FOSS4G 2021 - Open remote sensing data to analyze the effectiveness of Payments for Ecosystem.......

so we'll take we're taking just a minutes pause i feel like i should have um elevator music some view zach to to fill these empty spaces just so that we can get ourselves to the to the top of the hour to start in the next session for those of you that are looking forward to um juan miguel's talk we've not actually heard from him whether he's having difficulty joining us backstage or not we've not had him join us backstage so if somebody happens to know juan miguel gonzalez aranda and let him know that we're looking forward to him still joining us if he's able um and there's various um organizers ready to help with any tech issues that might be popping up so i think i think i'll go ahead and introduce you if i may flavia i see that you've got um a screen ready to go i don't know if you wanted to get powerpoint start we go we've got powerpoint fired up fantastic um flavia is uh a phd in a physical geography specializing in remote sensing from the university of uh gerting and in germany she has more than 14 years of remote sensing experience in renowned research institutes on optical and thyroid sensors and her expertise lies in forest agriculture and land use change in tropical forests she's currently gis and remote sensing analyst at the remote sensing solutions in munich and she's going to talk to us today about um open remote remote sensing data to analyze the effectiveness of payments for ecosystem services in the sofala community carbon project i'm really looking forward to hearing about this movie mozambique's case i hand over to you to tell us all about it thank you very much for your very nice introduction first of all i would like to thank you everyone and also thank you the force 4g and the support of the group of earth observation the geo to try to make this very nice event more gender diversity and also topic diversity because today in our session i was learning a lot and making a lot of notes because after of course after the session i'm going to ask a lot of questions to my colleagues to my speaker colleagues so um saying that i would like to introduce then the topic uh that i would like to to to introduce a little bit today it's a part of a part of the result of this the project that we have it's a multi-disciplinary project and the title of my presentation today as marketer already said it's open remote sensing data to analyze the effectiveness of payment for ecosystem service in the sofa community carbon project in mozambique but first let's talk a little bit about the project so it's a project collaboration project between the university of mapbook and also the agriculture institution in mozambique and the name it's impacted so the impact of terminated payment for ecosystem service on carbon stock deforestation collective actions and interesting motivations for conservation so as you can see in the title it's a little bit multidisciplinary so the the university of mabuk are responsible for of course uh leading the the whole project and also focus on the social economic analysis of the payment ecosystem service and then the group in mozambique are responsible for all the the fieldwork campaigns and also the the analysis that together with us and which who are responsible for the remote sensing part of this project in mozambique the analysis of this project in mozambique but what exactly payment for ecosystem service so payment for ecosystem service um i could call an instrument or maybe a market-based mechanism which try to help changing the practice of people who are doing damaged activities such as cutting down the trees for charcoal production so as you can see here a group of people who do a deforestation for cardone production so they sell of course everything that they defer state so instead of that this local community could receive incentives to protect to protect the forest in their area and also plant their go for a system so that's the idea the general let's say one one of the parts of the idea for the ecosystem service to to protect of course the the environmental area around this region but also gives some resource for this community which mostly i mean when you when you think about underdeveloping countries like brazil or like mozambique mostly are financially more complicated let's say than compare it to the region countries and i believe that the payment for ecosystem service could potentially also play an important role in the adaptation to climate change reducing reducing the vulnerability of the ecosystem so today we see a lot of news and we see also a lot of initiatives regarding the the the market of the carbon credit or some for example companies and also institution is studying and analyzing the supply chain of um of a company or for of a farmer and everything and so this is a as i think it's a very relevant topic for the actual days we just had a uh an event i think two weeks ago in new york to talk more about this paris agreement and so on so i believe that this is uh it plays an important role inside of all these contests that we are talking about climate change um and i would like to show you um a little bit more about the payment ecosystem service in the area where we are going to see so in general in this area there are three kind of sources so the first is the individual contracts which is um they do individual contracts they call agroforest contracts to with the farmers so they they pay the farmers to plant the agroforest systems and to keep the the these trees there and in preserving the trees doing the management and keeping them up to stock carbon other kind of source that this project also generated is related to jobs related jobs which are indirectly created by the project because you need to manage you need to do administrative administrative work and you need to to to any way to to to do all the infrastructure of the project and also you have the red plus component uh in this case for example they were they were paying a money to not