A global model for the uptake of atmospheric hydrogen by soils

HAL Id: hal-03206186

https://hal.archives-ouvertes.fr/hal-03206186

Submitted on 24 Apr 2021

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

A global model for the uptake of atmospheric hydrogen

by soils

C. Morfopoulos, P. Foster, P. Friedlingstein, P. Bousquet, I. Prentice

To cite this version:

C. Morfopoulos, P. Foster, P. Friedlingstein, P. Bousquet, I. Prentice. A global model for the uptake

of atmospheric hydrogen by soils. Global Biogeochemical Cycles, American Geophysical Union, 2012,

26 (3), �10.1029/2011GB004248�. �hal-03206186�

A global model for the uptake of atmospheric hydrogen by soils

C. Morfopoulos,

P. N. Foster,

P. Friedlingstein,

P. Bousquet,

and I. C. Prentice

Received 22 November 2011; revised 15 June 2012; accepted 8 July 2012; published 15 August 2012.

[1]

A simple process-based model for the consumption of atmospheric hydrogen (H

)

has been developed. The model includes a description of diffusion and biological processes

which together control H

flux into the soil. The model was incorporated into the

LPJ-WHyMe Dynamic Global Vegetation Model, and used to simulate H

fluxes over

the 1988

–2006 period. The model results have been confronted with field and laboratory

measurements. The model reproduces observed seasonal cycles of H

uptake at different

sites and shows a realistic sensitivity to changes in soil temperature and soil water content

in comparisons with field and laboratory measurements. A recent study, based on 3D

atmospheric model inversion, found an increase of the global H

sink from soils, with a

trend of

0.77 Tg a

for the 1992–2004 period (fluxes are negative as soils act as a sink

for atmospheric H

). For the same period, however, our process-based model calculates a

trend of only

0.04 Tg a

. Even when forced with drastic changes in soil water content,

soil temperature and snow cover depth, our model is unable to reproduce the trend found

in the inversion-based study, questioning the realism of such a large trend.

Citation: Morfopoulos, C., P. N. Foster, P. Friedlingstein, P. Bousquet, and I. C. Prentice (2012), A global model for the uptake of atmospheric hydrogen by soils, Global Biogeochem. Cycles, 26, GB3013, doi:10.1029/2011GB004248.

1.

Introduction

[2] Molecular hydrogen (H2) is the second most abundant

oxidizable gas in the troposphere after methane (CH4) with

an average mixing ratio of 531 ppb [Novelli et al., 1999]. H2

has recently attracted interest as hydrogen-based technolo-gies, which are sustainable, clean and transportable, are widely regarded as a future energy alternative to traditional fossil fuels [Larsen et al., 2004]. However, little is currently known about the possible environmental effects of wide-spread use of hydrogen for fuel. In a hydrogen economy, inevitable H2leakage into the atmosphere would contribute

to an increase in the mixing ratio of H2. On the other hand,

switching to hydrogen-based technologies could in principle reduce fossil fuel combustion, one of the main current H2

sources into the atmosphere, thereby contributing to a decrease in atmospheric H2[Warwick et al., 2004; Schultz

et al., 2003]. Hydrogen acts as a sink for hydroxyl radicals (OH); an increase in the atmospheric H2 burden would

therefore reduce OH availability, increasing the lifetime of CH4 and thus contributing to the greenhouse effect. An

increase in atmospheric H2would also increase water vapor

in the stratosphere through H2oxidation leading to changes

in stratospheric temperature and ozone chemistry [Tromp et al., 2003]. A further complication is the potential impact of climate change on the uptake of H2by soils, which is the

largest sink for atmospheric H2. The magnitude and even the

sign of this impact are essentially unknown.

[3] A simplified schematic of the atmospheric H2cycle is

represented in Figure 1. Two main chemical processes pro-duce atmospheric H2. The first is photochemical H2

pro-duction through photolysis of formaldehyde (HCHO), which is a product of the oxidation of methane (CH4) and other

volatile organic compounds (VOCs). This source represents around 50% of total H2 production. The second is

incom-plete combustion during biomass burning and fossil fuel combustion. This source accounts for about 40% of total H2

production. Biological N2fixation on land and in the ocean

constitutes a further, minor source of atmospheric H2.

Molecular hydrogen is removed from the atmosphere in two main ways. The dominant sink for H2is soil uptake due to

bacterial and/or extracellular enzymatic activity. Soil H2

consumption amounts to 70–80% of the total sink. Oxida-tion of H2by OH is the second major loss pathway. A small

part of tropospheric H2manages to reach the stratosphere,

where it is destroyed by reactions with OH and O(1D) radi-cals [Ehhalt and Rohrer, 2009; Constant et al., 2009].

[4] The detailed mechanisms by which H2is consumed

in soil are still unclear. Guo and Conrad [2008] focus on a family of extracellular hydrogenases of microbial origin. However, more recently Constant et al. [2010] have con-cluded that Streptomycetes (aerobic bacteria) are the main group responsible for H2soil uptake. Thus, the mechanisms

controlling H2 uptake in the soil are still subject to fairly

large uncertainties.

1

Division of Biology, Imperial College, Ascot, UK.

2_{Department of Earth Sciences, University of Bristol, Bristol, UK.} 3

College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter, UK.

4_{Laboratoire des Sciences du Climat et de l’Environnement, CEA,}

Gif-sur-Yvette, France.

5_{Department of Biological Sciences, Macquarie University, North}

Ryde, New South Wales, Australia.

