 All right, we're at the top of the hour, so let's get started. Greetings, and thank you all for attending this month's Science Seminar, presented by the NSF's National Ecological Observatory Network, or NEON, which is operated by Battelle. 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 very excited to have Tom Q here to present with us today, but before we turn it over to the speaker, I'd like to go through a few logistics. We have enabled optional automated closed captioning for today's talk, so if you would like to use that feature, please find the CC button in your Zoom menu bar. The webinar will consist of a presentation followed by Q&A. As you think of questions, please add them to the Q&A box. We also have a meeting chat, so you can use this to share links or other items of interest with the group, but do place your questions for the speaker in the Q&A. We'll facilitate discussion at the end, and there should also be opportunity to ask questions over audio by raising your hands. NEON welcomes contributions from everyone who shares our values of unity, creativity, collaboration, excellence, and appreciation, as outlined in our NEON Code of Conduct, and this applies to NEON staff as well as anyone participating in the NEON event. The full code of conduct is available via a link that I will drop into the chat in a moment, and also embedded in the Science Seminars webpage, which I am showing here. I believe the code of conduct is down here. This talk will be recorded and made available for later, reviewing on the same Science Seminars webpage. If you click on the talk for today, we will embed the video in the event page for this talk once it's available. And to complement our monthly Science Seminars, we host related data skills webinars on how to access and use NEON data. Registration for those is also available on the same webpage, which I will make sure to drop a link to the page in the chat shortly. You can see that our first webinar is coming up later in September, and we would love to have any of you join for that. Lastly, if you have ideas for a great talk, future talk for the seminar series, consider nominating yourself or a colleague today by filling out the form up here toward the top of the Seminars webpage. All right. Now I'm happy to turn it over to Laura Nagel to introduce today's speaker. Hi, everyone. Tong Chiu is an assistant professor at the Pennsylvania State University, where he leads the Spatial Ecology and Environmental Data Sciences or SEEDS Lab. His lab is dedicated to understanding how forest ecosystems are functioning under global change at scales ranging from individual trees to the entire globe. To achieve this, they use satellite and airborne remote sensing, monitoring studies and advanced statistical models to quantify the regeneration potential of global forests in a rapidly changing climate. His lab also seeks to determine how climate and habitat characteristics influence biodiversity change by integrating data from NEON airborne observatory platform and biological survey. Welcome, Tong Chiu. Hey, thank you very much, Laura. Let me share my screen. Screen one and the slides. Okay, here we go. Okay, thanks again for the warm introduction. This NEON science seminar has a really great series talk. I'm really happy to be part of it and share some of the results from synthesizing NEON AOP with biological survey data on the ground. Basically, today's talk focuses on two topics. As we know, NEON really offers amazing resources for ecological research by providing consistent and long-term data across a diverse ecosystem at a continental scale. Through its standardized data collection and state-of-the-art instrumentation, NEON really enables ecologists to understand and predict future environmental change. Today's talk gives two examples. There's a synthesized biological survey on the ground with NEON AOP, which is airborne observation platform, so NEON remote sensing approach, to better monitoring biodiversity change. So the first example focuses on the role of AOP-derived habitat characteristics. And the second example aims to quantify the fundamental role of food supply in quantified biodiversity change. So now let's start with the first example that estimates the response of ground beetles to the interactions between climate and habitat variables. So terrestrial insects provide important ecosystem services and functions that are critical to human well-being, such as pollination of fruits and vegetables, acting as biological control agents for weeds and agricultural pests, and contribution to nutrient cycling through the decomposition process. So while insects have many important ecological and economic values, there's a green evidence suggesting that insect population are facing severe threats under climate change. So in 2021, a special issue that includes 11 papers on PNS provides assessments of insect population decline and summarizes the environmental stressors that drive those trends. So climate change, including increasing temperature, more severe drought events, and more frequent extreme weather events, all contributed to these insect decline trends. Additionally, like habitat loss, fragmentation, and degradation have eliminated the essential environments that insects realize are. So examples include deforestation and the wide usage of pesticide from agricultural intensification, as well as pollutions from urban addition. So air pollution, light pollution, and chemical pollutions that are associated with urbanization process. So those stressors are directly linked to the decline of insect population, and the evaluation of these insects declined across biogeography scale relies on meta-analysis in the current literature. So the first meta-analysis is a cover article on science, which found that terrestrial insects are declining at a rate of approximately one percentage per year, as indicated by the prevalent negative slopes in this figure. So their results are primarily driven by European and North America data sets. However, in about six months, a paper on nature ecology and evolution found the contradicting results, where the overall trends of insect population are