Ecosystem Consequences of Soil Warming: Microbes, Vegetation, Fauna and Soil Biogeochemistry

Ecosystem Consequences of Soil Warming: Microbes, Vegetation, Fauna and Soil Biogeochemistry focuses on biotic and biogeochemical responses to warmer soils including plant and microbial evolution. It covers various field settings, such as arctic tundra; alpine meadows; temperate, tropical and subalpine forests; drylands; and grassland ecosystems. Information integrates multiple natural science disciplines, providing a holistic, integrative approach that will help readers understand and forecast future planetwide responses to soil warming. Students and educators will find this book informative for understanding biotic and biogeochemical responses to changing climatic conditions. Scientists from a wide range of disciplines, including soil scientists, ecologists, geneticists, as well as molecular, evolutionary and conservation biologists, will find this book a valuable resource in understanding and planning for warmer climate conditions.

• Emphasizes biological components of soils, plants and microbes that provide linkages to physics and chemistry
• Brings together chapters written by global scientific experts with interests in communication and education
• Includes coverage of polar, alpine, tropical, temperate and dryland ecosystems

• Language: English
• Publisher: Academic Press
• Release date: Apr 12, 2019
• ISBN: 9780128134948

Book preview

much.

Foreword and introduction: Past, present & future

Jacqueline E. Mohan, Odum School of Ecology, University of Georgia, Athens, GA, United States

To this day, I vividly recall the afternoon in 1995 when, as a new doctoral student at Duke University, I read for the first time the Harte and Shaw (1995) paper that, along with other work including Odum (1969), Tilman and Downing (1994), Tilman et al. (1996), Vitousek (1994), Vitousek and Reiners (1975), Bazzaz (1996), Schlesinger (1997), and Clark et al. (1998, 1999) directed the course of my academic career. After the earlier tundra work of Chapin and Shaver (1985), Harte and Shaw (1995) was one of the first to investigate how experimental climate warming might impact biodiversity in actual field settings, with relevance for future community dynamics and ecosystem functioning. Odum had earlier coupled community successional development with net ecosystem productivity and carbon storage. Schlesinger explicitly laid out global-scale carbon dynamics and a global carbon budget-terrestrial ecosystems included. Tillman investigated the role of biodiversity for ecosystem productivity and Bazzaz depicted how dissimilar species and plant functional groups responded very differently to changes in Earth's global environment. It was a heady time to be a graduate student interested in the merging of biodiversity research, evolutionary ecology, ecosystem functioning, and Earth system science, all in the context of profound anthropogenic global change.

Yet our sciences have progressed so much since the mid-1990s, particularly in addressing the role of Earth's terrestrial ecosystems for responding to and influencing our planet's changing climate conditions. Nucleotide sequencing methods are illuminating the roles of evolution and community structure for ecosystem-scale responses, especially for microbial life forms. Dramatic increases in sequencing throughput have enabled metatranscriptomic techniques to capture a more accurate picture of biodiversity-ecosystem linkages, and specifically of functional biodiversity signals via analyses including mRNA in addition to DNA. Bayesian hierarchical and other modeling techniques have become mainstream in ecology and other environmental disciplines (Clark, 2005, 2007, 2010, Ellison, 1996, 2004, Hilborn and Mangel, 1997). Changing land surface interactions with the atmosphere and oceans, including those of terrestrial ecosystems, are now incorporated into projections of Earth's future climate conditions and impacts (IPCC, 2013, 2018).

The importance of integrating multiple environmental disciplines has become increasingly apparent for understanding whole ecosystem responses to global change. To this end, here we present writings from over 70 authors covering soil-warming responses from biomes spanning the globe, and organisms ranging in size and generation time from microbes to fauna to trees. Together these chapters present what we know and what we have yet to discover about how Earth's ecosystems are responding, and may continue to respond, to our warming planet.

An emergent theme from many chapters of this volume is the importance of integrating soil-warming impacts on abiotic and biotic drivers to understand system responses to change. Another is the current general bias in the geographical distribution of soil-warming experiments, with relatively few from lower latitudes and the Southern Hemisphere. Clearly, results from one latitude or biome type may be very different from warming responses in other systems. While for higher latitudes and higher elevations the role of snowmelt and specifically its timing is emerging as a critical component of climate change, these findings are not applicable to warmer biomes. The role of increasingly frequent extreme weather events, including droughts and hurricanes, is becoming a crucial research need for understanding ecosystem consequences of warming. And the fundamental importance of long-term soil-warming research from the few sites where we have it-demonstrating how conclusions from later decades are often in discord with those of the initial years of investigation-represents a clarion call for scientific investment far beyond typical grant funding cycles (see Harte, Chapter 1 and Melillo et al., 2011, 2017).

Earth system science, biogeochemistry, and subdisciplines of ecology and evolution are no longer the largely separate fields they once were. Scientists and especially students are blending the lines and pushing for greater understanding of how exactly Earth's ecosystems and climate influence one another, including what the ramifications of these interactions are for our planet's biodiversity, humans included. Clearly, we need more warming experiments across larger spatial scales, particularly low-latitude ecosystems, spanning longer periods of time. Let us hope the pace of empirical and theoretical discovery only increases, better enabling us to maintain the health of the only planet where we know life exists, and upon which our species is utterly dependent. I thank the authors who so graciously agreed to share their world class scientific insights with readers of this book.

References

Bazzaz F.A. Plants in Changing Environments: Linking Physiological, Population, and Community Ecology. Cambridge Univ Press; 1996.

Chapin F.S., Shaver G.R. Individualistic growth response of tundra plant species to environmental manipulations in the field. Ecology. 1985;66:564–576.

Clark J.S. Why environmental scientists are becoming Bayesians. Ecol. Lett. 2005;8:2–14.

