CONCEPTS & THEORY A new paradigm of quantifying ecosystem stress through chemical signatures Ben Kravitz,1,† Alex B. Guenther,2 Lianhong Gu,3 Thomas Karl,4 Lisa Kaser,5 Stephen G. Pallardy,6 Josep Peñuelas,7,8 Mark J. Potosnak,9 and Roger Seco2 1Atmospheric Sciences and Global Change Division, Pacific Northwest National Laboratory, P.O. Box 999, MSIN K9-30, Richland, Washington 99352 USA 2Department of Earth System Science, University of California Irvine, 3200 Croul Hall Street, Irvine, California 92697 USA 3Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831 USA 4Institute of Atmospheric and Crysopheric Sciences, University of Innsbruck, Innrain 52f, A-6020 Innsbruck, Austria 5National Center for Atmospheric Research, P.O. Box 3000, Boulder, Colorado 80307 USA 6Department of Forestry, University of Missouri, 203 Anheuser-Busch Natural Resources Building, Columbia, Missouri 65211 USA 7CREAF, Cerdanyola del Vallès, 08193 Catalonia, Spain 8Global Ecology Unit CREAF-CSIC-UAB, CSIC, Cerdanyola del Vallès, 08193 Catalonia, Spain 9Department of Environmental Science and Studies, DePaul University, McGowan South, Suite 203, Chicago, Illinois 60604 USA Citation: Kravitz, B., A. B. Guenther, L. Gu, T. Karl, L. Kaser, S. G. Pallardy, J. Peñuelas, M. J. Potosnak, and R. Seco. 2016. A new paradigm of quantifying ecosystem stress through chemical signatures. Ecosphere 7(11):e01559. 10.1002/ecs2.1559 Abstract. Stress-­induced emissions of biogenic volatile organic compounds (VOCs) from terrestrial eco- systems may be one of the dominant sources of VOC emissions worldwide. Understanding the ecosystem stress response could reveal how ecosystems will respond and adapt to climate change and, in turn, quan- tify changes in the atmospheric burden of VOC oxidants and secondary organic aerosols. Here, we argue, based on preliminary evidence from several opportunistic measurement sources, that chemical signatures of stress can be identified and quantified at the ecosystem scale. We also outline future endeavors that we see as next steps toward uncovering quantitative signatures of stress, including new advances in both VOC data collection and analysis of “big data.” Key words: chemical signatures; ecosystem; stress; volatile organic compounds. Received 4 June 2016; revised 28 July 2016; accepted 15 August 2016. Corresponding Editor: Jason B. West. Copyright: © 2016 Kravitz et al. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. † E-mail: ben.kravitz@pnnl.gov Introduction natural ecosystem services, agricultural eco- systems and food production, the carbon cycle, Gaining understanding of how to quantify atmospheric chemistry, and biodiversity. ecosystem stress is intimately tied with predict- The concept of stress on an ecosystem scale ing the stress response of ecosystems and, ulti- is difficult to measure, quantify, or even define. mately, how ecosystems will respond and adapt One major obstacle in ascertaining a definition is to climate change. As climate change progresses, that the concept of stress is ambiguous at such ecosystems will undergo greater amounts of a broad scale (Ulanowicz 1978, Ryder 1990). environmental stress (Field et al. 2014). This will Kolasa and Pickett (1992) evaluate in detail the have far-­reaching implications for many climate various difficulties involved in assessing stress; and economic sectors of the world, including one of their arguments is that such assessments  v www.esajournals.org 1 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. depend upon recognizing changes in the ecosys- Each particular plant species, and hence each tem’s underlying architecture, although identify- ecosystem, has a different pattern of VOC emis- ing this so-­called minimum interactive structure sions that depends upon a wide variety of envi- is not without its own difficulties. Rapport (1992) ronmental factors. For example, Jud et al. (2016) argues that assessments of the health of ecosys- found that postillumination bursts of acetalde- tems are necessarily subjective. Ecosystems are hyde and green leaf volatiles from poplars had capable of adapting to changes in their environ- common, identifiable features. Drewniak et  al. ments, so it is difficult to argue that an adapting (2014) found a shift in dominant VOC emis- ecosystem is in poorer health than in a prior state. sions resulting from an expanse of red maple Due to these definitional challenges, attempts at at the expense of oak; this serves as an example quantifying ecosystem stress are scarce. of how VOC emissions can serve as markers of One of the first steps in quantifying stress climate shifts. The different factors that control would be to identify events that may cause stress. VOC emissions are numerous and complex; There are clear examples of stress events, such as untangling these myriad components has been a severe weather, droughts, and insect infestation, grand research challenge for well over a decade but even these events have subtleties. A defoliat- (Peñuelas and Llusia 2001). ing hail storm is a stressful event, but other forms Approximately 90% of total worldwide VOC of precipitation may not be. Droughts lack clear emissions are biogenic (Guenther et  al. 1995), beginnings, ends, and thresholds, and moreover, and Bergström et al. (2014) estimate that 50–70% drought severity does not necessarily correspond of all biogenic secondary organic aerosols in to well-­defined changes in precipitation, tem- Central and Northern Europe forests are due to perature, or soil moisture. The impact of drought stress-­induced VOC emissions. Quantifying all can be even farther removed from meteorology, sources of VOC emissions, as well as how those as it depends on ecosystem-­specific attributes emissions respond to biotic and abiotic factors, such as plant rooting depths. are crucial steps toward improving understand- One known response of individual plants to ing of the Earth system and our role in it. stress events is an increase in the emissions of Modeling