FluxLetter

The Newsletter of FLUXNET Vol. 1 No. 2 May, 2008

Highlight FLUXNET site Experimental Site of Puechabon, South of France Puechabon by Laurent Misson

History- In 1984, pioneer system that could extract water attracted many researchers for researchers François Romane at more than 5-m depth. Quercus decades. In This Issue: started long-term ecological ilex is a strong terpenoid emitter, First, studies at Puéchabon studies in a Quercus ilex with VOC production account- focused on forest structure, next to the village of Puéchabon ing for 1 to 2% of GPP. This biogeochemical cycles and post- near Montpellier, south of species covers millions hectares disturbance recovery. In the FLUXNET site “Puechabon” Laurent Misson……….….Page 1-3

Editorial: FLUXNET, Quo vadis Dennis Baldocchi ………...Page 4-5

What can dense Ameriflux site clusters say about spatial and temporal heterogeneity in carbon and water cycling? Photo 1: General view of the experimental site of Puechabon Ankur Desai…………..….Page 5-6 early nineties, with questions

France (Photo 1). Quercus ilex around the Mediterranean sea. concerning global change arising, FLUXNET graduate student have been long ago con- As such it’s a fascinating research detailed functional studies Antje Moffat...... …..Page 7 sidered as a paradigm for Medi- subject and the functioning of started to understand the effects terranean growing Quercus ilex ecosystems have of climate on the vulnerability of FLUXNET young scientist on hard limestone karstic . Hans Verbeeck...... …Page 8 It’s a highly adapted species to unpredictable environments Integrating airborne lidar with characterized by long summer covariance and beyond: droughts, storm events acting as new research within the Cana- resource pulses, and strong and dian Carbon Program frequent disturbances such as Laura Chasmer and Valerie Tho- fire. Quercus ilex life history- mas……………………..Page 9-13 strategies include some of the most prominent found in such ecosystems: it’s an evergreen Keeping an eye on the carbon species with thick sclerophylous balance: linking canopy devel- leaves; it has a diffuse-porous opment and net wood of high density providing exchange using a webcam resistance to cavitation; and it’s a network resprouter allocating vast Lisa Wingate et al….….Page 14-17 amount of reserve carbohy- drates to an overdeveloped root Photo 2: Flux tower at Puechabon FLUXNET SITE cont. on page 2 Page 2

Puechabon...south of France FLUXNET SITE cont. from page 1

Mediterranean forests. With the Infrastructure- The infra- the forest acted as a net carbon 1997 Kyoto protocol, European structures at Puéchabon include sink of -250 g C m-2 yr-1. Ex- “The database includes countries have been committed several experiments. First, an treme events such as spring long-term ecological data to report precise carbon ac- eddy-flux tower and two mete- droughts (2005, 2006), insect- on forest growth, above-

counting. Since 1998, Puéchabon orological stations record CO2, induced canopy defoliation ground and belowground is one of the reference sites of water and energy fluxes between (2005), and in a smaller propor- biomass, regeneration, the European network for meas- litterfall, functional traits urements of carbon and energy exchanges between the atmos- and phenology ” phere and the land surface, through the Medeflux (1998- 1999), Carboeuroflux (2000- designed to simulate the effect 2004), and Carboeurope-IP of an extreme climatic event on (2004-2009) projects. In 2003, the functioning and the vulner- several manipulative experiments ability of this ecosystem. A started with the MIND project rainfall shelter was installed financed by the European Union. above the canopy in order to Rain exclusion and thinning were simulate extreme drought sea- applied to study the effects of sonally (Photo 5). The rainfall changing precipitation amounts shelter is mobile and move on and management practices on two 60-m rails that are 15-m forest- carbon and apart. Four plots are defined: an water exchange. Theses experi- Figure 1: Relation between ecosystem annual NEE sum and rainfall during spring (March early drought plot on the south ments culminate in 2007 with a to June) side (spring drought) and, a late new project testing the effects of drought plot on the north side extreme seasonal drought in the atmosphere and the forest tion the 2003 heat-wave, have (fall drought), a standby plot in spring and fall on the vulnerabil- continuously since July 1998 the effects of greatly reducing the middle, and a control plot ity of Mediterranean forest eco- (Photo 2). A new paper reports this sink capacity (Fig. 1). nearby. The roof will stay over systems (Drought+ project, analyzes of seasonal and annual Second, a series of long-term the standby plot when it is not French National Research variation of carbon exchange continuous manipulative experi- raining to avoid disturbing the Agency). (Allard et al. 2008). On average ments started in 2003. These included 4 treatments: a con- trol, a throughfall exclusion, a thinning, and a throughfall ex- clusion x thinning treatment. Throughfall exclusion was achieved using gutters under the canopy, with the aim to exclude 30% of the throughfall continuously (Limousin et al. 2008) (Photo 3). Thinning re- duced the basal area by half. Scaffolds allow researchers to have access to two levels in the canopy for ecophysiological measurements of sun and Photo 4: Second level of the scaffold in shaded leaves (Photo 4). the continuous rain exclusion experiment Third, in 2007 a new manipu-

Photo 3: Continuous rain exclusion experiment: control and dry plot lative experiment has been FLUXNET SITE cont. on page 3 Page 3

Puechabon...south of France FLUXNET SITE cont. from page 2

Photo 5: Mobile rainfall shelter for simulation of seasonal drought (DROUGHT + project)

micrometeorology of the water cycle include all the main have been measured discontinu- drought plots. fluxes, the top water con- Literature ously since the 1990’s. Data- The database at tent and discrete measurements Allard V, Ourcival J-M, Rambal S, Joffre R and Rocheteau A. 2008. Seasonal Puéchabon is huge and goes back of soil water storage across 5-m For further information see: and annual variation of carbon ex- to the 1980’s. It includes long- depth since the 1990’s. Dry and change in an evergreen Mediterranean http://www.cefe.cnrs.fr/fe/ forest in southern France. Global term ecological data on forest wet nitrogen deposition started Change , 14: 1-12. growth, aboveground and below- in 2007 as a companion site of puechabon/index.htm Granier, A. et al. 2007. Evidence for ground biomass, regeneration, the Nitroeurope UE project. soil water control on carbon and litterfall, functional traits and Organs, soil and litter contact: water dynamics in European forests Serge Rambal during the extremely dry year: 2003. phenology. Data on the main and biochemistry have been Agricultural and Forest . biogeochemical cycles include described semi-continuously. (Carboeurope, Mind projects) 143:123-145. [email protected] eddy-covariance CO2 fluxes, soil Detailed ecophysiological data Rambal S, Joffre R, Ourcival J-M, and organ-level gas exchange and organ level VOC emission Cavender-Bares J and Rocheteau A. data. Measurements for the Laurent Misson 2004. The growth respiration compo- nent in eddy CO2 flux from a Quercus (Drought+ project) ilex mediterranean forest. Global [email protected] Change Biology. 10:1460-1469.

Joffre R, Ourcival J-M, Rambal S and Rocheteau A. 2003. The key-role of topsoil moisture on CO2 efflux from a team: Mediterranean Quercus ilex forest. Serge Rambal Annals of Forest Science. 60:519-526.

