Introduction
It is challenging to gain improved understanding of the
different mechanisms that drive land–atmospheric interaction. One of these
mechanisms is the contribution of terrestrial evaporation to local and remote
precipitation (i.e. moisture recycling). Early studies have used analytical
methods to estimate the amount of precipitation that recycles within a basin
or area of interest see
e.g.. However, this
field of study has advanced much with the introduction of atmospheric
moisture tracking methods to estimate moisture recycling
e.g..
Several studies have shown global maps of continental precipitation
recycling, indicating that about 40 % of the continental precipitation is
of continental origin, but this number is much higher in, e.g., China
.
included a wide variety of moisture tracers into a general
circulation model (GCM) to track water and its age through the atmosphere as
well as through the soil. It was found, for example, that, counting from the
moment of evaporation from the ocean, the mean water age of precipitating
water in north-eastern Asia could exceed half a year whereby a water particle
had been recycled on average twice. A comprehensive overview and
quantification of import and export of water vapour between countries was
given by .
While nearly all previous studies focused on the “recycled” part of
precipitation, also focused on the recycled part of
evaporation. They, for example, found that in evaporation recycling “hot
spots” such as eastern Africa and the northern Amazon about 60 to 90 %
of the evaporation returns as continental precipitation. When it comes
to moisture recycling as a metric for local land–atmosphere coupling, the
follow-up study of solved the problem of scale and
shape dependency of the regional moisture recycling ratios by converting
these to length scales of the recycling process (average travel distances
under local conditions). They showed that in the tropics and in mountainous
terrain these length scales can be as low as 500 to 2000 km.
Spatially distributed global maps of actual average travel distances to
precipitation were given by .
Partly these findings will be due to a passive role of moisture in the hydrological cycle e.g.,
but it is also suggested by the results of other studies that less (more) evaporated moisture
actively leads to a decrease (increase) in precipitation e.g..
Therefore, moisture recycling is believed to be of
significant importance for water resources, agriculture, and
ecosystems. Some studies have looked specifically at these issues. For
instance, studied the central United States
plains and concluded that local evaporative fluxes ensure
ecoclimatological stability through a continued moisture contribution
when advective fluxes diminish. Another example of ecosystem
importance is the study by , who found that air
passing over dense vegetation produces much more rain than air passing
over sparse vegetation. Regarding agriculture,
reported that reduced moisture recycling due to land-cover change may
lead to potential crop yield reductions of 1 to 17 % in the
world's breadbasket regions, while other studies have looked at the
positive effect of irrigation in increasing moisture recycling
e.g.. Considering moisture recycling as
something that could potentially be managed, proposed
the concept of the precipitationshed as a tool to assess the
vulnerability of a certain region to land-use changes in its moisture-contributing regions.
Land-use changes change not only total evaporation but also its
partitioning into its direct and delayed components. It is therefore
somewhat surprising that all moisture recycling studies have reported
their results in terms of moisture recycling due to total evaporation
only. It has been speculated, however, that interception (direct
evaporation) and transpiration (delayed evaporation) are likely to
play a different role in moisture recycling . This
has, however, never been quantified. A possible method would be to try
to link stable water isotope measurements to moisture recycling
e.g.. used the
deuterium excess (d-excess) value of stable water isotopes to estimate that 20–40 %
of the evaporative flux in the Amazon basin is fractionating the
isotopic composition. Theoretically, d-excess values in precipitation
from for example the Global Network of Isotopes in Precipitation
(GNIP) database could be combined with estimates
of moisture recycling e.g. to infer the
contributions of fractionating and non-fractionating
evaporation. However, the spatial and temporal resolution of available
isotopic data is rather limited. Another difficulty is the fact that,
while it is generally accepted that open-water evaporation is
fractionating and evaporation of transpired water is not, for
vegetation interception and floor interception the extent of
fractionation is less clear e.g..
Global land-surface models generally include a partitioning of
terrestrial evaporation into several direct and delayed
components. These components include evaporation from transpiration,
vegetation interception, floor interception, soil moisture, and open
water, although the names and exact definitions of these terms can
differ from model to model. In any case, information on these
individual components is not often reported and data are generally not
provided e.g.. This is probably the reason that,
to our knowledge, no studies applying numerical atmospheric moisture
tracking see have considered the
different components of terrestrial evaporation separately. In order
to obtain a tailor-made data set of partitioned evaporation,
, the companion paper, hereafter referred to as Part 1,
developed STEAM (Simple Terrestrial Evaporation to Atmosphere
Model). This is a global hydrological land-surface model, which is
specifically focused on realistic estimations of partitioned
evaporation and how this depends on vegetation and land use.
The goal of this paper is to investigate and quantify the importance
of the different components of evaporation in the hydrological cycle
over continents. We aim to present a new image of the global
hydrological cycle which includes quantification of partitioned
evaporation and moisture recycling as well as the atmospheric
residence times of the individual components. Furthermore, we aim to
provide spatially distributed global maps of different moisture
recycling metrics that describe the role of interception and
transpiration for local and remote moisture recycling processes in
time and space. This provides new information on the susceptibility of
regions to land-use changes. For example, if region A receives
precipitation from transpiration in region B's dry season, then region
A may experience increased dryness if region B were to be desertified.
