 Okay, so good afternoon everyone and also our friends online. We wait another three minutes, maybe people are still coming because of our previous session have been postponed a little bit. Okay, we will start at 3.35 on time, yeah. Thank you very much for coming here to our session, both in-city and also online. Yeah, dear friends, here it will be the session where we want to share some appearance and progress, especially for the first-of-the-version activity in China and also we have some international corporations between China and Europe and also China and Southeast Asian countries. So in this session, there will be five presentations. The first presentation will be given by Professor Li Zhengyuan. Professor Li comes from the Chinese Academy of Forestry. For the second one, we will be given by me myself. My name is Pan Yong and also for the third one, we will be given by Dr. Tianxin. And so all of our three, we are coming from the Chinese Academy of Forestry. So our first presentation will be given by Dr. Zhao Dan. He's come from the Chinese Academy of Sciences. And also for our last presentation, we will be given by Dr. Huang Shuo. He's come from the UK. Actually, we also have a very close cooperation during the past years. So through these early sessions of this GIFI meeting, we all know that for the first, brings more and more attention. And also people want to use some just special technologies, especially for the remote sensing data, to characterize the forest restoration, forest disturbance, and also degradation activities both for the amount change and also some quality changes like the quantitative parameters. So in this session, we want to share some of the appearance, especially for the activities during China, and also for the Chinese data sources. So let's welcome Professor Li Zhengyuan to give us first the presentation, especially using the Chinese GIFI satellite data for the first mapping activity in China. So I will... Thanks Professor Pan. Also, Li Zhengyuan, good afternoon. Here, and Pan said I would like to give you some ideas I mean about the forest and grassland mapping on national scale. Also for me, I used the Chinese satellite we call the GIF. GIF means the Chinese pin-in at Goufen. Goufen, I will mention later. That's several of the targets. First, as I mentioned, this is the Goufen satellite series. In the past ten years, Chinese government launched seven GIF satellites. GIF means Goufen. The meaning is high spatial, high temporal, and high spectral satellite. This is the meaning of the Goufen. Like here, so the Goufen 1 and the Goufen 6 molest same. I mean, this is very, very wide. It's 800 kilometers. Goufen 2 is a high spatial. Goufen 4 is a stationary satellite. So, I mean, the temporal is very high, even in seconds or minutes. So, also, we have another satellite we call the Goufen 3. Goufen 7 is your mapping satellite. Which is an amazing satellite. So, I'm from the Chinese Academy of Philosophy. This Academy belongs to the National Philosophy and Graduate Administration. So, for me, to work only focus on our administration. So, the duty of this administration is to manage forage, grassland, wetland, even target resources. Also, showcase of wide animal and plant resources. This is, I mean, the duty of this administration. So, in our administration, we use a lot of satellite data. So, satellite data is very important. So, go back to my talk. I mean, the objective of my work is forage to our classification system. Then, we use a Chinese satellite to mapping the forage and grassland on national scale. Also, we use a lot of reference data. Here, mentioned that for classification, certain construction. We used some national standard, some of the international standard. This is for the classification system. So, for classification saving, forage level, we have seven times. Second level, we have 38 clutches. So, here, I mean, like here, mentioned, forage level, second level, 10 clutches, including many failures. Only mixed swabs, bamboo, mangrove also, 10 clutches. For grassland, we have six clutches, like here, making like a physical grassland swamp, swabland. And natural grassland, this high-coverage, medium-coverage and low-coverage, three clutches. Right now, we have five clutches. I mean, here, it's amazing, like mangrove, forage and swamp, swabland. These four elements, we categorically write into forage land. Also, for grassland also, swamp and meadow, the counter-riding to grassland. Here, the direct sandy land, there are six clutches. So, areas, including cultivated land, construction land, and areas, like here, we are making this table for the second level. So, I mean, for this work, we use the Kauffman 1 and the Kauffman 6, and I say this is 16 meters, the lowest mountain aspect for satellite data. Like here, I mean, the swaths, this is very wide, at 800 kilometers. For year 2015, we used about 700 clutches. For year 2020, we used more or less about 638 clutches. In my work, I mean, we also use some supplementary data, like here, the National Provence Boundaries, DM, and also radios types. We use the data, like here, like radios types, I mean, to revise the primary types. The method, I mean, for classification, we used all data with automatic classification process. Also, I mean, manual cultivation to improve classification result, we also used it. The procedure is not to compare the data properization and segmentation. The important thing is the feature, optimized particular classification and polarization. This is the roadmap for the segmentation. I mean, here, the first is image segmentation, then the stratify classification. Finally, we get 38 clutches. Data precession is very simple, important thing for radio-metric collection and also rectification. Segmentation, we used multi-scale segmentation. Here, that is the formula for regional heterogeneity of the mod or delta. This is the formula. I mean, segmentation, like here, machine, multi-scale, I use multi-scale to, I mean, for the different segmentation scale, as well as for different land types. Also, we, plus, it had this, all that layer network, this is very, very important. So, I mean, when I need to print segmentation in scale, cannot effectively separate different land types that scale parameters to form a fine layer on top of previous one. In this way, the segmentation result at multi-scale scale can be obtained under an object-layer network. It is established. Feature selection, I mean, the algorithm, we used several, several ability and thresholds that to evaluate the degree of association between two categories on certain features. Also, the GM, the detent used for the measurement, the separability between categories. Now, here are several examples. This is one is the North Spectral or OLAP, another one is Parsley OLAP, another one is very heavy OLAP. So, part processing normally is very simple. I mean, the OLAP is operating blocks more than plots, small plots, cutting, I mean, trity areas, taking topology of plots and the land-line result. Here, I mean, for the metric information processing here, that we have several steps. For example, we get the result for year 2015. Then, I mean, for 2020. We put to segment the interest, I mean, for 2020. Then, combined with the result, I mean, map produced for 2015 to change the detection method to update to some 15th map into some 20th. This is my idea. So, here, I mean, for the most essential result, always we have to, I mean, to validate the accuracy. So, we also used a method called random stratify sampling method to evaluate our result. Here, I mean, sampling population is problems. So, use this method to get a lot of sample plots. First, we get the sample plots. Then, based on the random point, create a circle buffer, this area of 107 square meters. Then, I mean, check the certainty of the image properties of this circle, this circle, I mean, with the information trying to patrol problem place, then to calculate accuracy. So, this is in Monique for the interest of Gauphin satellite. This is a result for year 2015. Here, we have 38 classes for the whole time. I mentioned it for our classification system. This is for year 2020. Here, I mean, every lab type has one layer. Here, this is just sort of 40 lab types that are built in China. Then, the grassland types, wetland type, also the direct type. Also here, the distribution of cultivated land and construction land and others. As I mentioned, we used that method for the accuracy analysis. So, for year 2015 and year 2020, accuracy, I mean, is more or less about 85, even more than 85 percent. This is the area in China, different types, I mean, for the area of each type. So, this is my simple presentation that makes it not so complex. Okay, thank you for your attention. Thank you. Okay, so thanks to our professor Litz's presentation. Thank you, professor Litz's presentation. He especially introduced the Chinese