Richard Joffre Rambal S, Ourcival J-M, Joffre R, Jean-Marc Ourcival Mouillot F, Nouvellon Y, Reichstein M, Rocheteau A. 2003. Drought controls Michael Staudt over conductance and assimilation of a Florent Mouillot Mediterranean evergreen ecosystem: scaling from leaf to canopy. Global Laurent Misson Change Biology. 9:1813-1824. Alain Rocheteau

Jean-Pierre Ratte Hervé Bohbot

Raquel Rodriguez

Anne-Violette Lavoir Jean-Marc Limousin

Rita Giuliani

Christian Collin David Degueldre Violette Sarda

Photo 6: Puechabon team Page 4

Editorial: FLUXNET, Quo vadis Dennis Baldocchi

At this moment 510 sites are unprecedented dataset and a The long-term data records and $25M to $75M per year, if we registered on the FLUXNET large body of literature. To- future flux measurements will assume that the cost of each site database. In planning for the gether, they contain many excit- continue to be necessary for ranges between $50k and $150k future we must ask and assess ing findings about the biophysical validating land-surface schemes per year, based on salaries, in- what will FLUXNET look like in controls of carbon fluxes at daily in models that address problems strumentation, travel and dis- 3 to 5 years? For example, will to annual time scales, how car- associated with the climate, counting for teams who run growth continue, as we attempt bon fluxes differ across ecosys- ecosystems and vegetation dy- multiple sites. The cost of this to fill holes in the network in key tems and climate gradients, their namics and biogeochemical cy- global flux network is small if regions like the Arctic tundra, responses to disturbance and cles (water, carbon). Flux and one considers the value that India, Africa and Mexico, or will their interannual variability. meteorological data are also information on carbon cycling the network contract as funding Consequently, we face the added needed by those interested in will provide to emerging carbon becomes tighter and many of the burden of convincing peer- data assimilation and model trading markets and towards original research questions get review panels that continued parameter inversion schemes efforts to reduce carbon emis- answered? operation of a long-term flux and interpreting remote sensing sions and manage ecosystems for The continued operation of a measurement site will produce indices. And in the future, I carbon sequestration. For in- global network, like FLUXNET, more discoveries, rather than expect carbon flux data will be stance, to stabilize and reverse lies on the foundation of diverse produce incremental findings. used by ecosystem managers, greenhouse warming, efforts will sources of funding. Inevitably, Convincing arguments and ra- carbon traders and policy mak- be needed to reduce carbon like Death and Taxes, regional tionales are needed to ensure ers. At present these user emission 80% of current levels networks and independent re- continued funding and to help communities do not bear the by 2050, if not sooner. So a $25 search teams must periodically set the agenda for future re- cost of the network, but their - 75M annual investment in a submit renewal proposals that search priorities. Foremost, we activities will suffer if groups of global economy, dependent on are subject to peer-evaluation. can't lose sight of the fact that energy from fossil fuels and

Unfortunately, there is no guar- we are measuring how the worth 10’s of trillions of dollars, antee of future and continued Earth’s terrestrial is “Convincing arguments should be justified to better support for particular tower responding to an unprecedented and rationales are manage carbon sinks and sites, or regional networks. change in climate and land use. needed to ensure contin- sources. To sustain funding of regional This fact alone provides us with ued funding and to help This leads me to ask the ques- networks and independently a strong opportunity, motivation set the agenda for future tion ‘how big the network funded sites there are a variety and responsibility to continue should be and how long individ- of science and funding issues we making long term flux measure- research priorities ” ual sites should operate?’ The need to recognize and over- ments across the globe. As data answer depends on the scientific come. At present, funding records get longer, we will soon questions being asked of the across the environmental sci- have the opportunity to study towers quit operating. At best, network. Different configura- ences is highly competitive how carbon fluxes change as these stakeholder communities tions of the global and regional (success rates among many agen- ecosystems undergo natural should be encouraged to lobby networks provide different ser- cies in the U.S. ranges between succession. To emphasize this on our behalf for extending the vices and sets of information. A 10 and 20%) and research priori- point, I submit evidence that flux measurement record into small group (50 to 100) of long- ties change. No doubt, the trends in carbon fluxes are the future. term and intensive field sites proposal/review process and emerging from decade plus re- The investment in this global flux spread across the globe, like competition is good because it cords at Harvard forest, which is network may seem expensive, Hyytiala and Harvard Forest, forces us to re-evaluate our recovering from disturbance in but it is comparable to many may prove to be adequate senti- science and strive for evolving the late 1930s (Urbanski et al., projects in and it is nels for assessing the impact of and better projects that reflect 2007). very cost effective considering global change across a diverse evolving research priorities. But It is also noteworthy that the the value of carbon flux data to set of ecosystems. Conversely, a in competing for new funding we FLUXNET community continues many broad economic sectors larger network of cheaper and must be conscious of and avoid to produce a treasure trove of and policy. The investment in less intensive sites (300-500) becoming ‘victims of our suc- information to a wide number FLUXNET represents a yearly may be needed to produce sta- cess’. We have produced an and variety of stake holders. investment, globally, of about tistical models that are being

cont. on page 5 Page 5

FLUXNET, Quo vadis cont. from page 4 used in conjunction with remote lations (Hember and Lafleur, against science that is not directly Literature sensing data, neural networks, 2008; Magnani et al., 2007). aimed at hypothesis testing. At a inverse modeling and data as- This editorial has been moti- time when the planet is being pro- Ciais, P. et al., 2005. Europe-wide reduction in primary productivity similation schemes to perform vated because I suspect we may pelled by human action into an- caused by the heat and drought in 2003. continental and regional integra- be near a tipping-point regarding other climate regime with incalcula- Nature, 437(7058): 529-533. tions. A large network is useful sustained funding for the contin- ble social and environmental costs, Gu, L. et al., 2008. The 2007 Eastern US for capturing low probability ued operation of components of we cannot afford such a rigid view Spring Freeze: Increased Cold Damage in a Warming . BioScience, 58: climate and disturbance events the flux network. Consequently, of the scientific enterprise. The only 253–262. like regional droughts, we will need to work together way to figure out what is happening Hember, R.A. and Lafleur, P.M., 2008. storms, fires, and pest out- and provide decision makers, to our planet is to measure it, and Effects of serial dependence and large- breaks. An example of this func- potential funders, reviewers and this means tracking changes dec- scale tropospheric circulation on mid- latitude North American terrestrial tion has recently been demon- future users of the data with ade after decade and poring over exchange. Journal of strated by case studies associ- compelling reasons to keep the records. A point of diminishing Climate, 21(4): 751-770. ated with the 2003 heat spell and these networks and sites funded scientific returns has never been Keeling, C.D., 1998. Rewards and drought across Europe (Ciais et and operating. This is a lesson I realized in what is now known as penalties of monitoring the Earth. Annual Review of Energy & the Environ- al., 2005) and the late spring have learned from reading about the "Keeling Curve," the Mauna ment, 23(1): 25. frost across North America in the trials and tribulations of Loa CO2 record’. (Keeling, 2008) Keeling, R.F., 2008. ATMOSPHERIC 2007 (Gu et al., 2008). Data David Keeling (Keeling, 1998). SCIENCE: Recording Earth's Vital Signs. records on the order of a dec- He had to struggle continually to This sentiment is true for flux Science, 319(5871):1771-1772 ade or two will be needed to keep the Mauna Loa carbon networks too. Magnani, F. et al., 2007. The human detect if trends in fluxes are dioxide observation site running, footprint in the carbon cycle of temper- occurring and determine a point re-iterated in a recent FluxLetter invites the readers ate and boreal forests. Nature, 447 to contribute Editorials on (7146): 848-850. whether they are due to global issue of Science by his son, Ralph: emerging science and policy , natural succes- Urbanski, S. et al., 2007. Factors con- issues of interest to the wider trolling CO2 exchange on timescales sion or large-scale climatic oscil- ‘A continuing challenge to long-term community. from hourly to decadal at Harvard Earth observations is the prejudice Forest. J. Geophys. Res., 112: 10.1029/2006JG000293.

