 All right, I'll get started. We're at the top of the hour. Greetings, everyone, and thank you for attending this month's science seminar presented by the NSF's National Ecological Observatory Network, which is operated by Patel. Our goal with this monthly series of talks is to build community among researchers at the intersection of ecology, environmental science, and neon. We are excited to have Manoj Hari here to present today. But before we turn it over to Manoj, a few logistics. So we have enabled optional automated closed captioning for today's talk. If you'd like to use it, please find the C button in your Zoom menu bar. The webinar will consist of a presentation followed by a Q&A session. As you think of questions throughout the talk, please do add them to the Q&A box. That would be preferred over the chat box, whereas in the chat you can share links, introduce yourself, kind of use that as more of a networking space, and put questions for the speaker in the Q&A if you can. We'll then facilitate a discussion at the end, and there will also be an opportunity to unmute and ask questions over audio if you would like. Neon welcomes contributions from everyone who shares our values of creativity, unity, collaboration, excellence, and appreciation. This can be found in our code of conduct, which is found on our science seminars webpage. So let me do a quick screen share of that. So here is our webpage showing all of the talks included in the Science Seminar series. And if you go all the way down to the bottom of this page, you can find our code of conduct. I'll drop the link to the webpage in the chat in a moment. And so obviously this applies to anyone, all the staff at Neon and anyone attending our event. So thank you so much for following the code of conduct. This talk will be recorded and made available for later viewing on this very Neon Science Seminars webpage. We can see all of the previous talks up here with links to the recorded presentations, and this last talk will be, the recording will be added soon, probably end of this week or early next. To complement our monthly Science Seminars, we host related data-skilled webinars on how to access and use Neon data. We are complete with our data-skilled webinars for this year, but there will be more webinars starting up after a summer break. So look for more of those in the fall. And of course, we do have a venue for nominating speakers for the seminar series. And so if you can think of someone who might give a great presentation, please don't be shy to nominate them here. Okay, now I will turn it over to Dave Durden to introduce today's talk and today's speaker. Thank you, Samantha. So, I should be sharing my screen now. So I just wanted to give a kind of brief introduction to the Incarnion system and how it's kind of bridging the ecological and climate sciences. The idea behind the Incarnion system was to kind of remove the barriers to enable integrating Neon ecological observations with Incarn or the National Center for Atmospheric Research Earth System model. And through this, we believe that we can enable exciting scientific discoveries such as predicting changes in terrestrial processes, understanding changes at and across continental scales, and disentangling complex ecological interactions between the climate and the land service. We recently submitted a paper describing this system to, and it's currently in review at EGU Sphere. So we hope that we will provide a link to this so people can look into that if they're interested, but to kind of just give a little overview of this is kind of a conceptual model describing the Incarnion framework where we're taking Neon observations, gap filling and partitioning those to enable those to be ingested into the Community Land Model or the CESM Lab, which was created as part of this project. It's a containerized version of CLM that can be ran anywhere, even on your laptop. And then, ultimately using the simulated flux data and comparing that against the observed Neon data to both help improve our data quality and to help us better understand the model processes and model prioritization. Lastly, I just wanted to mention that, now we have these kind of accessible framework for CLM simulation, and you can run the Community Land Model even on your laptop. There's a lot of great tutorials available that can show you how to use this containerized system to run a CLM simulation and how to visualize and evaluate the model data in a Python Jupyter Notebook kind of interface. You can see the link here at the bottom of the slide. So with that, I would like to introduce to you Manoj Hari, he is a PhD candidate at the National Institute of Technology Rorkela in Odisha, India, in the Department of Earth and Atmospheric Sciences. For the last 10 months, Manoj has been a visiting doctoral researcher at NCAR in the Climate and Global Dynamics Group, working with Dr. Sannika Lombardazi. Manoj worked with Dr. Lombardazi on outputs from the NCAR Neon framework in conjunction with satellite remote sensing data. The title of his talk today is Synergies Between the Community Land Model and Satellite Proxies in Capturing the Terrestrial Carbon Dynamics at Neon Sites. And with that, I will stop sharing and turn it over to Manoj. Thanks, Dave. That was a wonderful introduction. Hi everyone, myself, Manoj and I'm a doctoral candidate from India. And I've been working with Dr. Bhishma Thiyadee over here in the Department of Earth and Atmospheric Science. And most of my work has been like entangled to