to the person or to the farmer straight but they were giving the money to um to like a budget of the community where they could use this money for example to build school to get a better health infrastructure and to change infrastructure in general for the people there so for the red plus they for example were doing initiatives in early burning so early burning usually practice they they do that to reduce the number and the heat intensity for example of wind or late season fires also petroleum for example so they were like uh going around the region to see if the forest that were there were city stand if there was anyone trying to cut down the trees so they were also doing let's say the safety of the forest and also fire breaks i will give you a very short overview on the project timeline so the project started in 2003 the pilot project and then some farmers were already doing contracts and then in 2006 they started the red component inside of this project and then in 2008 we have the end of the phylos at the the pilot phase uh that they started in 2003 then in 2009 was the last year actually 2009 2010 that uh the new contracts uh were signed and from 2011 we have no new contract sign and also after this year we started to not have so much information like when we were having with animal reports and and then in 2014 they decided to close the business unfortunately and then we decided to try to analyze what is happening between all these years before and also after the project therefore the objective of this big project impacted is to propose a project that will evaluate several long-term aspects so before during and after the project in the payment ecosystem service in mozambique in the sofala region so we have then two let's say main objectives uh one is the environmental effectiveness uh respect to deforestation for carbon sequestration land use change or the land use and the second one uh it's more a social economical analysis and that's what i really like about this project because of course we cannot explain everything with remote sensing and we are always missing not only few data but also economical and social aspects they make a huge difference and and make us to better understand all the of the context and the impact of our remote sensing analysis analysis but today i'm going to focus of course in the environmental perspective of this work so let's go to mozambique now not on the beach that i heard they're very beautiful but in the countryside of the country so and our study area it's more or less in the center of mozambique in the sofala region and here i'm gonna show you the area where we were really analyzing so you can see a sentinel to image from the from may 2019 so this whole area that you are seeing with the sentinel 2 image is the study area and the polygonal with the red lines is the sofala community carbon project where the the agroforest systems and where the project was implemented so we chose a big area of course to try to understand what is going on inside of the area of the carbon project and now so outside and see if we can see any impact so the remote sensing part has three main tasks so first of all is the land use and the land use change mapping in the year of 1996 2002 2019 and why we chose this this year is not only of course availability of image but principally because one was before one was when the project started and when one was when the project ended and in the second task which we are right now we are marrying mapping the agroforest systems in this region and the third task is to do the a carbon storage assimilation and try to estimate the carbon storage in once we identify all these agroforest systems that we are trying to map in the second task so now i'm going to talk about this first task so we were mapping diverse many different land use so as you can see normal forest agriculture also some distance because sometimes you have to understand for example that if you have a national park around the region the probability of having less deforestation in this case it's of course a little bit higher because the park is usually protected so you can see that at least in the border of the the the park you have you should have less deforestation and also distance to the closest world we know that uh i think amazon in brazil is the biggest example that we can really see these uh they call uh i think fine fish when you can really see a road and then you see the deforestation goings in the vertical way and then it and it's very clear that the road plays also a big a big role in this and also the urban areas and the second task as i said it's the agroforest system but today we're going to focus more on the land use change and the and the land use and land use change and mostly wood forest woodland forest and deforestation so for that we needed to have open source free data which thanks to landsat 5 landsat 7 landsat 8 that we were able to use to do all the time series and all the long-term impact analysis and that's here we have to highlight the importance of the open data and not only in a context for example of an institution uh governmental institution but also in the private sector we see that uh ever this this is a big advantage let's say of using uh this open data source to do this long term effect analysis um so we were um using we had three um image satellite image from mars so we tried to keep the same month as this region has a strong seasonality a savannah it's a it's a very rich biome with a strong seasonality with fires with with with periods of drought and and so on and we were doing the maps with a 30 meter spatial resolution so taking this data we did an object object baseland approach classification so and and usually they are hierarchical classification rule set and the classification for example was was able to differentiate nine classes and it's also based on decision rules so i think what is very interesting about object-based i know that uh there are some really new techniques with ai we are also using here and pixel based i think we