What can dense Ameriflux site clusters say about spatial and temporal heterogeneity in carbon and water cycling? Ankur Desai

Carbon and water cycling in parameter tuning/optimization, cover generated by fine-scale An alternative approach is to northern forests is spatially and biometric measurements, tran- topographic and hydrologic gra- promote interested parties temporally heterogeneous. Is spiration variability, and bound- dients, and a history of intensive through consortium develop- there any hope that we could ary layer dynamics? The answers forest management in an other- ment. The Chequamegon Eco- observe, explain, and model this are yes, maybe, and read the wise sparsely populated area. system-Atmosphere Study spatial and temporal variation in articles in a 2008 special issue of Significant climate change is pre- (ChEAS) was initiated with the light of understanding impacts of Agricultural and Forest Meteor- dicted for the region. Given the latter approach because of future climatic change and hu- ology to find out. complex interaction of water common interests from several man disturbances on regional One of the densest cluster or and carbon cycles, biogeochemi- groups, including Ken Davis’ lab ? Could high of Ameriflux sites is cal impacts from climate change at Pennsylvania State University, density observational coverage located in the northern portions are poorly constrained. Scott Denning’s Biocycle re- and high fidelity modeling of of Wisconsin and Michigan in Organizing a large research team search group at the Colorado ecosystems across such a land- temperate/sub-boreal forested to understand regional response State University, Paul Bolstad’s scape help? What do these ob- landscapes. This is a region typi- to climate change has been tried forestry and remote sensing lab servations and models say about cal of recently glaciated sub- (e.g., the BOREAS, MBL, at the University of Minnesota, the role of ecosystem stand age, boreal landscapes, with abundant NEESPI), when funding is avail- ecohydrology labs of Scott moisture sensitivity, phenology, wetlands, heterogeneous land able, with many lessons learned. Mackay at SUNY-Buffalo and

cont. on page 6 Page 6

Ameriflux site clusters cont. from page 5

Brent Ewers at University of , land cover assess- sue" (Volume 148, Issue 2, 13 several papers is the complexity Wyoming, micrometeorology ment, soil respiration, and at- February 2008). A review edito- that forest disturbance, stand (Ankur Desai lab) and forest mospheric chemistry. These rial in that issue (Chen et al., age, and wetlands impart on groups (Tom Gower efforts involve over a dozen 2008) provides an overview of spatial scaling, interannual vari- lab, Mark Schwartz lab) at Uni- Principle Investigators and doz- the research, needs and contri- ability, and climate sensitivity of versity of Wisconsin, Jiquan ens of post-docs, graduate stu- bution, and future work. ecosystem water and carbon Chen’s Landscape and Ecosys- dents, and field technicians Manuscripts in this issue include cycling. Other studies, however, tem Science Lab at the Univer- across many University and labs. those that focus on note that multi-year, regional sity of Toledo, Joe Berry’s lab at Over time, our lively and infor- 1) Quantifying ecological compo- scale investigations from multiple the Carnegie Institute of Wash- mal group, closely tied by many nent flux budgets by Gough et al. ecological and atmospheric per- ington at Stanford, Paul joint research agreements and and Tang et al. spectives that employ advanced Wennberg’s atmospheric chem- the annual ChEAS meetings and 2) Observing and modeling atmos- model-data fusion have great istry lab at CalTech, the Land workshops, have generated pheric surface and boundary layer potential to improve how we Cover and Land Cover Dynam- extremely productive collabora- properties by Denning et al. and Su understand this complexity. ics lab at Boston University, the tions leading to numerous publi- et al. One of the biggest challenges for Numerical Terradynamic Simu- the ChEAS group and other lation Group at University of similar groups across the Fluxnet Montana, the University of domain is to effectively continue Wisconsin Kemp Natural Re- our collaboration as long-term sources Station, the University data and lessons from this con- of Michigan Biological Station, sortium will be non-additively biogeochemical and micromete- increased. Toward this goal, orological research groups at firm commitments from both the National Center for Atmos- researchers and foundations are pheric Research, the NOAA needed, including continued and Earth Systems Research Lab increased funding for long-term Carbon Cycle and Greenhouse ecological observation and analy- Gas group, and the U.S. Forest sis (e.g., NSF LTREB), collabora- Service Northern Research tor networks (e.g., NSF RCN), Station, among several others. and cross-disciplinary investiga- Together, these groups have tions. ChEAS and the related been measuring land- Figure 1: Participants at the 8th annual ChEAS meeting held in June 2005 at the University groups intend to continue their atmosphere carbon and water of Wisconsin, Kemp Natural Resources Station, Woodruff, WI. (Photo credit: Qinglin Li, fruitful collaborations and look University of Toledo) exchange in the upper Midwest forward to extending to cross- for over a decade beginning cations, presentations, and re- 3) Investigating ecohydrological group and Fluxnet-wide coop- with the initiation of the Park search results funded by NSF, and carbon-water responses of eration as we enter an era of Falls, WI WLEF-television DOE, NASA, NOAA, USDA, ecosystems by Ewers et al. and analyzing the entire Ameriflux transmitter 447-m tall tower, a and several other groups. Noormets et al. and Fluxnet database. NOAA carbon cycle gases tall Twelve of these publications 4) Modeling ecosystem carbon and tower observatory and one of were recently jointly published in water cycles by Ryu et al. and Sun contact: the earliest 2008 in a special issue of Agricul- et al. Ankur R Desai flux towers in the Ameriflux tural and Forest Meteorology 5) Applying observational data [email protected] network. Since then, the groups titled "Chequamegon Ecosystem- assimilation for parameter optimi- http://flux.aos.wisc.wedu have erected nearly a dozen Atmosphere Study Special Issue: zation in models at multiple time- other flux towers with several Ecosystem-Atmosphere Carbon scales by Prihodko et al. and Ricci- Literature

still in operation along with and Water Cycling in the Tem- uto et al. many related observing sites for perate Northern Forests of the 6) Upscaling flux tower observa- Chen, J., Davis, K.J., and Meyers, T.P., 2008. Ecosystem–atmosphere carbon transpiration, biometric meas- Great Lakes Region - Great tions by Desai et al. and water cycling in the upper Great urements, ecophysiology, eco- Lakes Region Special Is- One theme that arose out of Lakes Region, Agricultural and Forest Meteorology, 148(2): 155-157. Page 7

Highlight Graduate Student Antje Maria Moffat

Nature has always been my data time series as a necessary network, I was glad to receive tractor at night. fascination, and I love hiking, first step. At the Gap Filling and funding from the Max Planck Currently, I am using neural canoeing or just being in the Partitioning Workshop in Society to organize a workshop; networks to explore the sensi- outdoors. Growing up, I became Viterbo, Italy, Dario Papale, it is so much easier and more tivity of ecosystem carbon fluxes increasingly aware how fragile Markus Reichstein and I initiated effective to have round table to climatic controls. This is really natural ecosystems are, and this a gap filling comparison which I discussions. The Gap Filling exciting because it employs a awoke a desire to understand took the lead on and extended Comparison Workshop was held purely mathematical algorithm to and protect them. But when I from the Carboeurope to the in Jena in 2006 and it was very capture the "real" ecosystem pursued my studies in , I Fluxnet community. positive that nearly all members processes and connects the specialized in surface science and Collaborating with almost of the comparison could come. model world back to plant physi- took an opportunity to build a twenty researchers meant to The successful exchange not ology. This year will be the last only helped finalize the gap filling year of my Ph.D. and I hope to comparison paper (Moffat et al., be able to continue research in 2007) but also led to further in the collaboration resulting in three future. companion papers (Desai et al., 2008; Richardson and Hollinger, contact: Antje M. Moffat