understand the background process of terrestrial carbon cycle at a regional state. But in terms of region, I just mostly focus over India. Okay, well, for the past 10 months, I've been working with Dr. Danika Lombardazi at National Center for Atmospheric Research with a full-grade project that I've been engaged over there where we were working on NCAR Neon framework where we tried to understand how carbon flux has been changing at different sites based upon different ecological and climatical background. So based on that one, we try to understand how satellite data has been used or infused within this model structure framework in order to understand the overall dynamics of carbon at each site level. So with that, I'll just jump into the research that we did. Next, please. So initially, I'll just give you a broad outline about the proxies like the satellite data that I've been using for this analysis. Satellite has a very broad facility in order to trace the carbon across the globe. The one facility that it promotes is like even though flux networks across the US or like globe has precise measurements of the flux of carbon and water flux, they are like site-confined. But when it comes to the satellite data, they do have synergies along with the flux networks, but it captures the whole spatial extent which the flux network cannot be done. So for this analysis, we try to understand a key factor for the carbon flux which is considered to be the gross primary productivity. Well, gross primary productivity is considered to be like the uptake of carbon by the plants and the key flux which is modulates these dynamics is considered to be the gross primary productivity. So when it comes to the satellite, there is no such satellite have a direct observations of GPV. So we do have like different proxies as a measurement of GPV. So we just like try to compare the observations along with different proxies. One such proxy is called NARV which is like a novelised, I mean, a reflectance for vegetation. But for this proxy, we derived all the characteristics from modus reflectance band. The spectral properties is like they try to matches with almost 68 percentage as of a GPV's observation. The spatial extent of this, sorry, the spatial resolution of this NARV is around 500 meter where we try to frame it on the footprint of the beach flux servers at different locations. Next please. The next advancement of proxy that we used is called KNDVA which is generalised, normalised differential vegetation index which is considered to be the advanced version of NARV where they have a sigma parameter which has dynamic infusion within the flux networks. I mean, sorry, within the reflectance product as they can able to trace the flux variations at like greater extent. Again, we try to extend this product with the same modus reflectance band and with the 500 meter resolution we try to confine it at like site level based analysis. The map over here, you can see that like, that's the global extent of KNDVA where we try to map on a global scale where you could see higher KNDVA values which is obviously unit list where you could see the higher KNDVA values on always on the tropics. Next please. So the other proxies which is considered to be much more closer to GPP is called SIF which is solar induced fluorescence where we try to extract this proxy from tropomy SIF which is having a higher resolution in order to trace SIF at site level. So the spatial extent for this SIF is about like 3.5 into 7.5 kilometers square like at the NADY level where they try to capture SIF at a daily scale. The advantage of having SIF is like they are considered to be the direct proxy compared to the earlier proxy that I told. Next please. So by using these three proxies we try to understand the variability at different neon sites all across over there in US. So we again used NCAR neon system in order to simulate site level simulations using the community land model which has a concept of biogeochemical concept and we try to simulate at a half hourly temporal scale at different sites. So the main theme around the tropics about this processes we need to compare how well CTSM I mean CLM the neon sites and the synergies I mean the proxies are capturing the carbon dynamics and how it has been varying at different timescale. Next please. So for this we framed structural analysis behind this theme. So the initial idea behind this analysis was to understand how satellite proxies has been like much more equivalent with CSM and in situ analysis in capturing the carbon dynamics. And we want to trace the pattern of GPP across different gradients based upon temporal based upon their vegetational background and their climatic background. And further we need to extrapolate this analysis from the point scale to the global scale. So based on this one we framed our workflow. Next please. We framed our workflow and different parameters structure. So initially we try to understand how well this proxies has been modulated by different global environmental parameters. And we classify these sites based upon different vegetational backgrounds which is called like PFTs. The PFTs is nothing but plan functional type categories where we just like lean towards the CLMs PFT types because in order to have an ambiguity across the global scale. And further in terms of climate. So we classify the sites based upon the carbon climatic classification. And further we are like for the whole data period we just temporarily build the data at different scales. Next please. So when