temperature of surface air and the temperature of condensation, can also affect the quality of reconstruction of
paleoclimates. For Antarctica, for example, the uncertainty in the reconstruction of temperature is estimated to
be of the order of 20–30% between glacial and interglacial periods, and recent work has produced evidence of
spatio-temporal variations in this relationship over the last
decades;
• Finally, it should be noted that in the tropics, it is not the
temperature of surface air that determines the isotopic
composition of rainfall. Indeed, both spatially and temporally (seasonal and inter-annual), the isotopic composition of rainfall is mainly related to its intensity (‘mass
effect’), because the intensity of the isotopic distillation
depends on atmospheric convective activity, irrespective
of the air temperature at the surface. The interpretation of
isotopic signals of tropical glaciers is therefore fundamentally different from that of polar ice. Note that at time
scales greater than a decade, it is still possible that tropical sea surface temperatures have a leading role on
tropical precipitation isotopic composition through their
impacts on atmospheric dynamics and convective activity
(see Chap. 20).
Assessing the stability of the relationships between the
isotopes of precipitation and meteorological parameters
under different climate conditions requires the exploration of
the processes associated with the three-dimensional atmospheric circulation and water cycle. This can be achieved
using general or regional atmospheric circulation models
implemented with the representation of the different isotopic
forms of the water molecule and the associated fractionation
processes. These modeling tools have been successfully used
to explore the processes affecting the isotope-temperature
relationship at glacial-interglacial scales. Current challenges
are related to the ability to perform long (multi-centennial or
longer) simulations using coupled ocean-atmosphere models
equipped with water stable isotopes to quantify the climatic
drivers of precipitation isotopic composition in different
regions and over different time scales (seasonal, inter-annual,
decennial, centennial etc.). Recently, new understanding has
emerged from in situ and remote sensing monitoring of
water vapor isotopic composition, which provides more
continuous insights than the sampling of precipitation. These
data are used to better understand the climatic drivers of
water vapor isotopic composition, at the scale of weather
events, but also to benchmark the ability of atmospheric
models to correctly simulate the origin of atmospheric
moisture.
Uncertainties Associated with Biological
Indicators
The commonly used methods to reconstruct climate from
biological assemblages are known under the term transfer
function (Imbrie and Kipp 1971). Their principle is based on
the expression of the relationship between the climate variable and the relative abundances of each taxon considered,
as if the climate were dependent on the assemblage. This is
an inverse approach, since the reality is that the assemblage
depends on climatic conditions. The direct problem is called
the response function. A few equations suffice to show the
drawback of such an approach. Note X, the assemblage, C,
all climate factors combined, D, all non-climatic factors that
may also influence X (e.g. soil), and R, the response function
of the assemblage to C and X = R(C,D). If climate dominates
over non-climatic factors, the relationship can be approximated as follows: X = R c (C). The transfer function may be
obtained by inversion: ^
C ¼ ^
R
À1
c X
ð Þ. But in general, the
number of variables included in the vector C is far fewer
than the number of X variables, and in this case, only a least
squares method can solve the system of equations by minimizing the deviation between C and its estimation which
expresses C as a function of X: ^
C ¼ ^
T X
ð Þ where ^
T denotes
the transfer function.
The ‘transfer function’ (TF) approach is based on several
assumptions which should be kept in mind:
1. climate conditions are the ultimate cause of any changes
observed in the data; human action which often modifies
the landscape is assumed to be negligible;
2. the ecological properties of the studied species have not
changed between the period of analysis and the present:
the relationships between species and climate are constant through time;
3. current observations contain all the information necessary to interpret fossil data: so, it is necessary that the
vegetation of the past, for example, survived somewhere
in the world and that we have the corresponding information. This third hypothesis, added to the second, may
be translated as the principle of uniformitarianism (the
present is the key to the past), which is implicit in any
paleontological approach.
It is clear that these three assumptions are quite strong.
The differences found between the various approaches often
stem from the fact that these assumptions are not always
entirely verifiable.