Corresponding author: C. Morfopoulos, Division of Biology, Imperial College, Ascot SL5 7PY, UK. (c.morfopoulos@imperial.ac.uk) ©2012. American Geophysical Union. All Rights Reserved. 0886-6236/12/2011GB004248

[5] A handful of studies have attempted to estimate the soil H2sink at a global scale. Three different methods have

been used. The first consists in prescribing H2 deposition

velocities. Deposition velocities (vdin cm s
1) are defined as

the ratio of flux (mol cm
2s
1) of a gas at a sink surface to its concentration in the atmosphere (in mol cm
3). Hauglustaine and Ehhalt [2002] used previous estimations of CO deposition velocities [Müller, 1992] and applied Yonemura et al. [2000a]’s linear relationship between CO and H2 deposition velocities (vdH2/vdCO = 1.5). NPP (Net

Primary Productivity) estimates were used to constrain the seasonal and geographical distribution of CO and H2

depo-sition velocities. Sanderson et al. [2003] used measured H2

deposition velocities and their variation with soil water content and ecosystem type to derive a bottom-up global estimate. Finally, Price et al. [2007] used a simple scheme with a constant H2deposition velocity of 3.94 10
2s
1

over non-snow covered grid cells. To take into account the effect of low soil temperature, H2 deposition velocity was

reduced by half below 0C and again by half below
15C. The second method consists in using atmospheric model inversions. Xiao et al. [2007] estimated global soil uptake from surface atmospheric H2 observations by using a 2-D

global transport model and state-space Kalman filter. Recently, Bousquet et al. [2011] used a new approach to estimate the total soil sink using an atmospheric inversion of global and regional fluxes based on a three-dimensional chemistry-transport model, a global network of flask obser-vations of H2 concentration (NOAA/ESRL and CSIRO/

CAWS networks) [Novelli et al., 1999; Steele et al., 1992], and prior information on natural and anthropogenic fluxes, in a Bayesian inversion framework. One of the findings of Bousquet et al. [2011] is a long-term trend of
0.77 Tg a
2

in the soil uptake between 1991 and 2004, suggesting that the soil sink has been increasing (fluxes are negative as soils act as a sink for atmospheric H2). The third method involves

process-based modeling such that soil H2uptake is simulated

as a function of diffusion and biological oxidation. Smith Downey [2006] modeled soil H2uptake in this way, with a

global estimate ranging between 59.8 and 73.2 Tg a
1. However, her model has not been extensively confronted with laboratory and field measurements.

[6] In this paper we describe a simple, globally applicable process-based submodel representing the diffusion of H2

into the soil and oxidation of H2in the soil and its

imple-mentation in the LPJ-WHyMe Dynamic Global Vegetation Model [Wania et al., 2009a, 2009b, 2010]. We present simulations of H2fluxes for the 1988–2006 period, forced

by observed variability in climate, and we compare the results with available experimental data. Finally, we re-examine Bousquet et al. [2011]’s finding of an increasing trend in soil H2uptake.

2.

Model Implementation

[7] We developed a simple diffusion-consumption model using the same approach that Ridgwell et al. [1999] and Curry [2007] used to model the soil consumption of meth-ane. A similar approach has been adopted by Yonemura et al. [2000b] and Smith Downey [2006]. An extended description of the model can be found in the auxiliary material.1 The model includes a description of diffusion, Figure 1. Simplified scheme of the H2cycle. The sources are represented by red arrows and the sinks by

green arrows. The numerical ranges come from various authors: Novelli et al. [1999], Simmonds et al. [2000], Hauglustaine and Ehhalt [2002], Sanderson et al. [2003], Rhee et al. [2006]; Price et al. [2007], Xiao et al. [2007], Ehhalt and Rohrer [2009], and Bousquet et al. [2011].

1

Auxiliary materials are available in the HTML. doi:10.1029/ 2011GB004248.

MORFOPOULOS ET AL.: A GLOBAL H2SOIL UPTAKE MODEL GB3013

based on the Fick’s first law, and biological processes which together control H2flux into the soil. Variables controlling

diffusion are soil texture, air filled porosity and soil tem-perature. The effect of snow cover is also taken into account. The biological uptake is primary controlled by soil water content and soil temperature [Smith Downey, 2006; Smith-Downey et al., 2006], and is calculated by modulating a maximal biological uptake, kmax, with functions taking in

account the effect of soil water content and soil temperature. A universal empirical value of kmax0.038 s
1was calculated

by adjusting our global mean uptake calculated for the 1991–2005 period to that obtained by Bousquet et al. [2011] using 3-D atmospheric inversion model. This value is higher than the kmax of 0.012266 s
1 found by Smith Downey

[2006] in laboratory experiments (N. Smith-Downey, per-sonal communication, 2008). We thus performed a sensi-tivity test with a kmaxof 0.012266 s
1(Tables 1 and 2). As

biological activity requires a minimum of soil organic con-tent to be activated, we used also a Net Primary Productivity (NPP) mask for the desert. We tested the impact of this NPP mask with sensitivity tests described in the auxiliary material and in Tables 1 and 2. As the model doesn’t take in account the first few centimeters of the soil, which might be dry enough to inhibit biological uptake [Smith Downey, 2006; Yonemura et al., 2000b], our model might overestimates H2

fluxes under very dry conditions.

[8] In order to estimate the long-term trend and the sea-sonal variations of the global soil sink of H2, the

diffusion-consumption model is applied on the top layer of the Dynamic Global Vegetation Model LPJ, LPJ-WHyMe (LPJ Wetland Hydrology and Methane) [Wania et al., 2009a, 2009b, 2010]. H2fluxes were calculated for the 1988–2006

period at a spatial resolution of 1by 1for the region 60S to 90N,

3.

Results

3.1. Model Evaluation

[9] The full range of H2 deposition velocities that the

model can calculate ranges between zero and 0.12 cm s
1. This range covers most of the typical measurements [Ehhalt and Rohrer, 2009] although a few studies have reported deposition velocities greater than 0.12 cm s
1[Conrad and Seiler, 1985; Ehhalt and Rohrer, 2009].

[10] Figure 2 shows the model sensitivity to changes in temperature and soil water content for a loamy sand soil (Table S1 in Text S1 in the auxiliary material). The response of the simulated H2deposition velocity to changes of soil

water content is mainly due to its effect on the H2diffusion

into the soil, except for low soil water content when the biological uptake is inhibited in the model. On another hand, the response to changes in temperature is mainly driven by the biological part of the model.

[11] We now compare the model against several sets of H2

uptake values in the literature including laboratory and field measurements as well as estimates based on atmospheric measurements.