it comes to the neon sites obviously neon has a lot of sites all across the US but our focus is mostly confined to the corners I mean the contiguous US where it has neon has around 47 sites. But 39 is considered to be the most prominent which is in corners and eight are in Alaska, Hawaii and the other islands. Next please. The background image over here is just like high resolution of CTSM simulations. The average annual period for the whole analysis period. So for this analysis we are just selecting 28 sites which is highlighted here which has a consistent data flow and has CTSM simulations enabled sites. So next please. Further these sites has been like categorized based upon their PFT based upon their climatic background. For example, if you could see the top left corner the square where it indicates the site which is having a background of warm summer Mediterranean climate. When it comes to the circle obviously it belongs to humid subtropical climate where the neon has 12 sites which is which comes under that category. The site that has direct marks is represented where the site doesn't have a proper data flow where we had an in situ data gaps in this site. So we just wanna mark this site. Further, next please. Further we again classify these sites based upon their plan functional types which is obviously based upon the vegetational characteristics. The buffer, the highlighted buffer over here each color represents a different climatic or sorry the plan functional types. For example, the blue one represents the crop land. The yellow represents the deciduous broadleaf forest the red one evergreen needle forest and the violet it's grassland and the green is open shrubland. So again, if you just take out the top left side that site represents a warm summer Mediterranean climate. Again, that sites where the PFD is being considered to be evergreen needle forest. So that's how we classify each site based upon their climatic background as well as on the PFD background. Next please. So in order to consider in order to give the results just for this presentation sake we just picked up very few sites in order to represent our overall analysis. So we selected the sites in such a way each site represents each of the PFDs and all the climatic backgrounds has been included in our analysis so that you could see the diverse nature of fluxes across different sites. Next please. So initially we started our analysis by understanding the diagonal flux across different sites. And you could see the green bins that has been represented here are like half hourly fluxes flux variations across that has been mean for the whole analyzing period where the red solid line represents the CTSM simulations where the blue solid line represents the neon observation mean. So the gray shaded region represents the day period and obviously the white one represents the night period. So in this way, like we structurally frame in order to see the diagonal flux variations across different climatic as well as the PFD background. Next please. If you could see the highlighted box over here these two sites represents the deciduous broadleaf characteristics of vegetation. However, they do have a different climatic background irrespective of the climatic background just like because of the PFDs you could see compared to all the other sites the CTSM simulations over these two deciduous broadleaf sites has been underrepresented than the neon observations. Next please. In case on the other hand, if you could see the evergreen needle forest which is highlighted in red or the other irrespective of the evergreen needle background these sites has different climatic background and each climatic background has a different flux variations at diagonal scale. This shows both PFDs as well as climate has a distinct variable influence in modulating the flux across different sites. Next please. So in order to understand the variability of environmental factors, we try to understand how the atmospheric dryness or like the BPD which is vapor pressure deficit is modulating the energy partitioning across these sites. So for this, if you could see the plot over here the solid black line represents the neon observations sorry, the BPD neon observations whereas the red one indicates the CTSM simulations where you could see each site has distinct representation of like having a different response in terms of BPD whereas next slide please. Whereas if you overlay the neon observation sites which is like highlighted black where there is a lot of mismatch between the simulations as well as the observation. Next please. If you could particularly see the highlighted plot irrespective of their PFD background you could see a lot of mismatch between the simulations simulated Latin heat and the observation Latin heat at different sites. This means that like the CTSM like the CLM has a background of algorithms in such a way that like it has normalized all the PFD backgrounds. So irrespective of the site's genetic nature. Next please. So in order to consider the influence of different environmental variables we try to understand like to understand the correlation between different environmental variables such as temperature, precipitation, Latin heat and BPD. So for this we are spatially just like correlated each site's GPP observations along with the observed temperature. The sites over here has been like has been highlighted over here depending upon the correlation that has been correlated with GPP neon as well as the temperature. Next please. The pie chart