You are on page 1of 10

Global and Planetary Change 69 (2009) 195–204

Contents lists available at ScienceDirect

Global and Planetary Change


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / g l o p l a c h a

The net carbon drawdown of small scale afforestation from satellite observations
Alvaro Montenegro a,⁎,1, Michael Eby a, Qiaozhen Mu b, Mark Mulligan c, Andrew J. Weaver a,
Edward C. Wiebe a, Maosheng Zhao b
a
University of Victoria, Department of Earth and Ocean Sciences, Canada
b
The University of Montana, College of Forestry and Conservation, USA
c
King's College London, Department of Geography, UK

a r t i c l e i n f o a b s t r a c t

Article history: Climate models indicate that warming due to increase in shortwave absorption from the lowering of albedo caused
Received 15 April 2009 by afforestation reduces and can even overcome, particularly at high latitudes, the cooling caused by the carbon
Accepted 23 August 2009 drawdown. We use high resolution (0.05×0.05° to 1×1°) global satellite observations to investigate the effects of
Available online 16 September 2009
afforestation. Results are markedly different from the coarser (~2.5×~2.5°) model-based studies. Between 40°S and
60°N afforestation always results in cooling. Many of the areas with the highest net carbon drawdown (drawdown
Keywords:
afforestation
after albedo effects) are at high latitudes. There is large zonal variability in drawdown and latitude is not a good
climate mitigation indicator of afforestation efficiency. The overall efficiency of afforestation, defined as the net carbon drawdown
surface energy balance divided by the total drawdown, is about 50%. By only considering the total drawdown and not considering albedo
effects, the Kyoto Protocol carbon accounting rules grossly overestimate the cooling caused by afforestation
drawdown.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction because snow covered open lands have much higher albedo than
snow covered forests.
Afforestation, the human induced conversion of crop or marginal In general, afforestation promotes greater evapotranspiration
lands into forests, is considered by the IPCC as one of the key climate leading to an increase in latent heat flux and cooling (Kleidon et al.,
change mitigation strategies available to the forestry sector (Nabuurs 2000; Govindasamy et al., 2001; Bounoua et al., 2002; Bala et al.,
et al., 2007). The United Nations Framework Convention on Climate 2007). Unlike the changes in albedo and atmospheric CO2, this direct
Change (UNFCC), through the Kyoto Protocol (UNFCCC, 2001), allows cooling is a local effect which does not significantly influence the
the atmospheric carbon drawdown generated by afforestation to be planet's radiative budget (as the extra latent heat leaving the surface
accounted as sequestered carbon and contributes to the emission will be returned to the atmosphere as the vapour condenses). We use
budget of the signatory nations. satellite derived evapotranspiration to estimate this local change.
Surface reflectivity (or albedo) depends on land cover. Croplands Changes in evapotranspiration can affect the global radiative
tend to be brighter and hence absorb a smaller fraction of the budget by their impacts on cloud cover and hence atmospheric
incoming solar radiation than forests which tend to be darker. A albedo. This can be an important factor in the climate response to
change in vegetation will, through the associated albedo change, large scale land cover change, particularly in the lower latitudes (Betts
modify the local radiation budget. It is estimated that the increase in et al., 2007), and an increase in cloud cover (and albedo) is predicted
albedo caused by deforestation since 1750 has had a cooling effect by models over afforested regions (Bala et al., 2007). The picture is not
with an average global negative radiative forcing of a − 0.2 W m− 2 ± so clear for changes at smaller scales. For spatial scales ranging from
0.2 W m− 2 (Solomon et al., 2007a). 106 to 107 m2 (10 to 100 ha), observations indicate that a decrease in
In order to determine the climatic effects of afforestation, the convection and cloud cover is to be expected over the afforested area
cooling resulting from the sequestered carbon should be “dis- (Rabin and Martin, 1996; Gash and Nobre, 1997; Durieux et al., 2003;
counted” by the warming induced by the decrease in surface albedo Chagnon et al., 2004) and recent satellite data analysis over the tropics
as land cover changes from crop to forest (Betts, 2000). This effect show no consistent difference in cloud coverage between nearby
has been shown to be particularly important at higher latitudes forested and deforested areas (Mulligan, 2008). Here we use global
cloud observations with 25 km2 resolution to analyze the relation
between cloud and surface cover.
In a study that used modelled albedo and observed regional esti-
⁎ Corresponding author.
E-mail address: amontene@stfx.ca (A. Montenegro). mates of potential sequestration, afforestation resulted in cooling at
1
Now at St. Francis Xavier University, Canada. mid-latitudes and net warming in some high latitude areas (Betts,

0921-8181/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.gloplacha.2009.08.005
196 A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204

