Can Arctic sea ice extent be used as a proxy for local methane emissions?

(1)

Can Arctic sea ice extent be used as a proxy

for local methane emissions?

Peter Bosch

26 June 2017

Report Bachelor Project Physics and Astronomy, size 15 EC, conducted between 03 - 04 - 2017 and 07 - 07 - 2017

University Vrije Universiteit Amsterdam (FEW) Universiteit van Amsterdam (FNWI) Student number 2564092 (VU) and 10741968 (UvA)

Project The relation between the methane emissions and the Arctic sea ice extent

Supervisor prof. dr. E.A.A. Aben Second assessor dr. R.J. Wijngaarden Daily supervisor dr. S. Pandey

(2)

List of Figures

1.1 Methane concentrations at stations above 63◦ _{North . . . .} ₂

2.1 Map of NOAA stations . . . 4

2.2 Methane growth rate . . . 5

2.3 Fluctuations of the ice extent . . . 6

2.4 Methane growth rate and ice extent . . . 7

2.5 Correlation coefficients of methane growth rates and ice extent 8 2.6 Correlation coefficients when Northern Hemisphere is sub-tracted . . . 9

2.7 Location of Barrow, Alaska . . . 9

2.8 Correlation coefficients methane with southern wind and ice extent . . . 10

2.9 Correlation coefficients of methane growth rate and tempera-ture . . . 11

2.10 Seasonal trend ice extent . . . 12

2.11 Seasonal trend methane growth rate . . . 12

2.12 Correlation coefficients of modelled methane emissions and ice extent . . . 13

A.1 Bands used for calculation global methane concentrations . . . 18

A.2 Methane concentrations at several stations . . . 19

A.3 Local methane concentrations . . . 19

A.4 Local methane concentrations in July . . . 20

A.5 Methane growth rate on several stations . . . 20

A.6 Arctic sea ice extent . . . 21

A.7 Ice extent in July with polynomial fit . . . 21

A.8 Correlation coefficients of methane growth rates and ice area . 22 A.9 Methane growth rate with Northern hemisphere subtracted and ice extent . . . 22

A.10 Correlation coefficients when Northern Hemisphere is sub-tracted and ice area . . . 23

(4)

A.12 Methane with southern wind and ice extent in August . . . 24 A.13 Correlation coefficients methane with southern wind and ice

area . . . 24 A.14 Modelled methane emissions time series . . . 25 A.15 Correlation coefficients of the LPJ model and the ice area . . . 25

(5)

Populaire samenvatting

In 2007 kwam veel methaan vrij in het Arctische gebied. Er was dat jaar ook meer ijs op de Noordpool gesmolten dan normaal. In dit bachelor pro-ject is onderzocht of dit met elkaar te maken zou kunnen hebben.

Op verschillende plekken op de aarde is de concentratie methaan gemeten. Verder hebben vanuit de ruimte verschillende satellieten de hoeveelheid ijs op de Noordpool gemeten. Met deze data is de uitstoot van methaan en het krimpen/groeien van het ijs op de Noordpool berekend. Daarna is er gekeken of er een relatie is tussen de methaanuitstoot en het krimpen/ groeien van het ijs op de Noordpool.

Er zijn verschillende manieren gebruikt om de methaanuitstoot te bepalen. Zo is er gekeken of selectie op windrichting uitmaakt. Ook zijn, naast de wereldwijde concentraties methaan, de concentraties methaan op het noor-delijk halfrond gebruikt om de invloed van niet-Arctische bronnen van methaan te onderdrukken. Alleen bij gemodelleerde methaan emissies zijn hoge correlaties gevonden. In alle andere gevallen waren de correlaties tussen de methaanuitstoot en het smelten van het ijs op de Noordpool laag.

(6)

Popular summary

In 2007, a large amount of methane was emitted in the Arctic region to the atmosphere. In the same year the Arctic sea ice melted more than nor-mal. In this thesis, a possible correlation between sea ice extent and Arctic methane emissions is investigated.