Clark J.S. Models for Ecological Data: An Introduction. Princeton, NJ, USA: Princeton University Press; 2007.

Clark J.S. Individuals and the variation needed for high species diversity in forest trees. Science. 2010;327:1129.

Clark J.S., Macklin E., Wood L. Stages and spatial scales of recruitment limitation in southern Appalachian forests. Ecol. Monogr. 1998;68:213–235.

Clark J., Beckage B., Camill P., Cleveland B., Hillerislambers J., Lichter J., McLachlan J., Mohan J., Wyckoff P. Interpreting recruitment limitation in forests. Am. J. Bot. 1999;86:1–16.

Ellison A.M. An introduction to Bayesian inference for ecological research and environmental decision-making. Ecol. Appl. 1996;6:1036–1046.

Ellison A.M. Bayesian inference in ecology. Ecol. Lett. 2004;7:509–520.

Harte J., Shaw R. Shifting dominance within a montane vegetation community: results of a climate-warming experiment. Science. 1995;267:876.

Hilborn R., Mangel M. The Ecological Detective: Confronting Models with Data. Princeton University Press; 1997.

IPCC. In: Stocker T.F., Qin D., Plattner G.-K., et al., eds. IPCC, 2013: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assesment Report of the Intergovernmental Panel on Climate Change. Cambridge, UK and New York, NY, USA: Cambridge University Press; 2013:1535.

IPCC. Summary for policymakers. In: Masson-Delmotte V., Zhai P., Pörtner H.O., Roberts D., Skea J., Shukla P.R., Pirani A., Moufouma-Okia W., Péan C., Pidcock R., Connors S., Matthews J.B.R., Chen Y., Zhou X., Gomis M.I., Lonnoy E., Maycock T., Tignor M., Waterfield T., eds. Global Warming of 1.5°C. An IPCC Special Report on the Impacts of Global Warming of 1.5°C Above Pre-industrial Levels and Related Global Greenhouse Gas Emission Pathways, in the Context of Strengthening the Global Response to the Threat of Climate Change, Sustainable Development, and Efforts to Eradicate Poverty. Geneva, Switzerland: World Meteorological Organization; 2018:32.

Melillo J.M., Butler S., Johnson J., Mohan J., Steudler P., Lux H., Burrows E., Bowles F., Smith R., Scott L. Soil warming, carbon-nitrogen interactions, and forest carbon budgets. Proc. Natl. Acad. Sci. 2011;108:9508.

Melillo J.M., Frey S.D., Deangelis K.M., Werner W.J., Bernard M.J., Bowles F.P., Pold G., Knorr M.A., Grandy A.S. Long-term pattern and magnitude of soil carbon feedback to the climate system in a warming world. Science. 2017;358:101–105.

Odum E.P. The strategy of ecosystem development. Science. 1969;164:262–270.

Schlesinger William H. Biogeochemistry: An Analysis of Global Change. 2nd edition Amsterdam: Academic Press; 1997.588.

TIlman D., Downing J.A. Biodiversity and stability in grasslands. Nature. 1994;367:363.

Tilman D., Wedin D., Knops J. Productivity and sustainability influenced by biodiversity in grassland ecosystems. Nature. 1996;379:718.

Vitousek P.M. Beyond global warming: ecology and global change. Ecology. 1994;75:1861–1876.

Vitousek P.M., Reiners W.A. Ecosystem succession and nutrient retention: a hypothesis. Bioscience. 1975;25:376–381.

Chapter 1

Reflections on 27 years of manipulated ecosystem warming in a subalpine meadow

John Harte    The Energy and Resources Group, University of California, Berkeley, CA, United States

Abstract

From analyses of data obtained in a 27-year climate manipulation experiment, insight has accumulated about the importance of: (1) combining manipulation experiments with both modeling and observational studies, particularly along gradients; (2) avoiding the temptation to make long-term predictions from short-term studies; (3) optimally designing manipulation experiments and measurement protocols to yield useful data; and (4) taking an evolutionary as well as ecological perspective on population responses to experimental warming. Those and other insights, along with a summary of the key findings of the experiment, especially findings relevant to climate-ecosystem feedbacks, are described here.

Keywords

Ecosystem warming; Subalpine meadow; Global warming; Plant community composition; Soil carbon feedback

Contents

Introduction

Overview of the Rocky Mountain biological laboratory climate warming experiment and auxiliary gradient studies

Major findings

Vegetation responses: Production and community composition

Vegetation responses: Phenology

Vegetation responses: Physiology

Community interactions

Plant species richness

Plant pathogens and herbivores

The response of a short-lived biennial

Biogeochemical responses: Nitrogen

Biogeochemical responses: Methane oxidation

Biogeochemical responses: Soil carbon

The decomposition-weighted productivity mass-balance model

Shift in surface albedo

Soil mesofauna response

Observing the signal of contemporary climate warming

Some considerations in the design of climate-warming experiments

The importance of long-term studies

Establishing sufficiently large plots

Selecting a site

Designing treatment conditions

Explanatory variables

Pluralistic approaches overcome obstacles to prediction

The space-for-time assumption

A population perspective

Feedback

Keeping it going

References

Further reading

Introduction

Even the casual observer can discern that a prolonged drought, a heavy late-spring frost, or a deluge will affect the vegetation in gardens, meadows, and forests. Moreover, there will be discernable impacts on the animals that inhabit these plant communities. With the evidence for anthropogenic global climate change now overwhelming and unrefuted, science and society are increasingly concerned about how ecosystems will respond as the planet heats.

There are three primary reasons why we should care. First, the human enterprise depends, in a tangible material sense, upon the health of ecosystems. When ecosystems are degraded, humanity loses. This is because ecosystems build and protect soil, regulate the flows and stocks of available water, cleanse air and water, control the pests that would otherwise devastate our crops, moderate the natural vagaries of climate, and provision us with the genetic prototypes from which all of our foods and many of our medicines derive. In a direct economic sense, nature is the steward of humankind.