actually around zeros in 15 long-term ecological research sites across the United States. The large inconsistencies between the two meta-analysis arise for several reasons. First, time series of insect population are usually noisy. It is difficult to generate reliable estimates of trends from these types of data. Even when we have long-term series, meta-analysis offers aggregated information that is challenging to capture the species specific process at individual scale. The aggregated data is also from studies that implement a different sampling strategy, which could introduce biases into the trends estimation. Additionally, the coverage of sites in those meta-analysis are not balanced over climate and habitat space. To resolve those challenges, we use ground beetles at the study system, which are monitored using the same particle and consistently monitored across neon sites. So ground beetles is one of the most diverse species group with 40,000 species globally and 2,000 species in North America. So if we want to interpret the trends at species level, it will be challenging because we have so many species, but traits can generalize species-level response to global change. Examples of traits are demonstrated in this figure. So for example, some species have really long legs and powerful minerals like the tiger beetles, making them formidable predators in the insect world. So they are fast runners and can easily chase down their prey. Some species are also borrowers and can feed on the insects in the soil. Some species can fly or climb on trees, pursuing prey in the canopies. Many species have a broad diet, eating not only other insects but also on seeds and plants. So all those traits, such as body size, running, flight, borrowing, and diets, make ground beetles an important biological control agent in the agricultural system because it can reduce weeds and insect pests in the crop fields. This study focuses on answering two questions. First, how do ground beetles traits influence their ability to thrive under global change? And how does habitat modulate the effect of climate change on ground beetles? As I mentioned earlier, NEON provides unprecedented opportunity to answer those research questions. So ground observations from NEON includes over 700 species and 2500 polyears that cover multiple habitat types, including boreal tundra to template carnivorous and deciduous forest and then to tropical forest. Most importantly, they use standardized methods and protocols for data traction across all sites. This ensures consistency and comparability of data, which is critical when we aim to understand which combination of traits and habitat influence the climate impacts on ground beetles. We also supplement NEON data with data from literature that implement a similar sampling protocol. So those new sites are colored by orange in this figure. A little background on the NEON sampling protocol. So ground beetles are collected using pitfall traps approach. The figures on the left shows the arrangement of traps within each plot. Three or four traps are installed within each distributed base plot of a given NEON sites. Pitfall traps are emitted bi-weekly during the active scoring season for ground beetles. Those species are identified by technician and then a subset of those species is also validated by expert economist. Example of two species at different habitat. So deciduous versus pasture and their bi-weekly time series from June to September is demonstrated in this figure. So even though NEON provides such amazing data, there are still some modeling challenges. Like many ecological data, there are massive zeros in the abundance which are difficult to handle in traditional models. If we plot the histogram of abundance for ground beetles, we can see there's a median of zero. And ground beetles are not only affected by climate variables, so they are not only responding to temperature or deficit, they're also responding to habitat change. So because their body size is generally very small, fine spatial scale habitat information are very important predictors, but they are rarely explored in the literature. So first of all, species live within communities where the abundance of one species depends not all on its environmental conditions such as climate habitat, but also on the abundance of other species. So in other words, species do not respond independently to climate and habitat, and it is important to model multiple species jointly. Finally, there is a climate habitat interaction, which will be demonstrated in the next slides. So the term climate habitat interactions describe the effects of climate that depend on the heterogeneous habitat. We can use an interaction between temperature and canopy gaps as an example. Canopy gaps is introduced by deforestation, disturbance, and forest damage. Where there are openings in forest canopy, there is an immediate increase in sunlight penetration to the forest fall. This higher solar radiation can increase the temperature within the gap compared to the shaded industry. For example, in this figure, temperature is cooler if forest density is higher. Therefore, there is a possibility that larger gaps can amplify the effects of climate warming. We would expect a positive interaction between temperature and the gaps if increasing temperature has a larger effect when canopy is more open. By contrast, larger gaps also means fallen logs, branches, and stumps that could offer shelters for ground beetles, protecting them from predators and environmental stressors. If fallen trees provide local refugees that buffer the impacts of climate warming, then we would expect a negative interaction between forest gaps and climate warming, where warming impacts are less severe in more open canopy than closed ones. Those would be the two hypotheses we have regarding to how habitat modulated climate impacts on ground beetles. Now, NEON Airborne Observation Platform, AOP, offers cutting-edge remote sensing approach that allows us to test those hypotheses. The first approach is light detection ranging LIDA, which uses laser and active remote sensing approach to record information start from the top