At certain locations on Earth, the concentration of methane in the atmo-sphere is measured. At the same time, satellites measured the amount of sea ice in the Arctic region. With these data local methane emissions and the slinking/growing of the Arctic sea ice is computed. Then, the correla-tion between the methane emissions and the slinking/growing of the Arctic sea ice is investigated.

Several approaches to determine the methane emissions are used. For ex-ample, the dependence of the correlations on the wind direction is stud-ied. Also the methane concentrations in the Northern Hemisphere, instead of the global methane concentrations, are used to remove the influence of non-Arctic sources of methane. When modelled methane emissions are used, large correlations are found. In all other cases mentioned above, a small correlation is found between the emission of methane and the Arctic sea ice melting.

(7)

Chapter 1 Introduction

The emission of methane (CH4) in the atmosphere contributes to global

warming (Cao et al., 1998). The methane concentration is about 200 times lower than the carbon dioxide concentration, but methane absorbs per molecule more light (Lashof & Ahuja, 1990). So in the end methane is re-sponsible for about 40% of the total global warming since the pre-indus-trial times. The concentration of methane in the atmosphere is measured by the National Oceanic and Atmospheric Administration (NOAA) at various fixed locations around the world (NOAA, 2017b). Source cate-gory wise, wetlands are the largest source of methane on Earth (see Table 1.1). Also in the Arctic region the largest source of methane are wetlands. Methane producing microbes live in the wetlands and are affected by tem-perature (Segers, 1998). So a correlation between the temtem-perature and the methane emission is plausible. In the remainder of this thesis, when we speak about methane emissions, the methane emissions from wetlands is meant.

The Arctic sea ice extent is used as a proxy for cumulative local tempera-ture variations. The amount of sea ice is measured by the National Snow and Ice Data Centre (NSIDC, 2017). In 2007, a big decrease in the amount of sea ice extent in the Arctic was visible. In the same year, large concen-trations of methane are observed in the Arctic (see Figure 1.1). Because the methane emission can’t be measured directly, methane growth rates are taken as representative of the methane emissions. Different approaches were then taken to isolate the methane emissions coming from the Arc-tic wetlands. A possible correlation between the methane emissions in the Arctic region and the year-to-year fluctuations in the amount of Arctic sea ice is investigated. Also the relation between methane emissions and local temperatures are investigated at one station.

(8)

Origin Amount of global emissions in Tg CH4 per year

Wetlands 217

Other natural sources 132 Fossil fuels 96 Domesticated animals 89 Landfills and waste 75 Rice cultivation 36 Biomass burning 35

Table 1.1: The global amount of methane emitted by sev-eral sources in teragram methane per year (AMAP, 2015). Wetlands are the largest source of methane.

Figure 1.1: The mean of the methane concentrations at sta-tions above 63◦ _North.

The hypothesis is that the methane emissions in the wetlands should in-crease with higher temperatures. At the same time, the Arctic sea ex-tent ice will melt with higher temperatures. So the expectation is an anti-correlation between methane emission and Arctic sea ice extent.

(9)

Chapter 2 Results

In this chapter, the data that is used, the different approaches to deter-mine Arctic methane emissions, and the result of comparing the different datasets are described.

2.1 From methane concentration to

methane emission

The methane concentrations are measured by NOAA (2017b). A map in-dicating these locations, is given in Figure 2.1. The measurements are per-formed in two ways: NOAA has a few permanent stations where the con-centrations are measured in-situ at the surface (the larger, blue circles on the map). At the flask sampling stations, glass bubbles are filled with air -at the surface - and transported to a central labor-atory, where the methane concentration is determined. These places are shown in red (ongoing mea-surements) and yellow (terminated meamea-surements). In this research, all stations above 63◦ _{north latitude are used to represent the methane in the}

Arctic region. These stations are displayed in green in Figure 2.1.