still have some advantage and i think we can always use also the object based or even merge both techniques and but what i really like about this is in the segmentation that you can for example in this program without programming without uh with a deep knowledge of writing codes you can really give a power to different layers so for example if i have the band one and if i have the band three or four or the any dvi layer let's say that they play the band one plays not such an important role in the forest classification compared to the nadi for example so i could give let's say a bigger power to the ndvi layer so the segmentation will try to delineate all the object take into account more layers that are more important have a higher contribution to the forest cover for example and there we go for the first results for the first task so here we have a land use a map from the 1996 you can see that the object-based classification was able to identify nine classes and here we see that also we have a good forest and woodland cover already a lot of croplands in the south but a good woodland cover so when we go to the year of 2002 we already have a big lost in the vegetation and also an increase in the cropland the grassland complex but when you go we go to the 2019 we even see a big lost a bigger lost so in the in the woodland and and forest area so what i have done was a just quick um analysis on this land use change that was happened so you can see here that 43 percent that we here we have a comparison between the year of 1996 and 2009 so the whole time series that your analysis we were analyzing and here you can see that 42 of the woodland of the the woodland forest areas in this in this year it still remains forest in 2019 but 46 of the forest area in 1996 turned it into cropland and nine into bushland so if we analyze then before the project was implemented to after when the project was implemented maybe we could see some difference right in the how the land use change maybe the land use change could have had an impacted uh with the project in this area so here we did an analysis so we could see that 52 of the forest in 1996 it still was a forest in 2002 but 38 of the forest in 1996 turning into cropland and nine into uh shrub and bushland so if we go to the comparison after the project the payment ecosystem service will install it we see that 79 percent of the forest in 2002 remained at forest in 2018 and 20 percent turn it into cropland and grasslands and one percent shrublands so here we already see some difference we can maybe started to draw some conclusion or started to feed our analysis but now to give a short conclusion the first task we did also a deforestation analysis here and we were analyzing and analyzing the deforestation rate per year and also in two different periods as you can see from 1996 to 2002 the blue in the legend and from 2002 to 2019 so the first one the blue is before the project and the orange one it's after the project and here we can see this graphic which shows the hectaries and also the different areas so as i said we were not only doing in the safala project area which in the image in the in the right side you can see by the red shape file but also the study area around so we could see that in both case we have a decrease of the the the deforestation let's say so 63 of the total deforestation that happens between 1996 and 2063 of the total deforestation happened between the 1996 and 2002 and 37 between 2002 and 2018 in the study area so in the biggest area and 52 of the total deforestation happened that happened between 1996 and 2002 in the study area as you can see the area with the with the red shape file and 48 between 2002 and 2019. even with the non-signing of the new contracts from 2011 and the finalization the end of the the project in 2014 it was still possible to see that there is let's say a downward trend in deforestation in the project region and also in the surrounding and now let's go to the last task the task two so here uh i'm going to talk what we are doing regarding agroforest systems so there is a lot of challenges regarding agroforest system mapping because as you know we have the the usually the agroforest systems in some regions they are very located they are spurs they are not so like it they are not monoculture so it's a little bit more difficult to to to identify and only with very high spatial resolution image for example from from maybe planet now that it's free we can probably try to use but we still need a long term of image but also drones for example so it's it's a big challenge in the community to try to identify the agroforest systems and in our community there in in mozambique and so far we have two most important um let's say um i go for a system which is the boundary system as you can see so they plant like let's say you can see like in the squares like a square of trees like making a wall so the trees are next to each other but they also have the dispersed inter planting in the machamba and the machamba is the is the property here of the person who is living there so with this challenge i mean what could be the next approach and then we thought that maybe we could work with trend analysis so as we have a such a huge time series for landsat we could do probably also a change uh trend analysis so try to identify if there was a difference in the in the nadvi in the pixel where the plant was where the tree was planted so in this case you can see it's it's very hard we have a big challenge because here we only have two points within this pixel but in some cases we could have a good result because then we have more points of agroforest systems so more trees that are planted so let's say that in 2002 the we have only the soil and then they plant the tree and probably in 2000 after 10 years depending on the species or eight years we can start to see an increase in the nadvi value so here for example it is is a is a script that is just sprinting very very nice person just printed he he was he does an amazing job with the google