2000). In earlier studies using climate models, the global temperature 2. Data and methods
response is cooling for low latitude afforestation, negligible change or
warming for mid-latitude afforestation and warming for afforestation Analysis consists of using a present day land cover data set in
at high latitudes (Claussen et al., 2001; Gibbard et al., 2005; Bala et al., conjunction with a potential vegetation data set to identify 5 × 5 km
2007). pixels where present day land cover is cropland and where the po-
The present UNFCC regulations on afforestation do not take the tential vegetation land cover is forest. Atmospheric carbon drawdown
albedo effect into consideration (UNFCCC, 2001). According to the and radiative effects of land cover change are then estimated based on
IPCC, while albedo effects should be taken into account, there are conversion of the entire 5 × 5 km pixel from cropland to forest. The
knowledge gaps on how the change in albedo will impact mitigation sections below describe the data and methods used for these esti-
by afforestation (Nabuurs et al., 2007). Here we quantify the most mates. As analyses are only performed on areas where present day
relevant climatic effects of afforestation using high resolution satellite cover is cropland and potential vegetation cover is forest, in the text
derived shortwave radiation flux, albedo, evapotranspiration, land “present day” vegetation is equivalent to cropland and “potential”
cover, cloud cover and snow cover data. vegetation is equivalent to forest.
Previous studies on this theme were either based on global
climate models (Claussen et al., 2001; Gibbard et al., 2005; Bala et al.,
2007) or required some input from global models (Betts, 2000). They 2.1. Potential vegetation
were unable, due to their relatively coarse spatial resolution
(≥ 2 × 2°), to provide results at a spatial scale relevant to individual The present day potential vegetation estimates come from the
afforestation projects. The models simulate afforestation by chang- reconstruction by (Ramankutty and Foley, 1999) with five minute
ing the cover globally or in whole latitude bands. These experiments spatial resolution and represent the vegetation that would most likely
are able to bracket the effects of maximum afforestation and provide exist in 1992 in the absence of human activities. Prior to analysis these
information on the climatic process influenced by large scale change data were interpolated using a nearest neighbor method into a 0.05 ×
in vegetation cover, but they do not offer a realistic representation of 0.05° grid. As albedo is estimated based on relationships between albedo
how afforestation could occur. Even if afforestation becomes a very and International Geosphere–Biosphere Program (IGBP) vegetation
prominent mitigation tool, significant portions of present agricul- types, the potential vegetation classes were converted into the IGBP
tural land would have to remain as such in the foreseeable future. vegetation classes (Belward et al., 1999) (see Table 1 in Appendix A).
Also, with the exception of (Betts, 2000), land cover is changed in
areas where presently no crops are found, that is, the effects of
afforestation are being considered over areas where no afforestation 2.2. Present day land cover
is possible.
Our analysis is global, but performed at a spatial resolution that is Present day land cover data are obtained from the Moderate
much closer to the scale of individual afforestation projects (~5 km2– Resolution Imaging Spectroradiometer (MODIS) Land Cover product
25 km2). It is only conducted on areas where afforestation could take (Friedl et al., 2002). The adopted values are the dominant vegetation
place. These are chosen as areas presently occupied by cropland and in 2001 from the MOD12C1 0.05 × 0.05° global gridded product with
that, according to estimates of potential vegetation, would be IGBP classification as provided by the Land Cover and Land Cover
occupied by forests if it were not for human activities. The assumption Dynamics group at Boston University (LCD-BU, 2008).
is that afforestation is not a viable mitigation strategy if the existence
of the forest requires artificial supply of water, nutrients or other type
of high intensity management. Another important distinction is that 2.3. Snow cover
our input is based on observed and not modelled results. The analysis
consists of comparing the radiative effects of the carbon drawdown Snow cover data come from the MODIS MOD10CM global monthly
resulting from the increase in land carbon storage to the radiative 0.05 × 0.05° gridded product as provided by the MODIS Land group
effects of changes in surface albedo, latent heat flux and cloud cover (Hall and Riggs, 2007; MLG, 2008). The data are used to generate a
due to afforestation. monthly climatology of snow cover areal fraction. The MOD10CM data
Our analyses provide estimates of the radiative effects of small were available for the years of 2000, 2001, 2003 and 2005 but cover-
scale afforestation and were not designed to determine the impacts age was not continuous. The number of months averaged to produce
of afforestation on atmospheric circulation nor the effects of the climatology ranges from 3 to 4, with the exception of June, with
regional or continental land cover change. That is, results should two values.
not be integrated into global values and much less, reversed and
used as estimates of large scale deforestation (cutting down the
Amazon for example). While the sensitivity of results to climate 2.4. Albedo
change is discussed, the goal is to provide a best estimate of the
effects of afforestation under the present climate. These limitations The albedo values are estimated based on the relationship be-
are imposed by our use of satellite measurements and not model tween vegetation type and snow free albedo obtained by Gao et al.
simulations. While some flexibility and the effects of some climate (2005) and the relationship between vegetation type and snow
related feed-backs are lost, the adoption of satellite data makes it covered albedo presented by Moody et al. (2007). Both of these use
possible to estimate the albedo effects of afforestation at an the IGBP vegetation classes and both present maximum and minimum
unprecedented resolution and, at the same time, avoid uncertain- albedo values for each class. The land cover–albedo correlation is a
ties associated with modeling of snow cover and albedo. The use of function of latitude for the snow free albedo and spatially constant in
satellite land cover in tandem with biome-based land carbon the snow covered case. Monthly maps of maximum, minimum and
density also provides high resolution estimates of carbon density average albedo are calculated through a simple weighted average:
change. α = Fαvs + (1 − F)αv, where α is the total albedo, F is the snow
Our approach generates estimates of the climatic effects of covered fraction, αv is the snow free albedo, αvs the snow covered
afforestation that take into account the effects of albedo at a resolution albedo and the subscript v indicates the fact that the albedo is a
pertinent to individual afforestation projects making them a viable function of the local dominant vegetation type as determined by the
tool in carbon policy decision making. potential vegetation or MODIS land cover data.
A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204 197