[12] Lallo et al. [2008] used a soil chamber technique to measure H2 deposition velocity in Finland. The

measure-ments were made at Loppi (6038′N, 2421′E) and Helsinki (6012′N, 25′3′E) between August 2005 and April 2007. At each site, measurements were done on two different types of soil. Figure 3 shows observed H2deposition velocities for

the period August 2005–December 2006 [Lallo et al., 2008]

Table 1. Description of the Different Sensitivity Tests Done With the Modela

Test Label Description kmax(s
1)

V_f(NPP)2 Test with a progressive NPP mask:

f NPPð Þ ¼ 0:1  ∑yNPP y ; ∑yNPP y < 10gCm
2a
1 1; ∑yNPP y ≥ 10gCm
2a
1 8 > > < > > : 0.0317

V_no_f(NPP) Test with no NPP-based mask 0.024

V_kmax_NSD Test with kmaxfrom N. Smith-Downey (personal communication, 2008) 0.012266

V_f(M)2 Test with a new soil water content function: f Mð Þ2¼ 0; M < 2% 7:69M
0:153; 2% ≤ M < 15% 1; M ≥ 15% 8 < : 0.0363 a

The maximum biological uptake (kmax) has been recalculated so the global mean uptake calculated for the 1991–2005 period matches the one obtained

by Bousquet et al. [2011], except for the test with a kmaxfrom N. Smith-Downey (personal communication, 2008).

Table 2. Calculated Global Soil H2Uptake Annual Trends Between 1992 and 2004 in Tg a
2for Model Tests as Described in Table 1

and for the Different Scenarios as Described in Table 4

Scenario Test Ref WF
1 WF
5 WF + 1 WF + 5 T
1 T + 1 S
5 V_f(NPP)2
0.042
0.25
0.29 +0.19 +1.28 +0.003
0.08
0.11 V_no_f(NPP)
0.038
0.23
0.21 +0.17 +1.2 +0.005
0.08
0.1 V_kmax_NSD
0.024
0.14
0.2 +0.11 +0.76 +0.001
0.05
0.07 V_f(M)2
0.04
0.28
0.89 +0.22 +1.36 +0.005
0.083
0.12

plotted against the global H2 model results for the same

period at the grid cell corresponding to the measurement site coordinates. Generally, the model outputs and the data are in good agreement but at the grid cell corresponding to the Loppi site, low levels of soil water content calculated by LPJ-WHyMe between June and September 2006 lead to smaller values of deposition velocities for this period. The model slightly underestimates maximum values of H2

deposition velocities compared to the field measurements (0.054 cm s
1versus 0.07 (0.006) cm s
1). Nevertheless, for the Helsinki site, the model attends to well reproduce the seasonal cycle with a correlation (r) of 0.84. For the Loppi site the correlation is low, however, mainly due to the

anomalously low values of deposition velocities simulated during the summer of 2006.

[13] Figure 4 shows the temperature dependency of H2

deposition velocity as measured by Lallo et al. [2008] during the field measurements and compared to the ones simulated for the same period at the corresponding grid cells. In gen-eral, the temperature response is similar between the data and the simulations. The drop in simulated H2 deposition

velocities at the grid cell corresponding to the Loppi site is due to low soil water content simulated by LPJ-WHyMe during the summer of 2006.

[14] Lallo et al. [2008] also performed laboratory chamber measurements on soil collected for the Loppi site to Figure 2. Response of the H2soil model to variation in temperature and volumetric soil water content

(M ¼fwater

f ) for a sandy loam soil. (a) Effect of temperature only on H2diffusion with a maximal constant

biological uptake. (b) Addition of the soil water content effect on H2diffusion. (c) Addition of effect of the

soil water content on biological uptake. (d) Full model sensitivity to variation of temperature and soil water content. The temperature range between
15 and 50 C, the soil water content between 0 and 1, and the deposition velocity between 0 and 0.08 cm s
1.

MORFOPOULOS ET AL.: A GLOBAL H2SOIL UPTAKE MODEL GB3013

Figure 3. Digitized H2 deposition velocities in cm s
1 as measured by Lallo et al. [2008] at Loppi

(6038′N, 2421′E) and Helsinki (6012′N, 25′3′E) between August 2005 and December 2006 and H2

deposition velocities simulated by the model for the same period at the corresponding grid cells.

Figure 4. Hydrogen deposition velocities as function of soil temperature. Digitized H2deposition

veloc-ities in cm s
1are from field measurement [Lallo et al., 2008] at Loppi (6038′N, 2421′E) and Helsinki (6012′N, 25′3′E) between August 2005 and December 2006 and the simulations were made for the same period at the corresponding grid cells.

investigate the response of H2deposition velocity to changes

in temperature and soil water content. Figure 5 shows the comparison between the laboratory result and our model results for the mineral soil. The model response to changes in soil water content is in good agreement with the data, with a correlation of 0.74. However, the data suggest that bio-logical uptake starts at lower soil water content. Therefore, we tested our model (test V_f(M)2, Table 1) with a new soil water content function such that biological uptake starts at a soil water fraction of 2% instead of 8%. The new soil water content function becomes:

f Mð Þ2¼ 0; M < 2% 7:69M
0:153; 2% ≤ M < 15% 1; M ≥ 15% ; 8 < : where M ¼fwater

f andfwateris the water filled porosity.

[15] Simulations with the new soil water content function (Figure 5) give better results, with a correlation of 0.90. This version of the model is tested further in section 3.3. Figure 5 also shows the model soil temperature response compared with the one measured by Lallo et al. [2008] under labora-tory conditions.

[16] Hammer and Levin [2009]’s estimates, based on atmospheric H2 and

222

Rn observations in the boundary layer, of H2 deposition velocity at Heidelberg, Germany

(4924′N, 842′E) were compared with the global model simulation at the grid cell corresponding to the site coordi-nates (Figure 6). The measurements are representative of a suburban area influenced by local atmospheric pollution and where a large proportion of soils are covered by asphalt, affecting soil properties. These particular conditions make the comparison between data and simulation difficult. However, the monthly mean seasonal cycle estimated by Hammer and Levin [2009] between January 2005 and July 2007 was compared with the monthly mean seasonal cycle estimated by the model between January 2005 and Decem-ber 2006, as no simulations have been performed for the year 2007. The H2 deposition velocities simulated by the

model are higher than Hammer and Levin [2009]’s estima-tions with a mean value of 4.6 0.1 10
2cm s
1compared to 3 0.7 10
2cm s
1. The model reproduces the seasonal cycle, with a minimum in February and a maximum in late summer with a correlation of 0.62. Nevertheless, the model has higher peak-to-peak amplitude compared to the esti-mates from atmospheric H2and 222Rn observations at the

Heidelberg site.