of VOC emissions is a fruitful biogenic volatile organic compounds (VOCs; approach to understanding and disentangling e.g., Karl et al. 2008, Kleist et al. 2012, Kaser et al. some of these environmental factors, particularly 2013). Volatile organic compound emissions are on an ecosystem scale. The Model of Emissions both indicators and drivers of ecological func- of Gases and Aerosols from Nature (MEGAN; tion. Plants emit VOCs in response to changes in Guenther et al. 2012) was developed for exactly light, temperature, and water availability (e.g., this purpose: to capture the behavior of biogenic Guenther et al. 1995). Harley et al. (1999) argue VOC emissions as a function of a wide variety of that isoprene-­emitting plants do so to ameliorate meteorological and environmental changes. As stresses associated with heat and high light; such an example, in unstressed conditions, MEGAN responses may be more common as the threat of can predict isoprene emission patterns on an climate change grows more severe (Peñuelas and ecosystem scale to within the uncertainties of Llusià 2003). Plants can also emit VOCs to selec- observations (Potosnak et  al. 2014). However, tively attract different pollinating insects (Schiestl MEGAN  has low skill in capturing changes in 2015, Raguso 2016), as a defense mechanism to VOC emissions due to stress; MEGAN and other deter infesting insects (Mentel et al. 2013), or to biogenic VOC emission models may underpre- entice predators of herbivorous insects (Kessler dict VOC emissions during acute abiotic ecosys- and Baldwin 2001), in some cases highly selec- tem stress events (e.g., storms) and overpredict tively based on a combination of wound damage emissions during chronic stress (e.g., droughts; and oral secretions of the herbivore (Turlings Seco et al. 2015). et  al. 1990, Kessler and Baldwin 2002). Higher This difficulty in representing the “stress emissions from some ecosystems can affect sur- term” gives an indication of one of the key gaps rounding ecosystems (Bamberger et  al. 2011), in linkages between atmospheric chemistry possibly as a form of chemical communication and ecosystem studies: the connection between (e.g., Shulaev et al. 1997, Ton et al. 2007). organism-­level stress responses and a broader  v www.esajournals.org 2 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. atmosphere-­scale perspective of the chemi- four times background levels for 2  d after the cal signatures and effects of ecosystem stress. event; MEGAN underpredicted monoterpene Specifically, although plant responses to stress emissions over this two-­day period by approxi- have been quantified on the scale of individual mately a factor of 20, and the monthly total was plants (e.g., Niinemets et  al. 2013), there has underestimated by approximately 30% (Kaser been little work to date in robustly quantifying et  al. 2013). One compounding difficulty with the stress response of ecosystems, especially het- measurements at this site is that repetitions of erogeneous ecosystems, on the ecosystem scale. similar stresses could conceivably cause differ- This shortage of previous work is because, until ent responses. For example, there was another this point, there was both a lack of measurements storm on 27 July, prior to the measured storm on and of techniques that can identify complex sig- 4 August (Kaser et  al. 2013). Monoterpenes can natures in the collected data. In the past few be stored as well as emitted immediately after years, there has been substantial progress made, production (Peñuelas and Llusia 2001), so the allowing quantification of ecosystem-­scale stress storm on 4 August could have induced lower to proceed in new, promising directions. In this associated monoterpene emissions than the 27 article, we identify a potential new paradigm in July storm due to depletion of the storage pool. quantifying stress signatures in ecosystem-­scale Alternately, after the first storm event, the plants measurements of VOC emissions. We also out- had to repair any damage caused and could have line the types of measurements and techniques exhausted their carbohydrate reserves; thus, after that would be needed to build these signatures. the second event, monoterpene emissions could have been substrate limited. Ecosystem-­Scale Studies of VOC Emissions Another ecosystem-­scale example was observed during the Canopy Horizontal Array In this section, we discuss several sources of Turbulence Study (CHATS) campaign, which evidence that VOC emissions can be a sensitive was performed for 12 weeks in the spring of 2007 indicator of stress events. Most of the examples at a walnut plantation near Davis, California presented here are associated with abiotic (Patton et al. 2011). The campaign was designed stresses, as the relationships between VOC emis- as a turbulent exchange study to see how forest sions and these types of stresses are far better canopies modify the exchange of gases between known than the relationships with biotic stresses. the boundary layer and the free troposphere. Nevertheless, there are relevant examples related One of the VOCs measured during this campaign to biotic stressors that we touch upon. was methyl salicylate, which is a known plant-­ One example of ecosystem-­scale stress-­induced signaling compound and is the volatile form of VOC emissions involves measurements taken compounds that are known to activate defense at the Manitou Forest Observatory, an experi- genes in plants (Karl et al. 2008). The campaign mental forest near Colorado Springs, Colorado, began on 15 March, which had relatively cold primarily consisting of Ponderosa pine. Ortega night time temperatures, and irrigation stopped et  al. (2014) describe the site and summarize for a period during the campaign (Patton et  al. measurements taken during the BEACHON 2011). Methyl salicylate emissions were sub- ROCS campaign, which was designed to study stantially increased over most of the campaign the influence of forest canopies on chemical