142
V. Masson-Delmotte and J. Guiot
paleoclimates. For Antarctica, for example, the uncertainty in the reconstruction of temperature is estimated to
be of the order of 20–30% between glacial and interglacial periods, and recent work has produced evidence of
spatio-temporal variations in this relationship over the last
decades;
• Finally, it should be noted that in the tropics, it is not the
temperature of surface air that determines the isotopic
composition of rainfall. Indeed, both spatially and temporally (seasonal and inter-annual), the isotopic composition of rainfall is mainly related to its intensity (‘mass
effect’), because the intensity of the isotopic distillation
depends on atmospheric convective activity, irrespective
of the air temperature at the surface. The interpretation of
isotopic signals of tropical glaciers is therefore fundamentally different from that of polar ice. Note that at time
scales greater than a decade, it is still possible that tropical sea surface temperatures have a leading role on
tropical precipitation isotopic composition through their
impacts on atmospheric dynamics and convective activity
(see Chap. 20).
Assessing the stability of the relationships between the
isotopes of precipitation and meteorological parameters
under different climate conditions requires the exploration of
the processes associated with the three-dimensional atmospheric circulation and water cycle. This can be achieved
using general or regional atmospheric circulation models
implemented with the representation of the different isotopic
forms of the water molecule and the associated fractionation
processes. These modeling tools have been successfully used
to explore the processes affecting the isotope-temperature
relationship at glacial-interglacial scales. Current challenges
are related to the ability to perform long (multi-centennial or
longer) simulations using coupled ocean-atmosphere models
equipped with water stable isotopes to quantify the climatic
drivers of precipitation isotopic composition in different
regions and over different time scales (seasonal, inter-annual,
decennial, centennial etc.). Recently, new understanding has
emerged from in situ and remote sensing monitoring of
water vapor isotopic composition, which provides more
continuous insights than the sampling of precipitation. These
data are used to better understand the climatic drivers of
water vapor isotopic composition, at the scale of weather
events, but also to benchmark the ability of atmospheric
models to correctly simulate the origin of atmospheric
moisture.
Uncertainties Associated with Biological
Indicators
The commonly used methods to reconstruct climate from
biological assemblages are known under the term transfer
function (Imbrie and Kipp 1971). Their principle is based on
the expression of the relationship between the climate variable and the relative abundances of each taxon considered,
as if the climate were dependent on the assemblage. This is
an inverse approach, since the reality is that the assemblage
depends on climatic conditions. The direct problem is called
the response function. A few equations suffice to show the
drawback of such an approach. Note X, the assemblage, C,
all climate factors combined, D, all non-climatic factors that
may also influence X (e.g. soil), and R, the response function
of the assemblage to C and X = R(C,D). If climate dominates
over non-climatic factors, the relationship can be approximated as follows: X = R c (C). The transfer function may be
obtained by inversion: ^
C ¼ ^
R
À1
c X
ð Þ. But in general, the
number of variables included in the vector C is far fewer
than the number of X variables, and in this case, only a least
squares method can solve the system of equations by minimizing the deviation between C and its estimation which
expresses C as a function of X: ^
C ¼ ^
T X
ð Þ where ^
T denotes
the transfer function.
The ‘transfer function’ (TF) approach is based on several
assumptions which should be kept in mind:
1. climate conditions are the ultimate cause of any changes
observed in the data; human action which often modifies
the landscape is assumed to be negligible;
2. the ecological properties of the studied species have not
changed between the period of analysis and the present:
the relationships between species and climate are constant through time;
3. current observations contain all the information necessary to interpret fossil data: so, it is necessary that the
vegetation of the past, for example, survived somewhere
in the world and that we have the corresponding information. This third hypothesis, added to the second, may
be translated as the principle of uniformitarianism (the
present is the key to the past), which is implicit in any
paleontological approach.
It is clear that these three assumptions are quite strong.
The differences found between the various approaches often
stem from the fact that these assumptions are not always
entirely verifiable.
142
V. Masson-Delmotte and J. Guiot