Methods
Data
The input data for our atmospheric moisture tracking model,
WAM-2layers (Water Accounting Model–2 layers) (see Appendix A), come
from STEAM (Part 1)
and the ERA-Interim reanalysis (ERA-I) . STEAM
evaporation data are also based on ERA-I (see Part 1 for details). The
output of STEAM is the total evaporation E over all terrestrial areas partitioned into
five components:
E=Evegetation _ interception+Efloor _interception+Esoil _ moisture+Einland _ waters+Etranspiration.
Vegetation interception is all water that is intercepted by the leaves, branches, and stems of
vegetation. We define floor interception as all water that is intercepted by the ground and
litter surface. Soil moisture evaporation is physical evaporation from the unsaturated zone.
Transpiration is water that evaporates from the stomata of leaves and draws from the unsaturated
zone as well (see Part 1 for details). In this paper we combine the direct (purely physical) evaporative
fluxes into one term Ei, containing evaporation from interception,
soil moisture, and inland waters:
Ei=Evegetation _ interception+Efloor _interception+Esoil _ moisture+Einland _ waters.
This term consists of the direct fluxes from vegetation interception, floor
interception, and soil moisture evaporation, which have a small storage
reservoir and short residence time at the surface (Part 1, Figs. 9 and 10).
As the relative global contribution from the soil moisture and inland waters
is quite small (Part 1, Fig. 2), this term mainly represents interception,
but regionally other components can dominate. Transpiration, the delayed
(biophysical) evaporative flux, on the other hand, provides a slow feedback
with a large storage reservoir, which is the other component that we track:
Et=Etranspiration.
From ERA-I we use precipitation and evaporation over the oceans. For
the terrestrial evaporation we use the partitioned evaporation fluxes
computed by STEAM (forced by ERA-I; see Part 1). Furthermore, we use
specific humidity and zonal and meridional wind speed from ERA-I. We
downloaded these data at model levels spanning the atmosphere from
zero pressure to surface pressure. Surface fluxes were downloaded at
3-hourly intervals and the other data at 6-hourly intervals. The data we use
are on a 1.5∘ latitude × 1.5∘ longitude
grid and cover the period of 1998–2009, but the results are presented
for 1999–2008, because we use 1 year as model spin-up for
both the backward and forward tracking. In Appendix A we show annual
average, as well as January and July figures for precipitation, direct
evaporative fluxes (interception), and the delayed evaporative flux
(transpiration). Appendix B contains further details about the
moisture tracking in WAM-2layers.
We consider STEAM and ERA-I as adequate data sources to perform realistic
moisture tracking, and their global estimates of evaporation and
precipitation fall well within the range of estimates given by other studies.
It was shown that ERA-I performs better in reproducing the hydrological cycle
than ERA-40 and even performs better in terms of water
balance closure than the other reanalysis products MERRA (Modern-Era
Retrospective Analysis for Research and Applications) and CFSR (Climate
Forecast System Reanalysis) . used both ERA-I
and MERRA as inputs for WAM-2layers and showed that global moisture recycling
patterns are not very different. However, it should be clear that the
moisture recycling metrics presented in Sect. 3 are in fact dependent on the
input data. In Sect. 3.6 we investigate the robustness of our main results.
Definitions of moisture recycling metrics
Here, we define moisture recycling metrics, each of which contains
different information about the moisture recycling process. First, we
start with the metrics related to continental moisture recycling,
which are measures for land–atmosphere coupling at continental
scale. Second, we define metrics related to the timescale of the
moisture recycling process. Finally, we define metrics that act as
measure for local moisture feedback.
Continental moisture recycling
In the context of continental moisture recycling see
also, precipitation on land P can be separated as
follows:
P=Po+Pc=Po+Pc,i+Pc,t,
where Po is the part that is of oceanic origin and
Pc is the continentally recycled part of the
precipitation (i.e. most recently evaporated from a continental
area). Pc can be split further into Pc,i
(i.e. the recycled precipitation that originates from
vegetation interception, floor interception, soil moisture and inland
waters) and Pc,t (i.e. the recycled precipitation that
originates from transpiration). The “continental precipitation
recycling ratio for interception” is defined as
ρc,i=Pc,i/P
and the
“continental precipitation recycling ratio for transpiration” as
ρc,t=Pc,t/P.
Also in the context of continental moisture recycling, we split land evaporation E:
E=Eo+Ec=Eo,i+Eo,t+Ec,i+Ec,t,
where Eo is the part of the evaporation that
precipitates on the ocean and Ec is the continental
recycling part (i.e. returns as continental
precipitation). Subscripts i and t denote the interception (Eq. 2)
and transpiration (Eq. 3) respectively. It also holds that Ei=Eo,i+Ec,i
and that Et=Eo,t+Ec,t. This also allows us to define
the “continental evaporation recycling ratio for interception” as
εc,i=Ec,i/E
and the “continental evaporation recycling ratio for transpiration” as
εc,t=Ec,t/E.