[17] Yonemura et al. [1999] carried out experimental field measurements in an upland experimental field of the National Institute of Agro-Environmental Sciences in Tsu-kuba, Japan (3601′N, 14007′E), using an open flow chamber method. The soil type in our model that Figure 5. Comparison between the H2relative deposition velocities between laboratory experiments on

mineral soils from Loppi (6038′N, 2421′E) [Lallo et al., 2008] and simulations. The mineral soil used in the experiment is a halpic podsol with a total porosity of 0.75. The model has been forced with loamy sand soil type for the same range of temperatures and soil water contents than in the laboratory experiment. Dig-itized data from Lallo et al. [2008] are shown as filled symbols, the model simulations as open symbols and solid line. (a) Soil water content (M ¼fwater

f ) dependency of hydrogen relative deposition velocities. The measurements and simulations where performed with a soil temperature of 20C. (b) Soil water con-tent dependency of hydrogen relative deposition velocities. The simulations were performed with a new soil water content function such as described in Table 1. The measurements and simulations where per-formed with a soil temperature of 20C. (c) Temperature dependency of hydrogen relative deposition velocities. The observations were measured in optimum soil water content condition and the simulations were performed with M ¼fwater

f = 0.15.

MORFOPOULOS ET AL.: A GLOBAL H2SOIL UPTAKE MODEL GB3013

corresponds most closely to the soil particle distribution described by Yonemura et al. [1999] is light clay (Table S1 in Text S1 in the auxiliary material). We performed simula-tions using light clay soil properties over a range of 22–38_C for soil temperature and a range of 35–45% of the volumetric soil water content, corresponding to Yonemura et al. [1999]’s climatic data during the field experiment between 18 and 21 August 1995. The deposition velocities calculated by the model under these conditions ranged between 0.054 cm s
1 and 0.064 cm s
1. These simulations are in good agreement with measured H2deposition velocities from Yonemura et al.

[1999], which range from 0.045 cm s
1to 0.07 cm s
1. [18] We compared measurements made by Smith Downey [2006] and Smith-Downey et al. [2008] on sand soil, in a forest ecosystem at San Jacinto Mountain Reserve (33.81N, 116.79W). When forced with a loamy sand soil type, which is the closest soil type to sand in our model, for the same range of soil temperature (1–28_{C) and the same range of} volumetric soil water content (0.05–0.22 cm
3_/cm
3_{), the}

model simulated H2 deposition velocities ranging from

0.018 to 0.056 cm s
1which correspond to fluxes ranging from 4 to 12 nmol m
2s
1 (calculated for an atmospheric concentration of 2.2 10
11mol cm
3: Table S2 in Text S1 in the auxiliary material) with a mean value of 0.037 cm s
1 (8.14 nmol m
2s
1). These results are in agreement with the values found by Smith Downey [2006] with a range 2– 14 nmol m
2s
1and a mean value of between 8 4.5 and 8.62 4.8 nmol m
2s
1depending on the sampling location. [19] The same method is applied to compare our simula-tions with enclosure measurements done at the Grenzhof field site in Germany [4924′N, 836′E] by Schmitt et al. [2009]. We forced the model with sandy clay loamy sand proprieties, for the same range of soil temperature (0–35_C) and volumic soil water content (0.05–0.3 cm
3_/cm
3_{) as in}

Schmitt et al. [2009]. We obtained a range of H2deposition

velocities between zero and 0.056 cm s
1, compared with the range of 0.008–0.082 cm s
1in the data. Our model fails to reproduce the higher values of measured deposition velocities but these correspond to only three measurements. The rest of the measurements are in the same range as our modeled values.

[20] Conrad and Seiler [1985] measured H2 deposition

velocities at three subtropical sites (Transvaal and Karoo in South Africa, Andalusia in Spain) using static and equilib-rium box techniques. The values of 0.033 and 0.01 cms
1 obtained at the Andalusia and Karoo sites at a soil temper-ature of 30C can be retrieved by our model with loamy sand type of soil for low soil water content ratio, respectively 10% and 8.5%. None of our simulations is able to simulate the high value of 0.13 cms
1found for the Transvaal site.

[21] Rahn et al. [2002a] measured H2 soil uptake at the

boreal site of Delta Junction in Alaska (6348′N, 14506′W), using a new method described in Rahn et al. [2002b], in July 2001. As no information is available on the soil type, or the soil temperature and soil water content, it is hard to make a com-parison with the model simulations. Nevertheless, for the lati-tude band of 63N the deposition velocities simulated by the model for July 2001 range between 0.031 and 0.059 cms
1, which correspond to the lower values of the range of values observed by Rahn et al. [2002a] (0.044 0.013 cms
1for the burn site and 0.073 0.015 cms
1).

[22] Finally, Guo and Conrad [2008] analyzed the effect of temperature on hydrogenase activity in soil suspension and soil extract. Soils were extracted from a deciduous forest near Marburg, Germany (5100′N, 0950′E). The response to temperature of soil hydrogenase activity measured in laboratory conditions is similar to the function we use to describe the temperature effect on biological oxidation rate, f(Tsoil), described in the auxiliary material, with an activity

about one half of the maximum activity at 0C, an increase between 0C and the optimum temperature, around 30C for the measurements, 40C for the model, followed by a decrease for temperature above the optimum.

[23] In the following section, we discuss a comparison with a global H2inversion [Bousquet et al., 2011].