2007; Richardson et al., 2008). [email protected] The main challenge for me has http://www.bgc-jena.mpg.de/ been to find a good balance ~antje.moffat/ between my work and my family. My daughter Anna is now five Literature

years old. She loves to play out- Moffat, A.M., Papale, D., Reichstein, M., side, creating with sticks Hollinger, D.Y., Richardson, A.D., Barr, A.G., Beckstein, C., Braswell, B.H., and strings and always collecting Churkina, G., R., D.A., Falge, E., Gove, treasures like empty snail shells J.H., Heimann, M., Hui, D., Jarvis, A.J., Figure 1: Antje with her kids Leo and Anna have twenty valuable inputs but or glittering stones. My son Leo Kattge, J., Noormets, A. and Stauch, V.J., 2007. Comprehensive comparison of scanning tunneling microscope as also twenty different (sometimes was born when the work on the gap-filling techniques for eddy covari- my Master’s thesis. Afterwards I opposing) opinions, and the gap filling comparison was at its ance net carbon fluxes. Agricultural and Forest Meteorology, 147: 209-232. started working my way up in major challenge was to collect, peak, and I spent the days with the semiconductor industry - organize, and synthesize every- him and the nights and weekends Desai, A.R., Richardson, A.D., Moffat, A.M., Kattge, J., Hollinger, D.Y., Barr, only to realize that this just was- thing into a coherent whole. working. He is now two years A.G., Falge, E., Noormets, A., Papale, n't what I wanted to do! So After doing this by email, with old and totally excited about D., Reichstein, M. and Stauch, V.J., 2008. Cross site evaluation of eddy covariance when my family and I moved to me as the central node in a star cars; he likes to sleep with his GPP and ER decomposition techniques. Jena, I took the chance to start Agricultural and Forest Meteorology, doi:10.1016/j.agformet.2007.11.012. an interdisciplinary Ph.D. in the Biogeochemical Systems Group Richardson, A.D. and Hollinger, D.Y., 2007. A method to estimate the addi- at the Max Planck Institute in tional uncertainty in gap-filled NEE 2003. resulting from long gaps in the CO2 flux record. Agricultural and Forest Meteor- My research topic is data as- ology, 147: 199-208

similation of eddy covariance Richardson, A.D., Mahecha, M., Falge, E., data with artificial neural net- Kattge, J., Moffat, A.M., Papale, D., Reichstein, M., Stauch, V.J., Braswell, works. However, since neural B.H., Churkina, G. and Hollinger, D.Y., network models can only get as 2008. Statistical properties of random CO2 flux measurement uncertainty good as the dataset used for inferred from model residuals. Agricul- training them, I soon focused on tural and Forest Meteorology, 148(1): 38-50. flux data processing, quality controls, and gap filling of these Figure 2: Antje’s “site”. Experimenting with artificial neural networks at her desk Page 8

Highlight Young Scientist Hans Verbeeck

My name is Hans Verbeeck. I (www.lsce.ipsl.fr). From the ecosystem models (e.g. Saleska would call myself an ‘ecosystem stand scale models of my PhD I et al., Science 2003). An unex- modeller’. I am one of the guys moved to the global model OR- pected high carbon uptake was “My goal is to optimise that uses other people’s data to CHIDEE. ORCHIDEE is a state measured during dry season, and the ORCHIDEE run a model … Besides that I of the art mechanistic global in contrast, carbon release was parameters using have been responsible for the vegetation model. It calculates observed in the wet season. FLUXNET data...using a eddy flux data collection at the the carbon, water and nitrogen In order to be able to mimic the Bayesian optimisation Brasschaat site (Belgium) for cycle in the different soil and seasonal response of carbon approach” three years. So I know how hard vegetation pools and resolves fluxes to dry/wet conditions in it is to collect reliable datasets the diurnal cycle of fluxes. OR- tropical ecosystems we want to without too many gaps… Brass- CHIDEE is built on the concept optimise the ORCHIDEE model chaat is a mixed coniferous- of plant functional types (PFT) to using eddy covariance data. By deciduous forest site with Scots describe vegetation. doing this, we will try to identify pine as dominant species in the I am using this model at site the underlying mechanism of this Literature fluxtower footprint. level, and my goal is to optimise seasonal response. Verbeeck H, Samson R, Granier A, the ORCHIDEE parameters Montpied P, Lemeur R. Multi-year using FLUXNET data. To do this, …and besides this information model analysis of GPP in a temperate beech forest in France. 2008. Ecological I am using a Bayesian optimisa- about my scientific activities, I Modelling, 210:85-103. tion approach (Santaren et al., can tell you very proudly that Verbeeck H, Steppe K, Nadezhdina N, 2007). Next month (June), I will since January I became father of Op de Beeck M, Deckmyn G, start working on my own Marie my first son called Jonas! Meiresonne L, Lemeur R, Čermák J, Ceulemans R, Janssens IA. 2007. Stored Curie project, called POLICE: water use and transpiration in Scots Parameter Optimisation of a contact: H.Verbeeck pine: a modeling analysis using ANA- FORE. Tree Physiology 27:1671-1685. terrestrial biosphere model to [email protected] Link processes to Inter annual Santaren D, Peylin P, Viovy N, Ciais P (2007) Optimizing a process-based variability of Carbon fluxes in ecosystem model with eddy-covariance European forest Ecosystems. flux measurements: A pine forest in During last year, I learned to use southern France. Global Biogeochemical Cycles, 21, GB2013. the ORCHIDEE model and the Figure 1: Hans Verbeeck with his son data assimilation system: OR- Saleska SR, Miller SD, Matross DM, et Jonas. al. (2003) Carbon in Amazon Forests: CHIS (Orchidee Inversion Sys- unexpected seasonal fluxes and distur- One year ago I finished my PhD tem). Recently, I used the OR- bance-induced losses. Science, 302, 1554-1557. at the Research Group of Plant CHIDEE model to conduct simu- and Vegetation Ecology, Univer- lations for several sites in the sity of Antwerp, Belgium Amazon in the framework of the (www.ua.ac.be/pleco). During my LBA-MIP project, which is a PhD I mainly focussed on model- model intercomparison project ling water and carbon fluxes in using data of the LBA sites temperate forests. I worked on (www.climatemodeling.org/lba- carbon fluxes at the Hesse beech mip/). The work on these tropi- site (France) (Verbeeck et al., cal sites drew our attention to a 2008) and on water storage in very interesting problem: eddy the Scots pines of the Brasschaat covariance measurements at site (Verbeeck et al., 2007). several tropical forest sites re- After finishing my PhD, I started vealed an unexpected seasonal Figure 2: View of fluxtower at Brasschaat as a post-doc at the “Laboratoire pattern in carbon fluxes which FLUXNET site des Sciences du Climate et de still can not be simulated by l’Environnement” close to Paris existing state-of-the-art global Page 9

Integrating Airborne Lidar with Eddy Covariance and Beyond: New Research within the Canadian Carbon Program Laura Chasmer and Valerie Thomas

The 3D structure of the canopy, chronosequence consists of a (nadir) and ±25o from nadir. cies type and age. Before lidar understory, and ground surface mature jack pine forest (OJP); an Airborne lidar is unique because, data were available, we could topography plays an important immature site harvested in 1975 unlike optical remote sensing gain some understanding of the role on the movement of mass (HJP75); a regenerating site har- methods, it can sample the canopy structural characteristics and energy through ecosystems. vested in 1994 (HJP94) and a structural properties of the and the health of vegetation Therefore, maps of canopy recently harvested site scarified canopy, as well as the under- using aerial photography or high structure both within and be- in 2002 (HJP02). The jack pine story and ground surface. An- resolution spectral remote sens- yond the source area of eddy chronosequence has been oper- other type of lidar that we also ing data. Unfortunately, these covariance (EC) systems will ating as part of the Boreal Eco- use is called a terrestrial or technologies are affected by improve our understanding of system Research and Monitoring “ground-based” lidar system. shadow and cannot characterize plant function, health, photosyn- Sites (BERMS) project. Other EC This works similarly to the air- the lower canopy. Alternatively, thesis, and transpiration. Air- sites within the Canadian Car- borne lidar but can be set up on we can spend weeks to months borne light detection and ranging bon Program that have had lidar a tripod and scans horizontally measuring the structural charac- (lidar) is a powerful new remote data collections include some into a forest plot. teristics of the forest using plot sensing tool used for quantifying sites in British Columbia, Que- measurements or transects. To the 3D structure of forest cano- bec, and roving sites in Ontario do this over large spatial areas is pies. Lidar is able to measure a (not discussed here). time consuming, expensive, and variety of ecosystem properties logistically prohibitive in remote at resolutions ranging from 0.35 What is Lidar? areas. Lidar, on the other hand, m to ~5 m. In Canada, we have Airborne lidar is an active re- is not influenced by solar radia- obtained airborne lidar data over mote sensing device that rapidly tion, and when surveys are opti- several EC sites that were part emits and receives discrete mally configured (e.g. Thomas et of the Fluxnet-Canada Research pulses of laser light (1064 nm) al. 2006; Chasmer et al. 2006b), Network (FCRN) and the cur- from an airborne platform to the they can provide exquisitely high rent Canadian Carbon Program ground surface (Figure 1). With resolution and accurate informa- (CCP). Here we provide a brief each laser pulse that is emitted, tion on all parts of the canopy, description of how lidar works the time of pulse emission to understory and ground surface, and why it has great potential for reception is recorded, along with providing us with a three dimen- understanding and quantifying the heading, pitch, and roll of the sional picture of the ecosystem canopy, understory, and topog- aircraft, and aircraft position. Figure 1: Optech Inc. ALTM 3100 lidar that was never before possible. used for surveys raphic influences on mass and Laser pulses may be emitted at The potential opportunities and energy exchanges. We also out- rates of 5 kHz to 160 kHz, and Impacts of Lidar Research new information that we can line some of the challenges and can record between one and and EC Integration glean by combining lidar, spectral limitations faced when using lidar four reflections (or “returns”) Many studies have shown that remote sensing data, and EC are