The two metrics in Eqs. (8) and (9) both carry information about their
relative contribution to moisture recycling as well as their relative
contribution to total evaporation. To study the recycling efficiency
of the individual partitioned fluxes, we define the “continental
evaporation recycling efficiency for interception” as
εc,ii=Ec,i/Ei
and the “continental evaporation
recycling efficiency for transpiration” as
εc,tt=Ec,t/Et.
Atmospheric lifetime of recycled moisture
Previous studies by and by
calculated the local depletion and restoration
timescales of atmospheric moisture, defined as the atmospheric
moisture storage over precipitation and evaporation
respectively. estimated the average timescale
over land to be around 9 days. However meaningful, these timescales
only provided local information; they did not indicate the
actual time spent in the atmosphere by a recycled water
particle. Therefore, we propose new metrics that describe the actual
time spent in the atmospheric. We define the “lifetime of continentally recycled
precipitation”:
τρ,c=NPc←Ec,
where N stands for the time spent in the atmosphere, or, in other
words, the age of the water particle. The lifetime of continentally recycled
precipitation τρ,c is a measure at
the point where a water particle precipitates and stands for the
average time spent between continental evaporation and continental
precipitation, or, in other words, the average age at the point where
a water particle precipitates. Note that τρ,c
only provides information on the recycled part of
the precipitation and not on the total precipitation (see
Eq. 4). Likewise we define the “lifetime of the interception that recycles on
land” as
τε,c,i=NEc,i→Pc,i
and the “lifetime of the transpiration that recycles on land” as
τε,c,t=NEc,t→Pc,t.
Both metrics in Eqs. (13) and (14) are defined at the place where
evaporation occurs at the land surface (Ec in Eq. 7)
and determine the average time an evaporated water particle that returns as precipitation
on land will spend in the atmosphere. For the calculation of these
lifetimes we included water age tracers in our model (Appendix B3).
Local recycling and the length scales of evaporated water
Besides the continental recycling metrics, we are also interested in
the feedback between evaporation and precipitation locally. For
a certain predefined region (e.g. a grid cell) we can split
precipitation and evaporation as follows:
P=Pa+Pr=Pa,i+Pa,t+Pr,i+Pr,t
and
E=Ea+Er=Ea,i+Ea,t+Er,i+Er,t,
where Pa is the part of the precipitation that comes
from moisture advected into the region, Ea is the part
of the evaporation that is advected away from the grid cell, and
Pr and Er are the regional recycling
parts (i.e. recycle within the same region). Subscripts i and t
again denote interception (Eq. 2) and transpiration (Eq. 3)
respectively. This also allows us to define the “regional
precipitation recycling ratio” as
ρr=Pr/P
and the “regional evaporation recycling ratio” as
εr=Er/E.
We should realise that these ratios are scale- and shape-dependent,
which is problematic as grid cells generally differ in scale and
shape. Some studies have tried to overcome this problem by scaling
such ratios to a common reference area
e.g.. However, such an
approach fails to properly take into account the shape of the region
and the orientation of the prevailing winds.
As an alternative, developed a method that
yields scale- and shape-independent measures for local
evaporation–precipitation interaction. Suppose we are following an
atmospheric water particle along a streamline in the same direction as
the wind direction. The streamline starts in point X0 and ends in
point X1, and the distance between X0 and X1 is
Δx. Based on and ,
we can write
ρX1Δx=1-exp-Δx/λρ
and
εX0Δx=1-exp-Δx/λε,
where ρX1 is the precipitation recycling ratio in
X1 and εX0 is the evaporation
recycling ratio in X0 (i.e. the fraction of evaporation in
X0 that returns as precipitation to the land surface along the
streamline). λρ represents the “local length scale of
precipitation recycling”,
λρ=SuhE,
and λε is the “local length scale of evaporation
recycling”:
λε=SuhP,
where S is atmospheric moisture storage (i.e. precipitable water) and
uh is horizontal wind speed. These length scales λ
have dimension length [L] and can be physically interpreted as the average
travel distances before precipitation if SuhE is
constant upwind, or as the average travel distance after evaporation if
SuhP remains equal downwind. However, it is
generally unlikely for these quantities to remain equal over a large
distance, so λ must be interpreted as the local process scale of
recycling. When we consider the distance Δx to be sufficiently small,
we can also obtain the areal average regional precipitation recycling ratio
(Eq. 17) by integrating Eq. (19), dividing by the distance and substituting
Eq. (21):
ρr=Δx+λρexp-Δxλρ-λρΔx.
The exact solution for λρ is
λρ=ΔxWexp1ρr-1ρr-1+11-ρr,
where Wa is the Lambert W function
e.g.. In this research, however, we are
interested in the local length scale for interception and
transpiration recycling. Using the fluxes in Eq. (16), we first define
the “regional evaporation recycling efficiency for interception” as
εr,ii=Er,i/Ei
and the “regional
evaporation recycling efficiency for transpiration” as
εr,tt=Er,t/Et.