in response to these temperature and drought exchange and boundary layer turbulence. The stresses, but once the site was irrigated, nearly all campaign lasted for approximately 1  month of the measured emissions were monoterpenes. during August 2010. On 4 August, during the Although these two campaigns give a strong campaign, there was a severe hailstorm (140 mm indication that VOC emissions are affected by of hail) that physically damaged needles and stress, it is difficult to quantify stress in a way branches of the trees (Kaser et al. 2013). Wound that would allow for robust assessment of eco- stresses are known to cause increased VOC emis- system health. During the BEACHON ROCS sions in a variety of plants, including pines (e.g., campaign, the stress event was clearly defined, Juuti et al. 1990, Crespo et al. 2013). Monoterpene but the campaign was too short to determine fluxes at this site were elevated to approximately baseline VOC emission patterns, and there may  v www.esajournals.org 3 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. be complications in measurements due to repeti- This index can be easily calculated using daily tion of the stress event. During the CHATS cam- meteorological data collected at the site, thus paign, the site was quite homogeneous, but VOC necessarily giving a more local representation of emissions were in response to multiple sources drought level than the North American Drought of stress, so individual stress signatures could Monitor, which is often calculated at mesoscales not be determined easily. or synoptic scales. It can also provide fine tem- The Missouri Flux Site (MOFlux) is a long-­term poral resolution using relatively robust measure- observation station that is part of the AmeriFlux ments. Use of this index requires a long enough Network. The site is primarily broad-­leaf decidu- time record such that KBDI  =  0 when the tabu- ous trees (dominated by isoprene-­emitting oaks), lation is initiated (i.e., the index is initiated in located in the Ozarks near Columbia, Missouri. In conditions that are clearly non-­drought). Fig.  6 2011, the site experienced a drought, concurrent shows a drought index calculated from the KBDI with the highest isoprene fluxes ever measured for the MOFlux site from 2004 to 2013. By this to that point (Potosnak et  al. 2014). In 2012, the index, the drought in 2012 was the most severe drought was more severe and widespread (Seco in this record period, indicating that for our pur- et al. 2015). Although drought stress is more diffi- poses, the KBDI reproduces known occurrences cult to quantify than wound stress, the advantage of drought at the MOFlux site and can thus of this site is that VOC emissions were recorded serve as a coarse indicator of drought stress. The for multiple years, and meteorological data have KBDI also gives a similar quantitative picture to been collected at the site for nearly a decade. the results in Fig.  4, in that it indicates that the There are currently multiple methods of quan- drought in 2012 was more severe and had an ear- tifying drought. The North American Drought lier onset than the drought in 2011. Monitor (Svoboda et al. 2002, NCDC 2014) indi- One possible hypothesis describing the physi- cated drought at the MOFlux site during the ological response of this ecosystem is that severe summers of 2011 and 2012 (Figs. 1–3), although drought stress can cause early senescence, or the the footprint of this index as calculated by NCDC point at which deciduous trees begin to shed (2014) is substantially larger than the MOFlux their leaves in preparation for winter (Fig. 5 pro- site. Fig. 4 shows different measurements taken vides strong anecdotal evidence for this hypoth- at the MOFlux site in 2011 and 2012 that are also esis). Fig.  7 shows that for the years 2008–2010, indicators of drought. Values from 2011 are the which had comparatively low maximum KBDI same as were reported by Potosnak et al. (2014), values over the course of the year (<300), senes- and values from 2012 are the same as those cence began after Julian Day 240–260, which is reported by Seco et al. (2015). All of the indicators roughly the month of September. The onset of (net ecosystem exchange, soil moisture, air vapor senescence here is defined as the point at which pressure deficit, and predawn leaf water poten- leaf area index begins to decline. For other years, tial) replicate the broad features of the results the onset of senescence is anticorrelated with the in Figs.  1–3, in that the 2012 drought was more maximum KBDI value calculated for that year severe and had an earlier onset than the 2011 (R2 = 0.73). drought. However, it is difficult to obtain a more These results clearly indicate physiological quantitative picture of drought from this figure, responses that correlate with our chosen index as some of the indicators are noisy. Moreover, of drought stress. However, this coarse compar- there is no clear indication of a threshold beyond ison can only indicate threshold events. That is, which there is ecosystem damage, yet Fig.  5 identification of sufficient stress to cause early shows that there are clear visible signals of stress senescence is only determined after senescence in the canopy; leaf necrosis was widespread by has begun. Ideally, signatures could be identified the end of summer in 2012. that serve as early warnings when ecosystems As an additional metric, we tested use of the are undergoing stress or could indicate more Keetch–Byram Drought Index (KBDI), which subtle changes in ecosystem health that may was originally developed for determining the not appear in standard assessments of drought. amount of biomass available to burn in forest Measurements of changes in biogenic VOC emis- fires (Keetch and Byram 1968, Alexander 1990). sions show promise in revealing those signatures.  v www.esajournals.org 4 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. Fig. 1. Monthly averages of the North American Drought Index (Svoboda et al. 2002, NCDC 2014) for June 2011 and 2012. At the Missouri Flux site (indicated by a green star), June 2011 shows no drought.  