engine was a was a making friend available for everyone so to see kind of a trend in the nadi so here we did a test with one of the pixels you can see in the graphics showing the any dvi values in the epsilon axis and in the in the x-axis you can see the ears so we could see let's say a trend in the in the increase of the nadvi depending from where we get but here we really have to take care because um we have a strong seasonality in savannah and and we really have to take care of where we get from where we get the data we know that the ndvi is very high sensitive for for uh when after we if we have a big burning a big fire usually the regrowth can make also the nadvi increase so what is increasing so if we have an increase of the new dvi what is the reason it's because it do there was a fire and now it's a regrow or it's because there was another tree was only a soil and now we have a tree planted in niagara fall system and that's why we have an increase of the ndvi and that's why we also decided to use the savi and also the nbr and the results of this uh mapping that we are still running are still not valuable because you are still doing some tests so if you want to follow up and to see the second task of this very interesting project please follow us on twitter and if you have any question i will be very very happy to answer you and thank you very much thank you so much slavia that was so much uh wonderful information that you shared with us those visualizations are stark for both good news and bad please we look to them for hope that we see that our actions have a have an impact do we have any um i'm just gonna check and see if we've got any questions in that in the channel for you i think a lot of people have been uh spellbound by everything you're sharing they haven't had a chance to type something in and i'm seeing lots of uh claps and thumbs up flying up through the through the public venulis channel for you as well that's very nice i really appreciate i think it says yeah it's we are talking so much about climate change lately right and and that's such a great initiative that we can really try to to try to to to protect the environment around of these areas but also help the people who are living there right so economically try to to develop the region and doing environmental initiatives right so that's that's fantastic i'm not seeing any questions pop up so maybe i can ask one for you myself and that is what do you want to do next yes so next uh and i think it was very complicated when we started this ago forest mapping as i said because we have a problem with this spatial resolution right they're very small and with the landsat we cannot really identify and delineate all this agroforest system so we have been trying to use harmonics i have been trying to use for example the back scattering in the texture from the sentinel one which can give very good information to try to separate them but none of this thing was very um let's say enough to to really separate what is an agroforest system and what is a forest so that's also the problem right how you can differentiate an agroforest system to a forest so that's why i think the trend analysis is helping a lot so we can try to extract the phenology effect which the forest has of course in the savannah and trying to really um select and really see the pixels where we have an increase of the nadvi not because of seasonality but because we have a new tree there planted so that's the next and the task three it's the carbon storage which will be even next of next so a little bit fantastic we have a question for you about the ground tree thing how was your ground truth data set built and and the visual photo interpretation or terrain data yes so if i understood well um we we were doing a field work there actually not me because of kovid of course our team in mozambique they were doing a field work last december to point out uh to try to take a gps we took i think 120 interviews with different farmers and 220 more or less agroforest system because it's very common that they have more than one agroforest system they usually have two or three contracts even because it's uh each agree for a system it's one contract and so and then now that we have these points then we know where exactly these aggrava systems are so we can finally try to um to see if the behavior of the pixel where this forest are has a different let's say trend compare it to the the forest which was always there but has this phenology effect and then for the carbon we're gonna do a few work i think next january or maybe december to try to do the the field estimation of the biomass and carbon and then we can do a correlation of course using different open source satellite image for example sarah has a very impressive um probability of an accuracy of estimating biomass and so on fantastic thank you i'm going to see if um some of our previous speakers are still here and they'd be willing for me to add them back on stage greg are you still here and able to join us and stephanie are you still here unable to join us we still got fantastic stephanie might have needed to pop out birds they're now watching from venulis even though she's still here backstage um do you have any questions for each other okay i think i have just happened um stephanie in my phd i was working with the fragmentation effect so the edge effect so it's well known that in the first 100 meters of a forest patch we have a decrease of around 30 percent of the carbon and the biomass and one of the things that we lack in the community is field work trying to see the impact of the let's say the fragmentation effect on the carbon and biomass because sometimes we do this relation between of course the field work and the satellite data but if you don't get few data in these areas surrounding the fragment you might be underestimating the fragment effect and also underestimating the biomass so is there you

Channel: FOSS4G

Views: 150

Keywords:

Id: ziOcu6Rpe8w

Length: 30min 11sec (1811 seconds)

Published: Mon Nov 22 2021