Analogous to Eqs. (23) and (24), the “local length scale of
evaporation recycling for interception” can be found by
λε,i=ΔxWexp1εr,ii-1εr,ii-1+11-εr,ii,
and the “local length scale of evaporation recycling for
transpiration” can be found by
λε,t=ΔxWexp1εr,tt-1εr,tt-1+11-εr,tt.
Note that both λε,i and λε,t are defined by SuhP
(Eq. 22), so they are only equal if evaporation from interception and
transpiration occur simultaneously. However, in many cases they will
occur at different times when the quantity
SuhP is different. As a result, λε,i
and λε,t are likely
to have different values and can be effectively used in revealing
their relative importance for local moisture feedback. In Appendix B4
it is explained how the variable inputs in Eqs. (24), (27), and (28)
were obtained in this study. Table 1 provides an overview of all metrics used in this paper.
Overview of moisture recycling metrics used in this study.
Moisture recycling metric
Symbol
Formula(s)
Meaning
Figures
Continental moisture recycling metrics
Continental precipitation recycling ratio
ρc
Pc/P
Fraction of P that comes from terrestrial E
2a
Continental precipitation recycling ratio
ρc,i
Pc,i/P
Fraction of P that comes from Ei
2b, 6a, 6b
for interception
Continental precipitation recycling ratio
ρc,t
Pc,t/P
Fraction of P that comes from Et
2c, 6c, 6d
for transpiration
Continental evaporation recycling ratio
εc
Ec/E
Fraction of E that returns as terrestrial P
3e
Continental evaporation recycling ratio
εc,i
Ec,i/E
Fraction of E that is interception
3a
for interception
which returns as terrestrial P
Continental evaporation recycling efficiency
εc,ii
Ec,i/Ei
Fraction of Ei that returns as terrestrial P
3b
for interception
Continental evaporation recycling ratio
εc,t
Ec,t/E
Fraction of E that is transpiration
3c
for transpiration
which returns as terrestrial P
Continental evaporation recycling efficiency
εc,tt
Ec,t/Et
Fraction of Et that returns as terrestrial P
3d
for transpiration
Lifetime of continentally recycled precipitation
τρ,c
NPc←Ec
Time spent in the atmosphere by Pc
4a
Lifetime of the interception that recycles on land
τε,c,i
NEc,i→Pc,i
Time Ec,i spends in the atmosphere
4b, 6e, 6f
Lifetime of the transpiration that recycles on land
τε,c,t
NEc,i→Pc,i
Time Ec,t spends in the atmosphere
4c, 6g, 6h
Regional/local moisture recycling metrics
Regional precipitation recycling ratio
ρr
Pr/P
Fraction of P that comes from E
–
in the same region
Regional evaporation recycling ratio
εr
Er/E
Fraction of E that returns as P
–
in the same region
Regional evaporation recycling efficiency
εr,ii
Er,i/Ei
Fraction of Ei that returns as P
–
for interception
in the same region
Regional evaporation recycling efficiency
εr,tt
Er,t/Et
Fraction of Et that returns as P
–
for transpiration
in the same region
Length scale of precipitation recycling
λρ
Suh/E ,
Atmospheric travel distance of P
4a
ΔxWexp1ρr-1ρr-1+11-ρr
under local conditions*
Length scale of evaporation recycling
λε,i
Suh/P ,
Atmospheric travel distance of Ei
4b, 6i, 6j
for interception
ΔxWexp1εr,ii-1εr,ii-1+11-εr,ii
under local conditions*
Length scale of evaporation recycling
λε,t
Suh/P ,
Atmospheric travel distance of Et
4c, 6k, 6l
for transpiration
ΔxWexp1εr,tt-1εr,tt-1+11-εr,tt
under local conditions*
* Note that this is not an actual travel
distance, but an indication of the local intensity of the hydrological
cycle.
Global hydrological cycle over land, i.e. all continents considered together (1999–2008).
Fin is the atmospheric moisture of oceanic origin that crosses the ocean–land boundary
and enters the atmosphere above land. Fout is the atmospheric moisture that leaves the
ocean–land boundary towards the ocean. Thus, X represents the atmospheric moisture of oceanic
origin that passes through the continental atmosphere, but never precipitates. Precipitation on
land P (set to 100 %) is composed of moisture evaporated from the ocean Po and
a recycled part Pc. On the land surface, water runs off Q, or evaporates through
direct evaporation Ei or through transpiration Et. Part of this evaporation is
lost to the ocean Eo, while other parts of the evaporation recycle Ec,i and
Ec,t. Evaporation data are from STEAM (Part 1), precipitation data are from
ERA-I, the recycled fractions and lifetimes are calculated by WAM-2layers (Appendix B), and
the other terms follow from the water balance.
Symbols are further explained in Sects. 2.2.1 and 3.1.