over here it represents the first pie graph represents PFD's ratio but at the second one, second pie graph represents climate backgrounds ratio where the divisions in the pie chart represents the ratio between each PFDs. That means for example, if you consider the yellow in the first pie chart all the deciduous broadleaf sites has been cumulative ratio in such a way that like it represents the sites has been represented in a single frame. So if you could see the ratio between other sites considerably GPP neon and having a very good relation especially in the deciduous broadleaf site factors. Next please. In such a way, we are again correlated with precipitation, BPD and Latin heat but unfortunately we couldn't able to establish proper relation with precipitation especially with the observed precipitation except for grassland where you could see a higher ratio of like correlation just for precipitation but other than that irrespective of the sites we couldn't able to establish a very good relation but when it comes to Latin heat except for the shrub land sites which is marked on the green it has a very good relation especially for the grassland as well as for the deciduous broadleaf site. So this shows each environmental factors is playing differently at different sites based upon their climatic and PFT factors. Next please. So further we just want to check whether the simulation is going along with the observation. So that's why we just like a simulator the observations as well as we simulated the GPP as well as the temperature and we just like correlated with along with different PFTs, sorry across sites. It seems that like again precipitation doesn't give a very good correlation at these sites but compared to the observations here the simulator of GPP and BPD has a very good relation compared to the Latin heat which we saw in the earlier slide. Next please. So in this way again CTSM is having a very differential opinion in compared to the observations. So now we are just like I'm just shifting my focus towards the satellite based observations. The plot here represents the daily bin of GPP observations which is obviously the x-axis whereas the y-axis represents the CTSM simulations where we normalized the bins in order to fit all the other proxies in the analysis. Where you could see the ever been needed part which is the first bar where you could see a very good relation between the neon observations and the CTSM simulations. The vertical lines over there in the bins that represents the standard error within the simulation. So that's I need to mention it over here. But next one please. But when we overlay the other proxy which is KNDVA where it has a dynamic relation at different PFTs. Again, we do have a mismatch in data gaps because like the data swap that has been provided by modus is around like eight days. So again, we do have data gaps in each site. For example, if you could see the panels in the middle where you could see where you couldn't able to establish a proper relation because of data gaps. But for the sites on the middle panel on the left side where you could see for the grassland it has a very good relation similarly over on the top right where you could see a very good relation with observations as well as the CTSM. Next slide please. Similarly, we again overlaid with CIF which is in blue bins, sorry green bins but compared to the KNDVA again, CIF is having like outrunning all the other proxies. They try to irrespective of their PFT background and the climate background. They try to always try to go along with CTSM at like in most of the sites but when they go with observations just for like a higher canopy architecture such as deciduous broadly for every green sites. Next please. So what we need, what we plan to do is like we try to annually like we just like average it in order to see the annual dynamics as well as the seasonal dynamics and how this fluxes has been changing at different sites and how it has been modulated at different proxy levels. So here we just like for the full analyzing period we just annually made the annually average in order to see the climatology and the black solid line again it represents the neon observations and the bar graph represents the seasonal dynamics of different seasons from winter, spring, summer and fall. Next slide please. You could see over here just over these three panels where we couldn't able to establish the proper annual or the seasonal dynamics. That's because again, we do these are the sites where we had the data gaps. So we couldn't able to establish a proper relation in terms of annual dynamics or seasonal dynamics. Next please. But when we overlay it with CTSN simulations like the simulations, we had a difference of like contrasting analyzing opinions especially for higher canopy structure as well as for the deciduous broadly see. Again, if you could see the middle top, the middle panel which represents a deciduous broadly for us where it has been underrepresented, under simulated compared to the neon observation. Similarly, the last panel in the middle sector where again, it is a deciduous broadly for us but compared to all the other sector where you could see the CTSN simulations overestimated compared to the neon observations. And similarly, this way again, the seasonality has been changed depending upon mostly over in the summer season. Next slide please. Surprisingly, when we overlay the SIF observations over these particular sites, we could see that like SIF has is going with different is in favor of