v www.esajournals.org 5 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. Fig. 2. Monthly averages of the North American Drought Index (Svoboda et al. 2002, NCDC 2014) for July 2011 and 2012. At the Missouri Flux site (indicated by a green star), July 2011 shows abnormally dry (D0) conditions. In contrast, the drought in 2012 is extreme (D3).  v www.esajournals.org 6 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. Fig. 3. Monthly averages of the North American Drought Index (Svoboda et al. 2002, NCDC 2014) for August 2011 and 2012. At the Missouri Flux site (indicated by a green star), August 2011 shows abnormally dry (D0) conditions. In contrast, the drought in 2012 is severe (D2) to exceptional (D4).  v www.esajournals.org 7 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. Drought stress indicators (µmol·m−2·s−1) 0 −15 −10 −5 NEE 0.20 0.30 0.40 0.50 10 cm (m3/m3) Soil moisture Air vapor pressure 4.0 3.0 2.0 1.0 9/3/2010 deficit (kPa) Predawn leaf water 0 potential (MPa) −3 −2 −1 2011 2012 120 140 160 180 200 220 240 260 Day of year Fig.  4. Indicators of drought at the Missouri Flux 9/2/2012 site in 2011 (blue lines) and 2012 (red lines). Values presented for net ecosystem exchange (NEE), 10  cm Fig.  5. Photographs of the forest canopy at the soil moisture, and air vapor pressure deficit are five-­ Missouri Flux site during September 2010 (a wet day averages of half-­hourly data. For NEE and vapor summer) and September 2012 (exceptional drought). pressure deficit, only daytime values (08:00 to 18:00 Non-­green leaves at the top of the crown were largely local time) are incorporated in the average. Note that necrotic, not showing their early fall color. Assessments by convention, NEE is presented as positive upwards, of forest dieback cannot be made from these so a drought is represented as increasing values. All photographs, as leaf scorch did not penetrate through values for 2011 are repeated from Potosnak et  al. the entire canopy. (2014). suggesting the ecosystems are undergoing suf- Fig.  8 shows daytime average VOC measure- ficiently high levels of stress to modify their ments from the MOFlux site during the summers physiological responses to their environment. of 2011 and 2012. Isoprene measurements from Analyses of Fig. 7 would not reveal this thresh- 2011 were taken using a chemiluminescence old behavior, especially in terms of predictive analyzer (Potosnak et  al. 2014), and isoprene, capability. Methanol shows no strong response to methanol, and monoterpene measurements in changes in KBDI, although methanol emissions 2012 were taken with a Proton Transfer Reaction are known to respond to other types of stress Mass Spectrometer (PTR-­MS) (Seco et  al. 2015). (Peñuelas et al. 2005, Wohlfahrt et al. 2015). Isoprene and monoterpene emissions clearly Fig. 9 shows that when forced with the mete- show threshold behavior at KBDI  ≈  500. Before orological data collected at the MOFlux site, this point, the VOC fluxes increase with KBDI as MEGAN clearly underrepresents isoprene and expected; isoprene and monoterpene emissions monoterpene emissions before the drought and correlate with temperature, and temperature overrepresents these emissions during and after increases are one of the primary contributors to the drought. Moreover, this underrepresenta- increased values of KBDI. However, for KBDI tion does not necessarily correlate with increases values >500, isoprene and monoterpene emis- or decreases in the KBDI. The MEGAN results sions decrease with further increases in KBDI, shown in this figure may differ depending upon  v www.esajournals.org 8 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. MOFlux drought 800 8 700 7 600 6 500 5 Drought level KBDI 400 4 300 3 200 2 100 1 0 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Time Fig.  6. Daily values of the Keetch-­Byram Drought Index (KBDI) for 2004–2013 as calculated from meteorological data collected at the Missouri Flux site. Colored shading shows the drought level; increases in drought level correspond to a 100-­point increase in KBDI, where the maximum possible value is 800. Dashed vertical lines indicate 1 January of a particular year. the emission factor used (which has uncertain- Developments in several main areas that repre- ties), but this cannot explain why, at certain sent recent advances in measurement technology times, predictions from MEGAN do not match have the potential to push this field toward a observed values of VOC fluxes. More succinctly, new paradigm in which robust signatures of eco- the “stress term” is currently poorly represented system stress can be identified and quantified. in MEGAN. Discovering biogenic VOC signa- The first step in quantifying ecosystem stress tures of stress will enable the research commu- is collecting sufficient data of VOC emissions. nity to better represent environmental influences The campaigns discussed in the previous section on VOC emissions in models. were not designed to collect data for identifica- tion of stress signatures, so all of the measure- Quantifying Ecosystem Stress Through ments discussed were serendipitous. Targeted Biogenic VOC Signatures measurement campaigns would involve long-­ term, multiyear deployment of VOC measure- Using measurements of biogenic VOCs, indi- ment instruments in all seasons so a baseline of cations of stress on the ecosystem scale can be emissions could be established and longitudinal targeted and quantified to a finer degree than, for data surrounding stress events could be col- example, using estimates of carbon uptake and lected. Similar measurements have repeatedly transpiration to quantify stress (e.g., Fig.  4). been justified for CO2 flux data collection (e.g., Moreover, they have the potential to be predic- Boden et al. 2013); such efforts could differentiate tive, providing an “early warning” before stress “normal” emissions from stress-­induced emis- causes serious damage to the ecosystem. sions and also identify shifts in seasonal patterns  v www.esajournals.org 9 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. 