2.5. Soil and vegetation carbon density surface. Our atmospheric drawdown is estimated to be 50% of the
increase in land carbon. We therefore assume that the change in
As with albedo, relationships obtained by previous studies are atmospheric carbon caused by afforestation will be inverse and
used to determine soil (Prentice et al., 1993; House et al., 2002) and symmetric to the increase in atmospheric carbon generated by
vegetation ((Olson et al., 1985; House et al., 2002) carbon densities as deforestation (Betts, 2000; House et al., 2002) with total carbon
a function of land cover. All studies provide a range of densities for drawdown given by:
each land cover type so that maximum, minimum and average soil
and vegetation densities maps are generated. Soil and vegetation TD = 0:5 ΔLc : ð2Þ
carbon densities are determined for croplands and five different forest
categories. The resulting 12 (six soil and six vegetation) fields are The albedo fields based on present day and potential vegetation
combined and averaged to generate three (maximum, minimum and data are used to generate monthly maps of average, minimum and
average) total carbon storage maps. This is done for both present day maximum absorbed shortwave radiation in W m− 2 (a) by: a = I(1 −
and potential vegetation (see Fig. 1 for cropland and forest average α) + 0.23(αI) where the second term on the right accounts for the
carbon densities and Table 2 in Appendix B for the conversions of the atmospheric absorption of the reflected short wave flux (Weaver
vegetation classes used by these studies into IGBP classes). et al., 2001), the original 0.3 is here adjusted to 0.23 to account for
absorption of the downward and upward shortwave fluxes). These
2.6. Incoming shortwave flux at the surface monthly fields are averaged into annual mean absorption maps and
the difference between potential and present day vegetation values
Surface downward shortwave fluxes (I) come from the Interna- provides the increase in absorption caused by afforestation. The
tional Satellite Land Surface Climatology Project (ISLSCP) Initiative II maximum (minimum) absorption change is obtained by subtracting
Data Collection. We used the All Sky Surface Shortwave Flux product the maximum (minimum) potential vegetation absorption from the
available online (Shashi et al., 1999; Stackhouse et al., 2000; Collection, minimum (maximum) present day absorption.
2008) as monthly averages gridded at 1 × 1° resolution. Monthly The increase in absorption caused by afforestation is expressed for
values from 1986 to 1994 are averaged into a monthly climatology and each grid box in terms of its effect on the global radiative flux by
interpolated into a 0.05 × 0.05° grid. Δa = ab/E, where b is the grid box area and E the area of the Earth's
surface. To facilitate comparison to the estimated drawdown, Δa is
2.7. Evapotranspiration converted from W m− 2 to kilograms of carbon per m2 (Betts, 2000).
First, the radiative forcing is expressed as a change in atmospheric CO2
Evapotranspiration (ET) values represent a 2000–2006 mean and concentration (ΔaCO2) in parts per million by volume (ppmv):
are calculated at 0.05 × 0.05° resolution using the procedure described  
Δa
by Mu et al. (2007), which estimates ET based on MODIS albedo, leaf ΔaCO2 = C0 exp5:35 −1 : ð3Þ
area index (LAI) and Enhanced Vegetation Index (EVI) and daily
meteorological data from the Global Modeling and Assimilation Office
C0 is the 2006 atmospheric CO2 concentration, set to 380 ppmv
(GMAO) reanalysis.
(Tans, 2008). Eq. (3) is based on a 3.7 W m− 2 increase in radiative
forcing for a doubling of atmospheric CO2·ΔaCO2 values are then
2.8. Cloud cover
converted to kg of carbon (Δac) by:

Cloud frequency data were obtained from the MODIS 1 km cloud Δac = ð2ΔaCO2 Þ = k; ð4Þ
climatology data set of Mulligan (2006a). Cloud frequency is a
measure of the proportion of observations in which clouds are present where k=2.13×10− 12 is the conversion factor from atmospheric CO2
in a given 1 km pixel. concentration into kg of carbon (Jr, 1990) and the factor of two comes
The climatology was produced by combining the entire MODIS from the assumption explained above that only 50% of emissions remain
MOD35 Cloud product for MODIS TERRA and MODIS AQUA for the in the atmosphere.
period 2001–2006. All of the swath data were converted to grids and Net afforestation drawdown is calculated by subtracting the
the data were analysed to derive annual average, monthly and diurnal radiative forcing from the decrease in albedo (expressed as a local
statistics on cloud frequency at 1 km resolution for the entire tropics carbon emission equivalent in kg m− 2) from the total drawdown
and at 250 m resolution for some tropical land areas. resulting from increase in land carbon:

2.9. Afforestation changes ND = TD−Δac : ð5Þ

2.9.1. Albedo and carbon density Given the uncertainties in total drawdown and albedo estimates we
The drawdown generated by afforestation is a function of the calculate maximum, minimum and average net drawdown scenarios.
increase in total carbon storage on land as cover goes from crop, or Maximum (minimum) net drawdown is obtained by subtracting the
present day vegetation, to forest, or potential vegetation: minimum (maximum) increase in absorption from the maximum
(minimum) total drawdown.
ΔLc = ðVCfo + SCfo Þ−ðVCcr + SCcr Þ ð1Þ
2.9.2. Evapotranspiration change
where ΔLc is the total change in density, VC is vegetation carbon While vegetation cover is an important contributor, it is not the
density, SC is soil carbon density and the subscripts fo and cr refer to sole determinant of LAI, EVI and albedo, which are also influenced by
forest and cropland respectively. Maximum (minimum) increase in other factors such as temperature, precipitation and plant physiol-
carbon storage is estimated by subtracting the maximum (minimum) ogy. This means that changes in ET related to afforestation cannot be
potential storage estimate from the minimum (maximum) present obtained by global substitution of typical crop values LAI, EVI and
day storage estimate. albedo by typical forest values of these parameters. In our analyses
Historically, about half of the anthropogenic emissions, includ- these three quantities are calculated every 8- or 16-day MODIS data
ing those related to deforestation, have remained in the atmo- composite time period for each grid point where afforestation is
sphere, with the remainder being absorbed by the oceans and land possible in the following manner (using a generic grid point Gi,j as an
198
A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204
Fig. 1. Mean carbon densities over areas of realizable afforestation in kg m2. Note difference in color bar ranges. Top left, cropland vegetation carbon density. Bottom left, cropland soil carbon density. Top right, forest vegetation carbon density.
Bottom right, forest soil carbon density.
A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204
Fig. 2. Results for the average drawdown afforestation scenario. Top left, total carbon drawdown (kgm− 2) due to increase in land carbon storage. Bottom left, change in radiative forcing expressed in terms of carbon emissions equivalent (kgm− 2). Top
right, net drawdown (total-emissions equivalent). Bottom right, drawdown efficiency (net/total). Note the difference in color scale between panels. Values are only plotted for areas where afforestation could take place (see text). For clarity, the figure
shows results interpolated to 0.1 × 0.1° resolution, twice as coarse as the actual results. The abrupt discontinuity south of 20° N in India is caused by a lack of albedo information for the IGBP “mixed forest” class below that latitude (Gao et al., 2005).