That is also true in a mental or spiritual sense. In a variety of ways we derive solace, insight, perspective, awe, and beauty from nature.

The third reason is a little less obvious. While current climate models are based on sound physics, predicting future climate entails more than just knowing about the physics of heat and light, air and water. Ecosystems are a big player as well. Vegetation influences the physical stage on which climate plays out, and microorganisms regulate the gases that control energy flow in the atmosphere. And vegetation and microorganisms are controlled not only by climate, but by each other as well. In this truly complex system, as ecosystems are altered by climate, the climate is in turn altered. These feedback effects (Lashof et al., 1997) can only be understood and reliably incorporated into climate models if we first understand how ecosystems respond to climate change.

It was a sense of the need to unravel this complexity, to characterize climate-ecosystem feedbacks, that motivated me in 1988 to gain empirical insight with a warming experiment. This chapter will describe what I learned from 27 years of manipulating climate in a subalpine meadow. I summarize results from this research into ecosystem responses and feedbacks to climate change, but primarily I focus on insights obtained concerning the design and limitations of controlled climate manipulations, and the advantages of combining such experiments with more traditional observational and mathematical modeling studies in ecology.

Overview of the Rocky Mountain biological laboratory climate warming experiment and auxiliary gradient studies

To understand the effects of climate change on plant species composition and on biogeochemical processes, we integrated two field study methods—experimental heating and analysis across natural climate gradients—with additional off-site manipulations, laboratory incubations, and analytical modeling. The experimental manipulations provide insights into causal mechanisms governing short-term responses to climate change; gradient studies help elucidate longer term phenomena; and laboratory incubations pinned down quantitative rate constants that feed into a process-based predictive mathematical model capable of extending insights forward in time (Dunne et al., 2004).

The warming experiment is conducted at the Rocky Mountain Biological Laboratory (RMBL), Gunnison Co., Colorado, United States (38°57′N, −106°59′E; 2920 m a.s.l.). In 1989, I established ten 3 × 10 m plots in an ungrazed montane meadow (the warming meadow). These plots are laid out side by side, oriented north-south, and have ~ 3 m between adjacent plots. Each plot is on an east-west incline, with the upper edge extending to a moraine ridgeline. Above five of the plots, overhead infrared radiators have been on continuously since January 1991, casting a downward heat flux at the surface of ~ 15 W/m² prior to June 1993, and 22 W/m² subsequently. Within each plot, three subplots span a 10-m microclimate gradient: the upper third (zone C) is flat and relatively dry, the middle third (zone B) is steeper and dry, and the lower third (zone A) is flat and more moist (Fig. 1).

Fig. 1 Schematic of the RMBL warming experiment.

Soil temperature and moisture was measured and logged every 2 h at 5, 12, and 25 cm depths in each zone. The microclimatic effect of experimental heating throughout the growing season was to warm the top 15 cm of soil by ~ 2°C, dry it by 10%–20% (gravimetric basis) during the growing season, and to prolong the snow-free season at each end by an average of 2–4 weeks (Fig. 2). Importantly, over the course of the experiment the time interval between when snowmelt occurs in the heated and control plots has increased.

Fig. 2 Calendar day of snowmelt in the control ( dotted line ) and the heated ( solid line ) plots. The ellipse encloses the 5 drought years.

Vegetation in the upper zone of the plots, which is where most of the results presented here were obtained, consists of the dominant woody shrub, Artemisia tridentata (sagebrush), ~ 40 species of perennial forbs, including Erigeron speciosus (fleabane) and Delphinium nelsoni (larkspur), and 16 species of graminoids, including Festuca thurberi (a bunch grass). Forb biomass in winter is entirely below ground, and so net aboveground growing-season forb production can be estimated from peak areal coverage; shrub production is also estimated from areal coverage data (Harte and Shaw, 1995).

To extend knowledge gained from the warming experiment, we also investigated ecological trends along a natural elevational and climate gradient within the same drainage area (the Upper East River Valley) as the warming meadow. At each of three sites, separated by ~ 200 m elevation, ten 16-m² plots were laid out in 1995 and instrumented with microclimate probes. Within these plots many of the same measurements made at the warming meadow were regularly carried out. The experimental manipulation provides insights into causal mechanisms governing short-term responses to climate change, while gradient studies help elucidate longer term phenomena.

Additional details about the warming meadow and elevational sites, including experimental design and methods, soil properties, grazing history, and microclimatic effects of heating have been described (Harte et al., 1995, 2014; Harte and Shaw, 1995; Shaw and Harte, 2001a; Saleska et al., 2002).

Major findings

Over the 27 years of manipulated warming we have documented the following ecosystem responses (in addition to the effects on soil temperature and moisture and timing of snowmelt, mentioned above).

Vegetation responses: Production and community composition

We observed a reduction in the net aboveground production of forbs in the heated plots, relative to the controls, and an increase in woody shrub cover. In year 1 of the experiment, annual forb aboveground cover was approximately four times the shrub cover in both heated and control plots. By year 10, annual shrub production overtook forb production in the heated plots, and by year 27 it was approximately three times greater than that of forbs in those plots (Harte and Shaw, 1995; Perfors et al., 2003; Saavedra et al., 2003; Harte et al., 2006, 2014; Figs. 3 and 4). Graminoid production showed no consistent trend (Rudgers et al., 2014).

Fig. 3 Peak aboveground biomass (AGB) of forbs in heated (H) and control (C) plots as a function of time.

Fig. 4 Peak aboveground biomass (AGB) of shrubs in heated (H) and control (C) plots as a function of time.