satellite constellation data to do the first share in the grassland classification for the whole China. So, I also think for these communities, it's also a very good compensation because here people talk much about the data from the US, from Europe, and also from some other constellations during the past year. Some other countries also have good, just special data, fundamental facilities, like China, like Brazil, they also launch a lot of satellites. I think now we come to questions. So, firstly, we go to the questions from the audience in the room, then we are going to lie. So, one thing I would like to mention, for, I mean, Kaofeng-1 and Kaofeng-6 satellite data, we shared already for GEO. Yeah, okay, that's awesome. Professor Litz, you're front and then to the back. Professor Litz, thank you very much indeed. That was a really interesting presentation, and the Kaofeng satellites are very impressive, very impressive spatial resolution. I had a couple of questions, if you don't mind to use. The first fairly is, of course, China is a very large country, and your climate and your land cover classes are very diverse across the country. Maybe I didn't quite understand, but I was wondering whether, do you need to stratify your algorithms for classification according to the climatic types across the country, or are you able to apply algorithms which are valid across the whole of China? I mean, normally for the English, we use, in North Korea, we normally use summer, I mean, the data in summer. For South Korea, normally winter, because in winter, the clouds are not so much. Also, for each year, we get a lot of interest. We selected, normally, how to say, cloud-free, how to go back. Now here, for each year, we get a lot of interest. We selected more like cloud-free interest fields. In fact, you've already anticipated my second question, which we're asking about the winter and the summer imagery. You were speaking about the cloud cover in the south. Well, I came from the UK and in Britain, the cloud cover is a big problem. What is the temporal frequency of the Galvan satellites? Two satellites, I remember, about three days, because two satellites, I mean, the weather is very, very wind, it's 100 kilometers, so every three days, we can get one full coverage. Thank you very much. Thank you for your presentation. I apologize, I arrived a bit late, so maybe you have covered this, but I now work for GEO4Y, but before that, at the beginning of my career, I did a lot of mangrove research. So I saw one of your maps that you mentioned, wetlands, and I'm just curious to know when you did map a forestry, where does mangroves fit for you based on the forest definition? Because we know based on height and also percentage of cover, they could be included in being reported on their wetlands or being reported on the forestry. So where does it fit for you and what techniques did you use? Because honestly, it's not the easiest. Yes. Out of the blue carbon world is not the hardest ecosystem to map because it comes seagrass, but mangroves could be challenging by themselves. I'm just curious how you did it. Thank you. Yeah, yeah, yeah, really a question. I mean, just for the definition of forest in China, between China and the FEO a little bit different. In China, I mean, the coverage is 0.3 in the forest. FEO, I remember, 0.2. So others, as Molly said, same structure, I mean, as you mentioned for mangrove, yes, mangrove is really different. But here, for mangrove, we have one class for mangrove. This is the definition for mangrove forest. Thank you. Yeah, yeah, really, really. Some classes, I mean, accuracy, repeat, okay, someone is not so good. I mean, with the cloud cover that you're dealing, especially in the coastal areas, then the chance of producing high quality data is, is it challenging to try this in tropical islands and it's not straightforward task. Yeah, yeah. Okay, so, yeah, thank you. Any more questions? Just find the questions online. Okay, so we are on time. Okay, okay, thank you, Professor Lee again. Thank you. Yeah, so maybe for the general information, Professor Lee mentioned that currently, for the Chinese GF1 and GF6 data, has already been shared internationally under the GEO umbrella. So, and also I think for the GFOI audience, maybe you also can access such data. Yeah, that'd be nice for the, for the, for the transparent or more work. Okay, so next I will give my presentation to you. First I will share my presentation. Okay, so next I will share some of our work related to the forest restoration projects, ecological restoration projects. So, we all know that during the past years and for the trees, you know, keeping change, but how to use the remote sensing data to characterize this change. You know, both these disturbance is for the degradation and also for the forest ecosystem restoration. So, my presentation will be contents to introduce some general information about the ecological restoration programs in China, and then followed by two typical forest ecological restoration projects, how we use the remote sensing data to characterize its benefits for forest resources. And then followed by a brief summary. In fact, during the past, in China, we have, on a larger scale, we have contributed, we have carried out nine main ecological programs. Here I showed from the first one, it's a natural forest protection project. In the first picture shows that the areas of the ranch, you can see, especially in the northern, north-east part and also in the central part, which is mainly in the two main rivers regions of the Yangtze River and the Huanghe River regions. And also for the second one, is the three north, first the Sherwood Belt project, mainly in the dry area. And for the third one, is the area surrounding the Beijing area. In the last 1990s, is the sandstorm and ecological degradation get worse and worse in China. So the China government has started for the Beijing changing sandstorm source control project, which you can see in the third picture. That is for the surrounding the Beijing area, especially in the north part of Beijing, where for the sandstorm storm comes from. And for the fourth one, is the conversational corp lander to forest the program. This is for the main corp land areas, provinces that have this program. Then for the fifth one, is the Sherwood Belt construction project of the Yangtze River basin and also for the Zhujiang River basin. Then for the seventh one, is for the conversion of grazing land to grassland concession. And then followed by the ecological protection and construction in the headwaters of the three rivers, mainly in the Tibetan Plateau area, in the Qinghai area. And also for the last project, is the controlling of the cast the rocky desertifications, mainly in the southwest part of China. So here I mainly focused on the two projects, because of the time limit, I focused on the natural forest protection project, and also for the Beijing Tianjin Sandstorm Source Control project. So overall for all of these ecological projects, for the first restoration objectives, to try to prevent the degradation of the ecological environment or to protect the biodiversity and to restoration vegetation, et cetera. I think the most similar, as a UN study for the first restoration since the last year. Because we also, we can see that all of these projects, they have some overlaps, especially. Especially even for some counties, they even have a file for ecological restoration project. So it is difficult to use remote sensing data to characterize them separately. So I think this issue, we also discussed yesterday in the primary meeting conference. So for such ecological project or even our UN project, we also, there might be some overlaps too. And also for the main activities, we try to ban the logging activities, try to convert the cropland over the area to forestry and to build a new natural reserves area and then to plant some new trees, especially for the ecological consideration purpose. So next, I will try to take just two ecological projects to see how we use the remote sensing data to characterize these projects, effects our forestry resources. For the data we use, we also use the modest product, the Lancet data, the Sentinel-2 data and also for the Chinese golfing data. Here we mainly use the 1, 2, and 6. We also use some airborne data and airborne hyperspecial data for training and for validation for some specific areas. For the indicators for remote sensing data, we generate the forest coverage and the fractional vegetation coverage over forest areas, also the MPP, also the carbon secretrations. For the methods we use for the general for the chain detection and the comparisons between the areas, gene and the site of the project and then do the trend analysis because it's like we have a long time series, remote sensing data, we can do the trend analysis. Here it shows for the large