Results and discussion
New image of the hydrological cycle over land
Figure 1 presents an image of the global hydrological cycle over
land. In contrast to traditional images of the hydrological cycle
e.g., we include a quantification of moisture
recycling, partitioned evaporation, and the lifetime of all these
components separately. Before precipitation falls on land, its average
atmospheric residence time is about 10 days. We estimate that about
38 % of continental precipitation P is transformed into runoff
Q, and the remaining part evaporates by direct (purely physical)
fluxes Ei and by the delayed (biophysical) flux
Et (see Part 1). A portion of this land evaporation is
advected to the oceans and precipitates there Eo. The
remaining part recycles over land, but, interestingly, interception
Ec, i and transpiration Ec,t do so in
different relative magnitudes. Of interception, 60 %
(Ec,i/Ei) recycles, while
transpiration recycles slightly less at 56 % (Ec,t/Et). The lifetime in the atmosphere of
evaporated water is on average more than a week, which is similar to
a previous estimate of 9.2 days . The recycled
part of evaporation, however, spends on average less than a week in
the atmosphere. We can also observe that (the recycled part
of) interception has a shorter lifetime in the atmosphere. Finally,
global continental precipitation recycling Pc is
estimated at 36 %, slightly less than the 40 % estimated in
a previous study using WAM-1layer and ERA-I evaporation
. This is mainly caused by the other forcing data, STEAM
instead of ERA-Interim, but about 0.5 % is due to the inclusion of the second
layer in WAM-2layers. Averaged globally, the recycling efficiencies
and atmospheric lifetimes are not very different for interception and
transpiration, but locally these differences can be large, which we
show in Sects. 3.2 to 3.6, where we discuss the spatial patterns of
the magnitudes and timescales of the recycling fluxes in the
hydrological cycle.
Continental precipitation recycling
(1999–2008). (a) Continental precipitation recycling ratio
ρc, (b) continental precipitation
recycling ratio for interception ρc,i, and
(c) continental precipitation recycling ratio for
transpiration ρc,t. The colour scale of
(b) ends at 0.41, which is the global average fraction of
direct evaporative fluxes (interception); the colour scale of
(c) ends at 0.59, which is the global average fraction of
delayed evaporative flux (transpiration). The arrows in (a)
indicate the vertically integrated moisture fluxes.
Continental moisture recycling
Figure 2 shows the annual average continental precipitation recycling
ratios for total evaporation (Fig. 2a), for interception (Eq. 5 and
Fig. 2b), and transpiration (Eq. 6 and Fig. 2c). While interpreting
the figure, it should be remembered that “interception” includes
evaporation from the vegetation, floor, soil, and inland waters
(Eq. 2). The areas that depend heavily on continental precipitation
recycling are potentially susceptible to (upwind) changes in land
use. Animations 1 to 3 (Supplement) illustrate how we obtained Fig. 2
with forward tracking runs of tagged terrestrial evaporation,
interception, and transpiration. They show the fraction of atmospheric
moisture originating from terrestrial evaporation, i.e. interception and
transpiration respectively, averaged for each day (actual model time
step is 15 min) and clearly show the seasonality of moisture recycling.
As a sidenote, some differences in patterns between Fig. 2a and Fig. 3
are caused by the inclusion of the second layer in our atmospheric moisture tracking method.
For example, the higher precipitation recycling values in North America
are likely caused by the inclusion of fast recycling, while the lower values
along the coast of western Africa and the southeastern coast of South America are
likely caused by accounting for vertically sheared winds. These findings are similar to those of Fig. 6.
Precipitation recycling due to transpiration shows higher values
(Fig. 2c) and is in the absolute sense more important than
interception (Fig. 2b). Although the patterns of Fig. 2b and c are
very similar, there are a few noteworthy differences, for which we can
think of two reasons: first, dominance of one type of evaporative flux
in a certain area and, second, dominance of one type of evaporative flux
during a certain part of the year with different prevailing winds. For
example, in South America, the hot spot of interception recycling
is situated more to the north compared to the hot spot of
transpiration recycling. This is explained by high interception in the
Amazonian rainforest (Fig. A1b), compared to transpiration being high
throughout the continent (Fig. A1c), and by transpiration being more
dominant during winter when the atmospheric flow is more directed to
the south (Fig. A2).
Continental evaporation recycling (1999–2008). (a)
Continental evaporation recycling ratio for interception
εc,i, (b) continental evaporation
recycling efficiency for interception εc,ii,
(c) continental evaporation recycling ratio for
transpiration εc,t, (d)
continental evaporation recycling efficiency for transpiration
εc,tt, (e) continental
evaporation recycling ratio εc, and
(f) εc,ii-εc,tt.
Grey values on land indicate no data, due to the fact
that the evaporative flux in question is 0. The arrows in
(a) and (b) indicate the vertically integrated
moisture fluxes.