But not all forbs responded in the same way to treatment (de Valpine and Harte, 2001). For example, while the relatively shallow-rooted forb E. speciosus, the dominant contributor to aboveground forb biomass in the control plots, consistently showed greatly reduced growth and physiological symptoms of moisture stress in the heated plots, the second largest forb contributor to aboveground biomass, the deep-rooted Helianthella quinquenervis, exhibited only a weak response to heating. Off-plot nitrogen and water addition experiments to understand these species-specific responses revealed that shallow-rooted forbs responded strongly and positively to water addition but not to nitrogen addition, whereas the deep-rooted species were only weakly stimulated by addition of nitrogen and water (de Valpine and Harte, 2001).

Vegetation responses: Phenology

We observed significant effects of warming on plant reproductive phenology, both for late blooming as well as early blooming species, both at the time flowering occurs and during the reproductive cycle (from bud formation to seed dispersal) (Dunne et al., 2003; Price and Waser, 1998). These treatment effects are largely driven by the influence of experimental warming on the timing of snowmelt; however, late flowering species responded more weakly to manipulated snowmelt date than early flowering species. Duration of flowering period was slightly extended under heating-induced earlier snowmelt, again with early flowering species showing the strongest responses. Flowering date also responds to differences in ambient climate, both along an elevational gradient and as a result of interannual variability, with the same response function it has to a manipulated change in climate (Dunne et al., 2004). Thus the space-for-time assumption provides a useful approach to extending our knowledge of phenological responses to climate change.

Vegetation responses: Physiology

Heating affected forb water potential, thermal acclimation, frost tolerance, photosynthesis, and transpiration; the direction and magnitude of the responses, however, are highly species-specific (Harte and Shaw, 1995; Loik and Harte, 1996, 1997; Loik et al., 2000, 2004; Shaw et al., 2000; de Valpine and Harte, 2001; Saavedra et al., 2003; Lambrecht et al., 2006). In contrast, heating consistently enhanced a variety of measures of physiological vigor (Shaw et al., 2000) in both the foliar (Harte and Shaw, 1995) and wood production (Perfors et al., 2003) of the shrub A. tridentata.

Community interactions

To better understand the community-level consequences of losing shallow-rooted species, a group that is adversely affected by experimental warming, we conducted an off-site species removal experiment (Cross and Harte, 2007). After 3 years of experimental species removal, tap-rooted forbs and grasses were able to fully compensate for the loss of shallow-rooted forbs with increased biomass production. Moreover, the remaining plant community yielded a larger biomass response to nitrogen addition when shallow-rooted forbs were removed, possibly because removal led to increased soil moisture. Although shallow-rooted forbs share a common negative response to warming, their loss did not affect community-level biomass. However, the loss of shallow-rooted forbs could result in an increased sensitivity to perturbations, like changing nutrient availability.

Plant species richness

To date, no change in species richness has been observed, but, because the meadow flora consists largely of long-lived perennial forbs, this is an expected observation. Moreover, direct observations across our control meadow plots and in 30 plots located along an elevation gradient (Saleska et al., 2002; Dunne et al., 2003) indicate that forb species richness is negatively correlated with both length of growing season and the extent of shrub dominance. Manipulated warming increases both these negative correlates of species richness (Perfors et al., 2003), motivating the hypothesis that species richness will decline under warming. Observed declines in the flowering success, productivity, and physiological vigor of shallow-rooted forb species in the heated plots (Loik and Harte, 1996, 1997; Loik et al., 2000), coupled with observed patterns of association between community composition and environmental parameters, like snowmelt date along the climate gradient, suggest that shallow-rooted forb species are likely to greatly diminish in abundance and species richness in the heated plots over the next decade.

Plant pathogens and herbivores

We quantified the damage caused by all observable (aboveground) pathogens and herbivores on six of the most common plant species (A. tridentata, H. quinquenervis, E. speciosus, Potentilla gracilis, Potentilla hippiana, and Lathyrus leucanthus) (Roy et al., 2004). We found that plants in the earlier melting plots generally were most damaged and were attacked by a larger number of species. Although the overall trend was an increase in damage with warmer temperatures and earlier snowmelt, some pathogens and herbivores performed better in cooler or later melting plots. The idiosyncratic response of each species to environmental conditions reinforces our finding that there are likely to be changes in community composition as the planet warms.

An additional study (Adler et al., 2007) focused on aphid infestation of A. tridentata. It found variable effects of warming. In particular, in dry summers, aphid infestation was greater on plants in the control plots, whereas in wet summers the heated plots showed greater infestation.

The response of a short-lived biennial

Nearly all the plant species in the warming meadow plots are long-lived perennials, with the exception of Androsace septentrionalis, or rock jasmine, with a variable lifespan that averages about 2 years. Thus over the 27 years of the warming experiment, A. septentrionalis has gone through at least a dozen generations under the heating treatment, and is thus a candidate species for the study of evolutionary adaptation to warming. In Gunnison County, A. septentrionalis inhabits a large elevational and climate gradient, from approximately 2000 to 4000 m. Along that gradient it exhibits a trend in morphological traits, such as stem height and the diameter of its basal rosette at maturity. In pioneering work (Panetta et al., 2018), combined demographic and trait analysis of A. septentrionalis revealed evidence for selection for individuals with traits characteristic of the low-elevation populations in the heated plots. Growth chamber studies of subsequent generations produced from seeds obtained from warming meadow individuals indicate that selection on genetically controlled traits, not a plastic response, has occurred. While the evidence for evolutionary adaptation of A. septentrionalis to warming suggests the possibility that the local population could endure under global warming, individuals in the heated plots exhibit roughly one-tenth the flowering success of individuals in the control plots, and thus the rate of adaptation does not appear to be fast enough to insure survival.