area how we do this use the remote sensing data. We just create tiles for the whole China. We for each tile contains 2.1 degree by 2.1 degree, so we try to bring all the available remote sensing data together and then do the pixel-level-based image composition for the cloud-free composition. Especially we improve the dismantled, especially considering for the local forest phenology characteristics. That means we try to characterize the green-scented peak data as a priority to do this cloud-free composition. After that, we try to integrate to bring all the available land-covered classifications to try to improve them for the special and time-consistence, temporal-consistence for the ecological areas. Here we also try to think, it is difficult for a large-area ecological program. How can we do this in evaluations? We try to compare for the inside and outside. That means for the areas, for the inside of forest ecological program and outside, we try to make them comparable. Then both domestic and also abroad in a typical international basins, rural basins. Here we see the armor basin. We see it in Russia and in the northeast of China. This yellow basin is bordered to China and North Korea. We also have the micro-rural basin in the southwest of China and southeast of the country, several countries. There are also several other international river basins. We try to make them comparable, especially. For the temporary, we try to use even longer time-service data to compare before and after implementation of these ecological projects. This is how we do for the whole natural forest restoration program. Here you can see that mainly for the projects, mainly in the northeast of China, northwestern and also in the two central parts of China, the Yangtze River and the Huanghe River up basins. We generate this special and temporal consistent data. Then we can do this very temporal comparisons and analysis. Here we can show how we do the image, the free image composition. These compositions are covered from the project starting in 2010 and 2020. All of these data we try to consider for the local trees phenology to make them more comparable. Then we can see how the forest changes from the first coverage areas. For each area, we can see how many forests have been increased. This is for the amount point of view. Then further, we use the fraction of vegetarian cover to do the comparison. How many forests have been increasing qualitatively from the coverage fraction view. Then further on, we can do this trend analysis. Also for the different areas, we can see that for the different areas, as mentioned in the northeast of China and some other different of this region, forest ecosystem regions, we can see that for most of the regions, over half forest, they have the significant increasing for the fraction of cover. Also, as I mentioned, we also can do some comparisons before the project start. I'm here as a project start since 2000. We go back further more. We use the data from 1986 to 2018. We can do comparisons on how the forest loss area happens and also how the forest gain area happens. We can see pretty clear trend is that for the forest loss area gets decreasing roughly about 2000. Actually, it's 1998. It started from this, for the northeast of China, this project started from 1998. Then also for the forest gain activities, also it's actually try to keep pretty stable previous 1998, but after that, we can see it moves to a higher level. That means it also can give us some information. Also, we use a longer time series, remote sensing time series data can characterize this phenomenon. Further, we do the comparisons according to different ecological zoom, which has to compare for the areas inside of this project and outside. We also can see here, for the bar, it shows for the inside of the project, right by its outside. We also can see that for the green line, it has a significant increase. Here, we can see that also, generally, for the inside of the project area, we have much more for the increasing. This is similar. We both do the FVC and the MPP comparisons. For this one, we do the comparisons for the international river basin. Again, in the left side, it's in the river basin within China, and for the right side, it's outside of China. It is comparable, but for some areas, for China, it gets better. For some river basins, for the neighboring countries, it gets better, but we are still working on this, trying to figure out to get more information for the analysis. Also, for the desertification control, for the Beijing surrounding areas, similarly, we also do for the land cover and the MPP and the AI comparison stuff. Furthermore, we also have a lot of information for this like annually product and then we use them considering for the local climate variables like precipitation, temperature, etc. We try to separate for the natural driving forces with the ecosystem restoration project efforts. We are trying to separate them for the different driving forces. Here we can see that overall, this one, this yellow color actually shows for the many contributions from this restoration project. Overall, roughly it's about 40% for the ecological restoration project contributing to the increase. Here's a brief summary for the remote sensing data set which provides a good data source for the evaluation for the ecological restoration programs. In general, both for the amount and also for the quality are increasing. Since the intention of this natural forest protection project and also for the Beijing Tianjin central control project. That means we are beneficial from this forest restoration programs. But the more systematically evaluations are still needed to do the integration, ground measurements and also with more observations. We can use more quantitative indicators to do this evaluation. Thank you. Any questions? Please. Thank you, Professor Peng. I have one question regarding you mentioned that you exclude the natural influence from the ecological restoration. How do you do that? Yeah, in fact, for this part of work I've been done by one of our PhD students using a geographical model called the geodetector for using that model you can make all the potential drivers especially to put them out of the variables randomly to find which contribute much more. Yeah, and also for such a method they also open source tools. Yeah. So you quantify those variables and then you do some statistical model to the Yes, it's simply yes, but in the back inside the model they also consider both statistically also for the geovariability, they also call for the geolocation or areas surrounding that yes, they have a kind of complicated considerations for the models for the variables, even individual variables or the several variables they have like intersections to do it. Thank you. Thank you very much, Pan-Yong. I was wondering, so you presented a really nice comparison of using the most sensitive data of the impacts of the protection and also the version initiatives compared with similar sites which weren't part of the initiatives and what I was wondering is how are the areas identified that are part of these initiatives. Do you also use the most sensitive data to to classify to identify those areas that will be suitable to be protected, for example? You mean to how to find these comparable sites, right? To actually identify which areas will be part of the protection programs or the conversion programs. Okay, actually... Yeah, actually it's a perfect question. You see what we are doing since two years ago because during the early stages as I mentioned for such for the restoration program it starts in some in the 1970s but more started in about 2000. At that time, both for the GIS and also remote sensing data we do have some accurate data set available, also tools. So at that time it's pretty general considerations. Okay, these several provinces and also some of hundreds of counties will be included for the program for the ecological program but definitely at that time they don't have the specific area for the polygons. So, now we are trying to bring back to use the remote sensing now try to include the previously ecological efforts try to put them on a higher resolution remote sensing data than to reconsiderations for the 2020 in China for each new ecological program or even for the new efforts within the previously ecological project we need to use to consider them very specifically. Thank you. Thank you, Professor Pung. Firstly, thank you for the presentations. It's really quite impressive work. I'm very glad you came on this. This question may have partially been asked but just wondering how successful you have been in I guess communicating your findings and results to decision makers, policy makers, land managers how successful have you been in getting your data used to drive change in how land is managed in China. So the use and the uptake of your