mathematics)Network topologyCASE <Informatik>Medical imagingNetwork topologyUniform resource locatorLine (geometry)Computer animation
Link (knot theory)Web pageForestTexture mappingState of matterFilm editingNetwork topologyTerm (mathematics)Video gameComputer animation
Inclusion mapSatelliteFreewareGroup actionFreewareGraph coloringOpen setWeb applicationComputer configurationDigital photographyCombinational logicSelectivity (electronic)Data storage deviceSatellitePoint (geometry)Service (economics)Bookmark (World Wide Web)Price indexThermal expansionRevision controlView (database)Medical imagingFlow separationForestNetwork topologyComputer animation
Integrated development environmentAmsterdam Ordnance DatumWeb pageMessage passingNeuroinformatikProcess (computing)Greatest elementCombinational logicMedical imagingPosition operatorGraph coloringSoftware developerForestSatelliteType theoryDigital photographyPhysical systemNetwork topologyComputer animation
Mortality rateSystem administratorWeb serviceIntegrated development environmentSquare numberComputer animation
Mortality rateForestGateway (telecommunications)Field (computer science)ForestWebsitePrice indexWindows RegistrySystem administratorCartesian coordinate systemFlow separationPresentation of a groupComputer animation
ForestGateway (telecommunications)Mortality ratePhysical systemCASE <Informatik>ForestBitData managementWindows RegistryWebsiteComputer animation
ForestGateway (telecommunications)Computing platformMessage passingOpen setData typeWeb pageInformationHybrid computerMathematical singularityWindows RegistryMedical imagingAreaForestVideo gameNetwork topologyProjective planeGraph coloringEstimatorEvent horizonWebsiteView (database)Texture mappingSpring (hydrology)State of matterLine (geometry)InformationSatelliteDifferent (Kate Ryan album)Decision theoryMultiplication signPlanningActive contour modelResultantSelf-organizationCondition numberBitComputer configurationSocial classWhiteboardSquare numberLevel (video gaming)Food energyTesselationConservation lawPresentation of a groupSlide ruleStrategy gameNeuroinformatikData managementGreatest elementLocal ringFreewareSystem administratorPoint (geometry)Expert systemNumber1 (number)Graph (mathematics)Digital photographyDaylight saving timeNumbering schemeDependent and independent variablesReal numberComputer animationLecture/Conference
InformationOpen setForestWindows RegistryHybrid computerWeb pageRaster graphicsDivision (mathematics)Digital signalData modelSurfaceData typeService (economics)InformationDifferent (Kate Ryan album)DigitizingSurfaceDynamical systemNetwork topologyForestData managementProcess (computing)Multiplication signPoint (geometry)Computer animationLecture/Conference
InformationWeb pageData typeRaster graphicsDivision (mathematics)Data modelDigital signalSurfaceForestWindows RegistryComputer-assisted translationMultiplication signComputer animationLecture/Conference
Transcript: English(auto-generated)
I'm local, I'm truly a ground person, I do in nature, I use GIS as a third program next to the Word document and Excel, so I used to use a lot of GIS.
Now thanks to our LAN port we got such application that most of the time I'm using a mobile application and I almost no go to go to any of the GIS programs, sorry. So forest, this is a trees or land which supposed to have trees in Estonia but it
doesn't have, sometimes happens and shit happens more and more frequently. Mostly it's actually a human who is cutting but this is supposed to be a positive side of the forestry but unfortunately there are many causes why trees die. Is it big events, whether fire or mammals or fungi or pests, so it's a
huge package who want to kill your trees and in real life this is, it can be positive, it can be negative but think about this stand. This was a primary exemplary stand of the best productivity of Estonia, 400 plus cubic meters per
hectare. It was supposed to be that every can go and enjoy the real accumulation of carbon and saving the climate. By the way, Estonia saved the climate until these years because we increased our forestry, forest area from 25 to 55 percent so if you have something good in nature, climate,
this was done by Estonia alone. And last year you see no trees, no back anymore, carbon sequestration, everything is gone. So all the work, all the efforts of last eight years, done. And the natural disturbance is
increasing. You see it's from National Forest Inventory from 20 to 45 percent. Last period is not published, it's just in a week to publish so we will know how bad it is currently. But there is an interpretation aspect. What is good, what is bad, and this is depending on the stakeholders.