of the canopy, through the canopy, and all the way to the ground. For example, LIDA will characterize forest structure along a transect to include the top of the tree crowns as warm colors, and anniversary and ground as the coat colors. This unique of data sets will enable the quantification of undershore nitrogen density defined as the density of those blue points near the surface. A dense undershore may influence ground beetles foraging efficiencies, but also reduce its exposure to predators, so we include that as an important predictor in the model. The second metric is gap fraction in a forest stand, where purple color are the gaps, and yellow are the tree crowns. As I mentioned in the previous slides, it is a primary contributor to the climate habitat interaction that we will test in this study. Finally, there is terrain roughness, which can be derived using the ground elevation that LIDA measures. It potentially influences the mobility of ground beetles when they are moving on the ground. NEON AOP also provides the hyperspectral remote sensing approach, which captures data at a very narrow and continuous along the spectral dimension of this data cube, allowing a detailed analysis of reflected light from vegetation. So through the collaboration with Phil Thompson and Kyle Kovach at the University of Wisconsin Medicine, we have estimated constitutions of nitrogen in the canopy. The figures on the right includes two NEON sites, the Joseph Jones Ecological Research Center in Georgia, which is characterized by lonely pond mixed with patches of deciduous forest. The bottom canal is under the University of Notre Dame Environmental Research Center, that is characterized by primarily deciduous hardwood. The hyperspectral remote sensing will be able to capture variations in nutrients concentration across crowns, where canopy nitrogen is lower in conifers compared to that in the deciduous, but therefore we would expect a higher nitrogen in the Joseph Jones Ecological Research Center than the one in the end compared to the one in Joseph Jones Ecological Research Center. So now this as a proxy of ecosystem productivity that derives canopy nitrogen is an important habitat variable. So higher canopy nitrogen could support a higher population of ground beetles that feed directly or indirectly on those plants. So which means a higher canopy nitrogen could support a higher population of caterpillars that feed on those plants, and then also will support a higher population of ground beetles which feed on those caterpillars if that's an indirect relationship. So those are the bottom up controls. On the other hand, higher canopy nitrogen productivity could also support bird and small mammals that are predators of ground beetles reducing their population. So these are top down controls. So we explore those dynamics by introducing canopy nutrients as a predictor in the model. So recall that there are multiple challenges in modeling ground beetles abundance. The first two challenges have been resolved using NEON and its airborne remote sensing data. So from the NEON offers a consumer sampling at a continental scale, and also we have a one meter spatial resolution habitat variables derived from NEON AOP. The other three will be addressed by the generalized joint attribute model, GGM. So GGM is developed by Dr. Jim Clark at Duke University. It is a Bayesian high-core model that handles median zero data with partitioning. The partition translates the observed discrete abundance into a continuous latent variable. So which means the median zero data will be able to handle in this Bayesian high-core model. Then the latent variable is quantified by the design metrics, coefficient, and the covariance metrics which enables the joint modeling of species. The corollaries are the predictors in this model. They include climate variables such as temperature and moisture deficit. There are also three light-up derived habitat variables including canopy gaps, industrial density, and terrain roughness. And also one hyperspectral derived canopy nitrogen. So finally there's an interaction between gaps and temperature. So and then we also model trace jointly with the species through a predictive trace modeling approach. So we are transforming trace by communities using a species by trace metrics. The results taking over all species across NEON sites, the light-up derived habitat variables canopy gap is the most important source of variation in species abundance, as indicated by the blue bar, as well as for trace, which is the purple bar. So by contrast, terrain roughness account for lower overall variations in the community. So this sensitivity are ranked across different predictors. So higher importance of predictors will be ranked higher in this plot. Now as for the climate variables, they are followed the canopy gaps including temperature and moisture deficit. So similar to canopy gaps, they are more important for predicting trace than species abundance. A traditional literature of ecological study will find that climate variables is more important than habitat. So however, through a kind of study, we found habitat variables is more important than climate variables in terms of explaining variations in trace and abundance. Regarding to the canopy nitrogen from hyperspectral data, so they are more important in explaining variations in species abundance than in predicting traits and the interaction between gap and temperature is just following the hyperspectral canopy nitrogen. This figure shows the coefficient metrics where each rule indicates a trait and each column is a predictor. So brown and pure color represent negative and positive coefficients respectively. The bounding boxes in this figure highlighted similarities within different groups. For example, the lower left box include a group of traits that are negatively related with canopy gaps and the moisture deficit, while traits in the lower right boxes exhibit a positive relationship with temperature and annulary density. So those have teal colors in this group and brown colors in this group. So the usage of these coefficient metrics really can help us identify different trait syndromes