data for land management how successful have you been? So far for this part of the work mainly being used for the government purpose. They used this to evaluate how they are successful or where are the drawbacks in these previously ecological programs. As I mentioned before with Jackie's questions currently we are trying because for the remote sensing data it gets better and better so currently we can get that goes to very specific regions so now for the so currently for the first restoration project design and also for the carry cloud we are starting to talk and to serve for the local stakeholders because previously if you have data like a land search or even more this it's difficult for individuals thank you Thank you for your presentation I have a question regarding the data that you used do you think there is an option to use different sets of data in order to be able to extract information about the species distribution that later being linked to a better understanding of if not only we are mapping that there is more forest which is good but we are getting even deeper understanding are we increasing biodiversity when we are increasing the amount of forests and biodiversity of the broad aspect here of tree distribution as well as wildlife and all is there any way to use different data one to understand but for the protected areas maybe trying high perspective in order to be able to extract the different signals into what different species might be coming back and then later on identifying are these species leading to a higher biodiversity on the ground which being now this geofoia planary we would like to also not talk only about more forest but what do these forest support how are we actually measuring for increasing biodiversity yes thank you for the question this is also very interesting and also we are trying to discriminate some of them in fact we already did some of work but only for some very specific area because for large area it's difficult to get as I mentioned to get a good data set previously maybe since like since five years ago we started getting more and more hyperspecial data and more hyperspecial data we can do now but previously it's difficult so yes this is for something maybe in the future we get more and more chance to do so but this is definitely very important actually for some of the ecological restoration program in China we have some appearances not that good because during past we put a lot of efforts but to plant too many for more not to be seeds for larger area so also get some some disadvantages and I guess some lessons for the forest for the trees and the intersex so now yeah we also consider yeah but this is very very important indicator thank you very much browser I want to ask you if the mine grove is going to be include in the restoration project in China or in different programs yes in fact in China for the mine grove I think it's just get more and more attention during recent several years during recent several years we launched several programs along the coastal areas that put more attention on the mangrove forest but it's also very challenging to use more sensing because in China for the mangrove forest only it's like already approaching the north edge our mangrove forest not goes that far that tall is not over very large areas so yeah there are some programs but only very locally because we only have a small percentage for the just special distribution okay so thank you guys I have to move on but I will stay here we can come back later okay so our next presentation will be by Dr. Qian Xin also he's my colleagues also come from the China Academy of Forestry he has to be Professor Qian also he also is leading several projects about the international collaborations so I think it's also a good opportunity to share with our GFI audience yeah please okay thank you Dr. Pong for having everyone I will give a brief introduction about this project sorry okay recently benefits from the continuous funding from the Ministry of Science and Technology of China and the APF-Net and the Goufen program China has been making ever greater confusion to GO and the GFI so just as Professor Li mentioned recently during the last decades we have launched a lot of your dataset of your satellites including the Chinese Goufen satellites so this picture actually the Goufen images with 16 16 meter resolutions the global project so recently more advanced satellites will be or have launched for example the Territory Equalization Carbon Monitoring Satellite and the Lu Tan one it's the Araban Sun will be launched soon so I would like to talk something about the Senegal Cooperation in 2003 the most and the ISA decided to launch a large scale cooperation project and your observation because as a drugging program so now drugging since the kickoff in 2004 drugging now is 19 years old and now it's drugging 5 so Professor Li is Chinese coordinators of this program so in drugging 5 you can see in the right corner that's a totally 734 scientists from 213 institutions join to this program so from 2004 to 2019 actually most and the ISA sponsored the drugging annual symposium in 10s one year in China one year and the next in Europe and one week training course have been carried out since the kickoff and tell us about more than 200 young scientists have been trained two European partners one of the China's highest international cooperation awards so based on the drugging under the framework of the drugging based on good cooperation in forest monitoring and we decided to most international cooperation project funding actually we successfully got the funding so the second topic is about general introduction for this project so this project actually concerns the GUIs the requirements and state of the art and the problems of the drugging for example GFY requires or needs fully coverage of the remote sensing data to do the annual forest monitoring but currently only the land set and the Santero was the main resource so there could be some data gaps so the trend is how to use this data set with the Chinese GoFund data to generate the intensive data set to fill the data gaps so for forest type discrimination or disturbance detections the high resolution and the automatic method depending on the small samples for example the AI technology AI or deep landing method is performed so for forest parameters reversing retrievals under the virus scenarios so that how to generate accurate what to work forest mapping map is always interested so the cooperation project namely the name is the cooperation project between China and Europe in earth observation on forest monitoring technology and demonstration applications so as I mentioned it's supported by China most so it concerns key issues the current poor and low spatial and temporal resolution of regional forest we were sensing products actually it aims to support GFI capacity building and product service for continuous observation for the regional forest resources so it involves four key techniques the first one is harmonizing the Chinese GoFund and the ISA Centennial Cellular project including the GoFund 1, GoFund 6 with the Centennial 1 and the GoFund Centennial 2 and the GoFund 3 so the second one is forest type identification and change and disturbance detection technologies multi-modal characteristics so the third one is the execution of professional vertical structure information and forest banners incorporating multi-frequency system the last one is estimation of regional forest banners based on LiDAR and a particular multi-angular stereo observation such as the GoFund 7 satellite then we were built a developed platform based on this technology and to do forest monitoring based on multi-source observation data and we will use this platform to do the application demonstration over China, the two side in China one in north west and one in southwest and three from Europe for example, French Guiana, Kelton parking in UK and west getland in Sweden so expected achievements, our project will generate one set of the method and guidelines based on the four key techniques and we will build a monitoring system maybe in English and it uses a system to do the demonstration so that's our team, actually we have seven institutions and you can see Dr. Huang is here and we have four Chinese four institutions from China and three from Europe so I would like to talk something about the technical new issues, the first one is harmonizing GoFund and Sentinel-1 Sentinel-2 GoFund has a similar special resolution to Sentinel-2 but it has a larger image width than the Sentinel and even the land set so harmonizing the GoFund GoFund-1 and 6 with the Sentinel could help to increase the observation frequency so that's very different we found only a little difference of the simulated spectrum reflectance I was found between these two sensors so the second one we would talk about some job about