different sites. For example, if you could see the middle, I mean the first panel in the middle, where you could see which represents a grassland where it goes with the neon, sorry, it goes with the CTSN simulations. So again, similarly, the sites that has lower canopy architectures such as Disney on the top right where you could see it goes along with the neon, sorry, CTSN simulations, both in terms of annual dynamics as well as in the seasonal dynamics. But apart from those, for the Evergreen and the deciduous broadly for us, the surprisingly, SIF tries to go capture the dynamics. It goes along with the neon observation but it doesn't go with the CTSN. So it means that like a SIF has an capability to capture in terms of higher canopy architecture. Whereas when it comes to the lower bottom panel, like none of this is matching with either with the observations or with simulations. So we couldn't able to establish a proper compatibility along with these different proxies. Next please. So these are all the Evergreen middle for us and the deciduous broadly for us where you could see that there has been like the SIF tries to go along with the neon observations. Next please. Over here, these are all the sites which is again grassland and cropland where it goes along with the CTSN simulations. Next please. So further, in order to find like how well it has been going with the different temporal bins, we classified our data from daily scale to inter-annual bins based upon like different temporal bins. So for this, like over here, we like normalized all the data in order to fit everything in the graph where the green bar represents the neon observations and the horizontal green line represents the mean neon observations. Where the red bar represents the CTSN simulations where especially for the DB of deciduous broadly for us, it's been underrepresented. Similarly, all the other proxies is overrepresented. Next please. Likewise, we try to do the analysis for daily, weekly, monthly, seasonal to inter-annual. We just bring the data at different temporal bins in order to see the analysis. Irrespective of the temporal bins, it seems that like CTSN is underrepresented at different bins, whereas there has been a lot of fluctuations in terms of proxies. As you could see, compared to daily bins, for the monthly bins of KNDVA, which is like the black bin, where you could see that there's a lot of fluctuations and it seems like it's trying to match with the observations. Next like this. Likewise, we try to capture for different PFTs, the crop land, evergreen, native grass, grassland and shrubland. Next please. So each has distinct representation of sites, but then our focus has been limited to three distinct PFTs, next one please, which is deciduous, evergreen and grassland. That's because you could see the red, you can concentrate just on the red bar where you could see there's been a lot of changes, especially from daily scale to inter-annual scale, from even if you could just take grassland irrespective of evergreen or deciduous broadleaf forest. This way we again find a pattern where there's been a lot of fluctuations happened, especially if you just go along with the temporal scale as well. Next please. Likewise, we compare the analysis based upon the climatic background where the major of this climatic fluctuations has been found in the DFB, which is Heart Summers Humid Climate. Next please. So we try to again restrict our focus just on CFA and DFB in order to see the analysis. But comparing the others, other climate background, these two sites has a major dominance of evergreen native forest and the deciduous broadleaf forest. And that's why you could see the CTSM has been underrepresented in DVF climate background. Next please. So again, likewise, we just like compare CTSM with different proxies at different temporal scales in order to have how structurally there's been diversified at different sites. So here you could see it has been diversified across different sites when we compare the CTSM simulations with the observation. Next one please. Similarly, we just did the analysis for daily bins just for KNDVA in ARD and CTSM at different proxies. Unfortunately, we couldn't properly establish a very good relation with different proxies, except for CTSM. So try to capture a very good, try to replicate along with the neon. But unfortunately, neither KNDVA nor NRV could able to, again, could able to replicate the observations. And they do have a site which has a negative correlation aspect. Next one please. But it has been marked here. But this is the particular site where it has a background of deciduous broadening site. And still we are puzzled then why we couldn't able to have a very good relation just for this site. And we are trying to analyze the environmental background for these sites as well. Next one please. Similarly, again, another deciduous broadening site has comparatively higher correlation compared to the observations as well. They do have a very good relation with CTSM observations and the proxies. So this is just for the daily temporal bins. Next one. Similarly, we did the analysis for seasonal and you could see obviously there is a lot of better correlations at different proxy level, especially for SIFT and KNDVA. Next one please. But when we go, like when we again increase the temporal bins, you could see it almost caught the neon's correlation. It seems that like it almost replicate both the neon and the CTSM in terms of SIFT as well as KNDVA. Next please. So they do