The complementary process of precipitation recycling is evaporation
recycling. The different metrics corresponding to evaporation
recycling (Eqs. 7 to 11) are shown in Fig. 3. Regions with high
evaporation recycling (i.e. high ratio and substantial evaporation) are important source regions for sustaining
downwind precipitation. Figure 3a and c contain information about
where the respective evaporative fluxes are important as well as to
which regions they supply the moisture. The sum of Fig. 3a and c
leads to Fig. 3e. The evaporation recycling efficiencies (Fig. 3b and
d) just contain information about the likelihood of a particle to
recycle after continental evaporation. From Fig. 3f it can be seen
that in most regions of the world interception evaporation (Figs. 3b
and A1b) is more likely to return as precipitation over land than
transpiration (Figs. 3d and A1c). This is especially the case in
regions with a relatively small continental mass (in relation to the
prevalent winds) and distinct wet and dry seasons, such as southern
Africa, India, and Australia, where transpiration in the dry season is
relatively likely to return to the ocean (see also the seasonal differences in moisture recycling metrics in Sect. 3.5).
In the Congo and northern Amazon regions, the continental evaporation
recycling efficiencies are high (Fig. 3b and d) and the
differences between relative interception and transpiration recycling
are practically zero (Fig. 3f), which indicates that, independent of
the type of evaporation process, each water particle is equally likely to return to the continent. This
indicates strong local recycling, or at least evaporative fluxes that
contribute to precipitation elsewhere on the continent, throughout the
year. However, Fig. 3f also indicates some regions in Eurasia where
transpiration is more likely to return to the continent (in
blue). This can probably be explained by the fact that in these areas
almost all evaporation in winter comes from interception (Fig. A2c),
which, for the most part, is subsequently advected over and away from
the relatively dry continent (Fig. A2a). In other words, the moisture
coming from interception has less opportunity to recycle, whereas
transpiration is present only in the wetter summer season and has more
opportunity to recycle (see also the seasonal differences in moisture recycling metrics in Sect. 3.5).
Average atmospheric lifetimes of recycled moisture
(1999–2008). (a) Lifetime of continentally recycled precipitation
τρ,c (defined at the point of
precipitation), (b) lifetime of the interception that recycles on land
τε,c,i
(defined at the point of evaporation), and (c) lifetime of the transpiration that recycles on land
τε,c,t (defined at the point of
evaporation). Grey values on land indicate no data, due to the fact
that the evaporative flux in question is 0. The arrows in
(a) indicate the vertically integrated moisture fluxes.
Atmospheric lifetime of recycled moisture
Figure 4 shows the time spent in the atmosphere by the moisture that
recycles over land. Figure 4a indicates the time that continentally
evaporated moisture has spent in the atmosphere until it precipitates
(Eq. 12). In other words, it is the time component of Fig. 2a. Note
that in places where ρc (Fig. 2a) is low the
corresponding regions in Fig. 4a contain little
information. Figure 4b (Eq. 13) and c (Eq. 14) indicate the time it
takes before direct (interception, soil moisture, and inland waters)
and delayed (transpiration) evaporative fluxes return to the
terrestrial land surface.
Figure 4b and c are the time components of Fig. 3b and d. We can see that in
general the direct evaporative fluxes (Fig. 4b) remain in the atmosphere for
a shorter period of time compared to transpiration (Fig. 4c). We can explain
this by the fact that the terrestrial timescales of the direct evaporative
fluxes are much shorter than those of transpiration (Part 1, Figs. 9 to 11).
The differences between Fig. 4b and c are less strong in the very wet
tropical regions around the Equator, as well as in the Andes and Himalayas.
This is probably caused by the absence of distinctively different
precipitation-triggering mechanisms throughout the year. On the other hand,
we see several regions where the atmospheric lifetime of interception
recycling (Fig. 4b) is much lower than that of transpiration recycling
(Fig. 4c). Many of these regions correspond to those identified in Fig. 3f
(e.g. southern Africa, India, and Australia). However, in contrast to
Fig. 3f, the lifetime of interception recycling is also shorter in northern
Eurasia, which is probably due to the fact that Fig. 4 just considers the
recycled part of the precipitation.
Interestingly, recycled precipitation (Fig. 4a) in North America has spent less time in the atmosphere than
in Eurasia. We think that this could be explained by a fraction of
evaporation in North America that passes over the Atlantic Ocean in
summer and precipitates in Europe, which obviously increases the
average atmospheric residence time. This phenomenon can also be
observed from animations 2 and 3 (Supplement). It seems that
transpiration (animation 3 and Fig. 4c) is a slightly larger
contributor to this cross-continental transport than the direct
evaporative fluxes (animation 2 and Fig. 4b).
Local length scales of the moisture recycling process
(1999–2008). (a) Length scale of precipitation recycling
λρ, (b) length scale of evaporation
recycling for interception λε,i,
and (c) length scale of evaporation recycling for
transpiration λε,t. Grey values
on land indicate no data, due to the fact that the evaporative flux
in question is 0. Note that lower values indicate higher moisture
feedback strength. The arrows in (a) indicate the moisture
fluxes in the lowest part of the atmosphere (approximately the
lowest 2 km of the atmosphere at standard pressure,
Eq. B5).