3.2. Seasonal and Latitudinal Variations

[24] Figure 7 shows the spatial distribution of H2

deposi-tion velocities simulated for January, April, July and Octo-ber 2000. The H2 velocities vary from 0 cm s
1 to

0.066 cm s
1. The Northern hemisphere shows a strong seasonal cycle, mainly due to variation in temperature and Figure 6. Digitized monthly mean H2deposition velocities in cm s
1estimated by Hammer and Levin

[2009] between January 2005 and July 2007 at the Heidelberg site (4924′N, 842′E) and monthly mean H2 deposition velocities simulated by the model between January 2005 and December 2006 at the

snow cover, with a minimum simulated H2soil uptake of

29 Tg a
1occurring in January–February and a maximum of 50 Tg a
1 occurring in July. Seasonal variations of H2

deposition velocities are weaker in the Southern hemisphere. A slight seasonality is simulated in the tropics due to soil water content variations. On global average, H2 global

uptake in the Southern hemisphere is about 22 Tg a
1. Over the period of simulation 1988–2006, the averaged global H2uptake is 61 Tg a
1with most of the uptake occurring in

the Northern hemisphere, which accounts for 65% of the total soil uptake.

[25] Although we used global estimations of the soil H2

sink from Bousquet et al. [2011] to scale the maximum oxidation rate, kmax (cf. section 2 above), our estimate of

seasonality and trends is independent of their techniques and we now compare the two. We calculated the H2soil uptake

fluxes for the same regions of the globe that Bousquet et al. [2011] used for their inversions. The region map (Figure 8) is derived from the region map defined in the TRANSCOM project (http://www.purdue.edu/transcom/). In this paper, we will focus on three big regions, as defined by Bousquet et al. [2011]: the Northern regions (North America, Eur-ope, North Asia), the Tropical regions (Tropical America, Africa, Tropical Asia) and the Southern regions (all other land areas).

[26] Figure 9 shows the mean seasonal cycle for these three big regions simulated by our model and calculated, with an inversion method, by Bousquet et al. [2011]. Our bottom-up model has a stronger seasonal cycle for the Northern regions with a maximum uptake of 35 Tg a
1in July and a minimum uptake of 12.9 Tg a
1 in January– February, compared to a maximum uptake of 43 Tg a
1in July and a minimum uptake of 22.3 Tg a
1in December for results of Bousquet et al. [2011]. For the Southern and the Tropical regions, our model has a mean uptake of respec-tively 22.5 Tg a
1and 10 Tg a
1and no seasonal cycle is simulated while results of Bousquet et al. [2011] show sea-sonality for these regions. The lack of seasea-sonality in our model might be due to the difficulty of the model in repre-senting drought effects on the soil. However, no long data series of observed deposition H2 velocity are available to

test the seasonality in these regions.

[27] The total seasonal cycle obtained by our bottom-up model and through an inversion method are in good agree-ment with a maximum uptake of 72.3 Tg a
1and 78.6 Tg a
1for our model and for Bousquet et al. [2011]’s results, respectively. In both cases, the maximum uptake occurred in July and the minimum uptake occurred in February for our model (49.5 Tg a
1) and in March for the inversion results (50.8 Tg a
1).

Figure 7. Global distribution of H2 deposition velocity in cm s
1 simulated for (a) January 2000,

(b) April 2000, (c) July 2000, and (d) October 2000.

MORFOPOULOS ET AL.: A GLOBAL H2SOIL UPTAKE MODEL GB3013

Figure 8. Map of the 12 regions (11 land regions and 1 ocean region) and of the 62 atmospheric stations used in the inversion. Different symbols represent different networks: circles for NOAA, triangles for CSIRO, and stars when both NOAA and CSIRO operate (ALT, CGO, MLO, and SPO).

Figure 9. Mean seasonal variations of H2soil uptake in Tg a
1 for the 1991–2004 period. The three

regions, defined by Bousquet et al. [2011], are northern regions (North America, Europe, North Asia), tropical regions (Tropical America, Africa, Tropical Asia) and southern regions (all other land areas). The region partition is based on the TRANSCOM map. Data from Bousquet et al. [2011] are in dashed line; H2model results are in solid line.

[28] In order to better understand the contribution of each variable controlling H2uptake in our simulations, we

per-formed a sensitivity test at three grid cells: one in boreal latitudes, one in the midlatitudes and one in the tropics, in a similar way to Smith Downey [2006] (Figure 10). At the grid point 60N, 25E, we can see that the maxima are mostly controlled by the effect of soil water content on the diffusion part of the model. The minima are controlled in a minor way by the effect of soil water content and temperature on the biological uptake, but mostly the snow cover effect drives the minimal values calculated at this grid cell. At the mid-latitude grid point (42N, 72W), the simulated seasonal cycle is mostly controlled by the snow cover effect. We also note that effect of soil water content on biological uptake has no effect, suggesting that the soil water fraction is always higher than 15% for this point. Finally, the simulated sea-sonal cycle in the tropical grid cell (7S, 57W) is controlled in our model by the effect of soil water content on the dif-fusion of H2.

3.3. Interannual Variability and Long-Term Trends [29] Our simulated interannual variability of global soil H2 uptake ranges between 60.9 and 62.2 Tg a
1 for the

1991–2005 period. This is lower than estimated in Bousquet et al. [2011]’s atmospheric inversion with a global uptake ranging between 54.3 and 67.9 Tg a
1 for the same period (Table 3). Also, our model simulates a trend in the H2uptake

of
0.04 Tg a
2over the 1992–2004 period. This trend is much smaller than that inferred by Bousquet et al. [2011] of
0.77 Tg a
2_{for the same period. Note that Bousquet et al.}

[2011] gives a value of
0.12 Tg a
2 for a bottom-up

approach. This estimate was based on a previous version of our model with an inappropriate parameterization for bio-logical uptake and a simpler model of H2diffusion through

snow. Also, the version used in Bousquet et al. [2011] had not been confronted against observations.

[30] A long-term trend in the H2soil uptake, if any, should

come from trends in the forcing climatic variables (Tsoil, soil

water content). As our model shows realistic relationships between H2 uptake and climate drivers at the local scale

(see section 3.1), we use the model to investigate what trends Figure 10. Sensitivity test of the model for grid cells in boreal, temperate and tropical latitudes. Light

blue line: effect of temperature only on H2diffusion with a maximal constant biological uptake. Dark blue

line: addition of the soil water content effect on H2diffusion. Red dashed line: addition of effect of the soil

water content on biological uptake. Green line: full model sensitivity to variation of temperature and soil water content, without the effect of snow cover. Black line: full model sensitivity to variation of temper-ature and soil water content, with the effect of snow cover.