260 2006 measurements. A more suitable approach in this 2007 case might involve drones or tethered balloons 240 2008 2009 (e.g., Greenberg et  al. 2014), either to support Peak LAI (Julian Day) 220 2010 lightweight samples directly or to lift tubing 2011 2012 that serves as an inlet for larger in situ samplers; 200 these are easily movable and rapidly deployable when stress episodes are located. In addition, air- 180 R 2 = 0.73 craft PTR-­MS measurements have been demon- 160 strated (Karl et  al. 2013) and could be valuable tools for targeting and characterizing short-­term 140 stress events. As a starting point, it may be easier to observe more stationary biotic stresses, such 120 200 300 400 500 600 700 800 as plant disease; Jansen et  al. (2011) found that Maximum KBDI VOCs can provide indications of disease, and the emission signatures may be distinct for different Fig.  7. Yearly peak leaf area index (LAI) timing diseases. Biotic stressors that can move rapidly (Julian Day) vs. yearly maximum Keetch-­Byram (such as insect infestations) may prove to be Drought Index (KBDI) for years 2006–2012. Peak LAI more logistically challenging. is defined as the last day of measurements before LAI In addition to field-­based measurements, there values begin to decrease for the winter season. The are promising new techniques in remote sensing maximum KBDI is the largest value of KBDI during that could be used to quantify VOC emissions on spring, summer, or fall of that year (winter is excluded a global scale. Unique high-­spectral-­resolution to avoid counting drought carryover from previous measurements have been used to isolate signals years). Black line indicates ordinary least-­squares of chlorophyll fluorescence and reflectance (e.g., regression through the plotted points. Gamon et  al. 1992, Peñuelas et  al. 1995), pro- viding real-­time estimates of vegetation status of emissions, such as changes in leafout and (Joiner et  al. 2011). A decline in photosynthetic senescence timing. activity, which is directly related to chlorophyll A useful instrument for conducting these stud- fluorescence, has been tied to long-­term drought ies is a Proton Transfer Reaction Time-­of-­Flight stress in tropical forests, serving as an early Mass Spectrometer (PTR-­TOF-­MS), a device that warning of loss of greenness (Zhou et  al. 2014). has become available only in the past few years Not only can these measurements be used to give (Graus et al. 2010). Previously, instruments used accurate estimates of gross primary productivity to measure VOC emissions could only measure a on a global scale (Guanter et  al. 2014), but they small number of compounds. The PTR-­TOF-­MS can also be used as an indirect estimator of iso- can differentiate a multitude of compounds with prene emissions at the ecosystem level (Peñuelas fewer sources of interference in near-­real time; et  al. 2013). As such, satellite measurements such an instrument is crucial for characterizing could improve ecosystem-­scale measurements of the atmospheric VOC budget (Park et  al. 2013). VOC emissions on a global scale, complement- Further advances in ease of maintenance of ing the higher spatial and temporal resolution of this instrument may be necessary for long-­term in situ VOC emission measurements. In turn, in (multiple seasons) deployment. In the meantime, situ measurements can be used to interpret and using simpler instrumentation that avoids the validate satellite measurements. price and complexity of the PTR-­TOF-­MS could Another promising source of data for quan- be useful in providing long-­term measurements tifying ecosystem stress concerns a relatively of isoprene, monoterpenes, and methanol (Rinne new idea called ecometabolomics (Peñuelas and et al. 2016). Sardans 2009, Sardans et  al. 2011, Rivas-­Ubach These requirements become somewhat more et al. 2013). This concept takes the approach that difficult and nuanced for biotic stressors. Such environmental conditions trigger reactions along stress events occur at outbreak sites that are specific metabolic pathways. Then, as a physi- not necessarily ideally located for flux tower ological response, plants will produce volatile  v www.esajournals.org 10 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. 200 300 200 300 2011 Isoprene (µmol·m−2·h−1) 2012 Isoprene (µmol·m−2·h−1) 150 250 150 250 Julian Day 100 100 200 200 50 50 150 150 0 0 20 300 10 300 2012 MeOH (µmol·m−2·h–1) 2012 MT (µmol·m−2·h−1) 15 5 250 250 Julian Day 10 0 200 200 5 −5 150 150 0 −10 0 200 400 600 800 0 200 400 600 800 KBDI KBDI Fig.  8. Daily average volatile organic compound (VOC) fluxes plotted as a function of the Keetch–Byram Drought Index. Color indicates Julian Day. y-­axis indicates the particular measured VOC and the year in which measurements were taken. Normalizing values to account for short-­term variation in temperature and light did not affect results (data not shown). metabolites (i.e., VOCs). This approach has the secondary organic aerosol formation and conse- potential to capture dynamical changes in eco- quent changes in cloud cover via aerosol–cloud system metabolomes, revealing ecophysiological interactions (Berg et  al. 2013, Joutensaari et  al. mechanisms that are activated to produce VOC 2015). emissions (Peñuelas and Sardans 2009). One of The amount of VOC emission data that would the major challenges is that the plant kingdom be required to identify signatures is sufficiently has on the order of 105 metabolites, and indi- large that advanced “big data” techniques may vidual metabolomes in a living population may become necessary to robustly identify signatures vary (Sardans et al. 2011). As such, explorations of stress. In particular, these techniques must be of ecometabolomics have thus far been