Local length scales of moisture recycling
We assess local moisture recycling strength using local length scales
of moisture recycling (Eqs. 19 to 28), which are a scale- and
shape-independent alternative to the often-used regional recycling
ratios (Eqs. 17 and 18) (see also van der Ent and Savenije, 2011,
Fig. 2). Figure 5a shows the local length scale of precipitation
recycling, where the importance of local evaporation for precipitation
is indicated by a lower value. Note that the arrows in the graph now
indicate the moisture fluxes in the bottom part of the atmosphere only
(Eq. B5) as this is where the fast recycling takes place. If the
values are similar over a large area, the local length scale is also
a proxy for travel distance (e.g. ∼2000 km in
sub-Saharan Africa), despite a possible underestimation due to local
moisture not reaching the fast-moving, upper layers of the
atmosphere. The precipitation recycling length scale is generally low
in the wet tropical regions, but increases with strong winds, such as
is the case in the northern Amazon and eastern Africa (Fig. 5a). Low
length scales are also present in mountainous regions (e.g. Rocky
Mountains, Andes, Alps, Caucasus, and Tibetan Plateau) and areas of
weak winds (e.g. throughout Russia).
Figure 5b and c show the length scales of evaporation recycling for
interception and transpiration respectively. They provide a proxy for
the distance an evaporated water particle travels before returning to
the land surface. In the world's deserts there is obviously very
little precipitation, and the probability of an evaporated particle
returning locally is very low given the high local length
scales. Ignoring the deserts, Fig. 5b indicates that direct
evaporation on most of the globe has a length scale of less than
2500 km (this corresponds to ∼2 % recycling within
100 km).
We have already seen that interception in general has a higher probability to
recycle over land (Figs. 1 and 3) and returns to the land surface more
quickly (Figs. 1 and 4). Consistent with this, the length scale of
interception recycling (Fig. 5b) is much shorter compared to that of
transpiration recycling (Fig. 5c). The difference in length scales between
interception and transpiration is quite striking, especially in the temperate
zones. This is similar to the finding in Fig. 4, but seems more pronounced.
The typical timescale of a wet spell is 1–5 days , while
evaporation from interception has a timescale at the surface of the order of
hours (Part 1, Figs. 9c and d and 10c and d) and transpiration has
a timescale of the order of weeks to months (Part 1, Figs. 9a and 10a). Since
interception takes place only during wet spells and transpiration takes place
regardless, it follows that interception recycling is much more local than
transpiration recycling. During wet and dry seasons similar contrasting roles
of interception and transpiration are expected, which we will investigate in
the next section.
Moisture recycling metrics for January (left column) and July
(right column). The arrows in (a) and (b) indicate
the vertically integrated moisture fluxes, which are most relevant
for panels (a–h). The arrows in (i) and
(j) indicate the moisture flux only in approximately the
lowest 2 km of the atmosphere, which is most relevant for
panels (i–l).
Seasonality of moisture recycling metrics
A selection of moisture recycling metrics for the months of January
and July is shown in Fig. 6. In summer, the land is warmer than the
ocean and continental precipitation recycling ratios are higher,
whereas is winter this is the opposite (Fig. 6a–d). Looking at the
Northern Hemisphere's temperate and polar climate zones, the lifetimes
and length scales in winter (Fig. 6e, g, i and k) are in most places
shorter than in summer (Fig. 6f, h, j and l). This means that
evaporation in winter generally returns to the land surface more
quickly than in summer. However, evaporation in winter is much lower
(Fig. A2) and is thus a less important contributor to precipitation
than in summer (Fig. 6a–d). In the tropics and subtropics, the
moisture recycling metrics are driven more by monsoonal periods, with
stronger feedback, i.e. shorter atmospheric lifetimes (Fig. 6e–h)
and shorter length scales (Fig. 6i–l) during the monsoon season.
The different roles of interception and transpiration in the
hydrological cycle become evident when we compare January and July
(Fig. 6), relative to the annual averages (Figs. 2 to 5). For example,
it is clear that, in the Northern Hemisphere's temperate and polar
zones in January, evaporation from interception is the principal
moisture recycling mechanism (Fig. 6a vs. c, and Fig. 6i vs. k). This
is explained by the near absence of transpiration (Fig. A2e).
However, near absence of transpiration is not a necessity for
interception to be the principal recycling mechanism, which we can see
from Australia and South Africa in January (summer). This is probably
explained by the relatively small dimensions of these land masses,
which cause transpiration outside of a wet spell to be advected to
the oceanic atmosphere more often than evaporated interception.
Whereas transpiration can compensate for a reduction of interception
in the wet season, the opposite is not true, making transpiration-dependent
regions more vulnerable. For example, coastal western Africa in
January and the La Plata basin (rivers contributing to the bay bordering
Argentina and Uruguay) in July are predominantly dependent on
recycled moisture from transpiration. For both these regions, this
transpiration recycling dependence is in a period with little rainfall
(Fig. A2a and b). However, this rainfall could be important for dry season
farming and drinking water supply, making these regions susceptible to
local and remote land-use changes. These regions are particularly
threatened by upwind deforestation, which could therefore lead to
reduced precipitation in western Africa and the La Plata basin in
general, but particularly during their respective dry seasons.
