Terrestrial Carbon Cycle Variability

Introduction

In today’s world, CO₂ concentrations have risen beyond 400 ppm, a level not experienced over the past 800,000 years¹. This rise in atmospheric CO₂ is mostly due to fossil fuel emissions² and is largely responsible for a 1.5°C increase in air temperatures over land since the 1880s³. Together, rising CO₂ concentrations and temperatures are causing the carbon cycle to experience greater year-to-year perturbations and trends than those experienced during the past deglaciation events^4,5.

Today, the constituent fluxes and pools of the terrestrial carbon cycle are widely out of equilibrium from pre-historical conditions owing to human activities. For perspective, atmospheric CO₂ increased from about 180 to 280 ppm since the last glacial period, adding about 220 Pg-C to the atmosphere over a 10,000-year period⁶; this pre-industrial change was associated with a positive trend in the net global carbon exchange rate of about 20 Tg-C y^-1. In comparison, atmospheric CO₂ is increasing at a rate of about 4.4 Pg-C y^-1, as fossil fuel and cement production release 9 +/- 0.5 Pg-C y^-1, land use change releases 0.9 +/- 0.5 Pg-C y^-1, and terrestrial ecosystems assimilate 3 +/- 0.5 Pg-C y^-17.

Net carbon exchange by terrestrial ecosystems is expected to be variable and changing in our warming and CO₂-enriched world. This expectation is based on the fact that the rates of photosynthesis are tied to CO₂ and temperature and that respiration is tied to temperature, photosynthesis, and the size of the carbon pools. From first principles, we know that photosynthesis will increase with CO₂ concentrations in a diminishing returns fashion, defined by Michaelis-Menten reactions associated with the carboxylation reactions between CO₂ and ribulose bisphosphate. In addition, increased temperatures accelerate kinetic rates of enzyme reactions, thereby increasing mitochondrial respiration of plants and microbes.

On a year-to-year basis, the secular warming of the Earth’s climate is causing different regions of the world to experience episodes of extremes such as wetness, dryness, hot, and cold, which can perturb the fluxes of carbon to and from the plants and soils of those regions^8–10. And, on annual to decadal time scales, changes in land use, phenology, greenness of the biosphere, fires, and nitrogen deposition are introducing additional variability and trends on components of the carbon cycle^7,11–13.

How variable is the carbon cycle?

To answer this question with confidence, we have to separate trends and induced variability from natural variability and random sampling errors. We are entering an era where the sampling record is starting to be long enough to separate signal from noise. We have a 50-year record of CO₂ concentrations in the atmosphere, providing a top-down constraint on carbon cycle variability⁷, and we have a 30-year satellite record, giving us bottom-up information on spatial/temporal variability of the greening of the land¹³. Consequently, there has been growing activity to quantify and understand variability of the carbon cycle based on these longer global-scale records. The objective of this review is to survey key literature on the variability of the terrestrial carbon cycle at the global scale published over the past 4+ years (2012 into 2016).

To assess any attribution in the variability of the terrestrial carbon cycle, one must consider the degree of differential and induced modulations of its three major carbon pools. The vegetation, soil, and atmosphere carbon pools have different sizes, different turnover times, and different responses to environmental perturbations, like light, CO₂ concentration, temperature, and soil moisture¹⁴. In other words, there is a disequilibrium between the gains and losses of carbon to and from the plant and soil pools, which can partly be explained by the relatively fast way CO₂ enters the biosphere via photosynthesis and the relatively slow way it leaves via plant, root, and soil respiration. Consequently, these carbon pools have different susceptibility to anomalous weather and climate variability at global and regional scales. Furthermore, the variability of carbon fluxes is dependent upon changes in such ecological factors as plant functional type, leaf area index, time since disturbance and stand age, nitrogen loading, the intensity and frequency of fires, soil erosion, and transport as dissolved carbon.