Table 3. Total Annual Uptakes in Tg for the Period 1991–2005 Estimated by Bousquet et al. [2011] and Simulated by the Bottom-Up H2Soil Model

Years Inversion Model Bottom-Up Model

1991 54.3 61.4 1992 61.6 61.4 1993 57.5 60.9 1994 55.1 61.2 1995 55.3 62.1 1996 57.2 60.9 1997 61.9 61.5 1998 62.4 61.8 1999 64.0 61.8 2000 63.5 61.5 2001 65.1 61.7 2002 64.3 61.7 2003 66.7 61.8 2004 66.9 61.8 2005 67.9 62.2 Mean 61.6 61.6

MORFOPOULOS ET AL.: A GLOBAL H2SOIL UPTAKE MODEL GB3013

in climate variables would be required to reproduce the soil H2uptake trend found in Bousquet et al. [2011]. We

per-formed four runs over the 1991–2005 period with an artifi-cial increase (decrease) of +1% per year (
1% per year) and +5% per year (
5% per year) in the soil water filled porosity, two runs with an artificial increase (decrease) in soil temperature of 0.07C per year, and one run with an artificial decrease in the snow depth of 5% per year. Table 4 summarizes these runs. The long-term trends of global H2

uptake are shown in Figure 11 and Table 5. None of these cases can reproduce Bousquet et al. [2011]’s global trend of
0.77 Tg a
2 _{for the H}

2soil sink. The strongest negative

trend is obtained for the case with a water filled porosity decreased artificially by 5% per year (WF–5), with an H2

uptake trend of
0.35 Tg a
2still less than half of the mean trend found in Bousquet et al. [2011], although compatible with the lower values of their large range (
0.25 Tg a
2 /
1.30 Tg a
2_{). However, this scenario is completely}

unre-alistic: 87% of the globe ends up with a water filled porosity under the wilting point by 2004, which means that the soils are too dry for vegetation to grow.

[31] To be sure that this finding is independent of the particular parameterization we chose for the biological response to soil organic content and soil water content, and for maximum biological uptake, we performed the same runs as described in Table 4 for different versions of the model as described in Table 1. The results of these runs and the trends associated are summarized in Table 2. Only one scenario, with a global trend of
0.89 Tg a
2, produces a trend as high as the results from Bousquet et al. [2011]. This scenario corresponds to the version with a new biological soil water

Table 4. Description of the Different Scenarios and Sensitivity Tests Used to Calculate Monthly H2Soil Uptake Anomalies Shown

in Figure 8

Scenarios

Label Description

B_2011 3-D inversion model estimations Bousquet et al. [2011]

Ref Standard run for H2model forced with

climatic data and LPJ-WHyMe outputs WF
1 Soil water filled porosity is artificially

decreased by 1% per year for the 1992–2005 period

WF
5 Soil water filled porosity is artificially decreased by 5% per year for the 1992–2005 period

WF+1 Soil water filled porosity is artificially increased by 1% per year for the 1992–2005 period

WF+5 Soil water filled porosity is artificially increased by 5% per year for the 1992–2005 period

T
1 Soil temperature is artificially decreased by 1C between 1992 and 2005

(0.07C per year) T+5 Soil temperature is artificially

increased by 1C between 1992 and 2005

(0.07C per year)

S
5 Snow depth is artificially decreased by 5% per year for the 1992–2005 period

Figure 11. Interannual variations of H2soil uptake for the scenarios summarized in Table 4 in Tg a
1for

the 1992–2004 period. Deseasonalized H2fluxes are plotted for three groups of regions and globally. The

three regions, as described by Bousquet et al. [2011], are Northern regions (North America, Europe, North Asia), Tropical regions (Tropical America, Africa, Tropical Asia) and Southern regions (all other land areas). The region partition is based on the TRANSCOM map. A 13-month running mean is applied to the monthly fluxes to calculate the deseasonalized values.

content dependency such as described in Table 1 and for the unrealistic scenario where filled porosity decreased artifi-cially by 5% per year (WF–5).

4.

Discussion

[32] We have developed a simple global process-based model of atmospheric H2consumption by soils. The model is

based on an exact solution of the one-dimensional diffusion-reaction equation in a unique soil layer and includes a parameterization of biological oxidation. The model was incorporated in a Dynamic Global Vegetation Model and simulations have been performed for the 1988–2006 period. The model results were tested against available experimental data and performed satisfactorily.

[33] The global seasonal cycle simulated by our model is due to the Northern hemisphere. No seasonality was shown for the Southern hemisphere. According to our model, 65% of the global H2uptake by the soil occurs in the Northern

hemisphere.

[34] The model outputs were compared with a recent inversion study [Bousquet et al., 2011]. More specifically, we tested our model with different scenarios in order to see if changes in climatic inputs could produce an increase of the simulated H2soil sink by 0.77 Tg a
2. As pointed out by

Bousquet et al. [2011], the large range (
0.25 Tg a
2 _to

1.30 Tg a
2_{) in their long-term trend of H}

2 soil uptake

indicates a weak robustness of their results, which appear to be sensitive to the inversion setup and more precisely to the long-term trend of the fossil-fuel and N2related sources of

H2. Only for one scenario is our model able to predict a

long-term trend of this magnitude. However, this scenario leads to a desert planet that obviously is not realistic. In general, our process-based model is unable to simulate the long-term trend calculated through the inversion method, even by forcing extremely the different climatic variables controlling H2uptake by the soils, confirming the lack of robustness of

Bousquet et al. [2011]’s long-term trend. Indeed, there is no long-term significant trend observed in atmospheric H2

concentration itself. The increase in H2soil uptake inferred

by the inversion method was compensated by a similar increase of the atmospheric H2source representing

fossil-fuel and N2-fixation related emissions. This too is

unsub-stantiated by data. Even if the economic growth in the mid-1990s to early 2000s could explain a proportion of this trend, the increase in fossil fuel emissions between 1991 and 2004 was only about 23% [Forster et al., 2007] while the increase of the fossil fuel and N2-fixation related H2source

found by Bousquet et al. [2011] is twice as large. In a recent study, Yashiro et al. [2011], using a two-layered soil diffu-sion/uptake model, could not simulate a long-term trend in the global soil uptake, corroborating our results.