the fusion of UAV and TRS point cloud data, you know in China the LiDAR has been more frequently used to do the fresh event amendments so we know that whatever UAV or TRS cannot get the complete forest structure parameters so we do this job to how to say rebuild the structure of the individual trees so that's very good for how to say to the calibrating for the validation for our satellite products in the labor amendments because with LiDAR we can get a small amount of original forest information we can call it as a ground truth so beyond this we also combined the forest structure features with UAV images to do the individual tree species discrimination so we use many method and very high result we got the classification accuracy is more than 98% so for the forest disturbance detection all the UAV partners develop some method for monitoring the UAV through spark beetles so early detection and large error mapping using the Sentinel-2 images was developed so we also did the long 10 forest disturbance in the over Tetside in the North East China in Genghe so we found that over the last 13 years forest have been transformed to the extent more than 12% but the disturbed area of generally showed a trend toward reduction especially after the commercial logging activities was a bound in 2014 so here we saw some relationship between forest disturbance and the main factors so for the forest hide we use we utilize the difference in forest penetration between long and short because we know we can use DSM substrate DTM could get the forest hide not but the critical issue is how to get the high accurate DSM and DTM the first one is how to get the high accurate DSM so we use developed and multi-level model to do the DSM compensation so you can see after the conversation the DSM accuracy improved a little bit and also you can compare with compared the first high-touching result so you can find that with a new development method P band DTM with X band DTM so the result also improved a lot so here is some illustration for what to say to estimate the kind of hide using our China space phone LiDAR under multi-angle stereo images so here we combine with the UAV under the album album images album result we can find very very good result and here is another case in the so we can also find we got a reasonable result so finally I would like to talk something about our other activities actually during past four years under the leading of Professor Li we carried out the several simultaneous space phone album ground based limited compance you can see the site distribution in the China maps and to validate the Gulfon products so here is our how to say our domestic sensor I mean the album sensor including multi-spectral LiDAR hyperspectral under CSAR here is some commercial UAV sensor ok in 2020 in the southwest China so we did a lot of field work including the DOM, DS, non-carver, non-use forest so moisture, water color and here is another test site in Genghe, in Mongolia in 2022 so here is our EU partners that filled our memories so finally I would like to see what we have we have a good guidance and performance of carbon picking and carbon utility goals so four functional and technical link connect to forest observations so understanding organization on China EGO because most recently applied more and more funding to support the China EGO so in this international cooperation mechanism and the platforms so we firmly believe that China can contribute more to the GFY so our joint effort thank you all ok thank you and I got Tian's presentation very good downtime so yeah questions so any questions ok I think that the prospects of the multi-angular satellite from Galfin 7 is a comparison with Ebonlida as you were saying but also Waveformlida from space like Jedi is really interesting is this something that you have done like you said you shared some examples from Sweden and from China do you what Galfin 7 the multi-angular so canopy height from that have you been able to do some comparisons for for the different countries this one I think so we use the LiDAR data and the stereo images to say you use LiDAR to detect the on the story parents ok use stereo images to get the DSM ok ok to get the DSM so that we get the canopy height by substract the DSM to the DTM so we use this data the data is from Galfin 7 so the prospects of being able to do this from space so from Galfin 7 is very interesting I was wondering if you are also able to compare with LiDAR data LiDAR no we didn't do this but recently we also launched another similar but more advanced the satellite we called as this one it have more angles and with say there is more power for LiDAR so maybe next in future we can incorporate this with several LiDAR satellite data but currently we only use the Galfin 7 satellite data thank you very much ok so let's move on our next presentation we are given by Dr. Jordan Dr. George is specialized in LiDAR application in forestry so he will talk maybe more about how to use LiDAR technology for the bi-mass automation so it's kind of quantitative indicators and also Dr. George coming from the Chinese Academy of Sciences ok please ok thank you Professor Pan and good afternoon everyone and I'm Jordan from aerospace information research institute Chinese Academy of Sciences my topic is about the recent improvements of remote sensing based forest or bi-mass estimation in China here is today's talk about the two parts the first one is China Bi-mass 2002-2020 and the second one is a new trend we call it a non-destructive observations first I want to give a very short background as publicly known existing carbon solicitation is very important and the forest is the largest terrestrial carbon in China as a result accurate estimation of bi-mass is a very important basis for scientific carbon sequestration and zinc enhancement actually I have carried out bi-mass estimation research works for nearly more than 10 years so in this years we still found there are many many limitations in the bi-mass estimation for example when we do the regional monitoring we found that it is still difficult to improve the multi-source data fusion and the modeling accuracy and also when we do the ground observations it is difficult to reduce the mirroring errors and also difficult to improve the adaptation of the geometric equations so based on these problems Chinese government and Chinese Academy of Sciences have started many carbon related projects these projects aims to develop a series of techniques including the methodology of carbon budget estimation and the remodelling models for monitoring the status of carbon sequestration of China and our team undertaken several projects of them and again many achievements mainly include land cover data based from 1990 to 2020 and the second one is about ground bi-mass estimation using high-special resolution data like the data and the signaling models and the third one is I just mentioned is the new method and equipment this is the framework of the Chinese forest bi-mass estimation from 2000 to 2020 first we estimate the ground bi-mass in the flight area based on field measurements and LiDAR data and based on that we estimate bi-mass in typical size based on the model combines Landsat and LiDAR based bi-mass and we import crown height estimation based on calibrated space on LiDAR and Modis BRDF and finally we build bi-mass estimation model project based on crown height modification indexes and the Landsat based bi-mass here is the materials China cover data is land cover data from 1990 to 2020 the resolution of the data from 1990 to 2015 are 30 meters and then 2015 and 2020 were 10 meters and we also to build the models in different districts we set we set six typical forest size in the whole China with LiDAR and very high resolution optical images data acquired in this start size and also we collected ground sample data in general forest start size to help to improve the to improve the estimation model in this district in the flight area we first instructed height and density parameters from LiDAR point of cloud and build model combines field about ground bi-mass stand height present held crown densities and then get the very high accuracy bi-mass in the flight area and based on that we combined the input from Landsat data and some other data like DM and first types we build the model combine the AGB result in the flight area with this inputs and to the canopy height we previously we use the canopy height calibration based on the LiDAR extracted tree height in the flight area and the models and then we use the calibrated tree height in the glass footprint to and the bonus BRDF to estimate the rasterized canopy height and finally in each district we first analyzing time series feature features of MOTIS and EVIs to reduce the influence of the vegetation index saturation and then build model based on time series analysis in different