199
200 A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204

example): On the present day vegetation map, a 2×1° (longitude× gridded originally at 1 km resolution but sub-sampled for this
latitude) area around Gi,j is searched for the forest type that would be analysis to 0.25° averages of the 1 km data. These data represent
present at Gi,j according to the potential vegetation data. This search fractional tree cover change (Mulligan, 2007), rainfall (Mulligan,
strategy is chosen as a way to maximize the possibility that the selected 2006b) and cloud cover (Mulligan, 2006a) derived from Hansen
forested pixels will have similar seasonal profile in LAI to the original et al. (2006). The analysis covers the area between ±40°.
pixel as they would be in a similar climatic thermal zone. If grid points
containing the pertinent forest type exist, their distance-weighted
mean LAI, EVI and albedo at up to four nearest points are used to 3. Results and discussion
calculate the afforestation ET at Gi,j. The assumption is that nearby
points containing the same forest type should have similar parameters. The net drawdown in the average scenario, while tending to be
If the box around Gi,j does not contain the suitable forest type, its less than half of the total drawdown, is positive for all latitudes.
afforestation ET is based on interpolation of the neighboring Contrary to previous studies (Betts, 2000; Claussen et al., 2001;
afforestation ET values. Gibbard et al., 2005; Bala et al., 2007), afforestation does not cause
These changes in ET are used to calculate the impact of affores- warming anywhere between 40°S and 60°N (Fig. 2). Although some
tation on surface latent heat flux. In general, the switch from crops to N–S structure is present in regions like the eastern USA or
forest results in greater ET and latent heat fluxes but in some areas, southeastern Brazil, the global spatial pattern of the net drawdown
possibly due to irrigation (especially over temperate east China), is closely related to the total drawdown and does not show a clear
afforestation brings about a reduction of these parameters. latitudinal distribution as some models have suggested (Claussen
et al., 2001; Gibbard et al., 2005; Bala et al., 2007).
The albedo emissions equivalent tends to be greater at higher
2.9.3. Cloud cover change latitudes but there is significant zonal variability (Fig. 2). In the mid
The impacts of afforestation on cloud cover come from databased to high latitudes of the northern hemisphere, the only region where
analyses of the impact of historic (satellite measured) land cover comparison is possible, the spatial distribution of albedo effects are
change on satellite measured cloud cover and rainfall for the 2000– in good agreement with previous model-based estimates (Betts,
2006 period. One cannot directly compare land cover with 2000). The large albedo impact at higher latitudes does not
climatological cloud frequency and hope to extract the impact of overcome the local concomitant total drawdown of some high
land cover on cloud statistics since so many other variables (that latitude regions. Many of the highest net drawdown values come
have little to do with land cover) control the rainfall and cloud from locations above 50°N, mainly due to the high forest soil carbon
generation processes. Moreover land cover is fundamentally content. The compounded effects of low total drawdown and
dependent on the prevailing climate, particularly the balance relatively high albedo emissions equivalent result in the smallest
between rainfall and potential evapotranspiration (solar radiation net drawdown values being found in some areas between ~ 20° and
mediated by cloud cover). In order to understand the potential ~ 30° of both southern and northern hemisphere. The large
impact of afforestation on cloud cover we therefore have to compare drawdown at some high latitude regions and small drawdown at
areas with significantly different land cover but in the same climatic some areas in the tropics and sub-tropics are different from previous
environment i.e. very near to each other. We use three data sets modelled results (Claussen et al., 2001; Gibbard et al., 2005; Bala

Fig. 3. Minimum, top and maximum, bottom, net afforestation carbon drawdown (kgm− 2). Note the difference in color scale between panels and the negative values (net warming)
for some areas in the minimum drawdown case.
A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204 201