on the right in this correlation metrics. Positive correlation, which are colored red in this new metrics, means that the groups of traits share a similar response to predictors in the model. So trait syndromes include large-bodied borrowing omnivores, graphite fires, forest carnivores, and other groups. So those traits evaluate how different species can survive under global change. So for example, these slides have one small body size frequent fire that live in the grassland, agonautas, conjectures. The other one, druskos-pensavinicas has larger body size, cannot fly, and primarily found in the forest. To compare their response to global change, we predict their abundance change between historical and future time interval under projected climate change scenarios. Now these slides provide the information that the small non-flyer species will increase throughout North America under future warming scenarios, as shown by the red color. Because the species are responding positively to increase in temperature, so it really enjoys an increase in temperature which can increase its abundance over time. By contrast, druskos-pensavinica are responding negatively to warming, which means that its abundance will decline throughout the map. However, the projected changes are actually spatially heterogeneous with no uniform trends. We see some blue points, but there are also some red points, which means they are increasing over time. The interesting pattern is caused by climate habitat interactions. On the left side of this equation, we have abundance change caused by temperature, which is delta w with subscript t, which equals to the summation of direct effects and interaction effects on the right side of the equation. The sum of those two gives the full effects of temperature, which is what we saw in the previous figure. Direct effects of temperature equals to the coefficient of temperature times temperature change, which is delta t. The effects of interaction, the second term on the right side of the equation, equals to the coefficient times the multiplication of delta t and canopy gaps. So for this species, druskos-pensavinica, the coefficient to temperature is always negative, but the interaction coefficient is positive. Now if we put those equations in the map, first we have temperature change delta t, with a gradient of red colors indicating different degree of warming between historical background and future time intervals. So the abundance of change caused by warming, which is delta w with subscript t, equals to the negative coefficient times the delta t. The whole product is the negative, and the abundance of ground beetles are projected to decline in the future across the continent. But the impacts of climate change will differ by canopy gaps. Yellow color in this figure are forest with, not just forest, but other types of ecosystem, with a higher fraction of gaps and can be found in the central grid plains, and blue are closed then, mainly in the northeast. Now if we go back to the original negative products, which is the direct effects of temperature, the second term in the equation, the effects of temperature that are modulated by gaps is a positive product. Therefore, it is capable to offset the declining effects from the directive of temperature change, and even reverse the sign when gaps is large, like the central grid plains of this map, so which can really result into a spatially heterogeneous response to climate change. So this heterogeneous response that results from climate, results from the canopy gaps can also be vitrized at landscapes in northwest forest, where they are both open and closed then. If we take a look at this landscape level variation, so they are both open and closed then, the background maps shows the local terrain information. Symbology of the points follow the previous maps, so red means increasing and blue means decreasing. So all of the red points in this figure, which means projected increase of abundance change in the future compared to the historical background. There are places with open stand where yellow are tree crowns and purple patches are gaps. So a larger canopy gaps is more likely to be related with a projected increase of abundance change. Similarly, the projected decline of abundance mainly occur at plots with a lower fraction of gaps. So this really summarizes this habitat can really modulate climate change on the abundance change of ground buildings. We have also developed an open-source online vitriolization platform for the scientific community and policy makers. The website is called pbgjam.org. So this website is trying to apply advanced statistical models and remote sensing approach to understand and forecast the effects of a changing climate on the abundance and deterioration of American wildlife. So users can select one of the 100 species and identify their projected abundance change across North America and their different climate change scenarios. So we have a couple of scenarios including more severe warming and one more if we control the CO2 constitution that are released from to the atmosphere. Landscape-level habitat conditions such as terrain and canopy gaps are available with user interactions through the collecting on each of the plots on the left of this figure. So this really can help the conservation biologists to understand how different habitats might influence a different species response to global change. We also generate a walk wall habitat stability for each species across the continent. So this is the underdevelopment but we are trained to leverage the NEOP habitat characteristics to better forecast a continental scale prediction of different species abundance. So right now we have trees, small mammals, bird, and also ground beetles. And those are the different scenarios of climate change and also the different intervals in the future and historical mean. Now if we go back to the two recent questions we have we are able to answer them through the synthesis of NEOP and biodiversity survey data. First we found tree syndromes emerged from climate habitat interactions. So remember there are four different syndromes. So there are large-bodied