unfortunately, we couldn't able to have a very good relation just for these every native for a site. But apart from that one, we have a very prominent results. Next one please. Just all across the BF sites and the problem sites where they had the maximum correlations compared to the neon observations. Next please. So again, in order to see how these proxies has been modulating at different sites, we've been at like annual level. So we've just been the sites that has having the same PFD background. So here we are representing the Desiree's broadleaf site for us, where you could see the first panel represents the neon mean observations. And the next one is CTSM and KNDVA and SIFT. We neglected NARV because it doesn't give proximal value corresponding the other proxies. So again, like you could see, SIFT, the last panel, it tries to replicate the neon observations. And like still we are like, we are wondering how it is possible for a KNDVA in order to have a proper replication of CTSM. We're just having again a steep towards April and May, but then again, it tries to capture the same seasonality. Next one please. Similarly, we just build different PFDs and we try to find the annual dynamics. Next please. So this is for prop land and this is for evergreen where we could see the seasonality mix them please. And this is for grassland and this one is for next please. This is for grassland, sorry, shrubland. Unfortunately, again, because of data constraints and there is a lot of noises between in the reflectance product, we couldn't able to properly establish a relation with the neon observations. Neither proxy nor the model could able to have a very good relation just for this shrubland. Next one please. Likewise, we've been based upon different climatic background and again we found that like a CSB which is a humid climate and DFB which is hot summer Mediterranean or humid climate has again, it dominated the whole ship where CTSM and KNDVA again, it matches and SIF it goes along with the neon. It shows that like SIF is trying to go capture the footprint of neon luxus whereas KNDVA is trying to go with the CTSM. Next one please. So we just like try to go on an overall correlation at different levels in order to see how each PFDs has been performing at different proxies. Without doubt, obviously this is probably first which is having a higher canopy architecture. It has a very good relation irrespective of the proxies especially for KNDVA and SIF just over neon and CTSM. Similarly, grasslanders have again grasslanders, grassland and cropland they do have a very good relation with SIF as well as with KNDVA irrespective of the proxies but unfortunately we couldn't able to have establish a proper relation with CTSM and KNDVA for the shrubland background PFD. Next please. Similarly, we did the same for climatic background we found that like CSE and BFC we couldn't able to have a better relation that because CSE is having sites that have data gaps and both represent every native forest and again BFC has been represented by another every native forest but we couldn't able to establish a proper relation but DFB which is a humans of tropical climate background where most of the sites has a background of deciduous broadly forest again has dominated the whole the proxies as well as with the observation. Next one please. So here over here we are trying to go like again we are bringing back the temporal points in order to go with different time scales where you could see the parallel over here we compare the correlation between CTSM and KNDVA and we observed that like a different temporal points like each bins represents different temporal points for example the gold and the blue represents the annual and the inter-annual things and that seems having the higher correlation especially when we are comparing proxies. Likewise we did the same next one please. We did the same for like different PFDs irrespective of the different proxies where you could see higher the proxies like the irrespective of the proxies the higher the temporal points the higher the correlations between these model simulations as well as with the satellite proxies. This shows that like irrespective of the proxies the satellite goes along with the model simulations and they go along well with the different PFDs as well next please. So you could see that like except for ever being needle for us which is in the middle panel all the other side has a very good relation especially for annual and inter-annual bits but for ever being needle for us we couldn't able to have a proper higher correlation that's because again we do have data gaps in the sites that we simulated but then again we do find not environmental noises over the back just for the site and we are trying to expand this analysis in order to understand how will this has been globally represented at different sites. Next please. Likewise again we classify these sites based upon the climatic background and again irrespective of the climate or the PFDs temporal points when you increase at higher temporal points it seems higher correlation with proxies observations and the satellites. Next please. So we like we try to expand our analysis on a global scale. So we just like simulated a global CTSM for the same period from 2018 to 2022 and similarly we collected the data for KNDVA, NARV and so on irrespective of sites we do have a lot of variable animal dynamics just in the tropics but then if you could look in the temporary region just over in Europe the different proxies has a difference