et al., 2007), which show smaller (and even negative) drawdown at Even in our worst case scenario, many areas above 50° N have posi-
high latitudes and larger drawdown at smaller latitudes. tive drawdown. The afforestation efficiency of the minimum draw-
Part of the discrepancy between our average scenario results and down scenario (not shown) ranges from ~ −20% to ~ 40%. The net
some of the earlier studies can be traced to our larger total drawdown. Our drawdown of the “best case”, maximum drawdown scenario follows
global mean carbon drawdown is ~15kgm− 2 compared to the 10kgm− 2 the spatial patterns of the average case but with values that are ~40%
adopted by Gibbard et al. (2005) and our high latitude (>50°N) mean to ~ 60% larger (Fig. 3) and with an afforestation efficiency (not
carbon drawdown is 18kgm− 2 compared to the 5.5 kg m− 2 modelled by shown) of ~50%.
Claussen et al. (2001). Our regional high latitude values range from
17 kg m− 2 in Canada and eastern Siberia to 20 kg m− 2 in eastern 3.1. Climate sensitivity
Europe, much higher than the respective 6 kg m− 2 and 10 kg m− 2
used by Betts (2000). As our snow cover climatology comes from a relatively short time
The global increase in land carbon calculated by our methodology series, we estimate the sensitivity of the mean case drawdown to
for the conversion of all present day crops into forests ranges from changes in snow cover. This is done adding ±0.1 to bins where snow
43 Pg to 209 Pg with a mean of 134 Pg. This last value is slightly (15%) is present in the monthly snow cover areal fraction fields. This
lower than the recent estimate of historical land use change emis- generates minimum local snow cover anomalies of 10%, in agreement
sions, largely the result of deforestation, of 158 Pg of carbon (Canadell with estimates of northern hemisphere interannual snow cover
et al., 2007). Given this good agreement, it is unlikely that the variability (Robson and Frei, 2000) and close to the 13% projected
differences between our total drawdown and those from previous decrease in northern hemisphere snow cover at the end of the 21st
studies are caused by systematic error (overestimation) in our carbon century (Solomon et al., 2007b).The impact of these new snow cover
density values. There are indications that old-growth forests still act fields on net drawdown is small, with a mean change of 1% above 40°N
as carbon sinks (Luyssaert et al., 2008). If this is the case, our and maximum values around 3% over eastern Asia and western N.
estimates, which are based on mature forests being carbon neutral, America (not shown).
could underestimate carbon density.
A smaller difference between crop and forest albedo also helps 3.2. Changes to evapotranspiration, latent heat flux and cloud cover
explain why our high latitude afforestation is more efficient than
some of the earlier studies. The mean albedo change we obtain for As expected, observations indicate that afforestation causes an
the average scenario high latitudes is ~ − 0.15 compared to − 0.2 to increase in latent heat flux over most areas (Fig. 4, note exceptions in
− 0.25 of other studies (Bala et al., 2007; Betts et al., 2007) (See small portions of southeastern Brazil and northeastern China around
Fig. 5). The decrease in snow free albedo we compute is similar to 30°N). The cooling generated by the increase in ET over afforested areas
the decrease obtained by those other experiments and the observed is a local effect and cannot impact the global energy budget significantly
differences are mainly related to differences in the albedo of snow (see Appendix D for a discussion of the maximum effects of these
covered vegetation and/or differences in snow cover areal fraction. changes in ET).
As noted by the author, the model used by Betts (2000) over- Observations at 1 km resolution averaged over 0.25° show no
estimates snow mass over Siberia. consistent relationship between forest cover and mean annual cloud
The much finer spatial resolution used here allows us to better frequency (see Appendix E). This is an indication that cloud frequency
restrict the analyses to areas where afforestation can occur. By is not primarily determined by absence or presence of forest at these
considering the effects of afforestation where no agriculture takes spatial scales and that the response of cloud cover to afforestation
place, past studies could have overestimated the decrease in albedo cannot be evaluated at the scales pertinent to individual afforestation
over higher latitudes. These artificial boreal croplands, with their projects. Based on these results we opt not to impose a change in
extensive, long lasting snow cover could “contaminate” results with albedo due to changes in cloud cover in our radiative budget estimates
unrealistic high pre-afforestation albedo values. of afforestation.
For the minimum drawdown scenario, computed using the maxi-
mum change in absorption and minimum increase in land carbon, 3.3. Missing feed-backs
afforestation results in net warming (negative drawdown) at some
high latitude regions in North America and Asia as well as portions of Changes to ET and surface roughness caused by afforestation can
southeastern China and South America between 20°S and the equator. influence climate by their effects on atmospheric circulation and
Most of Europe shows positive drawdown, with negative values water vapour transport. While there is evidence for this in the case
concentrated around the Mediterranean (Fig. 3). The general location of regional scale land surface change (Gedney and Valdes, 2000; Fu,
of the negative drawdown areas are in good agreement with the other 2003; R.B. et al., 2005; D'Almeida et al., 2007), taking the cloud
study that focused exclusively on radiative changes (Betts, 2000). response into consideration we expect these impacts to be small for

Fig. 4. Change in surface latent heat flux into the atmosphere due to afforestation.
202 A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204

changes on the spatial scales of interest to our analysis; nonetheless, Appendix B. IGBP to soil and vegetation carbon conversion
our method does not account for such impacts. Another missing
effect of afforestation is the potential decrease in snow cover due to Table 2 describes vegetation classes used for carbon density
local warming, which could lead to further increase in absorption converted to IGBP classes.
and further warming in a positive feedback loop. Our conclusions
are robust to changes in snow cover of up to 10%, but estimating the
actual impact of this feedback requires the use of a climate model. Table 2
Conversion of the forest and cropland classes from the adopted soil and vegetation
carbon density tables into IGBP classes. For the Prentice Prentice et al. (1993) estimates,
4. Conclusions cropland soil carbon comes from House House et al. (2002). HL, higher latitude, areas
poleward of ±30°, LL; lower latitude, areas between ±30°.