first type with some other parameters this is the final result of the forest about ground bi-mass from 2000 to 2015 in 2020 we collected more than 1,000 samples in around 40 sites to improve our models and also we improved GDI and ATLAS data to promote forest canopy structure monitoring accuracy until now we just have the validation in 2015 with more than 5,000 validation points and the R-square between the estimate data and the observation data is close to 0.6 the result shows that the total forest about ground bi-mass of China in 2020 is about 22.75 and after comparing with the result from other researchers we found that remodelsensing based estimations are generally larger than statistic based method because I think remodelsensing could reflect more originality and then I want to talk about the non-destructive observations actually this is a new national research and development project which lead by myself we proposed a density multiply volume framework include several non-destructive observation techniques and equipment for bi-mass between forest and grassland here we just to talk about forest in this in this framework in this framework we divided the bi-mass observation into two parts one is above ground bi-mass density and another one is the volume and we use multi-flocusing microwave features to detect the bi-mass density and then we use multi-angles scanning to detect the volume and then multiply the density with the volume we get the accurate individual tree bi-mass non-destructive to the bi-mass density bi-mass density we first analysing the relationship between the amplitude and the phase difference of multi-flocusing microwave attenuation and the density of different tree species and diameter classes and then we also by analysing the variation of the tree trunk density in vertical direction with height and diameter to build the vertical density distribution model that means when we want to marry the bi-mass density of a tree we just need to detect the density in just one position of the tree using our equipment and then we can with this model we can get the whole what you call density distribution of the tree for the volume we use UAV and some ground-based scanning 3D scanning method to do that first we need to to to fuse the air-ground multi-sensor data after that we can use this data to do some particle separation and then we can reconstruct a three-dimensional model of the tree and after that we can calculate the accurate volume from this model actually this this project has been started for half a year we have several primary improvement so first we developed a live wood density analyzer with the tree parameter observation software with this equipment and software we have carried out several observation experiment in the whole China for example in the Tibet we have used our equipment and software to measure the dbh height and biomass density of the spruce and the pine trees the result shows that the only our developed software can get the dbh and the height very well but the single frequency detection equipment still are not good enough to get the biomass density I think based on the high accurate dbh and height marine software and then we can combine it with the UAV and some ground based scanning method we can get very high accurate volume and if we can we can import multiple multi-frequency microwave I think the biomass density could improve too so here is a very simple outlook first no matter the regional monitoring or the ground observation we will import some very advanced algorithms like deep learning and artificial intelligence to the regional monitoring we want to improve our resolution because as previously said the land cover data in China has been promoted into 10 or 16 meters our biomass estimation also needed to improve so more regions and even globally biomass observation monitoring should be should be concerned I think to the biomass observation as I have just said the multi-frequency microwave and multi-sensor scanning should be considered perhaps in future some new framework of biomass estimation should be think about for example now the biomass estimation models in regional scale is difficult to expand to different timbre to different years so I think if we can combine some ecological growth model I think the multi-temporal biomass estimation should be improved okay fine thank you this is my presentation thank you thank you David Zhang we have time for a quick question maybe we just I just realized so okay so any questions okay Jack can you thank you very much it's really interesting you were presenting that you've used glass data in the past and then Jedi data have you yet incorporated the carbon monitoring satellite Chinese carbon monitoring satellite or are you planning to do so and could you say a little bit about the difference in the density of footprints maybe the size of the footprints that will maybe enhance your estimates because the characteristics of the permission actually glass data is very old data right and I think the Jedi data is just similar to glass data with similar footprint similar distance between the footprints yeah yes actually when we doing the biomass estimation in 2020 we found that atlas data actually better than Jedi data because Jedi data was usually influenced by the Terian too much and it is difficult to correct the Terian influence to the Jedi data and in China you know there are a lot of mountains in the south part so yeah and when when China has designed the recent carbon satellite they not only use the large footprint lighter sensors they also improve the they also improve the very high resolution images to rebuild the 3D model of the DSM in the forest and then use this to extract the tree height I think about the about some of the design I think we are we just use the Jedi atlas to compare the difference but we do not have any ideas about some design or yeah thank you thank you so I think we have to move on to our last questions last presentation after that we see if you have questions we still can't discuss because tonight they also have a city tour for some of us I don't want you to miss it I can ask you from there yeah you were mentioning Erwin Leider and I was wondering whether you know whether the Chinese government or any provincial government has got any plans to cover world to world the whole territory with Erwin Leider actually now there are some some provinces have carried out the whole province scale world to world Leider campaigns just in several provinces because they cost a lot a lot of money I think so yeah it's a big country sure I lost close yeah you choose there but I cannot see you here yeah I try to close yeah okay here we go okay thank you very much sorry for standing up in your way for refreshments so I'll try to go quickly over my presentation so I've been working in collaboration with Chinese Academy of Forestry and Professor Pang Jung for last eight years and also Jackie Rosette from University of Swansea because well we had common goals and we wanted to explore how the new technologies can offer more information or better information to support forest planning forest operations forest monitoring in both countries so I'm going to be talking about what is the situation in the UK I was going to do a quiz in here but I think we don't have the time but let me explain why I'm asking that according to NASA we've got 422 trees per person allocated this is our quota as a global population in the UK we've got only 45 so the British government wanted to solve that and is embarking on an ambition program of increasing quite substantially the forest cover in the UK in some of the some of the areas this will mean almost doubling the amount of forest that are in the in those regions so we're talking about Scotland we're talking about a third of the of the country but if we go to where England we're talking about 17% of the country which might not mean a lot for certain countries but for a country as densely populated as England especially England that might be a lot of conflicts of interest in terms of land use and the best way to combine forestry activities with all the food production industrial production and housing right these processes are also complicated by the fact that this is another matrix of complication that we are not only looking at increasing the forest surface in the UK but also we are trying to introduce new methods of management that are more in line with what will be or similarly as possible what will be a natural forest ok so we are progressively evolving from monoculture plantings to something like would be continuous cover forestry or natural forest etc the problem is all our models the models that we have been developing for the last 100 years have been designed for monocultures we have to start almost from scratch or find another methods to combine the knowledge encoded in those models and to