In the remaining part of this section, the process in which the methane concentrations are converted into the local methane emissions is described. The starting point are time series of the methane concentrations at the stations above 63◦ _{north. As an example, time series for four stations are}

given in Figure A.2. As we are interested in the local methane emissions, and not the global trend, the mean of the methane concentrations at all stations on Earth is subtracted from the local methane concentrations (see

(10)

Figure 2.1: Map showing locations where the National Oceanic and Atmospheric Administration (NOAA) mea-sures the concentration of methane (NOAA, 2017b). The in-situ stations are marked in blue. The flask sampling sta-tions are marked in red (ongoing) and yellow (terminated). In green the stations above 63◦ _{north, are marked.}

Figure A.3).1 _{The month-to-month fluctuations are higher than the}

year-to-year fluctuations. To make the year-year-to-year fluctuations more visible, the months are plotted separately (e.g. see Figure A.4). The growth rates, which represent the methane emissions, are computed by taking the deriva-tive of the concentrations (see Figure A.5). Finally, a mean time series from the derivatives of all stations above 63◦ _{north is obtained (see Figure}

2.2). Because the effects of all other sources are subtracted from the local concentrations, the variability of the resulting time series is assumed to be mostly driven by local Arctic emissions.

1_{To calculate the global mean of methane concentrations, the Earth is split up in} seven bands (see Figure A.1). Based on the sine of the latitude the bands are evenly divided. For each band the mean of all the station in that band is calculated. Then the mean of the bands, weighted with the cosine of the latitude, is calculated. This is used as the global mean.

(11)

Figure 2.2: The mean of the methane growth rate at the stations above 63◦ _{North in July. From the individual}

mea-sured methane concentration, the global methane concentra-tion is subtracted. The derivative of this in July gives the growth rate.

2.2 Computing sea ice anomalies

The Arctic sea ice data used in this thesis is obtained from NSIDC (2017). Both Arctic sea ice extent and Arctic sea ice area are analysed. The ice ex-tent is the area encompassing all the Earth surface pixels with at least 15% of ice cover while the ice area is the actual area covered with ice. When the ice area is used, the plots are given in Appendix A.

As in the case of methane, the month-to-month fluctuations of the ice ex-tent are larger than the year-to-year fluctuations (see Figure A.6). So also here the months are investigated separately (e.g. July data is shown in Fig-ure A.7). To see the year-to-year fluctuations, a quadratic polynomial best fit is subtracted from the time series (see Figure 2.3). The resulting time series is used in further analysis.

2.3 Comparison of methane and ice extent

Figures 2.2 and 2.3 can be put together to compare them (see Figure 2.4). Between 1992 and 1997 an anti-correlation exist between the methane growth rate and the ice extent fluctuations. But in other years a

(12)

correla-Figure 2.3: The fluctuations of the Arctic ice extent in July. This is computed by subtracting a polynomial fit from the ice extent time series.

tion seems to be present. To represent the correlation, the Pearson correla-tion coefficient is used. This is a value between minus one (anti-correlated) and one (correlated) and is calculated with

rij =

Cij

pCii∗ Cjj

(2.1)

where Cij is the i,j-th component of the covariance matrix of the quantities

the correlation is computed for (Lee Rodgers & Nicewander, 1988). The Pearson correlation coefficient for the methane growth rate and the Arctic ice extent year-to-year fluctuations in July is r = −0.24 (see Figure 2.4).

The correlation coefficient mentioned above, is for both the ice extent and the methane growth rate in July. Because there might be a time lag be-tween a change in sea ice extent and methane emissions, the correlation coefficient is also calculated for all other combinations of months (see Fig-ure 2.5). For the most combinations of months no clear (anti-)correlation is found. But overall, a small anti-correlation seems to be present.

2.4 Northern Hemisphere subtraction

So far, the global methane concentrations are subtracted from the Arctic concentrations. The global concentrations are also affected by several pro-cesses in the Southern Hemisphere. These propro-cesses could spoil the signal

(13)

Figure 2.4: The methane growth rate (blue) and Arctic sea ice extent fluctuations (red) in July. For the methane growth rate, a mean of the selected stations (see Figure 2.1) is used. In some years an anti-correlation is present (e.g. 1992-1997). In other years a positive correlation is visible (e.g. 2006-2008).

coming from the Arctic processes. To prevent this, the methane concentra-tions in the Northern Hemisphere are subtracted from the local methane concentrations instead of the global methane concentrations.2 _{The trend in}

the correlation coefficients are not significantly changed when the Northern Hemisphere is subtracted (see Figure 2.6).