Natural competition within Estonians for a seed. So the natural view of the forester is that you grew 68 years of timber and then you actually, you
would be happy to have just a firewood. Of course we can't compare the prices of the same timber. So it's actually very bad and if you leave in a forest, yes, it is a slow burning but it's still burning. So it's just a matter of timing, it's everything will burn. Now at the same time, if you look on
historical aspect and actually dead trees can be very beautiful. And this very famous, it was almost important to have in every office in the Soviet Union as three bears. It was actually drawn in Estonia in the Northeast. So think about it, this is Estonia. Yeah, but three small bears. The big one is
not counted, it's too big and dangerous. Actually the original name was morning in the forest but nobody liked the name of it. So but from a nature lover, a tree hugger point of view, dead trees have a lot
of positive side. There is a hanging location for lichens, there is a food storage for many birds, there is a lot of fungi, there are mosses, there are insects. So it's a lot of positive things. So one forester told me that there is nothing more alive than the dead tree. It can be true. But if we're
going in the pests, then surprise surprise, look at number two percent. Two percent of Estonian areas are only affected by pests as insects and there was like 45 as natural disturbances. But currently 99% of Estonian foresters, let's say, cry
is about pest outbreaks. So what we did. I got into land board, I'm ecologist so I just got there. And talk, talk, talk and finally I got someone who had free time at summer called Nikita and he just grabbed a little bit of orthophotos. He used LIDAR land cover and the topographic maps. So
everything available just in a back computer of his office so it was easy to use. Land port is managing all the data and it's free. But why to drag all the data into Tartu and then send everything back to the Italians, so why? Let's save the world. And the beauty is that when you have usual
orthophoto maps, orthophotos, you also get trees as gray. But every forester knows in a forest, forest color combination, composition, the trees are blue. So it's very much more easy to recognize and this is easy to learn
from a machine. So Nikita did it. He gave semi, let's say, semi-automatic or this imaging recognition system in Arcus, so sorry, land board paid for it. But Nikita did everything for free, just a few cups of tea, so it's still almost free. He filtered out by a forest mask and the impetuous
areas and made this huge image sequestration, made a classification using a very complex random forest model and then the object generation and blah blah blah. So physically it looks more like this image, segments, recognition objects and finding blue areas as potentially dead,
then removing clear-cut areas and so it was. But surprise, surprise, it was low and the machine didn't want to learn. It always had errors, had a lot of problems. And then next I got into Italian in the office. I got a few
cups of coffee before that, so I'm full of energy. And then we sit in behind Nikita's computer and try it another way. Just did very basic, simple image thresholding of objects, looking on just image and just try putting threshold levels. And we found the blue, we defined the segments, we vectorized, we
put object centroids and then just defined. Done. And I promise it was three times more faster, it was more efficient and they provided more treats.
So simple and easy and worked. So the difference is, you can see large segments, small segments, more sensitive, possible to do all neighborhood analysis and everything else, so it was quick and dirty. And the cost was only, for me, one cup of coffee and for Nikita another tea,
that's all. And one and a half hours, all the job done. Again free, so it's open. Unfortunately, when we started to put all these to day callers, that somebody would like to be interested about it, then we got into troubles, that somebody said, oh you have too many, oh you have still missing one here, or
there is something strange going on here. So we had the difference and just to combine all these half year of talking of different state owner, state forest, having a half of the forest managing, they said, whatever you do, we don't care, we do much better, we have five person working on it for
years, so we will do better. The other large companies said, okay, why not, we try, but we are not, we can manage without it as well. Then they most interested were small owners, but unfortunately they are not very active, they don't care, even if they know their territory, they will not go and manage it frequently. And then we have subsidy management and assessment
administration, they said, yes, it's nice, it's very useful, but we used to have our own person outside and checking everything, so yes, it's nice. And then finally, forest monitoring and reporting, some of them need only five year step, but we have four year step, so it doesn't fit for
them, just administrative problems. And nature conservation, they want to have everything, but it's nice to say that, you see, dead trees are everywhere, so you don't have to complain about nature conservation areas where more trees are dying, so that's all. So the big emotional output was that everybody understood, dead trees are everywhere, so at
least we don't have to complain anymore that there are dead trees. But okay, we tried to respond with Nikita on these all those complaints. We tried to remove age effects, we tried to remove thinner forest, we tried to remove road effects, so we have
different, so we have like, currently tried to this clear cut the area example, you see there are so many, just by spatial and more tree height information, we removed a bit more, then unfortunately, you don't see, here is also neighborhood analysis, but okay, the circles were not visible anymore. So it's here,
it's a thinned areas, or this is all dead trees, no, this is a ground. Unfortunately, like in Finland, we have a lot of lichens, but lichens look like the dead tree. And if you have not industrial forest, you see a lot of lichens which look like that. So we tried to find a way to remove it, and it