[35] In general, changes in climatic inputs such as soil temperature, soil water content and snow depth do not strongly affect the soil uptake in our model. However, our model does not take in account the effect of the first few centimeters of soil, which can create an inactive layer and decrease the H2soil uptake. This inactive layer can have a

non-negligible impact on soil uptake.

[36] In general, and in order to better constrain the H2soil

uptake models, more field measurements for long period are necessary, especially in the tropics where no data are cur-rently available. A better observational constraint on the maximum biological uptake (kmax), and the minimum soil

water content needed to activate the biological pump, would be also very useful for modeling.

[37] Acknowledgments. We thank Renato Spahni and Rita Wania for their help with LPJ-WHyMe. We thank Nicole Smith-Downey for the explanations she provided us about the implementation in CLM2 of the H2model she developed during her thesis. We also thank T. Aalto and C.

Yver for the valuable information they provided concerning their published data. P.N.F. was sponsored by the EU-Project HYMN (Hydrogen, Methane and Nitrous Oxide: Trend variability, budgets, and interactions with the bio-sphere; GOCE-037048). The research leading to these results (CM, ICP) has received funding from the [European Community’s] Seventh Frame-work Programme (FP7 2007–2013) under grant agreement [238366].

References

Bousquet, P., et al. (2011), A three-dimensional synthesis inversion of the molecular hydrogen cycle: Sources and sinks budget and implications for the soil uptake flux, J. Geophys. Res., 116, D01302, doi:10.1029/ 2010JD014599.

Conrad, R., and W. Seiler (1985), Influence of temperature, moisture, and organic carbon on the flux of H2and CO between soil and atmosphere:

Field studies in subtropical regions, J. Geophys. Res., 90(D3), 5699–5709, doi:10.1029/JD090iD03p05699.

Constant, P., L. Poissant, and R. Villemur (2009), Tropospheric H2budget

and the response of its soil uptake under the changing environment, Sci. Total Environ., 407, 1809–1823, doi:10.1016/j.scitotenv.2008.10.064. Constant, P., S. P. Chowdhury, J. Pratscher, and R. Conrad (2010),

Strepto-mycetes contributing to atmospheric molecular hydrogen soil uptake are widespread and encode a putative high-affinity [NiFe]-hydrogenase, Envi-ron. Microbiol., 12(3), 821–829, doi:10.1111/j.1462-2920.2009.02130.x. Curry, C. L. (2007), Modeling the soil consumption of atmospheric methane at the global scale, Global Biogeochem. Cycles, 21, GB4012, doi:10.1029/ 2006GB002818.

Ehhalt, D. H., and F. Rohrer (2009), The tropospheric cycle of H2: A critical

review, Tellus, Ser. B, 61, 500–535.

Forster, P., et al. (2007), Changes in atmospheric constituents and in radia-tive forcing, in Climate Change 2007: The Physical Science Basis: Con-tribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp. 130–234, Cambridge Univ. Press, New York.

Guo, R., and R. Conrad (2008), Extraction and characterization of soil hydro-genases oxidizing atmospheric hydrogen, Soil Biol. Biochem., 40(5), 1149–1154, doi:10.1016/j.soilbio.2007.12.007.

Hammer, S., and I. Levin (2009), Seasonal variation of the molecular hydrogen uptake by soils inferred from continuous atmospheric observa-tions in Heidelberg, southwest Germany, Tellus, Ser. B, 61, 556–565. Hauglustaine, D. A., and D. H. Ehhalt (2002), A three-dimensional model

of molecular hydrogen in the troposphere, J. Geophys. Res., 107(D17), 4330, doi:10.1029/2001JD001156.

Table 5. Calculated Soil H2Uptake Annual Trends Between 1992 and 2004 in Tg a
2for the Different Scenarios Described in Table 4a

B_2011 Ref WF
1 WF
5 WF + 1 WF + 5 T
1 T + 1 S
5 Northern regions
0.68
0.05
0.1
0.3 +0.06 +0.56
0.02
0.08
0.12 Tropical regions
0.17 +0.005
0.12
0.13 +0.11 +0.58 +0.01
0.002 +0.005 Southern regions +0.11 0
0.03 +0.08 +0.03 +0.18 +0.006
0.005 0 Global
0.77
0.04
0.26
0.35 +0.2 +1.3 +0.003
0.09
0.12

a_{The three regions as defined by Bousquet et al. [2011] are northern regions (North America, Europe, north Asia), tropical regions (tropical America,}

Africa, Tropical Asia), and southern regions (all other land areas). The region partition is based on TRANSCOM map.

MORFOPOULOS ET AL.: A GLOBAL H2SOIL UPTAKE MODEL GB3013

Lallo, M., T. Aalto, T. Laurila, and J. Hatakka (2008), Seasonal variations in hydrogen deposition to boreal forest soil in southern Finland, Geophys. Res. Lett., 35, L04402, doi:10.1029/2007GL032357.

Larsen, H., R. Feidenhans’l, and L. S. Petersen (2004), Hydrogen and its competitors, Risø Energy Rep. 3, Risø Natl. Lab., Roskilde, Denmark. Müller, J.-F. (1992), Geographical distribution and seasonal variation of

surface emissions and deposition velocities of atmospheric trace gases, J. Geophys. Res., 97, 3787–3804, doi:10.1029/91JD02757.

Novelli, P. C., P. M. Lang, K. A. Masarie, D. F. Hurst, R. Myers, and J. W. Elkins (1999), Molecular hydrogen in the troposphere: Global distribu-tion and budget, J. Geophys. Res., 104(D23), 30,427–30,444.

Price, H., L. Jaeglé, A. Rice, P. Quay, P. C. Novelli, and R. Gammon (2007), Global budget of molecular hydrogen and its deuterium content: Constraints from ground station, cruise, and aircraft observations, J. Geophys. Res., 112, D22108, doi:10.1029/2006JD008152.