Given the superposition of natural (by weather and fire) and human-induced variability (by climate change, increasing CO₂, changing land use, changing forest age distributions, pollution, and nitrogen deposition) on the carbon cycle⁸, can we detect signals or responses among simultaneous variation of numerous drivers? Secondly, are the short-term changes in carbon pools large enough to detect with current observation systems? In assessing if variability in the carbon cycle is occurring, we must consider the detection limit and sampling errors of these systems as they affect how precise or accurate they are.

One group of scientists has reported that the amplitude of the seasonal swing in atmospheric CO₂ is growing^4,15,16. To illustrate this point, we show publicly available data from the long-term monitoring station at Mauna Loa, Hawaii (Figure 1). This figure shows that the magnitude of the difference between the maximum and minimum values of CO₂ concentration for each year, after the time series was detrended with a moving filter, has increased by nearly 15% over 55 years. When the global fields of CO₂, from a network of CO₂ sampling sites, are considered, it has been deduced that the annual global carbon uptake doubled from 2.46 to 5.06 Pg-C y^-1 between 1960 and 2010, a rate of 52 Tg-C y^-25.

Figure 1. Variation in the amplitude of CO₂ at Mauna Loa, Hawaii.

Data are courtesy of NOAA Global Monitoring Division (http://www.esrl.noaa.gov/gmd/), Ed Dlugokencky and Pieter Tans, NOAA/ESRL (www.esrl.noaa.gov/gmd/ccgg/trends/).

What is the explanation for this temporal increase in CO₂ amplitude? The answers are manifold. Scientists have reported that this trend is a consequence of the greening of the biosphere¹³, stronger northern latitude photosynthesis^4,15, more photosynthesis by semi-arid ecosystems^17,18, agriculture and the green revolution^16,19, tropical temperature anomalies²⁰, or increased winter respiration²¹.

Evidence for the greening of the terrestrial biosphere is provided by an analysis of satellite remote sensing data. The scientists report that seasonally integrated leaf area index is increasing across a quarter to one-half of the planet’s vegetated area, mostly to the CO₂ fertilization effect¹³. Browning is detected too, but on only 4% of the vegetated land area.

Other scientists point to the northern latitudes, home of the world’s boreal forests, as the locale for a growing carbon sink. Graven et al.⁴ reported that a 30 to 60% change has occurred in the carbon exchange of boreal forests. Forkel et al.¹⁵ lend support to this hypothesis by reporting that “the latitudinal gradient of the increasing CO₂ amplitude is mainly driven by positive trends in photosynthetic carbon uptake caused by recent climate change and mediated by changing vegetation cover in northern ecosystems”. Others claim winter conditions may have a strong influence on year-to-year variations in the growth of the CO₂ concentration amplitude. Yu et al.²¹ argue that warmer winters have less snow, which causes soil temperatures to be colder, reducing winter respiration. They conclude that this mechanism explains 25% of the enhancement of the carbon sink across boreal forests.

In addition, not all scientists agree on whether the activity of boreal forests explains the variability of the terrestrial carbon cycle. One team²⁰ reported that 50% of the interannual variation of the CO₂ growth rate, between 1959 and 2011, was associated with tropical air temperature; a 1°C anomaly in tropical air temperature corresponded with a 3.5 +/- 0.6 Pg-C y^-1 anomaly in the CO₂ growth rate. Others attribute the trends in the global carbon sink to the growth of semi-arid vegetation^17,18; semi-arid ecosystems are experiencing a 0.04 Pg-C y^-2 trend in their carbon sink, which is about 57% of the global trend. Two other groups point to the role of agriculture as a modulating factor of the global CO₂ concentration record. Gray et al.¹⁶ and Zeng et al.¹⁹ attribute a significant part of the increase in the CO₂ seasonal amplitude (17 to 25%) to the agricultural green revolution; cereal production in the northern hemisphere increased by 240 percent over the 47-year period between 1961 and 2008, thereby increasing the net carbon uptake of crops by 0.33 Pg-C.

Deconstructing the global carbon cycle

To better understand how and why the terrestrial carbon cycle is experiencing variability, let’s examine the potential for modulation and or trends in the distinct plant and soil carbon pools.