2.5 Wind direction influences

Another factor that could influence the correlations is the wind direction at the stations where methane concentrations are measured. Methane is emit-ted mostly from wetlands (see Chapter 1). When the wind is in the right direction, the methane emitted by wetlands will register a higher signal in the measurements. When the wind is coming from the direction opposite of the wetlands, the effect of the methane emissions might be absent. In this section the effect of the wind direction on methane measurements made at Barrow, Alaska is investigated. The wetlands are located south of Barrow

2_{The Northern Hemisphere concentrations are calculated the same way as the global} concentrations in section 2.1, but now with only bands and stations above the equator. Three bands are used to divide the Northern Hemisphere.

(14)

Figure 2.5: The correlation coefficients of the methane growth rates and the ice extent. Overall, no clear (anti-) correlation is visible.

(see Figure 2.7), so winds coming from the south would register strongest signal.3

To select the measurements with south wind, the individual measurements at Barrow and meteorological data (NOAA, 2017a) is analysed. The mea-surements at the hours when the wind was coming from the south were se-lected. South wind is defined as wind coming from between 110◦ _{and 250}◦

on a scale where 0◦ _{is north. Out of the selected methane concentrations,}

the methane growth rates are computed. With these growth rates the cor-relation coefficients are computed again (see Figure 2.8). Also with this approach, no clear correlation is found.

2.6 Correlation between methane emissions

and local temperature

In section 2.5, the methane growth rate on Barrow is compared with the Arctic ice extent. Until now in this study, the Arctic ice extent was used as a proxy for cumulative local temperature variations. As now we focus on single locations, the temperature measurements (at 10 meters height)

3_{Wind coming from the south is very rare at Barrow. As can be seen in Figures} A.11 and A.12, somethings big gaps up to 5 years appear in the time series. This will increase the error bars for the correlation coefficients.

(15)

Figure 2.6: Same as Figure 2.5, but now the Northern Hemisphere is subtracted from the measured concentrations.

Figure 2.7: The location of Barrow, Alaska

(maps.google.com). Barrow is located on the most north-ern part of Alaska. Wetlands are located south of Barrow.

at Barrow can be used directly for correlation analysis. The correlation coefficients in this case are more extreme than in section 2.5 (see Figure 2.9).

2.7 Seasonal changes over time

Until now, the long-term trend in the ice data is removed, by subtracting a polynomial fit (see section 2.2), as we were focused on short-term year-to-year variation in ice extent and any relations with similar scale varia-tion in Arctic methane emissions. However, there might exist a relavaria-tion between the long-term seasonal trend of ice extent and local methane emis-sion changes. The question is if the difference between the long-term trends

(16)

Figure 2.8: The correlation coefficients of the methane growth rate at Barrow and the ice extent. For the methane growth rate only concentrations when the wind was coming from land are used.

of summertime and wintertime sea ice extent (if present) is consistent with a similar difference in the CH4 growth rate trends.

These trends are visualised in Figure 2.10 and 2.11. It is noteworthy that the summer/winter period for sea ice extent is defined differently than that for methane growth. This is done under the assumption that the methane emissions from wetlands would have a few months lag from the seasonal temperature changes. A small difference in the trends is observed. How-ever, their significance needs to be tested with proper statistical analy-sis. Visually, the differences in the trends are in opposite direction. The summertime ice extent trend is stronger than the wintertime; for methane growth rate, it is opposite.