of opinions in terms of compared to CTSM similarly over there in the boreal forest this shows that like again sites in boreal forest do have like difference in variability which we are trying to expand our analysis in the future research. Next please. When we try to correlate it with the CTSM observations in a global scale we tried a lot of negative correlations has been observed just for the tropics that because like again these are the regions where you have tropical coral with like higher canopy architecture wherever you have higher canopy architecture the region seems to have a lot of variabilities that has been observed throughout this whole presentation but again it seems that like except NARV SIP seems to have a good correlation with CTSM over boreal and the tropical and the temporary regions over there in Europe but NARV is the only proxy where we couldn't have a proper having a negative correlation just over there in the boreal forest. Next please. So with that I'll just like summarize the work that we did. So initially we tried to understand how environmental factors have been playing their role at different structure. We consider that Latin heat and temperature has been driving GPP at like mostly over this year's broadleaf site and grassland but as there's been a lot of biases between Latin heat the observed Latin heat and the simulation one because just because they are having a higher canopy architecture and they have larger bias but when it comes to precipitation it just goes having a good correlation for the grassland site but other than that we don't have established a proper significant correlation for the climate exhaust but other than that temperature and VPD have a very good agreement for neon observations whereas Latin heat having a very good observation with the CTSM simulations. Similarly see grassland is only a site which is having a stronger relation irrespective of different proxies even though K and D we have a lesser representation for higher canopy grassland is a particular site where it had a stronger consistent correlation for different proxies but when it comes to PFT gradients you could see that like CTSM has been like always outperformed by neon but CTSM is underrepresented almost like just for the DVF sites. So irrespective of the climatic background except for DVF the CTSM has been like overrepresented whereas they are having a good relation with CTSM and the very good proxy that has been observed over this has been considered to be SIP. So they have a stronger seasonal correlation as well as annual relation except for grassland in the propland and when it comes to the shrubland neither model nor proxy could be able to capture a proper dynamics neither seasonal nor the annual dynamics just for the shrubland sites. When it comes to the climate gradient the CFA type and the DFB which is subtropical from the climate have exhibited a significant correlation irrespective of the proxies across different temporal gradients and similarly CSA and DFB pins has having a stronger disagreements with different proxies. When it comes to the temporal gradient aspect we again find that like the lower the temporal range it seems like we can be able to establish a proper significant correlations between proxies and observations but when we increase the temporal range it seems that like correlation seems to having an upscale especially on the higher temporal range and with that I would like to conclude my talk I'll take up some questions. Thank you, Manos. Everyone feel free to add your question to the question and answer box and zoom interface. I can go ahead and start with one question though. So how do you think the plant functional type impacts the modeled outputs? Do you feel that using something like FATES and where you can provide percentages of plant functional types across the site to kind of better characterize the site could improve this? No, again, since CLM is using the consistent data I mean PFT types which is again along 78 PFT types it is hard for us to again go with the percentage. Again, we could differentiate each site based upon their ecological background. Again, NEON has its own vegetational characteristics for each site but then we didn't go based upon the NEON. We just thought in order to go with on an ambiguity scale so we just try to stick with the CLM based PFT types but then based upon the ground variations again these results may change and obviously it will change because again we do obviously found some changes within the site level with a different PFT background as well. I mentioned it in the chat but just if anyone wants to make a comment or ask a question over audio there is a raise hand button in your Zoom menu and so if you do that we can unmute you and allow you to join the conversation. I can ask another question. So kind of following up on the importance of vegetation because that was interesting that you were pointing out that certain satellite proxies were better at matching the observations depending on whether there was essentially like short stature like not structurally complex vegetation versus tall structure structurally complex like forests and I wondered if you could speak more like why do you think that is? Why does NDVI, I think it was better in grasslands with SIF or vice versa in forests like what is going on there that interplay between what the satellite is picking up and the structure of the vegetation? Yeah, the characteristics of SIF is just like they are capturing just the particular wavelength of the