limited to able to reconstruct signatures that depend upon narrow, controlled cases. Nevertheless, if these multiple sources of variability and are convolved difficulties can be overcome, ecometabolomics with other signatures, all in the presence of noise could be used in conjunction with targeted field and potential missing data. As an example, plant measurements in revealing not only the missing emissions of isoprene are affected by the inter- “stress term,” but also how stress-­induced VOC active effects of CO2 concentration and leaf tem- emissions may change with climate change. perature (Potosnak 2014); sophisticated analysis In addition to direct measurements, proxy tools are required to untangle these complex measurements may be useful in locating and relationships. providing qualitative indicators of stress. For In recent years, a wide variety of tools that deal example, large insect outbreaks in boreal for- with complex data have become available. An ests have been tracked through observations of example of such tools are genetic algorithms that  v www.esajournals.org 11 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. COLOUR Fig.  9. Volatile organic compound (VOC) fluxes (y-­axis) as a function of Julian Day (x-­axis). Green lines indicate the maximum daily value of VOC emission. Blue lines indicate calculations of VOC emissions from Model of Emissions of Gases and Aerosols from Nature when forced with local meteorological measurements. Shading indicates the drought level; an increase in drought level corresponds to a 100-­point increase in Keetch-­ Byram Drought Index. robustly identify key parameters that explain for example, divide-­and-­recombine data pro- features in complex data, as well as identify cessing approaches (e.g., Hafen et  al. 2013) to latent variables that are missing from explana- discover features, analyze and compare data, or tions of signatures (e.g., Heredia-­Langner et  al. potentially perform real-­time analysis of data 2013). Another example involves machine learn- streams to adaptively target field measurements. ing techniques, such as support vector machine learning or Bayesian compressive sensing (e.g., Future Directions Cacciatore et al. 2014) to distinguish unique fea- tures and identify commonalities in signatures, Establishing robust, quantifiable, predictive sig- providing answers to two crucial questions: (1) natures of ecosystem stress has numerous applica- Can signatures associated with different types tions to a wide variety of sectors. From an of stress be distinguished from each other? and atmospheric science standpoint, the ultimate goal (2) Are there commonalities among signatures is to characterize the radiative and chemical impacts of the same stress type in different ecosystems of changes in biogenic VOC emissions in the face of with potentially different dominant plant phe- anthropogenic climate change. One major compo- notypes? Understanding complex, multidi- nent of this goal involves improvements in repre- mensional relationships can be aided by recent sentations of the “stress term” by MEGAN and advances in visualization techniques (Poco et al. other biogenic VOC models; in turn, these models 2014) in concert with parallel processing using, are being incorporated into Earth System Models  v www.esajournals.org 12 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. such as the Community Earth System Model the U.S. Department of Energy under Contract DE-­ (CESM; Hurrell et al. 2013) to understand the inter- AC05-­76RL01830. The National Center for Atmospheric actions with and feedbacks between biogenic VOC Research is sponsored by the National Science emissions and climate change. Foundation. Josep Peñuelas’s research was supported by the European Research Council Synergy grant SyG-­ An additional application of these techniques 2013-­610028 IMBALANCE-­P. is in the field of precision farming. In the cases of agricultural and bioenergy crops, the optimal growing strategy will involve maximizing pro- Literature Cited duction while minimizing environmental impact. By understanding early warning signatures of Alexander, M. E. 1990. Computer calculation of the stress through VOC emission measurements, Keetch-­Byram Drought Index – programmers be- ware! Fire Management Notes 51:23–25. farmers can alter irrigation, planting timing, Bamberger, I., et al. 2011. Deposition fluxes of terpenes and even planting location to minimize plant over grassland. Journal of Geophysical Research stress and increase productivity. This would 116:D14305. simultaneously decrease aggregate VOC emis- Berg, A. R., C. L. Heald, K. E. Huff Hartz, A. G. Hallar, sions, thereby reducing biogenic environmental A. J. H. Meddens, J. A. Hicke, J. F. Lamarque, and S. impacts. On a somewhat different note, there Tilmes. 2013. The impact of bark beetle infestations exists the potential to breed crops that naturally on monoterpene emissions and secondary organic emit antifungal or herbivore-­repellent VOCs to aerosol formation in western North America. At- enhance resilience without having to resort to mospheric Chemistry and Physics 13:3149–3161. pesticides (Stenberg et  al. 2015); using observa- Bergström, R., M. Hallquist, D. Simpson, J. Wildt, and tions of such plants as a control case could help T. F. Mentel. 2014. Biotic stress: A significant con- tributor to organic aerosol in Europe? Atmospheric better identify signatures of stress. Chemistry and Physics 14:13603–13647. There remains a great deal of work in seeing the Boden, T. A., M. Krassovski, and B. Yang. 2013. The ideas presented here to fruition. Measurement AmeriFlux data activity and data system: an evolv- of ecosystem-­scale chemical signatures will be ing collection of data management techniques, the result of a confluence of new foci in research tools, products and services. Geoscientific Instru- efforts with new technologies and techniques. mentation, Methods and Data Systems 2:165–176. Concerted effort and investment by the research Cacciatore, S., C. Luchinat, and L. Tenori. 