2.8 Using a forward approach

Until now, only an inverse approach is used: the emissions are computed starting from the measured concentrations. In this section, a forward ap-proach is used. Here, the methane emissions are modelled with the Lund-Potsdam-Jena Wetland Hydrology and Methane (LPJ-WHyMe) model driven by the Climate Research Unit (CRU) meteorological dataset. The LPJ-WHyMe model simulates peatland hydrology, permafrost dynamics,

(17)

Figure 2.9: The correlation coefficients of the methane growth rate and the temperature at 10 meters height at Barrow. Only the measurements taken when the wind was coming from the land are used.

peatland vegetation, and methane emissions (Wania et al., 2010).

The methane emissions computed with the LPJ-WHyMe model are given for all longitudes and latitudes starting from 1980 until 2015 in 0.25◦ _x

0.25◦ _{grids. A time series of the mean of all longitudes has been computed}

for the emissions above 63◦ _{north (see Figure A.14). When the mean is}

computed, the values from a grid closer to the poles are contributing more than the values from a grid closer to the equator, which will have a larger area. Therefore, each value in the mean should be multiplied with the co-sine of the latitude, like

X =

P

all long., lat.

Xlong., lat.∗ cos (lat)_{long., lat.}

N (2.2)

where X is the modelled methane emission and N is the total number of grids used. For the modelled methane emissions and the ice extent the cor-relation coefficients have been computed (see Figure 2.12). This figure are a bit blue, so a small anti-correlation is present.

(18)

Figure 2.10: The seasonal trend of the ice extent in the win-ter (mean of ice extent from October till March) is plotted in red and in the summer (April till September) in green.

Figure 2.11: The seasonal trend of the methane growth rate when the Northern Hemisphere is subtracted. In red the trend in the first six months is plotted and in green the trend in the second six months is plotted.

(19)

Figure 2.12: The correlation coefficients of the methane emissions modelled by the LPJ-WHyMe model and the ice extent. Overall the figure is more blueish, so a (small) anti-correlation is present.

(20)

Chapter 3 Discussion and conclusion

In this thesis, the correlation between Arctic methane emissions and Arc-tic ice extent is investigated. The correlation is computed for each month of methane emission and each month of sea ice extent for the last thirty years. There are several combinations of months with a clear anti-corre-lation. In some occasions, an extreme value up to r = −0.93 is found (see Figure 2.9), but in other occasions an extreme value - of opposite sign - up to r = 0.49 is found (see Figure 2.5).

An anti-correlation is found between the modelled methane emissions and the fluctuations in the Arctic ice extent (see Figure 2.12). Therefore, Arctic sea ice has the potency to be used as proxy for methane emissions. But this could not be proven based on the use of the methane growth rates yet. A possible explanation is that the signal of the Arctic methane emissions could not be sufficiently isolated to see the anti-correlation.

In section 2.6 the relation between the methane emission and tempera-ture on Barrow, when the wind is coming from land, is studied. In futempera-ture studies, this can be done for more stations. Also another definition of sum-mer/winter could be tested for the seasonal trends in section 2.7.

(21)

Abstract

In 2007 the Arctic methane growth rate measured by NOAA was higher than normal. In the same year, the sea ice extent in the region was smaller than normally. In this project I have searched for a possible correlation be-tween these quantities. Both forward and inverse approaches were used to quantify the methane emissions in the region. In the forward approach, the Lund-Potsdam-Jena model (LPJ-WHyMe) with the CRU meteorological dataset was used. In the inverse approach, the methane growth rates, that were used to estimate methane emissions, are computed for the NOAA sta-tions above 63◦ _{north. The global methane concentrations are than}

sub-tracted to eliminate the influence of non-arctic methane sources from the time series. Alternatively, also

• the methane concentrations in the Northern Hemisphere are sub-tracted instead of the global concentrations. By doing so, the non-Arctic changes in methane concentrations are largely eliminated, • the effect of the wind direction,

• the correlation of the methane growth rate and the temperature, and • seasonal effects

have been investigated. An anti-correlation is found between the modelled methane emissions and the fluctuations in the Arctic sea ice extent. But this anti-correlation is not found between the methane growth rates and the fluctuations in the Arctic sea ice extent.

(22)

Acknowledgements

In my opinion, this project was great. I learned a lot and had a great time in these last months. First of all, I would like to thank Sudhanshu for help-ing me durhelp-ing the project. Every time I asked, you could make some time for me, irrespective of other things you were working on. Thanks!