reflected band that has been entered from plants. So they could able to capture even though if they have a sparse or like structurally composed vegetational characteristics but when it comes to modus which is like we just having a multi optical band. They don't have a proper algorithm in order to differentiate between the background or the vegetation or structure over the particular region. And that's why there has been a lot of noises especially for shrubland varieties of types where neither of the satellites like even though they do have a different background neither could able to have a proper relation for especially for the shrubland type but when it comes to shrublands SIF and NDVI tries to match with the same but then they do have different personalities because the reflectance where SIF captures are only 750 to 800 micrometer whereas the KNDV is just like an indices that we develop. So again, we do expect a lot of similarities between these two proxies. I'll also add that the different spatial resolution of the SIF related to the NIRV could potentially have a player role as well. I was wondering if we were able to get a better kind of spatial resolution cut out around the neon tower it might improve that relationship, right? If we were able to, yeah. I mean, you mentioned shrubland so we could talk a little more about that because I thought that's really interesting that drylands that tend to be shrubby were very poorly simulated. It seemed across the board like neither the proxies nor the models seemed like they were doing a great job. Could you elaborate more on what you think is going on there? Like were you just sort of alluding to the fact that there's not as much leaf biomass? It's like a lot of woody biomass or why are we having such trouble and trouble and what do you think we could do about that? Still shrubland is a mystery for us because like even neon would be able to establish a proper annual or seasonal dynamics. Again, we do have a lot of data gaps especially for shrubland side. So like they are mostly kind of like sub tropical climate. So again, we do consider like climate might influence a larger background on the perspective but then since like we do have a lot of data gaps in neon we couldn't establish a very good significant results in order to portray what's happening in reality. But then if you could compare SIF, if you just take SIF, they do capture a double hump seasonality like in the early spring like in spring and in summer. So it might capture, it may capture the structure unless until we do have a proper observation of shrublands we won't be able to like validate it with SIF. So still we are in a mystery phase. I will point out that there are some issues with precipitation data products at grasslands sites in particular. And it's partly due to the fact that there's not replicate data streams. So a lot of the gap filling was performed using a replicate data stream structure and relationships between these replicate data streams. And so at these grasslands sites we have a known issue that we're working on. And one potential answer is being worked on by Keegan King at NCAR. So she's working on developing some data using Prism. Yeah, to be able to initialize the model using the Prism data as well. Yeah, Prism will do show a strong correlation with these sites, but then still again there's been a lot of new observations as well. Any more questions from the crowd? Quiet crew today or any comments from Manoj? I think he's still working on writing this up, so. These are just like the preliminary results we are still working on. And we are just like we were confused how to portray our results because it's just like we need to represent each site and have to represent all the results. So just like against finding an effective way to represent the results. We're finding the story. Yep. Lot of interesting work though. It's an exciting way to see the application of neon data and neon combined with model simulations. I think it's really cool. Dave, did you have anything else? Otherwise maybe we'll end a few minutes early. No, I think it's really interesting work Manoj. And I'm curious too if maybe we could adapt some of the work you've done to work with AOP data or AOP remote sensing proxies at some point. I think that would be very interesting as well. That would be great. Sure. AOP is the airborne observation platform. It's our neon airborne remote sensing payload. We definitely don't cover the whole globe. They give us a very fine scale hyperspectral remote sensing at the neon site. So some cool avenues for future research there. Absolutely. Yeah, we do plan to use the neon drone observation that we tried to, but then none of them could be able to, like we couldn't be able to properly match temporarily the observations that we manage. So we thought of excluding the analysis over there. Well, great. Thank you, Manoj, for taking the time to give this great presentation. And thank you, Samantha, for organizing. I guess with that, we'll wrap up a little bit early. Give people back some time. Yeah, this was our last science seminar for kind of this academic year, if you will. We'll be taking the summer off and then we're gonna start back up in September with another series of hopefully very interesting and stimulating neon science talks. It'll be same time and a monthly schedule. And we're also gonna ramp up another, the paired series of data skills webinars. So look for some more communications on that. Thank you all so much and have a great rest of your day.