Biogeochemical responses: Nitrogen

During the first 5 years of the experiment the net nitrogen mineralization rate increased in the heated plots, but that effect turned out to be transient and became virtually undetectable 5 years later (Shaw and Harte, 2001a).

Biogeochemical responses: Methane oxidation

Net methane oxidation, not production, was detected in the plots. When the measured net methane oxidation rate is plotted against soil moisture, data from both heated and control plots fall on a common unimodal curve; drier soils exhibit a low rate, presumably because dryness inhibits methanothrophic activity, and in very wet soils methane diffusion is inhibited (Torn and Harte, 1995).

Biogeochemical responses: Soil carbon

Over the first 5 years of the experiment, organic carbon levels in the top 10 cm of soil in the heated plots dropped by about 25%; control plot soil carbon levels showed no change (Fig. 5). Over the next half decade, soil carbon remained roughly constant in the heated and control plots. Then, during the second decade, a slow increase was observed in the soil carbon of the heated plots and a slow decrease was observed in the control plots. During the third decade of the experiment, carbon levels in the two treatments were indistinguishable (Saleska et al., 2002; Harte et al., 2014).

Fig. 5 Percentage of soil organic carbon averaged over five replicates in heated and ambient plots, plotted against year. Standard errors for ambient plots and heated plots in each year average ~ 0.40% and 0.21%.

The constraint of mass balance, coupled with some laboratory measurements, allowed us to construct a conceptual model that made sense of the phenomena summarized above. Our initial assumption was that warmer soils were driving the loss of soil carbon via increased soil respiration. However, a series of laboratory soil incubations using soil from the plots, in which we measured respiration rates under all 25 combinations of 5 temperature and 5 soil moisture treatments (Saleska et al., 2002), allowed us to generate a contour map of the response of soil respiration to microclimate. A regression model then informed us that the effect on respiration of soil warming nearly canceled the effect of soil drying. Direct flux measurements supported this conclusion. Hence we turned to the only other conceivable explanation: shifts in plant production and community composition.

Measurements revealed that in the heated plots, the impairment of forb production led to a decrease in litter input to the soil, causing a rapid decrease in soil carbon in the heated plots. Increased shrub production was accompanied by only about a third as much litter input as was lost by the forb response. However, shrub litter has a significantly higher lignin-to-nitrogen ratio than does forb litter, and litter bag experiments revealed that soil carbon derived from shrub litter has a significantly lower decomposition rate than does forb-derived soil. Thus the more refractory soil carbon derived from increasing shrub dominance explains the slow recovery of soil carbon in the heated plots. In summary, a litter quantity effect (from reduced total primary production) could explain the short-term drop in soil carbon, while a quality effect (resulting from an increase in the proportion of refractory soil organic matter) could explain the longer term recovery (Shaw and Harte, 2001b; Saleska et al., 1999, 2002; Harte et al., 2014). To pin this explanation down, and project future soil carbon, we constructed a mathematical model.

The decomposition-weighted productivity mass-balance model

The decomposition-weighted productivity (DWP) mass-balance model treats the soil carbon pool as the sum of carbon components, each associated with a broad group of plants. For the meadow-warming experiment, the plant groups are forbs, shrubs, and graminoids; in other ecosystems, different groupings can be used. For each of the three components of the soil carbon pool Ci, let Pi equal the annual production of plant type i, and let Di equal the annual mineralization of carbon component Ci. Hence:

   (1)

DWP takes Pi to be proportional to the peak aboveground biomass value of plant type i:

   (2)

where pi is the ratio of annual litter production of plant category i to its peak aboveground biomass. All variables are per unit area.

The parameter pi incorporates three kinds of information. First, it includes the contribution to annual soil carbon production from root turnover as well as aboveground turnover. Second, while only a single species of shrub dominates (A. tridentata) throughout the growing season, and thus its peak aboveground biomass provides a reasonable surrogate for total annual aboveground production, in contrast there are successive cycles of growth of the forb species, and so forb AGB at peak biomass underestimates total growing-season aboveground production. The quantity pforb is a correction factor that takes account of this. Third, because AGB is in units of grams of biomass per square meter and Pi has units of grams of carbon per square meter, pi reflects the conversion from biomass to carbon.

The rates Di describe the flow of carbon out of the soil, per unit area for each SOC type-i. We assume Di depends on three things: soil microclimate, a quality factor expressing the lability of the soil carbon derived from each plant type, and the size of the SOC pool for each SOC type:

   (3)

Here T and M are appropriately averaged values of soil temperature and moisture and ki is proportional to the inverse turnover time of Ci. Numerical values of the function μ(T,M) derive from the laboratory incubations mentioned above; the value of μ varied negligibly from plot to plot. The values of ki were estimated from in situ incubations to determine litter decomposition rates and from measurements of lignin-to-nitrogen ratios of plant litter (Shaw and Harte, 2001b). The estimates of pi are based on measurements of daytime and nighttime ecosystem carbon fluxes throughout the growing season in locations with differing shrub-to-forb ratios (Saleska et al., 2002).

In steady state, Ci(t + 1) = Ci(t); summing across plant functional types and rearranging gives:

   (4)

To turn the proportionality here into an equality, however, we regress SOC against DWP, and thereby derive a specific prediction for SOC in each plot based on measured DWP and the regression parameters β0 and β1:

   (5)

This regression effectively partitions SOC into a background constant component common to all sites (estimated by the intercept β0), and a vegetation-climate-influenced variable component depending on DWP through the proportionality constant β1.

A test of the DWP steady-state prediction for SOC (Eqs. 4, 5) carried out along a 12 km climate and vegetation gradient spanning a 400 m elevation change that includes the warming meadow is shown in Fig. 6 (Saleska et al., 2002; Harte et al., 2014).