Global gross primary productivity

The carbon cycle starts with the input of carbon through gross primary productivity. What is the value of global gross primary productivity? There is no single sensor or method to perform this assessment perfectly²². Consequently, scientists are using networks of carbon flux measurements and meteorological stations with remote sensing and machine learning techniques to produce spatially resolved flux maps on monthly scales that can be summed to produce a global estimate^23,24. These empirically based, machine learning estimates of global gross primary productivity range from between 119 +/- 6 and 123 +/- 4 Pg-C y^-1.

Given this uncertainty, how variable is global gross primary production on a temporal basis? One could assume that, to a first order, global photosynthesis may experience little year-to-year variation at the global scale because it is primarily a function of solar radiation, and the solar constant is relatively stable at 1360 +/- 0.5 W m^-225.

Yet other studies beg to differ and report significant variation and trends in global photosynthesis. For example, a synthesis of 10 modeling approaches for assessing global gross primary productivity showed a range of between 112 and 169 Pg-C y^-1, a mean of 138 +/- 17 Pg-C y^-1, interannual variability on the order of 2.64 +/- 1.12 Pg-C y^-1, and a sensitivity of the long-term trend of 33 +/- 23 Tg-C y^-226,27.

Temperature is bound to have different effects on photosynthesis on a regional basis. Piao et al.²⁷ report that global gross primary productivity has a negative relationship with temperature in the tropics and a positive relationship with temperature in the boreal regions owing to a longer growing season. At the global scale, they found that interannual variability in gross primary productivity is not correlated to its global temperature anomalies. Year-to-year variations in air temperature can also affect photosynthesis indirectly by its impact on phenology and the length of the growing season^28,29. Emergent properties like the acclimation of photosynthesis to temperature must be considered too³⁰ when contemplating the change in gross primary productivity to warmer temperatures.

Regional drought can limit photosynthesis by causing physiological stress^10,31. Integrating these drought effects globally, carbon cycle models reveal that average global gross primary productivity increases by 4.1 +/- 2 Pg-C y^-1 per 100 mm of precipitation²⁷. In humid tropical regions, moderate drought can enhance photosynthesis since it leads to reduced cloud cover and produces more incident sunlight³².

Extreme climate events may have a disproportionate effect on global gross primary productivity. One study produced a 30-year record of the variability of global gross primary productivity using remote sensing data from a global flux network and a set of carbon cycle and dynamic vegetation models³³. The authors reported that a few extreme events explain most of the interannual variability in global gross primary productivity; 78% of the global anomaly in gross primary productivity is associated with 200 extreme events in the tenth percentile³³.

The examination in temporal trends in global gross primary productivity, inferred from long-term satellite records, has the potential to detect if the terrestrial biosphere is experiencing a response to rising CO₂ on global carbon assimilation. A new study³⁴ assessed the large divergence between satellite and Earth system models regarding the CO₂ fertilization effect on the carbon cycle. The authors found that net primary productivity derived from satellites increased by 2.8% from 1982 to 2011. In contrast, estimates of global net primary productivity, the difference between gross primary productivity and autotrophic respiration, derived by Earth system models increased by 7.6% over 30 years. Smith et al. conclude that Earth system models may be oversensitive to CO₂ effect if the satellite inferred method is correctly sampling the response of the biosphere to a secular increase in CO₂. A comment on the Smith et al. paper³⁵, however, points out that the satellite estimates used explicitly exclude the direct effect of CO₂ on gross primary production and suggests that the opposite conclusion is a more appropriate interpretation of the Smith et al. data: that remote sensing estimates likely under-predict the response of gross primary production to CO₂.

Another set of studies infer that long-term trends in global primary productivity may be emerging owing to enhancement by elevated carbon dioxide and fertilization by nitrogen deposition. Schimel et al.¹² asked, “what is the effect of rising CO₂ concentration on the carbon cycle?” They argued that theory predicts that the enhancement of photosynthesis by CO₂ should have a tropical maximum. They evaluated data over the 2000 to 2010 timeframe using nine process models. They concluded that there was significant tropical uptake. Their results suggest that up to 60% of the present-day terrestrial sink is caused by increasing atmospheric CO₂. As for hard numbers, they report that the best estimate of the tropical + southern CO₂ enhancement effect was a sink of -1.4 ± 0.4 Pg-C y⁻¹; negative values indicate a loss of carbon from the atmosphere and a gain by the biosphere. Such debates highlight the large unknowns that remain regarding how global photosynthesis is responding to changes in atmospheric CO₂.