data, and we make suggestions from buildings, the ground sur- CO2, water, and energy fluxes almost endless! for getting the most out of your face, and within tree canopies. vary spatially and temporally What are some of the interest- lidar dataset. The examples dis- Some lidar systems can also within forested environments. ing things that we can do with cussed include data collected record the full waveform of all The variability of these mass and lidar data to better understand over: 1) a mature mixed-wood reflections within the canopy. A energy exchanges are due to a ecosystem variability and mass forest located near Timmins, scanning mirror is used to dis- myriad of environmental, ed- exchanges? Some examples in- Ontario (the Groundhog River tribute the emitted pulses across aphic, and vegetation drivers, clude: Flux Station (GRFS)); and 2) a the landscape in a saw-tooth including the spatial variability of 1. Mapping canopy physiology jack pine chronosequence lo- pattern, known as a “scan line”. forest structure and biomass, and foliar biochemistry, as re- cated north of Prince Albert, The angle at which pulses are previous history (e.g. distur- lated to species and health. For Saskatchewan. The jack pine distributed can vary between 0o bance), health, topography, spe- example, a combina- cont. on page 10 Page 10

Integrating Airborne Lidar with Eddy Covariance cont. from page 9

Figure 2: Examples of lidar point clouds within an individual tree, at the plot level, and within an ecosystem

tion of airborne lidar and hyper- budgeting by optimizing desirable ple, Chasmer et al. (accepted) (Thomas et al. 2006) and tempo- spectral remote sensing data species mixtures for carbon used lidar to model GPP at the ral change (Hopkinson et al. were used to predict and map uptake under a variety of envi- jack pine chronosequence. These 2008). foliar chlorophyll and carotenoid ronmental conditions. were compared with EC, and From these few research sugges- concentrations based on reflec- 5. Mapping variability in LAI and pixel level estimates of GPP from tions, one can begin to see many tance properties of the canopy the fraction of photosynthetically the Moderate Resolution Imaging research possibilities! We have (derived from hyperspectral active radiation absorbed by the Spectroradiometer (MODIS). not even mentioned the use of data) in combination with esti- canopy (fPAR) (e.g. Thomas et The influences of within-pixel lidar for forest hydrology, energy mates of canopy height and den- al. 2006b; Hopkinson and Chas- heterogeneity were then exam- balance modeling, evapotranspi- sity (derived from lidar data) mer, 2007). LAI and fPAR are ined by scaling lidar-derived GPP ration, snow distribution map- within a mixedwood stand important inputs used in many from the model at resolutions of ping, understory canopy map- (GRFS) (Thomas et al. 2008). ecosystem and hydrological 1 m to 25 m, 250 m, 500 m, and ping, laser return intensity as a Maps of canopy chlorophyll models. Imagine the possibilities 1000 m (Figure 4). potential for mapping ground generated with this approach for understanding ecosystem 7. Forest inventory assessment cover and surface soil moisture, reveal distinct spatial patterns processes if you combine eco- within the 1 km radius surround- system models with high resolu- ing the flux tower (Figure 3). tion spatial estimates of LAI! 2. Improving our understanding 6. Up-scaling of canopy informa-

of topography and CO2 drainage tion (such as LAI, fPAR, or gross on EC measurements in complex ecosystem production (GEP) environments. from high resolutions obtained 3. Examining the influence of from airborne lidar for compari- within site variability in canopy son and validation of low resolu- structure and topography on tion satellite products. This last

CO2 and water exchanges using component is required to bridge a footprint model of the source the gap between the local scale flux areas (e.g. Chasmer et al. of measurement possible at a 2007). flux tower and the large scale 4. Improving forest management predictions from continental- practices as related to carbon scale carbon models. For exam- Figure 3: Average leaf chlorophyll concentration at GRFS, August, 2004 (modified from Thomas et al 2008 cont. on page 11 Page 11

Integrating Airborne Lidar with Eddy Covariance cont. from page 10

and the influences of urban for- survey will cover their person- same or less than sending two or sensing data (e.g. Figure 2). Spe- ests on greenhouse gases. nel, the cost of the equipment, more students or scientists out cial software (e.g. Bentley Micro- the cost of the airplane, and will for a few weeks of field meas- Station, Terrascan, and Merrick The Challenges provide some profit to the com- urements. Organizations that MARS, etc.) may be required to With all of the great things that pany. Depending on the survey have publicly funded lidar sys- do some of the classifications of we can use lidar for, there must parameterization, a lidar survey tems include the AGRG and the the point data (e.g. separating be challenges with using the data. could set you back a minimum of Canadian Consortium for Lidar out ground returns from non- But what are these challenges? 10,000 US$ per hour of flying Environmental Applications Re- ground returns, and conversions First of all, lidar data can be time. That would cover one or search (used to collect lidar data from LAS binary format to AS- expensive to obtain because the perhaps two EC sites (local to at a number of CCP sites: Hop- CII). Typical remote sensing lidar itself is an expensive piece each other) and their surround- kinson, et al. http:// software packages such as Ar- of equipment. Lidar systems can ing source areas (excluding tall agrg.cogs.nscc.ca); the National cGIS, ENVI, QT Modeller, and cost between 500,000 and 2 towers), but would require an Center for Airborne Laser Map- Surfer, and coding environments million US$, so any service pro- airport nearby. Many have sug- ping in the USA (NCALM; (e.g. IDL, C, etc.) may also be vider will want to ensure that gested that the cost of a lidar www.ncalm.ufl.edu); and the used for analysis following classi- the amount charged for the data acquisition is about the National Environmental Re- fication. As lidar data becomes search Council in the UK more popular, software is adapt- (NERC; http:// ing, making data much easier to www.neodc.rl.ac.uk). These work with. All in all, we have groups usually work collabora- seen large reductions in the tively on research projects and costs of lidar data acquisitions, may charge less for lidar data computing resources, and soft- than industry rates if the areas ware making lidar data more surveyed are small and funding is accessible than it was ten years limited (so as not to compete ago. with industry). Another challenge is that lidar Getting the Most Out of data can be difficult to work Your Lidar Dataset with. The data volumes can be In all aspects of lidar data plan- very large, from tens of gigabytes ning, collection, and analysis, it is to terabytes, often requiring very important to first understand good computers and hard drive how your data can be or were space to store and work with collected. This can have an im- lidar data. The expense of good portant impact on how the laser computing resources is often not returns are distributed within considered, and in some cases the canopy. A few things to lidar data are “shelved” because consider are: 1. Lidar systems the users do not have the re- can obtain multiple returns in sources. Despite these issues, forested canopies, but can only those who process high fre- obtain one return at distances of quency EC data probably have less than approximately 1.5 m. the computing resources to This is important because lidar work with lidar data. will be unable to record a return Lidar data can also be difficult to from within short vegetation and work with because the laser the ground surface (Hopkinson return data are comprised of an et al. 2005). Therefore if you Figure 4: Influences of pixel scaling on lidar modeled GPP within MODIS pixel areas. Lower resolution pixels (e.g. 25 m) were subtracted from 1 m pixels at each site, assuming that 1 m “irregular” network of points were interested in knowing the resolution is most accurate. Positive differences indicate that lower resolutions underesti- which are unlike typical remote height of grasses mate GPP compared with 1 m resolution and vice versa for negative differences. cont. on page 12 Page 12