Fig. 6 Observed SOC (in gC/m ² ) versus model-predicted SOC, including least-squares regression lines for lower, middle, upper, and warming meadow control plots separately; the 1:1 line; and the R ² across all plots combined. SOC levels are corrected for bulk density variations. Further details can be found in Saleska et al. (2002).

To run the model in nonsteady state, to predict soil carbon levels in the control plots in the future, we extrapolate the snowmelt trend shown in Fig. 2 and make use of empirical regressions between snowmelt date and forb and shrub production. The output matches the available data (circled output in Fig. 7) and makes an interesting prediction for soil carbon levels in the control plots over the coming 75 years (Fig. 7). In particular, SOC in the control plots should continue to decline until approximately 2050 and then recover, possibly overshooting its original value at the start of the experiment.

Fig. 7 100-year soil carbon simulation. Further details can be found in Harte et al. (2014).

The key finding that SOC first declines and then slowly recovers is the consequence of two facts. First, the increase in productivity of the plants that were favored by warming (sagebrush) is less than the decrease in productivity of the plants whose growth was slowed by warming (forbs). Second, sagebrush produces more recalcitrant litter and SOC than do the forbs. It is likely that in other types of ecosystems, with plants having different traits, these same two governing influences (rate of litter input to the soil and litter quality) will result in different outcomes for SOC.

For example, another warming-induced transition under a warming climate record is from spruce-fir-dominated forest to pine domination, as suggested by pollen records (Anderson et al., 2000); this may be an example of a transition from high production rate of long-lived litterto lower production rate of shorter-lived litter, and thus, under continuing warming, the DWP model would predict both short- and long-term SOC decline. A quantitative prediction would require acquisition of additional data to determine the parameters in the model, and this has not been performed, but the robust qualitative conclusion that can be drawn is that such a climate-induced forest transition would likely result in a positive carbon cycle feedback in the near and longer term.

Shift in surface albedo

Surface shortwave albedo decreased from roughly 12% in the control plots to 7% in the heated plots during the growing season because of the shift from forb cover to shrub cover (unpublished data). At the spatial scale of montane meadows, this albedo difference translates into an additional absorbed energy flux of approximately 10 W/m² averaged over night, day, and over the growing season. By comparison, the global averaged forcing due to a doubling of the atmospheric level of carbon dioxide is approximately 5 W/m² and thus the local-scale to landscape-scale climatic consequences of a forb-to-shrub transition, driven by a change in surface albedo, could be significant (Harte et al., 2014). Shrub albedo may also play an important climatic role around snowmelt time when the overwintering foliage protrudes above the shrinking snowpack and accelerates the melt process (see Feedback section).

Soil mesofauna response

Experimental heating enhanced both soil mesofaunal biomass and diversity during cool, wet summers. In contrast, in warmer, drier summers heating depressed diversity and biomass in the drier zone of the plots and diversity in the moist zone, but enhanced biomass in the moist zone. Biomass and diversity were positively correlated across plots and, under conditions of moisture stress, both positively correlated with soil organic matter. Data from closed cores indicated that the effects of heating on mesofauna were due to the in-plot effects of heating rather than to heating-induced movement of soil organisms (Harte et al., 1996). The sensitive response of soil mesofauna to an altered microclimate, and the dependence of that response on a variety of environmental factors, suggest both promise and limitations to using either experimental manipulations or natural correlations between soil microclimate and mesofaunal populations to forecast the effects of climate change.

Observing the signal of contemporary climate warming

By year 15 of the experiment it appeared that our control plots were showing trends in several of our measured variables that matched the much more dramatic trends in those same variables in the heated plots (Harte et al., 2014). These variables included snowmelt data (Fig. 2), aboveground biomass of forb and shrub production (Figs. 3 and 4), and soil organic carbon (Fig. 5). However, these trends were not statistically significant. By year 23, however, we had a sufficiently long time series to be able to establish the statistical significance of all these control plot responses. The fact that the directions of the control plot responses matched those in the heated plots, albeit at a slower pace, reinforces our confidence that our heat treatment of the plots provides a realistic preview of real global warming.

Comparison of the slopes of AGB over time showed that the control plot trend was not significantly different from the heated plot trend in forbs. The difference in the slope of the melt date data in the control and heated plots (Fig. 2) may be caused by an albedo effect—the dominant shrub, sagebrush, has aboveground overwintering stems and foliage, with relatively low albedo, that extend above the snowpack in spring during snowmelt. It is reasonable to suggest that the enhanced shrub growth actually promoted earlier melt in the immediate vicinity of the sagebrush plants, possibly explaining the marginally significant (P = 0.066) difference in slopes in Fig. 2, and in turn, through a positive feedback, contributing to the enhanced shrub growth shown in Fig. 4.

Importantly, had we just documented the control plot changes and not carried out heating manipulation, we would not be able to confidently attribute the observed responses to contemporary climate change. Many other types of disturbance afflict the subalpine meadows in this section of the Rockies, including acid deposition from coal-fired power plants, increased intrusion of cross country skiers leading to trampling of snow, and increased dust deposition from droughts in the southwest blowing into the region and from increased local traffic on nearby dirt roads. In addition, although not documented, there could well be multidecadal natural cycles in the dominance patterns of vegetation. Our controlled experiment allows us to attribute, with great confidence, the observed changes to contemporary climate change.

Some considerations in the design of climate-warming experiments

The importance of long-term studies

Throughout the 27 years of this investigation, numerous transient phenomena were observed. Had we stopped the experiment after year 5, we might well have written a paper concluding that subalpine meadows on a warmer planet will have increased rates of net nitrogen mineralization. By year 10, we realized that the observed increase in that process was a short-lived transient response to the onset of warming. Had we stopped the experiment in year 8 we might well have concluded that on a warmer planet, as shrubs replace forbs in subalpine meadows, soil carbon levels will be permanently depressed by about 25%. But by year 15, it was apparent that this too was a transient response and that soil carbon levels eventually recover. Had we stopped the experiment in year 12, we would not have had a sufficiently long time series of data to enable us to see that snowmelt date, plant species composition and cover, and soil carbon in our control plots were tracking, at a slower rate, those same variables in the heated plots. If we could run the experiment for a further 27 years we might even conclude that some of the conclusions we have recently reached (e.g., Harte et al., 2014) are not applicable on century timescales.