Rahn, T., J. M. Eiler, N. Kitchen, J. E. Fessenden, and J. T. Randerson (2002a), Concentration anddD of molecular hydrogen in boreal forests: Eco-system-scale systematics of atmospheric H2, Geophys. Res. Lett., 29(18),

1888, doi:10.1029/2002GL015118.

Rahn, T., N. Kitchen, and J. Eiler (2002b), D/H ratios of atmospheric H2in

urban air: Results using new methods for analysis of nano-molar H2

sam-ples, Geochim. Cosmochim. Acta, 66, 2475–2481, doi:10.1016/S0016-7037(02)00858-X.

Rhee, T. S., C. A. M. Brenninkmeijer, and T. Röckmann (2006), The over-whelming role of soils in the global atmospheric hydrogen cycle, Atmos. Chem. Phys., 6, 1611–1625, doi:10.5194/acp-6-1611-2006.

Ridgwell, A. J., S. J. Marshall, and K. Gregson (1999), Consumption of atmospheric methane by soils: A process-based model, Global Biogeo-chem. Cycles, 13(1), 59–70, doi:10.1029/1998GB900004.

Sanderson, M. G., W. J. Collins, R. G. Derwent, and C. E. Johnson (2003), Simulation of global hydrogen levels using a Lagrangian three-dimensional model, J. Atmos. Chem., 46, 15–28, doi:10.1023/A:1024824223232. Schmitt, S., A. Hanselmann, U. Wollschläger, S. Hammer, and I. Levin

(2009), Investigation of parameters controlling the soil sink of atmo-spheric molecular hydrogen, Tellus, Ser. B, 61, 416–423.

Schultz, M. G., T. Diehl, G. P. Brasseur, and W. Zittel (2003), Air pollution and climate-forcing impacts of a global hydrogen economy, Science, 302, 624–627, doi:10.1126/science.1089527.

Simmonds, P. G., R. G. Derwent, S. O’Doherty, D. B. Ryall, L. P. Steele, R. L. Langenfelds, P. Salameh, H. J. Wang, C. H. Dimmer, and L. E. Hudson (2000), Continuous high-frequency observations of hydrogen at the Mace Head baseline atmospheric monitoring station over the 1994– 1998 period, J. Geophys. Res., 105(D10), 12,105–12,121, doi:10.1029/ 2000JD900007.

Smith Downey, N. V. (2006), Soil uptake of molecular hydrogen and remote sensing of soil freeze and thaw, PhD dissertation, Calif. Inst. of

Technol., Pasadena. [Available at http://resolver.caltech.edu/CaltechETD: etd-08182006-105638.]

Smith-Downey, N. V., J. T. Randerson, and J. M. Eiler (2006), Temperature and moisture dependence of soil H2uptake measured in the laboratory,

Geophys. Res. Lett., 33, L14813, doi:10.1029/2006GL026749. Smith-Downey, N. V., J. T. Randerson, and J. M. Eiler (2008), Molecular

hydrogen uptake by soils in forest, desert, and marsh ecosystems in Cali-fornia, J. Geophys. Res., 113, G03037, doi:10.1029/2008JG000701. Steele, L. P., E. J. Dlugokencky, P. M. Lang, P. P. Tans, R. C. Martin, and

K. A. Masarie (1992), Slowing down of the global accumulation of atmo-spheric methane during the 1980s, Nature, 358, 313–316, doi:10.1038/ 358313a0.

Tromp, T. K., R.-L. Shia, M. Allen, J. M. Eiler, and Y. L. Yung (2003), Potential environmental impact of a hydrogen economy on the strato-sphere, Science, 300(5626), 1740–1742, doi:10.1126/science.1085169. Wania, R., I. Ross, and I. C. Prentice (2009a), Integrating peatlands and

per-mafrost into a dynamic global vegetation model: 1. Evaluation and sensi-tivity of physical land surface processes, Global Biogeochem. Cycles, 23, GB3014, doi:10.1029/2008GB003412.

Wania, R., I. Ross, and I. C. Prentice (2009b), Integrating peatlands and permafrost into a dynamic global vegetation model: 2. Evaluation and sensitivity of vegetation and carbon cycle processes, Global Biogeochem. Cycles, 23, GB3015, doi:10.1029/2008GB003413.

Wania, R., I. Ross, and I. C. Prentice (2010), Implementation and evalua-tion of a new methane model within a dynamic global vegetaevalua-tion model: LPJ-WHyMe v1.3, Geosci. Model Dev. Discuss., 3, 1–59, doi:10.5194/ gmdd-3-1-2010.

Warwick, N. J., S. Bekki, E. G. Nisbet, and J. A. Pyle (2004), Impact of a hydrogen economy on the stratosphere and troposphere studied in a 2-D model, Geophys. Res. Lett., 31, L05107, doi:10.1029/2003GL019224. Xiao, X., et al. (2007), Optimal estimation of the soil uptake rate of

molec-ular hydrogen from the Advanced Global Atmospheric Gases Experiment and other measurements, J. Geophys. Res., 112, D07303, doi:10.1029/ 2006JD007241.

Yashiro, H., K. Sudo, S. Yonemura, and M. Takigawa (2011), The impact of soil uptake on the global distribution of molecular hydrogen: Chemical transport model simulation, Atmos. Chem. Phys. Discuss., 11, 4059–4103, doi:10.5194/acpd-11-4059-2011.

Yonemura, S., S. Kawashima, and H. Tsuruta (1999), Continuous measure-ments of CO and H2deposition velocities onto an andisol: Uptake control

by soil moisture, Tellus, Ser. B, 51, 688–700.

Yonemura, S., S. Kawashima, and H. Tsuruta (2000a), Carbon monoxide, hydrogen, and methane uptake by soils in a temperate arable field and a forest, J. Geophys. Res., 105, 14,347–14,362, doi:10.1029/1999JD901156. Yonemura, S., M. Yokozawa, S. Kawashima, and H. Tsuruta (2000b), Model analysis of the influence of gas diffusivity in soil on CO and H2