Integrating Airborne Lidar with Eddy Covariance cont. from page 11

surrounding your EC system, infrared radiation of laser pulses of canopy structure from lidar

this might be difficult, but the resulting in lowered laser pulse also reduce time and costs asso- Lidar and Remote Sensing Publica- height of trees > 1.5 m (or energy return (or intensity) or ciated with in situ validation, and tions at BERMS and GRFS

shorter trees with large spacing no return at all from the ground in most cases, lidar data can be Chasmer, L., McCaughey, H., Treitz, P., between them)? No problem! 2. surface (in areas of standing available within a few days to a Barr, A., Black, A., Hopkinson, C., Shashkov, A. (accepted). Scaling and If you have any say in your lidar water). month of the survey (depending assessment of GPP from MODIS using a data collection and planning, ask on the size of the area flown). combination of airborne lidar and eddy covariance measurements over jack for 50% overlap of scan lines. pine forests. Remote Sensing of Envi- This ensures that both sides of ronment.

each individual tree are ade- Chasmer, L., Barr, A., Black, A., Hopkin- quately measured by the laser son, C., Kljun, N., McCaughey, H., Treitz, P., 2007. Using airborne lidar for returns and minimizes the assessment of canopy structure “shadowing” or blocking of re- influences on CO2 fluxes. Peer reviewed proceedings publication: ISPRS Laser turns on the sides of trees that Scanning 2007 and Silvilaser. September are not facing the direction of 12 – 14. Espoo, Finland.

travel of laser pulses. 3. Flying Thomas, V., Treitz, P., McCaughey, J.H., lower (e.g. 1000 m as opposed Noland, T., Rich, L. 2008. Canopy chlorophyll concentration estimation to 3000 m) and setting the scan using hyperspectral and lidar data for a angle to around ±13 to 18o on boreal mixedwood forest in northern either side of “nadir” will typi- Ontario, Canada. International Journal of Remote Sensing, 29(4), 1029-1052. cally ensure even coverage of returns from the top of the Thomas, V., Treitz, P., McCaughey, J.H., Morrison, I., 2006. Mapping stand-level canopy to the base and many forest biophysical variables for a mixed- returns from the ground surface wood boreal forest using lidar: an examination of scanning density. Cana- (in open canopies). Increasing Figure 5: Laura Chasmer, Chris Hopkinson (AGRG) and John Barlow (U. of Saskatche- dian Journal of Forest Research, Vol. 36, wan) working on the AGRG lidar system during surveys in the sub-arctic, summer o 34-47. the scan angle to > ±20 will 2007

reduce the time of your survey, Thomas, V., Finch, D.A., McCaughey, but will also shift pulses towards The Take Home Message: The software tools and comput- J.H., Noland, T., Rich, L., Treitz, P., 2006b. Spatial modelling of the fraction the top of the canopy. This is Lidar has become an exciting ing resources required to exam- of photosynthetically active radiation beneficial if you are interested in and relatively new tool for the ine lidar data are also becoming absorbed by a boreal mixedwood forest using a lidar-hyperspectral approach. canopy height, but will result in assessment of high resolution, less expensive and increasingly Agricultural and Forest Meteorology, some lack of data within the 3D canopy structural character- user-friendly. Finally, there are a Vol. 140, 287-307.

centre of the tree crowns. If you istics. The research that has and number of online resources and are more interested in ground will continue to be generated short courses that have been

surface topography (e.g. for from the integration of lidar and made available so that users can References and Other Good Re- sources hydrological modeling), lower EC has many important implica- efficiently understand and work scan angles are best (e.g. < ±12o) tions in several areas of carbon with their lidar datasets. We Chasmer, L., Hopkinson, C., Treitz, P., 2006. Investigating laser pulse penetra- because the laser may be able to cycle science and forest manage- hope that lidar will continue to tion of a conifer canopy through the penetrate through grasses to the ment. These include forest in- make headway into carbon cycle integration of airborne and terrestrial ground surface. A review of all ventory assessment, mapping science, and will be used increas- lidar. Canadian Journal of Remote Sensing. 32(2), 116-125. survey parameterization influ- canopy physiology and biochemi- ingly as a linkage between the ences on the estimation of vege- cal constituents, improving for- spatial characteristics of the Chasmer, L., Hopkinson, C., Smith, B., Treitz, P., 2006b. Examining the influ- tated canopy properties is pro- est management practices, un- vegetated landscape and meas- ence of changing laser pulse repetition vided in Hopkinson (2007). Fi- derstanding the influence of urements made by EC and mete- frequencies on conifer forest canopy returns. Photogrammetric Engineering nally, it is best to plan lidar sur- canopy and understory structure orological . and Remote Sensing. 17(12), 1359-1367. veys after a period of dry on disturbance, forest hydrology, Hopkinson, C., Chasmer, L., Hall, R., 2008. Using airborne lidar to examine weather. Standing water and scaling, and satellite model vali- forest growth. Remote Sensing of very wet soils absorb the near dation. Accurate measurements Environment. 112(3), 1168-1180.

cont. on page 13 Page 13

Integrating Airborne Lidar with Eddy Covariance cont. from page 12

Hopkinson, C., Pietroniro, A., Pomeroy, ning LiDAR. Canadian Journal of Re- using airborne scanning laser in a boreal J., (eds.) 2008. Hydroscan: Airborne mote Sensing. 31(2), 191-206. nature reserve. Remote Sensing of Laser Mapping of Hydrological Features Environment. 76, 298-309. Advisors, site PI’s, and col- and Resources. Environment Canada laborators: H. McCaughey, P. and the Canadian Water Resources Lefsky, M.A., Cohen, W.B., Acker, S.A., Omasa, K., Hosoi, F., Konishi, A., 2007. Treitz, T.A. Black, A. Barr, C. Association. ISBN 978-1-896513-36-2. Parker, G.G., Spies, T.A., Harding, D., 3D lidar imaging for detecting and Hopkinson, and N. Kljun 376 p. 1999. Lidar remote sensing of canopy understanding plant responses and structure and biophysical properties of canopy structure. Journal of Experimen- Hopkinson, C., Chasmer, L., 2007. Douglas-Fir Western Hemlock forests. tal Botany, 58(4), 881-898. Using discrete laser pulse return inten- Remote Sensing of Environment. 70, sity to model canopy gap fraction. 339-361. Finnish Journal of Photogrammetry. 20 contact: L. Chasmer (2), 16-26. Lim, K., Treitz, P., Wulder, M., St-Onge, B., Flood, M., 2003. LiDAR remote [email protected] Hopkinson, C., 2008. The influence of sensing of forest structure. Progress in flying altitude, beam divergence, and . 27(1), 88-106. pulse repetition frequency on laser contact: V. Thomas pulse return intensity and canopy fre- Morsdorf, F., Kotz, B., Meier, E., Itten, [email protected] quency distribution. Canadian Journal of K.I., Allgower, B., 2006. Estimation of Remote Sensing. 33(4), 312-324. LAI and fractional cover from small footprint airborne laser scanning data Hopkinson, C., Chasmer, L., Sass, G., based on gap fraction. Remote Sensing Creed, I., Sitar, M., Kalbfleisch, W., of Environment. 104(1), 50-61. Treitz P., 2005. Assessing vegetation height and canopy volume in a Boreal Naesset, E., Økland, T., 2002. Estimating wetland complex using airborne scan- tree height and tree crown properties Page 14