Unfortunately, there are major obstacles to running long-term experiments. The willingness of both government agencies and nongovernmental sources of research to initiate projects generally exceeds their willingness to sustain them beyond just a few funding cycles. Unless a climate manipulation experiment is linked to an LTER site, or perhaps to a NEON site, funding for a truly long-term (more than two decades) experiment is exceedingly difficult to sustain. Moreover, even if funding is available, sustaining interest and willingness to carry out the repair and replacement of equipment and routine monitoring, over a period of time comparable to the duration of a PI's career, poses a barrier. The best way to ensure a long-term commitment is probably to have substantial institutional involvement in the project.

Establishing sufficiently large plots

Related to longevity is the issue of experimental scale. If the study is to endure for decades, with annual monitoring of soil and perhaps plant material, the plots must be large enough to prevent possible damage from the repeated removal of material. In our experiment, we estimate we have removed about 2% of the soil in the upper zone of the plots over the course of 27 years of field sampling. If instead of having plots of 30 m² we had plots of 1 m², after only a few years we would have exceeded the possible duration of time in which innocuous annual sampling could take place.

Ant diversity and abundance provide useful examples of response variables that cannot be reliably studied even in plots as large at 30 m². Such response variables bear no resemblance to patterns observed along an elevational climate gradient, a discrepancy that is probably an effect of the spatial scale of the experimental warming. Ants respond to experimental warming in complex ways due to the physical location of their nests and their foraging area. This is a concern for warming experiments, and one that is hard to address for species that cover even modest areas during foraging (Menke et al., 2014).

Selecting a site

In ecosystems like the warming meadow where the timing of snowmelt can affect plant productivity, phenology, and much more, it is critical that experimental plots are hydrologically isolated from the surrounding landscape. For example, if we had located our experimental plots on the side of a hillslope, then some of the runoff from snowmelt on the slope above our plots would likely have infiltrated our heated plots, thereby washing out the effects of earlier snowmelt. For that reason, we located the upper elevation of each plot on a ridgeline, preventing delayed runoff at snowmelt from entering the heated plots from above.

Inadvertently we also benefited from having our 10 plots span a small but detectable aspect gradient (Fig. 1) as well as each plot having an elevational gradient. The existence of gradients across or within the plots comprising an experiment can augment the capacity to infer causal mechanisms. In particular, if the control versus treatment response of say shrub productivity to heater-induced earlier snowmelt matches the same response to earlier snowmelt in the more south-facing plots, that both reinforces the conclusion that snowmelt is a controlling variable and also bridges the timescale for response to heaters (years to decades) with the timescale for response to aspect (millennia).

Ecosystem heating experiments on plots that span a local ecotone can detect changes in community composition that arise as a consequence of local infilling. In the warming meadow plots, the huge observed increase in sagebrush cover did not occur because that species migrated in from distant lower elevation sites. Rather, small and somewhat isolated individual plants within the plots at the outset thrived under the heating, filled in the spaces between individuals, and became the dominant cover. Although current efforts to predict the future of vegetation communities mostly rely on species distribution models, the dominant changes in community composition may be driven less by migration from a far to more suitable climates and more by local infilling.

Designing treatment conditions

When I was designing plot treatment conditions I faced several choices and had many discussions with other scientists as to which options would be most fruitful. From the outset the issue of how best to simulate global warming on the plots was of prime concern. Some existing research had been carried out in the 1980s with open-top chambers but there was growing criticism of that technique (Shen and Harte, 2000). Moreover, I wanted to heat all year round because of the likelihood that warming-induced earlier snowmelt would be a dominant driver of effects on vegetation and soil. In 1988, at a dinner in an outdoor restaurant in San Francisco, the overhead heat lamps that repelled the cold evening fog triggered in me the idea of using downward-facing infrared lamps in the meadow. A search through farm equipment catalogs led me to KalGlo Electronics Co., Inc., Bath, Pennsylvania, United States. They made rugged outdoor heat lamps designed to keep piglets and chickens warm in northeast US winters. Remarkably, our heaters have now endured 28 Rocky Mountain winters without failure.

A more controversial choice had to do with the actual treatment conditions. At the time there was a growing consensus that using thermostats to achieve a fixed soil temperature increment (relative to controls) would simplify the task of deriving dose-response information from data and allow cross-experiment comparisons. But, on the other hand, a fixed increment of greenhouse gas in the atmosphere does not yield a fixed temperature increase; global warming does not come with a thermostat. Indeed, under real global warming, the incremental rise in soil temperature can differ hugely across the seasons and is strongly dependent, through the Bowen ratio effect, on how moist the soil is. Trying to achieve, say, a steady 2°C rise in very wet soil could push electric heaters beyond their capacity. Hence, I opted for a fixed level of infrared radiation flux to the ground, mimicking I hoped the relatively constant infrared flux from the great big heater in the sky, otherwise known as incremental greenhouse gases. The variations in incremental soil temperature would, I expected, be realistic, and in retrospect they were.