The climatic effects of small scale afforestation based on the IGBP Olsen House Prentice
analysis of high resolution satellite data are markedly different from Vegetation C Soil/Veg C Soil C
previous climate model-based results. Results show no simple relation
Evergreen Conifers Boreal forest Cool conifer
between afforestation efficiency and latitude. In our best estimate, the
needleleaf forest forest/taiga
conversion of cropland to forests brings about cooling in all areas of Evergreen Trop./subtrop. Tropical Tropical rain forest
the planet where afforestation is viable, including areas above 50°N. In broadleaf forest Broad-leaved forests
fact, many of the areas with the highest afforestation induced carbon humid forest
Deciduous Conifers Boreal forest Cool conifer
drawdown are found at high latitudes. Also contrasting with earlier
needleleaf forest forest /taiga
studies is the relatively low afforestation drawdown over some low Deciduous Mid-latitude temp. Temperate Temperate
latitude areas. Changes in cloud cover could be an important com- broadleaf forest Broad-leaved forest forest deciduous forests
ponent of the radiative response to afforestation, particularly at low Mixed forests — HL Mid-latitude Temperate Cool mixed forest
latitudes, but at the scales pertinent to individual afforestation mixed woods forest
Mixed forests — LL Mid-latitude Temperate Broad-leaved
projects observations demonstrate that cloud cover frequency is not
mixed woods forest evergreen/warm
controlled by the presence or absence of forests. mixed forest
While the overall picture is one where afforestation appears to be a Croplands — HL Cool or cold Croplands –
more viable mitigation option than previously believed, uncertainty farms, towns
Croplands — LL Warm/hot farms, Croplands –
in the data still point to a worst case scenario where afforestation
towns, cool
generates very low net carbon drawdown or even net warming at grass/scrub
both high and low latitudes. Crop./nat. Veg. Cool or cold Croplands –
The efficiency of afforestation is location dependent and has an Mosaic — HL farms, towns
average of about 50%. By not considering albedo effects, the current Crop./nat. Veg. Warm/hot farms, Croplands –
Mosaic — LL towns, cool
carbon accounting system of the Kyoto Protocol grossly overestimates
grass/scrub
afforestation drawdown. The use of satellite data allows the impact of
albedo changes on afforestation to be calculated at a resolution of
~ 5 km. It also circumvents risks associated with the complex
modeling of snow cover, vegetation and surface albedo. Our results Appendix C. Average change in albedo
demonstrate that analyses like the one conducted here can generate
reliable, site-specific estimates of this very important but presently Fig. 5 shows the change in albedo for the average scenario.
ignored process.
Appendix D. Evapotranspiration and latent heat flux
Acknowledgements
The cooling generated by the increase in ET over afforested areas is a
We are grateful for research funding from CFCAS. We would also local effect and cannot impact the global energy budget significantly.
like to thank the two anonymous reviewers for improving the Nevertheless, as this effect is more pronounced over low latitudes, where
manuscript and Mahendra K. Karki for help with the evapotranspi- our analysis contradicts previous results, we have decided to test the
ration MODIS data extraction. Funding for some of the work was maximum impact of this ET cooling assuming (unrealistically) that all the
provided by UK NERC, DfID and ESRC in their ESPA programme. extra latent heat flux caused by afforestation would escape into space.
Following the same procedure applied to the shortwave flux results, the
change in latent heat flux caused by afforestation is converted into a local
Appendix A. IGBP to potential vegetation conversion carbon drawdown (or emission in a few areas) and added to the net
carbon drawdown of the average afforestation scenario (Fig. 6).
Table 1 describes how the potential vegetation classes were As expected, the addition of ET effects results in increased
converted to IGBP classes. drawdown over most areas. The largest increases are seen in south
and southeast Asia, Australia and southern South America. Even under
Table 1 this unrealistic assumption, which greatly overestimates the effect of
Conversion of the potential vegetation forest classes into IGBP forest classes.
evapotranspiration on global cooling, it is clear that the efficiency of
Potential vegetation Converted to IGBP afforestation is not a simple function of latitude and that low latitude
Ramankutty and Foley (1999) afforestation does not necessarily result in larger drawdown than high
Tropical evergreen forest Evergreen broadleaf forest latitude afforestation.
Tropical deciduous forest Deciduous broadleaf forest
Temperate broadleaf evergreen forest Evergreen broadleaf forest
Temperate needleleaf evergreen forest Evergreen needleleaf forest Appendix E. Changes to cloud cover
Temperate deciduous forest Deciduous broadleaf forest
Boreal evergreen forest Evergreen needleleaf forest Observations at 1 km resolution averaged over 0.25° show no
Boreal deciduous forest Deciduous needleleaf forest consistent relationship between forest cover and mean annual cloud
Evergreen deciduous mixed forest Mixed forests
frequency (Fig. 7, top 3 panels). When results are averaged over 2° the
A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204 203

Fig. 5. Change in surface albedo caused by afforestation in the average scenario.

Fig. 6. Maximum⁎ impact of changes in latent heat flux on net afforestation carbon drawdown (kgm− 2) in the average scenario. (⁎Assuming all the extra latent heat from change in
evapotranspiration escapes into space.)

correlation is still very heterogeneous in space, but some regional with the opposite true for the central-west portions of the Amazon
patterns tend to appear (Fig. 7, bottom). Forested areas in south and (see (Mulligan, 2008) for a more comprehensive analysis of these
southeast Asia, for example, tend to have less frequent cloud cover data).