adapt them to the new situation ok that requires sometimes to go to a very fine level of detail and this is what I would like to explain in here so a part of identifying how many trees you may have there are other big questions I have selected 5 mostly because I am biased towards forest production this is my background but all the ideas of interest like ecology or flooding or managing risk or whatever I mean there is a large list of stakeholders might have other questions everybody has got a W that is absolutely critical ok so I am going to focus on these 5 right first of all we need to know which species we have got and where do we have them in the forest inventory but there are a lot of uncertainties when you are talking when you are looking at private land about 45% of the forest areas in Britain are public the other 55% is private hands we don't always know what they are planting what species they have got so we need to find out by other means exploring ways of classifying these species trying to discriminate the species so far we are very good at the genus level but not at the species so we need to improve that so this is an application developed by some of my colleagues at forest research that is combining satellite imagery of different characteristics radar and optical data so we got about 90 55% accuracy at the genus level the species level is still a long way to go but it is very interesting and this is connecting with what I was saying about trying to get the larger amount of data the highest level of detail and this is coming from national LiDAR programs in England and Wales it is called as still debating where to have this world LiDAR coverage this is an example of what we are doing with drones as well how we are extracting point clouds using photogrammatic methods and how we are combining well let me explain that we are extracting either by LiDAR data or by using certain areas we use drone analysis we can get three dimensional views of the forest cover that for our foresters they can get from reality so just by visually assessing this data they can get a lot of information and do some planning okay if we apply different techniques we can discriminate trees at individual tree levels so we can assign them an ID and create a tree lease for all these these areas that would be probably too much a kind of a data intoxication for the average manager but if we are able to repeat those measurements in time especially when we are using drones you can start thinking about trajectories of change of those trees or at least parts of the forest so that will give us another way of trying to modify or adapt our existing models at the standard level and try to refine them to allow for the combination of species when you have different species not only a monoculture and also when you have different age types okay this data is combined with terrestrial laser scanning and that again is a huge opportunity because we can get fully three dimensional models of trees we can not only create a tree lease but also we can see we can provide estimates like stem profiles we can get canopy dimensions we can get tree height we can get deviates basically the most fundamental variables that have been used in our models so the combination of both techniques are quite powerful right right this is what I mean so we can see trees from above but we cannot see what is from below using airborne systems only in certain areas if we combine that with this case mobile laser scanning we can see the whole tree profile okay right also we can create a better estimates of volume because we can reconstruct the stem profiles and as my colleagues from China will say you can classify different branches and different parts of the canopy so you can start thinking about how biomass is distributed in each tree across different species and also you can start refining these models again okay upscaling those results is a different discussion I'm not going to get into that yet right then the next question is how can we monitor I mean how can we monitor changes especially when there are large disturbances in the forest so in the past we explored the adaptation of the time sink tool that is available in Google Earth Engine now to look at the trajectories of change and monitor those dramatic changes that will be associated with clear felling and also with wind damage this is another way of looking at changes, disturbances is the classification of wind damage in the most recent storm that affected the east part of the country east south east of Scotland north east of England we were using radar imagery to do that we assigned some polygons that had the highest probability of wind wind damage we assigned that with our LIDAR estimates obtained from the national LIDAR program and then we could provide a report to government about volume loss in the different species in different areas right then we can start looking at forest dynamics and that is very interesting again this is a new possibility using time series analysis satellite data when you have satellite data from a time perspective you can start thinking about monitoring phenology, phenological cycles you have a lot of signal in here and then you can detrain those data sets what will be a seasonal trend what will be a random effect mostly attached to problems with the instruments very interestingly you can see positive, negative or negative trends in the canopy once you have phenological cycles you can extract different metrics like the length of the growing season and the top of the growing season and that is very interesting this has been used by one of my students in Ethiopia to look at how the different metrics compared to different land use classes they were changing in time that was a fantastic application in here we were using a totally different approach, we were using time series of data and we were using principal components trying to model what would be the phenological cycles in each pixel and then looking at new acquisitions to look to model what would be or try to detect what would be an anomaly based on the distance the difference between what the model will say and what the new acquisition is saying well if you look at the intensity of this distance from the outliers you can start thinking about dramatic changes like completely structural destruction of the forest has been clear fell or has been wind damage or has been burned basically has disappeared or very important monitor the onset of different pest and diseases and then this is when people when forest is on the ground can start doing something about that this is where they can implement some techniques that can be successfully can stop this problem that is an opportunity that is very interesting because satellites, there are millions of satellites now there is a lot of images and we can have daily images if we are thinking about planet for example we have got two images anywhere in the country every day so we can detect as soon as possible any problem right this is another word that we are doing with a student with Jackie myself and this is part of this Dragon 5 project we are looking at how different clones will be reacting or different provenances will be reacting to drought stress you will see the different behavior of some of them and that is giving us this is under control experiment in a polytunnel but the idea is to upscale these results and start doing a large phenotyping program that will allow us to not only select the best species for each location but also to tender them properly to see what are the critical values we should be taking care of and also managing the forest in a much better way because we must consider that at least in Britain any new forestation program is not a lifetime but at least a one generation investment you better do it properly and you better take care of it because having a second opportunity is a big mistake so you have to do it right from the very beginning all the countries might be it will take three generations if you go to Scandinavian countries every decision is it takes three generations to produce something you must be even more careful there right, monitoring forest dynamics very quickly when we are looking at trees we are looking at living entities so they are dynamic entities they change over time they grow, they die they grow in different ways they respond to climatic stressors they respond to to abiotic or biotic stressors is something going on all the time these are two applications we did one is to monitor directly this is a paper did 80's ago with a guy in the University of Ohio and we are looking at canopy delineation of light data and using different light acquisitions see the trajectories of chains or growth or how they were dying of different dominant types then this is another application looking at how can we use this canopy delineation as a baseline data for individual tree