Finally, I had to choose a heater setting. How much downward IR flux would be required to mimic the effect of, say, a doubling of carbon dioxide in the atmosphere? The difficulty arises because doubling global atmospheric carbon dioxide warms global air as well as soil, but on small plots there is not sufficient IR absorbing capacity in the air between the heaters and the ground to overcome advection. Because soils cool faster in cool air than in warm, to achieve a 2oC soil warming I couldn't simply use the estimated value of ~ 5–10 W/m² from climate models. A back-of-the-envelope calculation suggested a flux of ~ 15 W/m² would be appropriate. Later, in 1993, I raised the output to ~ 22 W/m² because general circulation models were, by then, predicting stronger feedback enhancement of the direct effect of doubling atmospheric carbon dioxide.

Explanatory variables

Attempts to unravel the dependence of ecological and biogeochemical events on climatic variables must confront the fact that there are numerous climatic variables that could potentially be explanatory. These include mean annual temperature or soil moisture, mean annual growing-season temperature or soil moisture, date of snowmelt, degree days from date of snowmelt to the time of the event in question, and many others. Because most climate variables exhibit some degree of cross correlation, it is not a straightforward task to identify the most explanatory variable or combination of variables. For example, a study of the ability of warming experiments to accurately predict observed regional trends in phenological responses of vegetation to warming relied upon mean annual temperature as an all-purpose fitting variable. This led to the incorrect conclusion that warming experiments systematically fail to predict phenological responses, a conclusion that is contradicted if date of snowmelt is used instead (Harte and Kueppers, 2012).

Pluralistic approaches overcome obstacles to prediction

There are three basic approaches to predicting how ecosystems will respond to climate change: observations of trends along climate gradients in space or time, controlled experimental climate manipulations, and mathematical models. Each has merits and disadvantages (Dunne et al., 2004).

Observational studies along spatial climate gradients, or over periods of time long enough to capture responses to temporal trends in climate, have the advantage of studying real ecosystems, some of which can be quite large. The latter consideration is particularly important because studies in small plots cannot capture the consequences of long-distance dispersal of plants and the movement of animals. On the other hand, drawing conclusions from such studies is difficult because of the multiple exogenous forces that typically influence ecosystems. That, and the absence of controls, can confound the ability to draw conclusions about mechanisms and causation. Moreover, the timescale over which spatial patterns are created on the landscape in response to spatial climate gradients is typically of the order of thousands of years, thus reducing the usefulness of those studies for inferring how ecosystems will respond to climate change over the coming decades.

Warming experiments have the indisputable advantage of controls and differing treatments, allowing insights into causation and mechanisms to be obtained. By allowing identification of causal mechanisms governing ecosystem responses to climate change, warming experiments can provide the information needed to extrapolate findings to ecosystems that are not directly the subject of intense observational or experimental study. On the other hand, such experiments are necessarily limited in spatial extent and thus may fail to capture responses, like the interactions of local vegetation communities with large herbivores, that are intrinsically of large scale.

Moreover, whatever warming technique is deployed, it will inevitably not reproduce all the physical details that characterize actual anthropogenic climate change, and thus may impose unrealistic environmental conditions (Shen and Harte, 2000). Finally, few warming experiments endure long enough to provide assurance that observed responses are not simply transient effects.

Under the broad category of experimental manipulations, the use of reciprocal transplants and common garden experiments along natural spatial climate gradients should be included. As discussed above, these can provide useful auxiliary information that reduces the spatial size limitation of active heating experiments.

Models can be designed to investigate single- or multiple-factor disturbances and, at least in principle, to overcome spatial and temporal limitations. Of course, in practice the output from models is only as good as the assumptions about dominant driving mechanisms, and the input empirical data. Without the insight and data obtained from observational and experimental studies, models are at best exploratory heuristic tools.

To derive the greatest advantage from warming manipulations we conclude from our research that combining the results from warming experiments, carried out at several sites along an elevational or other climate gradient, with data from purely observational studies along gradients in space and time, and combining the insights from those approaches with mathematical models, is the most effective way to make reliable predictions about future climate-ecosystem interactions (Dunne et al., 2003; Harte and Kueppers, 2012).

The space-for-time assumption

Warming experiments are expensive. A far cheaper alternative is application of the space-for-time assumption, allowing inference of climate-driven changes over time from variability along climate gradients. Indeed, we can conclude from our studies that a straightforward space-for-time (or more accurately space-for-treatment) assumption holds for some response variables, including snowmelt date, plant phenology, and shrub and forb aboveground biomass, but it fails for others, and in particular for soil carbon. Nevertheless, the DWP model provides a way to relate soil carbon responses over time to its variability in space (Saleska et al., 2002; Dunne et al., 2004; Harte et al., 2014). Without data from the warming experiment, however, there would be little grounds for confidence in that model.

A population perspective

Although much of the literature on the effects of climate change on plants and animals has focused on species as the unit of analysis, there are many reasons to put more effort into the study of population-level responses. If the populations that comprise a species are distributed along, say, a temperature gradient, and each population is adapted to its local climate conditions, then this can have major implications for the survival of the species under warming. Let us suppose that individuals cannot migrate fast enough to reach suitable climates, and that rates of selection of traits that could allow local adaptation are also too slow. Then each population could be at risk of extinction as the climate warms (Harte et al., 2004). If each population is driven to extinction, then so is the species. In contrast, if populations are ignored and research is focused solely at the species level, then one might well conclude, using a standard species distribution model under the conditions described above, that only individuals at the warm end of the range were at risk (Harte et al., 2004), leaving the majority of individuals in the species, and thus the species itself, intact. This is illustrated in Fig. 8.

Fig. 8 Illustration of why population distribution models, and not just species distribution models, are critical to understanding the fate of species.

More generally, a failure to investigate population-level effects can seriously contribute to environmental myopia. Virtually all examinations of extinction rates consider species extinctions, yet populations within species are becoming extinct at rates that are orders of magnitude higher than those of species (Ceballos et al., 2017). Since it is those populations that deliver ecosystem services that support civilization, species-focused studies lead society to vastly underestimate the threat of the sixth mass extinction episode we are now