was a lot of in a cost of positive observations. So this is wetland rewetting. So the ditches are closed, locations are here. And unfortunately, you see it's many, many trees trees are dead, and everybody is complaining, total waste of
money. Sadly, because of this removing ground and other effects, we had to remove quite many points on such very tense areas. Also, there is a left but the indication was still preserved. So for managers, it was still available, just we have to know that some trees are not
lost. So after that, we decided it goes open, you saw in the morning sessions, you saw the open data source database of Estonia, I promise it's huge, and I never find anything from that. But thanks to that, there is other services of different application of network, GIS,
thematic maps in Estonia. Sorry, there is no English version of this fullness. So I have to show that there is like a huge list of different topics which they have produced. And there is also like actually rural GIS or countryside GIS. And as I was a little bit like owner
of this thematic map, then I managed to put ask to add this dead trees into the system. And you see there is like as Estonia is flown by plane in by quarters, we have this to emphasize that the dead trees are known
in different times. But you can select here and I promise if you go there is English version available so you can try. The borderline was very important to know because you see the trees are here, but locations are here. But this is a case where you have border line of two flights. And this is a lighter is from 22, image is
from 23. Just one is from north another one from south. So this is exactly on this border. So it's there are missing cases. But in a big picture, you see Estonia is full of dead trees, full of life, I understand in ecological terms. And these yellows are officially
declared to be dead tree rich stands. And you see it's definitely an administrative play around. And you see frequently it's a little excuse for state or B companies to show that there is a dead tree collection. And now we can cut down and clear cut everything, not just the smallest piece. So they just
need excuse to cut. So and thanks to extra options, there is also now available satellite Sentinel to selection in countryside in expertise services. And the
same you can see it now that you have auto photo or satellite Sentinel to forest diversion, the same color combination. And you can see that here is expansion of the dead tree group. And what is nice in this case, you can make a bookmark of everyone that's easy to
put into somewhere in documents or and it will preserve for several years until landlord gets bored and they just delete all the bookmarks. So it's that there are three points are very useful indication that go and check it for future. Then we have another web application free for use or open
for use the satellite image lab. Look, okay, warehouse store storage. And there is very nice there is auto photos available, there is a running package of satellite images. And the same of
course, you have this for various types of combination of colors. And luckily, we could also now the trees into the system. And that way, it's also you can use it the same way in their mobile telephone. And now there is job positioning. So if
you want to not to use XGS, you can use also satellite hub and then just go into forest and find out everything. So it's everything is just take your mobile and go and do it. So it's you don't need a big computer. And this is what is very important for end users. So finally, now, now we
can emphasize that all this, what we did is more like, in the morning, who was in the talk, you got the message that we need bottom up and top down development. This was currently bottom to bottom one from one side to another side of bottom until sorry, very special page up. Okay,
until here. Now, to get it more official, there is another portal environmental portal, which is managed, used by clerks and administrators of environment. And surprise, surprise, to add the
same layer into the this service, which is just square mass, everybody can take it and use it. We needed the approval of the ministry to add it. We finally got it after several months of presentations and talks. And surprise, surprise, surprise, then it was just it person managing the
site who was pushing back and we took extra months to explain her that yes, we need it's true. It's a check. It's working. You can use it. Now she was pushing that we have already these administrative checked sites in the field. And this is good enough. It's very good indication. You don't need more. So it was a
huge fight to get it into the through the top back to down bottom. But okay, finally, we have it. Finally, we have one extra portal. We have many applications. And this is forest registry, the most important portal for
foresters. And can you imagine the IT system supporting everything and also a little bit manager of the registry to say that maybe we don't like we still don't like it. Yes, there are a few cases and maybe but we're still fighting for it to get it also public in this official really end official of site. So it's
really fun that you really have to fight to get into the public. And most important, every stakeholder has its own opinion about your open data. And to draw the line, find the IT
solutions which will save the planet and world and climate. So don't use these ones which work three times longer or 10 times longer. And AI might use even more energy. So we have to plant more trees. Bottom to bottom can be done quite easily. And you see it took only a
few hours, a few cups of coffee and tea and a little bit of free time and summertime when there was no harsh to do everything else. So it was that way it was low cost, no project, no support, nothing. Except that when you try to give it stakeholder, they find as it's free, is it easy? No, we don't need it. Easy come, easy
go. We want to pay and now currently, big companies and the state order project with half a million to get the same results. Just because they pay, they believe more. It looks real. Even if it's based on the same data, the same land board data material. So nothing
special new. Just you can pay and then your data get valued. So please, always ask money for your data or results. Otherwise, nobody will take you seriously. And then maybe you'll find the right stakeholders to appreciate your work. Thank you.