growth models and this is very interesting because we can see this is CITCAS blues and combined with a Canadian model called TAS we can see the trajectories of chains of every individual in the canopy in two scenarios one is business as normal and two there was a product of wind damage that affected partially some of those plots and the trajectories are completely different in terms of carbon allocation in different parts of the canopy and in total okay so getting accurate data to run to be used as a baseline data to run models at a high level of resolution and monitoring changes all the time to change this baseline observations can give you a very good estimate of what you should be getting in the future and you can take decisions about that this is again another idea you are getting models light up based models based on top height okay you calculate basal area and volume what will be the likely situation in a few years time this case 10 years time and that will be review every year over few months by new acquisitions of data that will monitor the canopy cover in there right finally that is fine but also it will be interesting to have to develop our own capabilities as a forest service to fly a wheel to capture data a wheel with the drones we can get an area legally speaking in the UK 500 meters around mass but it is a possibility to get protocols that are approved will allow us to cover large areas with our drones okay so we are collaborating with a company in Scotland that is designing this fixed wing planes and for example just to illustrate the possibilities that this urban systems can do nobody is using their space in Britain between 2 o'clock and maybe 5 o'clock in the morning at 100, 120 meters height you can fly entire districts with LiDAR and you can do that again and again a very very reduced cost okay so you can capture huge amounts of data every year you know probably you can monitor growing seasons with that and this is an enormous possibility okay because all the analysis can be automated you can process vast volumes of data and see how are the forest dynamics in the forest district also you can combine again when the people at CAF are using the LiDAR system we've got another system there's a head wall system that combines LiDAR with a hyperspectral so the combination of structural and optical data can create a very powerful combination okay and I'm going to leave it in there sorry for keeping you here okay okay thank you very much so any questions so Juan? very specific and a very nice work okay please thank you for your presentation very interesting I'm just curious here and maybe this question has been answered before but I looked at your slide when we were using indices that were created many many years ago like NDVI and whatnot and now we have saturated with data data with much higher spatial resolution temporal resolution and whatnot are there any new indices out there because we can't be moving forward with so much data but not really utilizing what we have gotten from the data point of view to start having more precise indices that we can actually differentiate from species and blind the LiDAR data on the ground UVA and whatnot but are there new indices that we are producing or is it a little bit of a stagnation on that side? sorry I don't understand the last part was that I'm wondering if scientists like yourself and the scientific community is utilizing the new data to have indices that actually better depict what's on the ground instead of using NDVI that was created many many years ago now we have much better data much higher resolution spatial that potentially with new indices we can have better results is this happening or is this something that is an area that we need to think about there are a few possibilities here if you look at the volume of hyperspectral data cubes are huge are you going to use all this information? no you don't use that but there are different parts of the spectrum that are quite interesting and they are relevant for what you are monitoring you should be able to detect that and construct a fine-tuned miniaturized sensors that would only capture this information this is what we are trying to do so we are doing a lot of polytarnal experiments we are trying to look at the relevant part of the spectrum and see whether we can just capture 5 or 6 bands in different combinations that we are going to use I don't want to have data that terabytes and terabytes and not be able to use not even 1% of what is in there this is the thing coming back to these are indexes that have been in the literature for a long time people know that but there are other indexes you can create by looking at any particular process so we are doing metabolomic analysis for example and trying to establish a link between what would be a metabolite that is being generated in that particular and a particular stressor especially pathogens and see whether we can detect that and then we will create sensors maybe ground-based sensors that will be able to detect that so this is uncharted land in many ways but I think this is the way forward because otherwise we are capturing a lot of data that perhaps somebody will use in the future but that will be a lot of information we will get intoxicated but I don't even try to get this kind of data to our foresters in the districts because they don't know what to do with that and some time ago I remember using hyperspectral data to map seagrasses and that's a challenge on its own so we had to exactly do this find the exact wavelength where it depicts seagrasses with a water column on top of it there is a publication but that just sits in somebody's shell if it's not out there to say this is probably what you should be using part of the spectrum instead of digging through the entire hyperspectral length and then we probably should have gone a step further and said okay maybe think about this is what we can create but the research stopped there there was no more money to implement more so there you go you stop at that level and didn't really progress to anything else although we all know especially people in the wetlands community mapping seagrasses is a challenge and we are still not doing all that great so that's why I was wondering are we moving forward, what's needed I mean this is a map why countries are struggling this way also what could be done, should we think we're going to move forward with this I remember a very interesting conversation about four or five years ago at the University of Glasgow with some people from the optical engineering department they were creating these filters for creating the signal and they said if you tell me you only need a part of the spectrum we will do that for you that costs a lot, no nothing very little it's possible to do that we're sometimes not connecting the right yeah, yeah exactly okay because that's the question now you have the question I just want to know do you have some current research combined uiui lyda and hyperspectral data combined together because that's the last we are starting now we got this sensor relatively recently but also we had the pandemic in the middle and things like that, now we are going to the field and I've got a student PhD student that is working on that she's going through the nuts and bolts of trying to link both data sets and Jackie has got the similar system in Swansea so we've got two of them and she's trying to do other experiments so this year we're going to do some field data collection and we are going to work on the combination that doesn't mean that every point in the point cloud will have all these bands of data because the sensors are radically different when it's a push room or the other one it's a lyda but we will try to see what we can do in terms of combination okay at the moment we are exploring okay thank you again okay so now we are almost approaching for the close session so thank you very much for coming so through all these five presentation I think for the data, for the method for the large scale or very specific scale I think we bring some of concept ideas and how we can use the remote sensing data how we can use remote sensing technology to characterize our forests you know both across time scales and also across the different special scales I think this brings some of our thinking and for this session if it's more contributions from China work and also some international collaborations but in the future we want to I think there will be much more opportunity to get more involved in this communication different in this community and also so you guys are welcome to China and also we want to get more involved in the GFI community from both of these data method even code so you can try to to make more contribution and also to make this forest research monitoring more transparent okay so thank you thank you all the audience and also our lectures okay thanks