borrowing omnivores, there are forest carnivores, and grassland fires which we were able to quantify through the climate habitat interactions. So ground beetles response are species and tree specific. A small frequent fire is projected to increase under future warming change scenarios while a larger non-flyer species shows spatial heterogeneous patterns because gaps provide a buffer against climate warming. So here the interaction between climate change and gap fraction is different from how the kind of gaps can amplify the warming impacts. The gaps actually provide the refugees needed to buffer climate warming. So now we understand the important rules of habitat such as atmospheric density, gap fraction in modeling biodiversity. We should look further than just considering habitat factors. So the availability, quality, and diversity of food supply can have profound effects on the distribution and abundance of species. The second part of this talk focuses on integrating fecundity monitoring across the neon plots and linking it with crown nutrients from neon LOP data. So the amount of seeds, fruits, and nuts are the foundation of forest food web, particularly for the mass consumers like birds and mammals. Even though ground beetles cannot directly, does not directly participate into this mass system, they are still influenced by the birds and the mammals because they are their predators and they are like a top-down control on their population. So the mass team inference forecasting, the mass data network is led by Dr. Jim Fock and involves collaborations with over 100 scientists from five continents. So we have assimilated data over 12 million three-year observations from 1200 species, 2,000 long-term plots, and more than 3,000 anatomist observations, which are citizen science component. This huge amount of data served as a foundation to quantify global seed supply and how they can influence forest food webs. So mass team network also include neon sites. So fecundity monitoring is not part of the neon sampling protocol. So mass data supplements neon's data and bring more observations for tree dominant sites. So we are not trying to sample all the neon sites, but we are focused on sites that have more trees. There are two different data tasks. The first data is crop count, which involves observing the number of crops each tree produce in each year with a vernacular. The second one is seed trap that has a modeling component to link seeds from trees to traps and also seed counts in a trap within a mapped forest stand. So the synthesis of massive data at neon plots with neon AOP enables the fecundity mapping at individual tree crown scale. This figure includes a large, this is just dominant tree plot at Bartlett experimental forest in the northeast, so it's in the upper panel. And they also have a colorful dominant small plot at narrow range in the lower panel. So background is a canopy hat model from Leida, where green means high and brown means low values of tree heights. Each polygon is an individual tree prongs that are and they are colored by species. So fecundity is mapped on the red panel. We are higher transparencies in this figure indicate a low fecundity values. So we are interested in the environmental factors that control this large variation of fecundity from tree to tree and from species to species at landscape scale. So we are trying to understand this question, what controls fecundity variation? And then one of the potential factors in driving fecundity variation are nutrient levels. So our previous synthesis found that species level of fecundity would be related with the species capacity to explore the nutrients, such as foliar nitrogen in the x-axis and foliar phosphorus in the y-axis. So each point in this figure is one species and it is colored by its leaf habit. So interestingly, species level of fecundity decreases for all the magnitude from the lowest, which has a green color in the background, to the highest, which are the purple color in this figure, along the phosphorus gradient. By contrast, foliar nitrogen has limited effects. The contour line is almost parallel to the x-axis, indicating that fecundity does not change very much along the nitrogen gradient. However, regarding to the relationship between individual fecundity and the tree nutrients, studies are still limited to a few locations and a few species due to the large investment needed for field measurement on both nutrients and fecundity variables. You literally have to go to the field to quantify the amount of seeds that a tree produces in each year and also measure the nutrients access to those trees. So it's really a large investment. In horticultural practice, proper amount of fertilization would stimulate crop yield. On the other hand, trees growing under more fertile sites would potentially grow faster and do not reproduce too much. In other words, increased nutrients might lead to an extensive vegetative growth at the expense of tree fecundity. So there are really two pathways here. So one can stimulate crop yield and the other one can increase the amount of vegetation growth instead of the tree fecundity. So to test those hypotheses, we call that hyper-spatial data from NEON-AOP can generate kind of a netrogen, so through the collaboration with Phil Thompson and Cal from UDAR Medicine. So we can also produce other nutrients like phosphorus, potassium, magnesium, and calcium that are potentially important for fecundity. So we modeled the fecundity using those five nutrient variables, following the same generalized joint attribute model framework Gdian in the ground beetle study. So this figure, similar to the ones in ground beetle study, shows the coefficient matrix where each row is a tree species and each column is a nutrient predictor. Bronze teal color still represents negative and positive correlations, really, respectively. So first we found a prevalent association between high foliar phosphorus concentration and low individual fecundity in many species, which is consistent with our previous species level synthesis, so where phosphorus can lead to a decline of fecundity about four orders of magnitude. So other