Land use change can affect regional and global photosynthesis both positively and negatively. Land use change and rates of land use change affect the extent or direction of the change (deforestation or reforestation) in the green land area and the number of leaves intercepting photons¹³.

Questions remain as to the certainty of inferring small changes in global photosynthesis with confidence given the degree of measurement and sampling errors that are associated with upscaling photosynthesis to the global scale. One can estimate the 95% confidence interval that random trends in global gross primary productivity must exceed. We assumed that the uncertainty about global gross primary productivity with an empirical artificial neural network method²³ is on the order of +/- 4 Pg-C y^-1. Next, we drew a random population (n=1000) of a 30-year time series, about this error, using a Gaussian random number generator. We then fit linear regressions through each of these randomly probable trends and produced a histogram of their slopes (Figure 2). We found that trends in global gross primary productivity must have a slope exceeding +/- 5.3 Tg-C y^-2 to exceed the 95% confidence interval of the randomly sampled trends. If we assume the interannual uncertainty in global gross photosynthesis is on the order of 2.64 Pg-C y^-1, as shown, the 95% confidence interval of the slope is bound within +/- 3.4 Tg-C y^-2.

Figure 2.

a) Trends in global gross primary production (GPP) of a 30-year-long time series derived from a population of random numbers that were sampled from a Gaussian distribution of +/- 4 Pg-C y^-1 1000 times; b) histogram and 95% confidence interval of the slopes derived from the populations of slopes computed in part a). The standard deviation is 0.0856 Pg-C y^-2 and the 95% confidence interval is +/-5.3 Tg-C y^-2.

Clearly, more work and longer datasets are needed to resolve the contrasting conclusions derived from the satellite-based inferences and the model upscaling methods of global gross photosynthesis. New efforts are underway to produce independent estimates and constraints on global photosynthesis with sun-induced fluorescence³⁶ and to expand existing global networks of land-atmosphere CO₂ exchange, which should shed new light on this question³⁷.

Concluding remarks

The variability of the global carbon cycle is changing, as the amplitude of seasonal CO₂ concentration and the net land sink are increasing. The causes of these trends in the global carbon cycle remain uncertain and a challenge for future research.

In this review, we show that different analyses have different explanations for why the seasonal CO₂ amplitude and land sink are increasing. We advise investigators of future studies to test whether or not reported trends are significantly different from random errors associated with the uncertainties in modeling and measuring gross and net carbon fluxes.

Photosynthesis is the starting point of the carbon cycle. Yet, despite decades of research, our ability to produce an estimate of global photosynthesis with narrow confidence intervals and high accuracy remains elusive. Only with more accurate estimates of photosynthesis will we be able to better resolve and understand differences between data-driven and process models and to better simulate responses between net and gross carbon fluxes to temperature, precipitation, CO₂, and nitrogen.

The response of global photosynthesis to CO₂ fertilization remains unresolved, as the variations in the assimilation fluxes are small relative to the uncertainties of the drivers and information systems, using bottom-up and top-down methods. Indirect effects can be important too. For example, lengthening growing seasons can increase gross primary productivity on regional scales⁴³. Vegetative regrowth may be offsetting losses of carbon by preceding fires. Thawing of the permafrost may expose and release vast stores of carbon that have been decoupled for millennia, while warming northlands are getting greener. Many experiments show acclimation effects of respiration to temperature⁴⁴. This is a process that is not well considered by models, though some are starting to consider this factor³⁰.

In closing, we need to continue the collection of long-term and multi-faceted products that are used to assess the global carbon cycle. Additional data from new satellites and longer time series from eddy covariance and CO₂ concentration networks will reduce uncertainty in sampling and modeling and improve our ability to answer the questions associated with carbon cycle variability and trends.