Fig. 7. Percentage difference in mean annual cloud frequency between forested and non-forested areas. Negative values are present when forested areas have less frequent cloud
cover than neighboring non-forested areas and positive values indicate areas where forested areas have more frequent cloud cover than neighboring non-forested areas. Top three
panels are examples at 0.25° resolution for regions within Equatorial South America (left), Equatorial Africa (center) and Southeast Asia (right). Bottom panel show the same results
averaged over 2° bins. The color bar applies to all panels.
204 A. Montenegro et al. / Global and Planetary Change 69 (2009) 195–204

References MLG, 2008. Modis-land-group, modis snow cover data set. http://modis-land.gsfc.nasa.
gov/snow.htm.
Moody, E.G., King, M.D., Schaaf, C.B., Hall, D.K., Platnick, S., 2007. Northern Hemisphere
Bala, G., Caldeira, K., Wickett, M., Phillips, T.J., Lobell, D.B., Delire, C., Mirin, A., 2007. five-year average (2000–2004) spectral albedos of surfaces in the presence of
Combined climate and carbon-cycle effects of large-scale deforestation. Proceed- snow: statistics computed from Terra MODIS land products. Remote Sensing of
ings of the National Academy of Sciences of the United States of America 104 (16), Environment 111 (2–3), 337–345.
6550–6555. Mu, Q., Heinsch, F., Zhao, M., Running, S., 2007. Development of a global evapo-
Belward, A.S., Estes, J.E., Kline, K.D., 1999. The IGBP-DIS global 1-km land-cover data set transpiration algorithm based on modis and global meteorology data. Remote
DISCover: a project overview. Photogrammetric Engineering and Remote Sensing Sensing of Environment 111, 519–536.
65 (9), 1013–1020. Mulligan, M., 2006a. MODIS MOD35 pan-tropical cloud climatology. Version 1.
Betts, R.A., 2000. Offset of the potential carbon sink from boreal forestation by decreases Database: http://www.amiotek.com/clouds.
in surface albedo. Nature 408 (6809), 187–190. Mulligan, M., 2006b. Global Gridded 1km TRMM Rainfall Climatology and Derivatives.
Betts, R.A., Falloon, P., Goldewijk, K.K., Ramankutty, N., 2007. Biogephysical effects of Version 1.0. Database: http://www.ambiotek.com/1kmrainfall.
land use on climate: model simulations of radiative forcing and large-scale Mulligan, M., 2007. VCF_change: a Google Earth visualisation of MODIS VCF tree cover
temperature change. Agricultural and Forest Meteorology 142, 216–233. change globally (2000–2005). Version 1.0. Database: http://www.ambiotek.com/tree.
Bounoua, L., Defries, R., Collatz, G., Sellers, P., Khan, H., 2002. Effects of land cover Mulligan, M., 2008. Annex 4, page 120 in: Regulation of the climate system related
conversion on surface climate. Climate Change 52, 29–64. ecosystem services in the Amazon, R. Porro et al., eds. UK NERC-DfID-ESRC ESPA
Canadell, J.G., Quere, C.L., Raupach, M.R., Field, C.B., Buitenhuis, E.T., Ciais, P., Conway, T.J., Programme. Challenges to Managing Ecosystems Sustainably for Poverty Alleviation:
Gillett, N.P., Houghton, R.A., Marland, G., 2007. Contributions to accelerating Securing Well-Being in the Andes/Amazon. http://www.ecosystemsandpoverty.org/
atmospheric CO2 growth from economic activity, carbon intensity, and efficiency of wp-content/uploads/2008/05/annexes_final_small.pdf.
natural sinks. Proceedings of the National Academy of Sciences of the United States Nabuurs, G., Masera, O., Andrasko, K., Benitez-Ponce, P., Boer, R., Dutschke, M., Elsiddig, E.,
of America 104 (47), 18866–18870. Ford-Robertson, J., Frumhoff, P., Karjalainen, T., Krankina, O., Kurz, W., Matsumoto, M.,
Chagnon, F., Bras, R., Wang, J., 2004. Climatic shift in patterns of shallow clouds over the Oyhantcabal, W., Ravindranath, N., Sanchez, M.S., Zhang, X., et al., 2007. Climate
Amazon. Geophysical Research Letters 31. doi:10.1029/2004GL021188. change 2007: mitigation. In: Metz, et al. (Ed.), Contribution of Working Group III to the
Claussen, M., Brovkin, V., Ganopolski, A., 2001. Biogeophysical versus biogeochemical Fourth Assessment Report of the Intergovernmental Panel on Climate Change.
feedbacks of large-scale land cover change. Geophysical Research Letters 28 (6), Cambridge University Press, Cambridge.
1011–1014. Olson, J., Watts, J., Allison, L., 1985. Major world ecosystem complexes ranked by carbon
Collection, I. S. L. S. C. P. I. I. I. D., 2008. (http://www.islscp2.sesda.com/ISLSCP2_1/ in live vegetation. Tech. Rep. “NDP-017”. Carbon Dioxide Information Center, Oak
html_pages/islscp2_home.html). Ridge National Laboratory, Oak Ridge, Tennessee.
D'Almeida, C., Vörösmarty, C., Hurtt, G., Marengo, J.A., Dingman, S., Keim, B., 2007. The Prentice, I., Sykes, M., Lautenschlager, M., Harrison, S., Denissenko, O., Bartlein, P., 1993.
effects of deforestation on the hydrological cycle in Amazonia: a review of on scale Modelling global vegetation patterns and terrestrial carbon storage at the last
and resolution. International Journal of Climatology 27, 633–647. glacial maximum. Global Ecology and Biogeography Letters 3, 67–76.
Durieux, L., Machado, L., Laurent, H., 2003. The impact of deforestation on cloud cover Rabin, R., Martin, D., 1996. Satellite observations of shallow cumulus coverage over the