bonding boxes in this figure highlight similarities within different groups. For example, the lower left box includes a group of species that are positively related with calcium and magnesium, and upper right boxes include a group of species that are positively related with nitrogen, phosphorus, and potassium. So those different communities with a coherent response to nutrients really emerge on the right in the correlation matrix. Positive correlation, which are colored red in the new matrix, means that the group of species share a similar response to nutrients. So there are five distinct communities. First one here, and then second one here, and then a bunch of other threes in the upper of this matrix. If we put those communities on the map, the northeastern mountain forest are associated with a higher amount of phosphorus and low calcium. By contrast, the eastern temporary species are negatively correlated with all nutrients. The forests in the west contain species associated with higher nitrogen, phosphorus, and potassium, but lower calcium and magnesium. These relationships have broader implications when we apply the fitted model from the mustive and the neon process to force imagery data and produce a continental scale prediction of focundity. So right now we are using a graded soil fatality product, which is cut exchange capacity, which is a spatially-caused product to predict continental focundity on the right. So there's a focundity hotbox in the southeast part of the United States. However, this continental scale prediction does not benefit from the neon AOP and its connection to canopy nutrients. As demonstrated in the top-bottom panels, the AOP-developed nutrients variable will help us understand and predict focundity change at a much finer spatial scale. So we'll be able to track individual tree crowns focundity and their influence on forest flue labs from individual trees to the entire continent. So in summation, so we are able to answer the questions through the synthesis of neon AOP and the focundity data. So we found a prevalent negative association between focundity and the crown phosphorus, while other nutrients have mixed effects. The biogeographic patterns also emerged from this focundity and nutrients relationship and can inform continental scale prediction of focundity. Now I want to wrap up this talk by acknowledging the collaborators at Duke University, including the P.I. Jean Clark, co-P.I. Jennifer Swenson, and the P.I. two students, Len, Maggie, Arinata, and the technician Jordan. So they are really critical critical roles in collecting data and putting together this massive network. Also, I want to acknowledge the collaborators from Inray in France, including Valentine, George, and Benoit, and then Phil and Kyle provided the neon AOP-derived canopy nutrients, which are critical in understanding focundity and the ground beetle's abundance. And then Aaron Bell is a co-author on the ground beetle's paper, and he provided other data that supplements neon observations. So funding for this project, including National Science Foundation grant, the DEB, a microsystem biology grant, and also two NASA grants from the AIST. I also want to acknowledge the neon for making all the data openly available so that researchers will be able to answer questions at a continental scale. And then these are my contact information, and I'm happy to take any questions that I may have. Should I stop here? Sure. Thank you so much for a wonderful presentation. That was so much amazing information in there. Just a reminder to everyone, we do have a Q&A box, so as you think of questions for the speaker, and either the first or the second part of the talk, pop them in there. I see that we have a few. Laura, are you interested to read those out? Yes. Yeah, we can go in order. First question is from Alyssa Wilson, and she wants to know, when you were beginning these projects, did you start with asking what question can I ask using neon data, or did you ask what data can I use to address this research question? Or is it something in between those two paradigms? Oh, that's a great question. So I also have been wondering that question when I was a PhD student. So when I was a PhD student, I always asked my advisor, so should I just look at those data available, and they're open access data, and then should I just use those data to explore some questions? My advisor told me it's kind of dangerous if you don't have a question in mind, and if you don't have a couple of things that are related with those questions. So I would stress that you came up with a good question that makes the best usage of neon data, and then quantify this answer those amazing ecological questions at a continental scale, and then using this neon sampling approach. So that would be my answer. So I would have a question in mind first before using the neon data. The next one is from Shashi Kunduri. I'm curious to know how you calculated the understory density using neon LiDAR data. LiDAR systems have a range resolution, which makes it challenging to discriminate between two objects, if they are too close to one another in a vertical profile. In this case, differentiating low vegetation LiDAR returns from the ground may be challenging due to these range resolution limitations. Even the newest LiDAR system used by neon has a range resolution of 67 centimeters. So how do you calculate the understory density using the neon LiDAR data? That's also a very good question, and so let me go back to my slides. I think it would be useful if I can have the slides open. Now I need to go back to the slides. I should try a different approach. Where is it? Oh, it's here. Okay. So we try to actually a bunch of approaches. So the understory density, they influence the mobility of ground beetle species. So we try to use several thresholds. So we try the understory density between 0.15 meters to two meters. So like a lower end of those blue points and divided by the from 0.15 meters to the entire kind of heights. So that's one approach we tried. We also tried from 0 to 0.15, and then divided by 0 to 2. So one is called normalized understory density. The other one is called