Keeping an eye on the carbon balance: linking canopy development and net ecosystem exchange using a webcam Lisa Wingate, Andrew D. Richardson, Jake F. Weltzin, Kenlo N. Nasahara and John Grace

Why observe phenology input to the forest floor. phenological networks and flux occurred between the Tharandt within FLUXNET? Phenology is a robust integrator monitoring networks for the International Phenological Gar- Phenology is the study of the of the effects of climate change purpose of understanding pat- den (also one of the 24 GPM timing of lifecycle events, espe- on natural systems (Schwartz et terns and processes controlling gardens) and the nearby Car- cially as influenced by the sea- al., 2006; IPCC 2007), and it is carbon budgets across a broad boeurope-IP site Anchor Station sons and by the changes in recognized that improved moni- range of scales, explicit activities Tharandt over the past 12 years weather patterns from year to toring of phenology on local-to- to assess the impact of phenol- (Niemand et al., 2005; Grünwald year. The oldest phenological continental scales is needed. ogy on ecosystem carbon bal- & Bernhofer, 2007). Using the records, observations of cherry Historically, phenological obser- ance are still somewhat lacking standard observations from both flowering at the Royal Court in vations were a pastime of ama- within the carbon cycle commu- networks it was demonstrated Kyoto date back to 705 AD, and teur naturalists (e.g. the Mar- nity. The reasons are clear: long- that the appearance of the Mait- are still maintained to this day sham family) and reliable records term observations, otherwise rieb (May shoot) for Norway across Japan where the Japanese were often dependent on the called ‘monitoring’ are not popu- spruce is correlated with annual Meteorological Agency use these skills and effort of the observer. lar with those that sponsor re- estimates of ecosystem gross data to provide weekly forecast The increased demand for inter- search in this area; three or five primary productivity (GPP) and maps of expected blooming national co-operation and stan- year projects are the norm, net ecosystem productivity dates (http://www.jma.go.jp/jma/ dardisation in this area led to the when in practice much longer (NEP) (with the exception of the en/News/sakura.html). Robert creation of many large-scale records are required to detect extreme drought event of 2003 Marsham, the father of modern phenological monitoring net- long-term trends and their rela- in Europe). This indicates that phenological recording, was a works such as the International tionships to climatic drivers. the earlier appearance of shoots wealthy landowner and amateur Phenology Garden (IPG) pro- There is however, evidence for a potentially increases the length naturalist who recorded gram (http://www.agrar.hu- shift in attitudes. Keeling’s meas- of the growing season, leading to

"Indications of spring" in Nor- belin.de/struktur/institute/pfb/str urements of atmospheric CO2 a greater annual carbon seques- folk, England, beginning in 1736. uktur/agrarmet/phaenologie/ipg) concentrations, that began in tration. Mean March-April tem- His family maintained these re- (founded in 1957), the Global 1958, are an outstanding exam- peratures were correlated with cords until the 1950s. In the Phenological Monitoring (GPM) ple of the value long-term moni- the data of May shoot, indicating modern era, phenology has program (http://www.agrar.hu- toring represents in the context a potential scalar for GPP and gained a new impetus, as people belin.de/struktur/institute/pfb/str of a changing world (Nisbet, NEP when coupled to longer realize that such records, if sus- uktur/agrarmet/phaenologie/gpm 2007). Moreover, continuous time-series from such IPG re- tained over many years, can ) (established in 1998) as well as eddy covariance measurements cords. Similarly, an analysis cou-

reveal how plants and animals the recently-established USA- of CO2 fluxes began in the early pling budburst observations and

respond to climate change. National Phenology Network 1990s at a handful of sites. Every CO2 flux measurements at the Moreover, phenological events (USA-NPN) year, more and more sites have Howland (since 1996) and Har- such as the spring leaf-out and (http://www.usanpn.org) and been added to FLUXNET, and vard (since 1992) AmeriFlux the autumn fall exert a strong associated regional networks many of these are now providing sites indicated that earlier bud- control on both spatial and tem- (e.g., http://www.nerpn.org). useful long term data not only burst resulted in greater spring- poral patterns of the carbon These networks have focused on with regard to spatial patterns of time GPP (5 g C m-2 per 1 day cycle. Phenology also influences developing standardized proto- carbon uptake and release, but advancement of budburst date), hydrologic processes, as spring cols for phenological observa- also in relation to the influence but these increases in carbon leaf-out is accompanied by a tions, and ensuring overlap be- of phenology on carbon seques- uptake were offset by increases marked increase in evapotranspi- tween plant species found across tration. in springtime ecosystem respira- ration, and nutrient cycling as locations. Although there are One example of a synergy be- tion (RE), resulting in an uncer- autumn senescence results in a obvious advantages in creating tween phenology and flux moni- tain effect (not significantly dif- flush of fresh litter (nutrient) explicit linkages between these toring networks in Europe has ferent from zero) on springtime

cont. on page 15 Page 15

Keeping an eye on the carbon balance cont. from page 14

NEP (Richardson et al., in prepa- steadily, and already represents balance and phenology. resources (that are typically ration). some 58 site-years of combined The opportunity presented to us scarce) and, as a consequence, flux and webcam data. A large is clear: webcam measurements these observations are not pur- Phenological Activities number of these sites are in Asia at FLUXNET sites will reveal the sued at the majority of flux sites. within FLUXNET where the Phenological Eyes link between phenology and Several of the phenology net- At FLUXNET sites around the Network was set-up in 2003 to carbon uptake; they will also works include a substantial vol- world that overlook forests, create a much needed validation provide much-needed ground unteer or “citizen science” com- pastures, and wetlands, we have platform for remote sensing verification of phenology prod- ponent, wherein trained observ- the opportunity of establishing products such as NDVI ucts derived from satellite re- ers track the response of plants precision measurements of (http://www.asiaflux.net/newslett mote sensing (e.g., MODIS). using standardized protocols, on- phenological events by simply er/no21_2007.pdf). It is also line data entry forms and visuali- mounting networked digital promising to learn that this num- The role of the phenology zations designed and streamlined cameras (‘webcams’) and re- ber should continue to grow network and citizen scien- for the more casual observer cording daily (or even hourly) with the addition of sites in the tists (e.g., UK Nature Watch, US images of the vegetation canopy, US National Ecological Observa- Within FLUXNET a protocol for Project BudBurst and USA-NPN, as recommended by Baldocchi et tory Network phenological observations was and the GLOBE project (Gazal et al. (2005). A recent FLUXNET (www.neoninc.org). However, also created to harmonise al., in press). These networks of survey just as phenological gardens phenological observations across observers represent a potential (http://www.geos.ed.ac.uk/homes must commit to observations in flux sites bridge between phenological and /lwingate/webcam.html) has excess of ten year periods it is (http://www.fluxdata.org/DataInf flux observations, in that data uncovered at least 26 such web- also necessary for this webcam o/Dataset%20Doc%20Lib/FLUX collected by such “citizen scien- cams already ‘keeping an eye’ on activity to maintain a long-term NET_phenophase_protocol.pdf). tists” can be used to (a) increase canopy development (Figure 1). perspective, especially when it However, initiation of such long the density of observation sites Although this network is in its comes to unravelling the rela- term monitoring requires a sus- and species, (b) collect informa- infancy, it appears to be growing tionships between forest carbon tained commitment of human tion on presence/absence of

Figure 1 Global distribution of flux sites with webcams (Agarwal et al., 2008) cont. on page 16 Page 16