Thank you very much Jan. That was a really good start to the session and also keeping it on time. So well, but you did it this time. So if you have questions, please raise your hand and wait for the microphone. Hi, thank you for the
presentation. It was very nice. Do you have plans to update the results? I saw the Estonia map was split into four, probably due to time. And there are two aspects. Nikita is
working with one of the practical to get the same, let's say extraction from satellite image. But unfortunately, as I is managing with colors square by square and they mess up the color scheme and the quality level. So Estonia
is made into small tiles, which have different view and critical levels. So it's a pain in ass currently to just make it even to work for a small country like Estonia, thanks to ESA. And another problem is that thanks to Russians, there we have such a huge mess up of GPS signals
that this summer there is no flights. We have some flight in spring, but currently they said no, no photos, no rider this year. So maybe when Nikita will manage to find out the solution for satellite image, then we get update. Otherwise, sadly, we have to wait until
your, your flying is safe. Okay. Thank you. Actually, the flying is safe anyway. This lighter is just messed up. Any more questions? We have plenty of time. Thank you. It's very
interesting. I'm I have a question maybe about notification. I'm from country and we are involved in disaster management and we help humanitarian organization with first response. And sometimes when you estimate consequences
of a disaster, you have results like this event for, for, for, for the area of special event, you could collect a number of people that was affected by it. And sometimes if you share it
with unprepared people, they have a panic. How are you going to solve this problem? If we will talk about a big amount of dirt trees.
It's nice to have many trees. You can have a lot of fire food. Yeah. You, you know it because you work with it, but it's a really sadness actually. But if you, usually people will, will see on your map, they will, they can make a
very wrong conclusion about a bad situation. And do you have a special strategy? No strategies that people have information they can decide what they do. And as Estonia has four or five big owners and 10,000 small owners, then there are so many different decisions, options. And as
the life shows, they have all different behavior. And currently this was my point that at the administrative levels, land forest owner levels, conservation level, everybody has different emphasis on the same dead tree. And that way you can't have single right decision.
And every, every piece, every forest, everything should be decided at site and it's local conditions. Maybe as an option, you can add additional layer when you will show an estimation from the expert from forest management who knows how many trees should be
died in normal for this area. That's a real problem. Estonia has historically had the minimum level 20, 25% of forest land. So now we have 55. So you understand that the most of it is planned and not naturally managed. Most of it is not naturally managed anyway, it's
artificial. So you have owned artificial stand. If you want to, ah, I can show the real forest. We have very few quarters which have been poorly managed last hundred years. Like you can go, which is, oh, just, just the last minute. So
I had the slide about this, just, I had
when you are behind the computer and you don't took your glasses with you, you, you feel a bit different. So we have all the background information. You want to have digital surface layer. And so this is the quarter we consider as
old growth and almost like a semi virgin forest. And you look, it's a cheese. The real forest is actually coverage. The typical management dynamics of the forest is coverage. So dead trees should die and get disturbed all the time. But the point is, it's a continuous process. And here it's
almost started already. And currently you see it's, it's almost like cheese. There are, okay, actually there is more dead trees, but unfortunately most of them already fell down. So currently you can't move there anymore. It's worse than these three bears picture. I promise. Okay. We ran out of time. Thank you again,
Jan, for this motivating and encouraging talk. And I welcome you to approach Jan then later in the social bites, et cetera. Thank you all.