relative understory density. So we tried both, and we use one that has the highest prediction and accuracy. And then so I understand also neon has a bunch of accuracy spatial resolution issues. So those won't affect the, if you normalize the data across the range of heights. So if you normalize those across the range of heights, you'll be able to get a standardized understory density across all neon sites. So that's the approach we use to quantify the understory density. I hope that answered your question. If you send me an email, I can send you the code I use to derive understory density, which I think can be easily adapted to other LiDAR systems such as USGS LiDAR. The next question is also from Alyssa Wilson. How do you think spatial processes play into the environmental drivers of beetle communities? For example, gaps that are closer together may change the microclimate at the forest floor more than isolated canopy gaps. Is there any plan to incorporate space into the GJAM site? That's also a really good question. So basically, let me go back to this figure. So we found that canopy gaps really provide a buffer for ground beetles. So normally, if they are microclimateologies, you would think larger gaps may be amplified, impacts kind of warming because they are just so many solar radiation that can penetrate through the forest gap. So the way we handle that is we try to incorporate the spatial variations of canopy gaps at a kind of scale and put all those variations into one single model. So we are trying to model every single neon sites jointly in a model. So we are not trying to compare different gap levels across time. So the space information is kind of already incorporating the GJAM because we model all species jointly and including a bunch of sites across the neon domain. So if we put the interaction between gaps and warming, that can tell us questions on how gaps can modify or mitigate or amplify kind of warming effects. We can also do a similar approach. So if we use a climate modeling approach and just change the level of gaps and then detect how this level of gaps can influence local climate, we won't be able to directly incorporate gaps in the prediction. So we can adjust the change. We can just remove the gap variable in the model but change the temperature variables. That's another way to isolate the effects of gaps from microclimate. So I hope that answers question. Wonderful. Next is Michael Kaspari. Pitfall traps do not measure abundance but activity density. So more like individuals moving through a finite area over time. Roughly active density is a function of abundance times velocity. The question is could your results be showing not changes in abundance but that gaps and warmer neon's tower temperatures are associated with more corroborate movement, not necessarily corroborate abundance? Yeah, I'm reading the question that there are also some hypotheses that are related on this question. Okay, that's really nice hypothesis. Yeah, so one thing we can do is I'm thinking, so the caribate movement, yes, we do try to quantify the temporal trends of caribate movement. So there's another version of GGM called dynamic GGM which incorporates the movement of a species, the species interaction and also climate variables like this static version of GGM. So we're trying to use that dynamic version of GGM to predict the response of ground beetles to climate and habitat variables. So and we get a really noisy response. So it's hard to tell if the ground beetles can move from an open canopy to closed stand because the gaps is closer there and maybe kind of influence the microclimate. So we didn't find any meaningful results. For two reasons. The first one is the neon time series is still very short for the ground beetles and it's also very noisy because it's bi-weekly sampling. And we hope if neon has accumulated more data in the future, we'll be able to re-implement that model to understand the movement of ground beetles. Regarding to the, okay, I think that answers the question, I hope. Yeah, we are not using the temperature data at neon parts. We are using the graded temperature from DMED products. So maybe we can use the neon tower temperature can better predict the model. So I can try that for sure. Yeah, thank you for your suggestions. The next one is from Courtney Meyer. I noticed that gaps appeared to have strong influence on ground beetle abundance, but core stone wood did not have much effect. I would have thought that core stone wood abundance and gaps would be strongly correlated. What do you think might be going on here? And did you use neon core stone wood data? Yes, I did. So the cost woody debris is a very important source that neon offers so much data size. Yeah, it's hard to not use all of them. Anyway, so the cost woody debris, we use the volume of tree logs in those sites. And then, so I'm not sure why we didn't detect signal of cost woody debris on the ground beetle abundance. So one potential reason is that there are some data gaps in this cost woody debris. So some sites does not have this information on which might influence this ability to explain variations in trees and abundance. But definitely cost woody debris should be a very important predictors in wildlife species because they just simply offer shelters against climate warming. So here we found the cost woody debris is just following kind of the nitrogen and above gaps, the interaction between gap and temperature. So we tried the interaction between temperature and the cost woody debris and that effects has a lower prediction accuracy compared to the gap, the interaction between temperature and gap. So that's why we focused on temperature gap instead. And yes, I also checked the correlation between gaps and cost woody debris. There's not a strong correlation between those two. So it's safe to put both of them in the model. If that's also your wondering. We have one more right now for Roland K's first of all, great talk. But the question is how did you get the tree level fruit fecundity to match up with the LiDAR data at neon sites? Let's see. Okay. So here's the approach we are using. So we basically, so in the