2014. Knowl- community in this new paradigm of quantifica- edge discovery by accuracy maximization. Pro- tion has the potential to transform a broad range ceedings of the National Academy of Sciences USA of communities that depend on understanding 111:5117–5122. stress at the ecosystem level. Crespo, E., M. Graus, J. B. Gilman, B. M. Lerner, R. Fall, F. J. M. Warren, and C. Warneke. 2013. Volatile ­organic compound emissions from elephant grass Acknowledgments and bamboo cultivars used as potential bioethanol crop. Atmospheric Environment 65:61–68. We thank two anonymous reviewers for their help- Drewniak, B. A., P. K. Snyder, A. L. Steiner, T. E. Twine, ful comments. We also thank the members of the and D. J. Wuebbles. 2014. Simulated changes in Signature Discovery Initiative team for their support biogenic VOC emissions and ozone formulation and guidance throughout this process. Special thanks from habitat expansion of Acer rubrum (red maple). go to Vanessa Bailey, Nathan Baker, George Bonheyo, Environmental Research Letters 9:014006. Ryan Hafen, Alejandro Heredia-­Langner, Jenna Field, C. B., et  al. 2014. Technical summary. In C. B. Larson, LeeAnn McCue, Trenton Pulsipher, Landon Field, et al., editors. Climate change 2014: ­impacts, Sego, Yannan Sun, Mark Tardiff, and Tim White. A adaptation, and vulnerability. Contribution of portion of this research was supported by EMSL, a Working Group II to the Fifth Assessment Re- DOE Office of Science user facility sponsored by the port of the Intergovernmental Panel on Climate Department of Energy’s Office of Biological and Change. Cambridge University Press, Cambridge, Environmental Research and located at Pacific UK and New York, New York, USA. Northwest National Laboratory. A portion of this Gamon, J. A., J. Peñuelas, and C. B. Field. 1992. A research was conducted under the Laboratory Directed narrow-­waveband spectral index that tracks diur- Research and Development Program at PNNL, a mul- nal changes in photosynthetic efficiency. Remote tiprogram national laboratory operated by Battelle for Sensing of Environment 41:35–44.  v www.esajournals.org 13 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. Graus, M., M. Müller, and A. Hansel. 2010. High bursts of volatile organic compounds in isoprene-­ ­resolution PTR-­TOF: quantification and formula con- emitting and non-­emitting poplar. Plant, Cell, and firmation of VOC in real time. Journal of the Ameri- Environment 39:1204–1215. can Society for Mass Spectrometry 21:1037–1044. Juuti, S., J. Arey, and R. Atkinson. 1990. Monoterpene Greenberg, J. P., et  al. 2014. A tethered-­balloon emission rate measurements from a monterey pine. PTRMS sampling approach for rapid surveying of Journal of Geophysical Research 95:7515–7519. landscape-­scale biogenic VOC fluxes. Atmospheric Karl, T., A. Guenther, A. Turnipseed, E. G. Patton, and Measurement Techniques 7:2263–2271. K. Jardine. 2008. Chemical sensing of plant stress Guanter, L., et al. 2014. Global and time-­resolved mon- at the ecosystem scale. Biogeosciences 5:1287–1294. itoring of crop photosynthesis with chlorophyll Karl, T., P. K. Misztal, H. H. Jonsson, S. Shertz, A. H. fluorescence. Proceedings of the National Acade- Goldstein, and A. B. Guenther. 2013. Airborne my of Sciences USA 111:1327–1333. flux measurements of BVOCs above Californian Guenther, A. B., X. Jiang, C. L. Heald, T. Sakulyanont- oak forests: experimental investigation of surface vittaya, T. Duhl, L. K. Emmons, and X. Wang. 2012. and entrainment fluxes, OH densities and Dahm- The Model of Emissions of Gases and Aerosols köohler numbers. Journal of Atmospheric Science from Nature version 2.1 (MEGAN2.1): an extend- 70:3277–3287. ed and updated framework for modeling biogen- Kaser, L., et al. 2013. Undisturbed and disturbed above ic emissions. Geoscientific Model Development canopy ponderosa pine emissions: PTR-­TOF-­MS 5:1471–1492. measurements and MEGAN 2.1 model results. At- Guenther, A. B., et al. 1995. A global model of natural mospheric Chemistry and Physics 13:11935–11947. volatile organic compound emissions. Journal of Keetch, J. J., and G. M. Byram. 1968. A drought index for Geophysical Research 100:8873–8892. forest fire control. Tech. rep., U.S. Department of Ag- Hafen, R., L. Gosink, J. McDermott, K. Rodland, K. K. riculture – Forest Service, Washington, D.C., USA. Van Dam, and W. S. Cleveland. 2013. Trelliscope: a Kessler, A., and I. T. Baldwin. 2001. Defensive function system for detailed visualization in the deep anal- of herbivore-­induced plant volatile emissions in ysis of complex data. IEEE Symposium on Large- nature. Science 291:2141–2144. Scale Data Analysis and Visualization. Kessler, A., and I. T. Baldwin. 2002. Plant responses to Harley, P. C., R. K. Monson, and M. T. Lerdau. 1999. insect herbivory: the emerging molecular analysis. Ecological and evolutionary aspects of isoprene Annual Review of Plant Biology 53:299–328. emission from plants. Oecologia 118:109. Kleist, E., et al. 2012. Irreversible impacts of heat on the Heredia-Langner, A., K. H. Jarman, B. G. Amidan, and emissions of monoterpenes, sesquiterpenes, phe- J. G. Pounds. 2013. Genetic algorithms and classifi- nolic BVOC and green leaf volatiles from several cation trees in feature discovery: diabetes and the tree species. Biogeosciences 9:5111–5123. NHANES database. The 9th International Confer- Kolasa, J., and T. A. Pickett. 1992. Ecosystem stress and ence on Data Mining (DMIN13). health: an expansion of the conceptual basis. Jour- Hurrell, J. W., et  al. 2013. The community earth sys- nal of Aquatic Ecosystem Health 1:7–13. tem model: a framework for collaborative research. Mentel, T. F., et  al. 2013. Secondary organic aerosol Bulletin of the American Meteorological Society formation from stress-­induced biogenic emis- 94:1339–1360. sions and possible climate feedbacks. Atmospheric Jansen, R. M. C., J. Wildt, I. F. Kappers, H. J. Bouw- Chemistry and Physics 13:8755–8770. meester, J. W. Hofstee, and E. J. van Henten. 