over the Amazon arc of deforestation. Remote Sensing of Environment 86, 132–140. central United States: an exploration of the land use impact on cloud cover. Journal
Friedl, M.A., McIver, D.K., Hodges, J.C.F., Zhang, X.Y., Muchoney, D., Strahler, A.H., of Geophysical Research D3 (101), 7149–7155.
Woodcock, C.E., Gopal, S., Schneider, A., Cooper, A., Baccini, A., Gao, F., Schaaf, C., Ramankutty, P., Foley, J.A., 1999. Estimating historical changes in global land cover:
2002. Global land cover mapping from MODIS: algorithms and early results. croplands from 1700 to 1992. Global Biogeochemical Cycles 13 (4), 997–1027.
Remote Sensing of Environment 83 (1–2), 287–302. R.B., J., Jobbágy, E., Avissar, R., Roy, S.B., Barrett, D., Cook, C., Farley, K., leMaitre, D.,
Fu, C., 2003. Potential impacts of human-induced land cover change on east Asia McCarl, B., Murray, B., 2005. Trading water for carbon with biological carbon
monsoon. Global and Planetary 37, 219–229. sequestration. Science 310, 1944–1947.
Gao, F., Schaaf, C.B., Strahler, A.H., Roesch, A., Lucht, W., Dickinson, R., 2005. Modis Robson, D.A., Frei, A., 2000. Seasonal variability of northern hemisphere snow extent
bidirectional reflectance distribution function and albedo Climate Modeling Grid using visible satellite data. Professional Geographer 52, 307–315.
products and the variability of albedo for major global vegetation types. Journal of Shashi, K., Gupta, S.K., Ritchey, N.A., Wilber, A.C., Whitlock, C.H., Gibson, G.G.,
Geophysical Research-Atmospheres 110 (D1). Stackhouse, P.W., 1999. A climatology of surface radiation budget derived from
Gash, J., Nobre, C., 1997. Climatic effects of Amazonian deforestation: some results from satellite data. Journal of Climate 12, 2691–2710.
ABRACOS. Bulletin of the American Meteorological Society 78, 823–830. Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K., Tignor, M., Miller, H.
Gedney, N., Valdes, P., 2000. The effect of Amazonian deforestation on the northern (Eds.), 2007a. Climate change 2007: the physical science basis. Contribution of
hemisphere circulation and climate. Geophysical Research Letters 27 (19), 3053–3056. Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on
Gibbard, S., Caldeira, K., Bala, G., Phillips, T.J., Wickett, M., 2005. Climate effects of global Climate Change. Cambridge University Press, Cambridge. Summary for Policymakers.
land cover change. Geophysical Research Letters 32 (23). Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K., Tignor, M., Miller, H. (Eds.),
Govindasamy, B., Duffy, P.B., Caldeira, K., 2001. Land use changes and northern 2007b. Climate change 2007: the physical science basis. Contribution of Working Group I
hemisphere cooling. Geophysical Research Letters 28 (2), 291–294. to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change.
Hall, D.K., Riggs, G.A., 2007. Accuracy assessment of the MODIS snow products. Cambridge University Press, Cambridge. Ch. 10, Global Climate Projections.
Hydrological Processes 21 (12), 1534–1547. Stackhouse, P.W., Gupta, S.K., Cox, S.J., Chiacchio, M., Mikovitz, J.C., 2000. The WCRP/
Hansen, M., DeFries, R., Townshend, J.R., Carroll, M., Dimiceli, C., Sohlberg, R., 2006. GEWEX surface radiation budget project release 2: an assessment of surface fluxes
Vegetation Continuous Fields MOD44B, 2001 Percent Tree Cover, Collection 4, at 1 degree resolution. In: Smith, W.L., Timofeyev, Y.M. (Eds.), International
University of Maryland, College Park, Maryland, 2001. Radiation Symposium, St. Petersburg, Russia, July 24–29, 2000: IRS 2000: Current
House, J., Prentice, I.C., Quéré, C.L., 2002. Maximum impacts of future reforestation or Problems in Atmospheric Radiation.
deforestation on atmospheric CO2. Global Change Biology 8, 1047–1052. Tans, P., 2008. Globally averaged marine surface annual mean data. National Oceanic
Jr, F.O. (Ed.), 1990. Carbon Dioxide and Climate, third edition. Carbon Dioxide and Atmospheric Administration. Earth System Research Laboratory. www.esrl.
Information Analysis Center, Oak Ridge, Tennessee. noaa.gov/gmd/ccgg/trends.
Kleidon, A., Fraedrich, K., Heimann, M., 2000. A green planet versus a desert world: UNFCCC, 2001. United Nations Framework Convention on Climate Change. (2001) Decision
estimating the maximum effect of vegetation on the land surface climate. Climatic 11/CP.7:Land use, land-use change and forestry. Publication FCCC/CP/2001/13/Add.1.
Change 44 (4), 471–493. Weaver, A.J., Eby, M., Wiebe, E., Bitz, C.M., Duffy, P.B., Ewen, T.L., Fanning, A.F., Holland,
LCD-BU, 2008. Land cover dynamics group at Boston university, modis land cover data M.M., MacFayden, A., Matthews, H.D., Meissner, K.J., Saenko, O.A., Schmittner, A.,
set. http://www-modis.bu.edu/landcover/. Wang, Z., Yoshimori, M., 2001. The UVic Earth System Climate Model: Model
Luyssaert, S., Schulze, D., Borner, A., Knohl, A., Hessenmoller, D., Law, B., Ciais, P., Grace, description, climatology and application to past, present and future climates.
J., 2008. Old-growth forests as global carbon sinks. Nature 455, 213–215. Atmosphere-Ocean 39 (4), 361–428.

You might also like