background, there's the canopy had models from LiDAR, which is the difference between the canopy surface model minus, sorry, a digital surface model minus a digital elevation model. So the difference gave you the canopy height. And they were trying to implement a crown delineation of those tree canopies. And we will be able to quantify different polygons. So there'll be a tree tops and there'll be a crown that is surrounding these tree tops. So that's how we get the individual tree crowns. We also have the location of the trees where we have the fecundity estimation from the masking inference and focusing model. And then we link those tree fecundity estimate with the crowns. So that's why I can see some of the trees has no prediction of fecundity, because they are simply small, not big enough to produce any seeds, or there's a mismatch between the crown and maybe there's a coordinate error that cause a mismatch between observed fecundity from masking model and also the tree crowns from AOP LiDAR. So that's the approach we use to match those two data sets. And then, so yeah, that's how, yeah, I think that answers the question, the tree level froze with LiDAR data at the outside. I could take this moment to ask one because I was going to ask about fecundity. Do you know why so many species when they have higher foliar phosphorus, there's the negative trade-off with the country that seems interesting and a bit surprising. What do you think is going on there? Yes, that's a good question. So I can only speculate on those because I haven't done any mechanism modeling on those approach. So this slide summarized two happenrences. So if we, so in like in horticultural applications, if you apply fertilizer, so it can potentially increase the crop yield in the agricultural field. So one of the, especially potassium can increase the crop yield. So however, if you apply too much fertilizer, so the crop will just grow crazy and there'll be lots of vegetation growth instead of fecundity instead of a crop yield. So basically, there's a trade-off between reproduction and growth. So basically, we found where basically we are exploring the coefficient of fecundity versus foliar nitrogen and phosphorus. So it's actually not that causation between those two. So we are not sure why, we just observe a negative relationship between phosphorus and fecundity. We are not sure if just the increase in phosphorus can reduce fecundity. It's just if you increase phosphorus can introduce lots of growth instead of reproduction. So there's the potential trade-off between reproduction and growth, by the way, I'm not sure if those signals are pointing to those directions because that's not causation. It's just the simple coefficients of the response to phosphorus. Right. That's so interesting. Maybe Mastiff needs to do some targeted fertilization experiments or something. Try that. Anyway, thank you for that. Yeah, thank you very much. Looks like there's one more question from Hanxi Chen. Could you share your experience on how to set up the hypothesis reasonably and how you chose the related variables regarding the biodiversity research? Oh yeah, that's also a good question. So I would suggest to read the literature to find the yearly which variables I use to model biodiversity change. And you know, it's hard to find literature that focus on food supply because it's still a, even though it's a fundamental rule of biodiversity, but it's still, there's no information on how much seeds are tree produced in a given year. So that's why we are trying to incorporate the food supply information. But there are many variables that are important in regulating biodiversity change, such as the one I mentioned earlier, the canopy height models, sorry, the industry density canopy heights and terrain roughness. And when you are trying to quantify which variables are important for biodiversity change, try to think about the mechanism that controls. Sorry, let me go back to the slides. So like for the interaction between gaps and the temperature, there's two potential mechanisms. So like one, this is from a traditional view of a micro climatologist. So basically larger gaps might amplify the impacts of climate warming. And then another hypothesis will be for traditional ecologists, so they can provide local refugees. So my recommendation will be just to look at the literature that try to provide as many hypotheses as possible. Okay. There's also one question in the chat. I'm not sure. Is that is that my chat or Yeah, no goal. I was going to point that out that someone was asking about sharing software and sharing code. Tom, do you want to, I'm sure some of the work that's published, maybe it's already publicly available or what would you say on, did you say that? Yes, there's a GitHub report that is associated with Gromitos modeling. It's hosted by the ecological focus initiative. So yeah, if you look at that GitHub, you'll be able to find the code. And as for the, so that's for the industry, as for the nutrients. So Phil Thompson has built this really amazing efforts to model continental scale prediction of nutrients. Their data is also publicly available. So you can download their equations and then apply the equation to the Neon AOP data and get the nutrients variables. Send me an email if you couldn't find the link. And the graphs, what's the meaning of the graphs? Oh, you mean the draw the graphs of the figure? I think that's also in the EFI GitHub repository. Okay, I hope that answers the question. Thanks, so well, unless there's any other questions, this seems like a good time to wrap up. Thanks again to our speaker for a fantastic talk. A lot of thought provoking results shown. We're so excited to see all the exciting things you're doing with Neon Data. And for everyone else, please consider joining us. We've got two upcoming events. We have our first data skills webinar is going to be September 26. You can sign up for that on the science seminars webpage. And our next seminar, speaking of FE, is going to be on the ecological forecasting and the Neon Forecasting Challenge October 10. So please consider joining for that. Until then, take care. Thanks so much.