2011. NCDC. 2014. North American drought monitor. Detection of diseased plants by analysis of volatile https://www.ncdc.noaa.gov/temp-and-precip/ organic compound emission. Annual Review of drought/nadm/ Phytopathology 49:157–174. Niinemets, Ü., A. Kännaste, and L. Copolovici. 2013. Joiner, J., Y. Yoshida, A. P. Vasilkov, Y. Yoshida, L. A. Quantitative patterns between plant volatile emis- Corp, and E. M. Middleton. 2011. First observa- sions induced by biotic stresses and the degree of tions of global and seasonal terrestrial chlorophyll damage. Frontiers in Plant Science 4:262. fluorescence from space. Biogeosciences 8:637–651. Ortega, J., et al. 2014. Overview of the Manitou Experi- Joutensaari, J., et  al. 2015. Biotic stress accelerates mental Forest Observatory. Atmospheric Chemistry formation of climate-­relevant aerosols in bore- and Physics 14:1647–1709. al forests. Atmospheric Chemistry and Physics Park, J. H., A. H. Goldstein, J. Timovsky, S. Fares, 15:12139–12157. R. Weber, J. Karlik, and R. Holzinger. 2013. Active Jud, W., E. Vanzo, Z. Li, A. Ghirardo, I. Zimmer, T. D. atmosphere-­ecosystem exchange of the vast major- Sharkey, A. Hansel, and J. P. Schnitzler. 2016. Effects ity of detected volatile organic compounds. Science of heat and drought stress on post-­illumination 341:643–647.  v www.esajournals.org 14 November 2016 v Volume 7(11) v Article e01559 CONCEPTS & THEORY Kravitz et al. Patton, E. G., et al. 2011. The canopy horizontal array Rivas-Ubach, A., M. Pèrez-Trujillo, J. Sardans, turbulence study. Bulletin of the American Meteo- A. ­Gargallo-Garriga, T. Parilla, and J. Peñuelas. 2013. rological Society 92:593–611. Ecometabolomics: optimized NMR-­based method. Peñuelas, J., I. Filella, and J. A. Gamon. 1995. Assess- Methods in Ecology and Evolution 4:464–473. ment of photosynthetic radiation-­use efficien- Ryder, R. A. 1990. Ecosystem health, a human percep- cy with spectral reflectance. New Phytologist tion: definition, detection, and the dichotomous 131:291–296. key. Journal of Great Lakes Research 16:619–624. Peñuelas, J., I. Filella, C. Stefanescu, and J. Llusià. 2005. Sardans, J., J. Peñuelas, and A. Rivas-Ubach. 2011. Caterpillars of Euphydryas aurinia (Lepidoptera: Ecological metabolomics: overview of current de- Nymphalidae) feeding on Succisa pratensis leaves velopments and future challenges. Chemoecology induce large foliar emissions of methanol. New 21:191–225. Phytologist 167:851–857. Schiestl, F. 2015. Ecology and evolution of oral volatile-­ Peñuelas, J., and J. Llusia. 2001. The complexity of mediated information transfer in plants. New Phy- ­factors driving volatile organic compound emis- tologist 206:571–577. sions by plants. Biologia Plantarum 44:481–487. Seco, R., T. Karl, A. Guenther, K. P. Hosman, S. G. Pal- Peñuelas, J., and J. Llusià. 2003. BVOCs: Plant defense lardy, L. Gu, C. Geron, P. Harley, and S. Kim. 2015. against climate warming? Trends in Plant Science Ecosystem-­scale volatile organic compound fluxes 8:105–109. during an extreme drought in a broadleaf temper- Peñuelas, J., G. Marino, J. Llusia, C. Morfopoulos, ate forest of the Missouri Ozarks (central USA). G. Farrè-Armengol, and I. Filella. 2013. Photo- Global Change Biology 21:3657–3674. chemical reflectance index as an indirect estimator Shulaev, V., P. Silverman, and I. Raskin. 1997. Airborne of foliar isoprenoid emissions at the ecosystem lev- signalling by methyl salicylate in plant pathogen el. Nature Communications 4:2604. resistance. Nature 385:718–721. Peñuelas, J., and J. Sardans. 2009. Ecological metabolo- Stenberg, J. A., M. Heil, I. Åhman, and C. Bjöorkman. mics. Chemistry and Ecology 25:305–309. 2015. Optimizing crops for biocontrol of pests and Poco, J., A. Dasgupta, Y. Wei, W. Hargrove, disease. Trends in Plant Science 20:698–712. C. Schwalm, R. Cook, E. Bertini, and C. Silva. 2014. Svoboda, M., et al. 2002. The drought monitor. Bulletin SimilarityExplorer, a visual inter-comparison tool of the American Meteorological Society 83:1181– for multifaceted climate data. In H. Carr, P. Rhe- 1190. ingans, and H. Schumann, editors. Eurographics Ton, J., M. D’Alessandro, V. Jourdie, G. Jakab, D. Kar- Conference on Visualization (EuroVis) 2014. Vol- len, M. Held, B. Mauch-Mani, and T. C. J. Turlings. ume 33. 2007. Priming by airborne signals boosts direct and Potosnak, M. J. 2014. Including the interactive effect of indirect resistance in maize. Plant Journal 49:16–26. elevated CO2 concentration and leaf temperature Turlings, T. C. J., J. H. Tumlinson, and W. J. Lewis. in global models of isoprene emission. Plant, Cell, 1990. Exploitation of herbivore-­induced plant and Environment 37:1723–1726. odors by host-­seeking parasitic wraps. Science Potosnak, M., et  al. 2014. Observed and modeled 250:1251–1253. ecosystem isoprene fluxes from an oak dominat- Ulanowicz, R. E. 1978. Modeling environmental ed temperate forest and the influence of drought stress. Pages 1–18 in J. H. Thorp and J. W. Gibbons, stress. Atmospheric Environment 84:314–322. ­editors. Energy and environmental stress in aquat- Raguso, R. A. 2016. More lessons from linalool: in- ic systems. Technical Information Center, U.S. sights gained from a ubiquitous oral volatile. Cur- ­Department of Energy, Washington, D.C., USA. rent Opinion in Plant Biology 32:31–36. Wohlfahrt, G., et  al. 2015. An ecosystem-­scale per- Rapport, D. J. 1992. Evaluating ecosystem health. Jour- spective of the net land methanol flux: synthesis nal of Aquatic Ecosystem Health 1:15–24. of ­micrometeorological flux measurements. Atmo- Rinne, J., T. Karl, and A. Guenther. 2016. Simple, sta- spheric Chemistry and Physics 15:7413–7427. ble, and affordable: towards long-­term ecosystem Zhou, L., et  al. 2014. Widespread decline of Congo scale flux measurements of VOCs. Atmospheric rainforest greenness in the past decade. Nature Environment 131:225–227. 509:86–90.  v www.esajournals.org 15 November 2016 v Volume 7(11) v Article e01559