I would like to thank Ilse for organising this project and correcting my the-sis. Rinke, thank you for being my second assessor. Sander, I would like to thank you for the work you did for me.

Last but not least, I would like to thank everyone else who helped me dur-ing last three months.

(23)

Bibliography

AMAP. 2015, AMAP Assessment 2015: Methane as an Arctic climate forcer. (Oslo, Norway), Arctic Monitoring and Assessment Programme Cao, M., Gregson, K., & Marshall, S. 1998, Atmospheric environment, 32,

3293, Global methane emission from wetlands and its sensitivity to cli-mate change

Lashof, D. A. & Ahuja, D. R. 1990, Nature, 344, 529, Relative contribu-tions of greenhouse gas emissions to global warming

Lee Rodgers, J. & Nicewander, W. A. 1988, The American Statistician, 42, 59, Thirteen ways to look at the correlation coefficient

NOAA. 2017a, Downloaded from

ftp://aftp.cmdl.noaa.gov/data/meteorology/in-situ/brw/ NOAA. 2017b, Downloaded from

https://www.esrl.noaa.gov/gmd/dv/iadv/ NSIDC. 2017, Downloaded from

ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/north/monthly/ data/

Segers, R. 1998, Biogeochemistry, 41, 23, Methane production and methane consumption: a review of processes underlying wetland methane fluxes Wania, R., Ross, I., & Prentice, I. 2010, Geoscientific Model Development,

3, 565, Implementation and evaluation of a new methane model within a dynamic global vegetation model: LPJ-WHyMe v1.3.1

(24)

Appendix A

Figures

From methane concentration to methane emission

Figure A.1: The bands used for calculation global methane concentrations are given with the red stripes (worldmapson-line.com).

(25)

Figure A.2: The atmospheric methane concentrations mea-sured at Alert, Barrow, Ny-˚Alesund and Ocean Station M.

Figure A.3: The methane concentrations at Alert, Barrow, Ny-˚Alesund and Ocean Station M after subtracting the global methane concentration.

(26)

Figure A.4: Same as Figure A.3 but only July measure-ments are shown.

Figure A.5: The methane growth rate in July. This is com-puted by taking the derivative of the concentrations in Fig-ure A.4.

(27)

Computing sea ice anomalies

Figure A.6: The raw data of the Arctic sea ice extent.

Figure A.7: The Arctic ice extent data in July. A quadratic polynomial is fitted through the data.

(28)

Comparison of methane and ice extent

Figure A.8: The correlation coefficients of the methane growth rates and the ice area. Overall, no clear (anti-) correlation is visible.

Northern Hemisphere subtraction

Figure A.9: In blue the methane growth rate is given. The Northern Hemisphere methane concentrations are sub-tracted from the measurements instead of the global con-centrations as in Figure 2.4. In red the Arctic sea ice extent is shown.

(29)

Figure A.10: Same as Figure A.8, but now the Northern Hemisphere is subtracted from the measured concentrations.

Wind direction influences

Figure A.11: The time series of the methane growth rate and ice extent in July at Barrow. Only the methane data which is measured when the wind was coming from south where wetlands are located, is used. The correlation is neg-ligible (r = −0.03).

(30)

Figure A.12: The time series of the methane growth rate and ice extent in August at Barrow. Only the methane data which is measured when the wind was coming from south where wetlands are located, is used. A (strong) anti-correlation of r = −0.72 is observed.

Figure A.13: The correlation coefficients of the methane growth rate and ice area at Barrow. For the methane

growth rate analysis, only measurements when the wind was coming from south where wetlands are located, are used.

(31)

Using a forward approach

Figure A.14: The modelled wetland emissions of methane and the ice extent in July. The LPJ-WHyMe model is used.

Figure A.15: The correlation coefficients of the methane emissions modelled by the LPJ-WHyMe modelled methane emissions and the ice area.

Can Arctic sea ice extent be used as a proxy for local methane emissions?