Keeping an eye on the carbon balance cont. from page 15

snow, flowers or foliage unde- ucts. This is especially the case if webcam network presents a way variability in forest function and tectable to remote sensing plat- we are to understand the appar- to directly link on-the-ground its potential impact on ecosys- forms, and (c) ground-truth ent contradiction in findings observer records to remotely- tem carbon balance in response

observations from ‘near’ (e.g., between the CO2, phenology sensed data, and moreover to to long-term changes in climate. camera, or eddy correlation) or and remote sensing communities link these to ecosystem physiol- This multi-scale monitoring of ‘far’ remote sensing platforms with respect to the timing of ogy measured with flux towers. phenology and net ecosystem

(e.g., AVHRR, MODIS). Devel- canopy green up and senescence The growing webcam network exchange of CO2 will enrich our opment of a volunteer program and how this relates directly to now represents a novel opportu- understanding and efforts at

for FLUXNET sites would changes in the atmospheric CO2 nity to implement both regional modelling not only the impacts greatly strengthen tools available record, especially during spring and global monitoring of phenol- of climate on phenology but also for the interpretation of eddy and autumn in the northern ogy at flux sites. Thus efforts to the impact of phenology on flux data. hemisphere (e.g., Piao et al., extend the spatial coverage of climate through feedbacks on 2008). This webcam network phenological observations at flux the carbon and energy cycle of Towards an international could soon be in a position to sites through the simple addition the planet. Moreover, it has the

canopy phenology camera test whether the start, maximum of cameras on towers are now potential to link the CO2 flux network and end of the growing season required within FLUXNET. In community to the thousands of In situ phenological observations, derived from satellite NDVI data time this network will not only amateur observers, many of gas exchange and radiometric really correspond to the actual establish an archive of images them school-children who will signals at the same flux sites are start, maximum and end of the documenting seasonal and inter- become the next generation of currently required for compari- growth period of plants as ob- annual changes in forest phenol- scientists. son with remotely sensed prod- served in flux sites. Thus the ogy, but also capture associated As we have illustrated above, webcams are an important way BOX 1— Time-lapse animations of canopy development and net ecosystem exchange can illus- of tracking canopy phenology. trate the degree of coupling between the two signals and the additional influence of understorey The digital images when col- and snow cover on flux measurements. A number of such animations can be observed below and lected at such regular intervals at the following website (http://www.geos.ed.ac.uk/homes/lwingate/webcam.html). These observa- can be easily assembled into tions were taken at flux sites within the Carboeurope-IP, Ameriflux and Asiaflux networks. time-lapse movies such as those in Box 1, providing an important product for raising public aware- ness on phenological and carbon cycle research. The color infor- mation of these very same im- ages can also be analyzed to retrieve information on canopy development in both deciduous and evergreen forests as de- scribed in Box 1 and 2. If you plan to mount a camera at your flux site in the near future and have any queries for the net- work please do not hesitate to contact us and we will do our best to help get you started.

Acknowledgements Many thanks to all those that Figure 2 : Time-lapse movie links for the Howland Ameriflux site (http://www.forest.sr.unh.edu/richardson/Howland2007b.avi), a Betula ermanii Cham. at the Takayama Asiaflux site (http://pen.agbi.tsukuba.ac.jp/~TKY/summary/dc/dc_2007_digest_TKY__y18bb/) took the time to respond to the and the Hainich Carboeurope-IP site (http://xweb.geos.ed.ac.uk/~lwingate/Hainich_forest_Flux_phenology.avi) recent survey on the phenology

cont. on page 17 Page 17

Keeping an eye on the carbon balance cont. from page 16

BOX 2— Coupling webcam technology with flux observations can help us understand sea- FluxLetter sonal changes in forest physiology. Recent advances in camera technology and the interpreta- tion of digital imagery now allow the quantification of plant canopy development at flux sites auto- The Newsletter of matically and without the inherent subjectivity of observer-based systems (Richardson et al., 2007). FLUXNET Besides the obvious information on snow or foliage presence at our flux sites we can also perform image analysis on the red, green and blue (RGB) colour channel brightness to obtain the informa- Vol.1 No.2 May, 2008

tion about the timing and rate of canopy green-up and senescence. Richardson et al. (2007) evalu- ated relationships between indices derived from RGB colour information and radiometric measure- FluxLetter is produced ments of the fraction of incident photosynthetically active radiation absorbed by the canopy (fAPAR) quarterly at the FLUXNET and NDVI, as well as the canopy-level photosynthetic capacity (Amax) derived from eddy covariance Office with support from the measurements at a deciduous forest (Figure 3). This study showed that webcams, although they National Science Foundation. are not calibrated radiometric instruments, could provide valuable insights into canopy development and function. A more recent analysis (Fig. 3b) has shown that a “green excess” index (2 x G% - R% - B%; Richardson et al. 2007) tracks the seasonal variation in tower-based estimates of GPP at an evergreen conifer forest, offering the possibility that “near” remote sensing can provide additional insights into canopy-scale physiological activity.

(a) (b)

Figure 3a: Red, green and blue image analysis of canopy development.

Figure 3b: Measured ‘green excess index’ from image analysis of webcam images of an evergreen conifer canopy alongside daily GPP estimates derived from eddy flux measurements. Data are from the Howland AmeriFlux site in Maine, USA; a movie of the camera images is available online (http://www.forest.sr.unh.edu/richardson/Howland2007b.avi). Also indicated are observed average budburst

activities within FLUXNET. We teachers, and scientists demonstrate Nisbet, E., 2007. Earth monitoring: This issue of FluxLetter was also thank Technical Computing variable differences between urban and Cinderella science. Nature 450: 789- edited, designed and produced by: rural leaf phenology. Global Change 790. Microsoft (www.microsoft.com/ Biology, doi:10.1111/j.1365- science). 2486.2008.01602.x. Piao, S. et al., 2008. Net carbon dioxide Dennis Baldocchi losses of northern ecosystems in re- Grunwald, T. and Bernhofer, C., 2007. sponse to autumn warming. Nature, Rodrigo Vargas Literature A decade of carbon, water and energy 451(7174): 49-52. flux measurements of an old spruce FLUXNET Office, 137 Mulford Agarwal, D., Humphrey, M., van Ingen, forest at the Anchor Station Tharandt. Richardson, A.D. et al., 2007. Use of Hall, University of California, C., Beekwilder, N., Goode, M., Jack- Tellus, 59B: 387-396. digital webcam images to track spring Berkeley, CA 94720 son, K., Rodriguez, M. and Weber, R. green-up in a deciduous broadleaf ph: 1-(510)-642-2421 "The Fluxdata Data and Collaboration IPCC. Climate Change 2007: The Physi- forest. Oecologia, 152: 323-334. Fax: 1-(510)-643-5098 Server (http://www.fluxdata.org)," cal Sciences Basis: Contribution of

Berkeley Laboratory Technical Report, Working Group I to the Fourth Assess- Schwartz, M.D., Ahas, R. and Aasa, A. May, 2008. Available at http:// ment Report of the Intergovernmental 2006. Onset of spring starting earlier We plan to make the FLUXNET bwc.berkeley.edu/. Panel on Climate Change. (Cambridge across the Northern Hemisphere. newsletter a powerful information, University Press, Cambridge, 2007). Global Change Biology, 12: 343-351. networking, and communication

Baldocchi, D.D., et al., 2005. Predicting resource for the community. If you the onset of net carbon uptake by Niemand, C., Kostner, B., Prasse, H., want to contribute to any section or deciduous forests with soil temperature Grunwald, T. and Bernhofer, C., 2005. contact: L. Wingate and climate data: a synthesis of FLUX- Relating tree phenology with annual propose a new one please contact the NET data. International Journal of carbon fluxes at Tharandt forest. Mete- [email protected] FLUXNET Office. THANKS!! Biometeorology 49: 377-387. orologische Zeitschrift, 14(2): 197-202. Gazal, R. et al., 2008. GLOBE students,