Observations already show a general decrease (with some edge effects) in
precipitation over forest-to-non-forest transitions due to deforestation in
the Amazon basin . Our results suggest that reduced moisture
recycling could propagate the decline in precipitation further downwind.
showed how the northern part of the Amazon, which is wet all
year round, depends on recycled moisture and as such is vulnerable to
deforestation as well. Our results suggest that deforestation in this
northern part would mainly lead to reduced interception recycling.
Potentially, other evaporative fluxes may compensate for the reduction in
interception evaporation (Part 1, Table 5), and other well-managed vegetation
would not necessarily lead to dramatic rainfall reductions. For the southern
part of the Amazon and the link with the La Plata basin, however,
deforestation could be a much bigger problem, as reduced transpiration
recycling could lead to a drier dry season. It must be noted, however, that
the magnitude of the reduced moisture recycling effect depends on the land
use that replaces the forest. Irrigated agriculture or open water could
theoretically maintain high evaporation rates as well, but most other
land-use types would not be able to produce high evaporation rates during the
dry season.
Global differences in moisture recycling behaviour for direct evaporation (mostly interception) and transpiration for different evaporation data.
Default
Transpiration-plus
Interception-plus
STEAM change
Interception storage capacity
100 %
50 %
150 %
Unsaturated zone storage capacity
100 %
120 %
80 %
STEAM output
Total evaporation E
73 900
73 200
74 200
km3year-1
km3year-1
km3year-1
Ei/E
41 %
36 %
46 %
Et/E
59 %
64 %
54 %
Global average results
εc,ii
60 %
60 %
60 %
εc,tt
56 %
57 %
56 %
Atmospheric lifetime of Ei
8.1 days
8.2 days
8.0 days
Atmospheric lifetime of Et
9.1 days
9.1 days
9.1 days
τε,c,i
5.9 days
6.0 days
5.8 days
τε,c,t
6.8 days
6.8 days
6.8 days
Robustness of the results
We can conclude from the previous results that the roles of interception and
transpiration for moisture recycling are different due to the fact that they
have different magnitudes during wet and dry spells, and due to the fact that
they are dominant during different seasons. However, the question could be
raised of whether the different moisture recycling characteristics for
interception and transpiration would also be true if other partitioned
evaporation were used as input data. Therefore, we repeated our moisture
recycling analysis with two different input data sets: one where the
parameterisation of STEAM favours more transpiration, and another where the
parameterisation favours interception (Part 1, Table 5). The effects on the
moisture recycling efficiency and the residence times for Ei and
Et are shown in Table 2. It can be seen that the results are not
very sensitive to the parameterisation. The efficiencies of interception and
transpiration are almost the same for each of the scenarios. The residence
time in the atmosphere of (recycled) direct evaporation in the
interception-plus scenario is slightly lower than in the default scenario.
This is probably explained by the fact that in the interception-plus scenario
Ei consists to a greater extent of vegetation interception, which
is the fastest feedback process (Part 1, Table 5 and Fig. 11). The opposite
is the case for the transpiration-plus scenario. Overall, the differences
between the scenarios are minor, and thus we consider the moisture recycling
differences found for interception and transpiration to be robust results.
Summary, conclusions and outlook
The objective of this paper
was to assess the role of the different components of evaporation in the
hydrological cycle over continents. We have used the atmospheric moisture
tracking model WAM-2layers to track direct (purely physical) and delayed
(biophysical) evaporative fluxes, as computed by STEAM (Part 1). By direct
evaporative fluxes we mean the water evaporated from vegetation interception,
floor interception, soil moisture, and inland waters. Interception is what
largely dominates direct evaporation (Part 1, Fig. 2). By delayed
evaporative flux we mean transpiration.
We summarise our findings about the different roles of
interception and transpiration in the hydrological cycle as follows:
(1) 60 % of direct evaporation returns to the land surface, whereas
this is 56 %, and thus slightly less, for transpiration; (2) the
residence time of direct evaporation in the atmosphere is 8 days (6
for the recycling part only) and 9 days for transpiration; and (3) the
local length scale of interception recycling is on average much
shorter than the length scale of transpiration recycling. We attribute
these results to the fact that interception has a small storage
reservoir and therefore occurs mostly during wet spells. Transpiration,
on the other hand, draws from a large storage reservoir and can occur
during dry periods, when evaporated moisture is more likely to be
advected over large distances, as well.
Therefore, the results found are particularly useful from a landscape
resilience perspective. Regions that receive precipitation from continentally
recycled evaporation are vulnerable to upwind land-use changes. However,
a region that receives precipitation originating from interception is more
resilient to land-use changes in their source region than a region
that depends on transpiration. A land-use change could for example reduce
interception capacity, but during a wet period this is likely to be
compensated by other evaporative fluxes. Regions that receive precipitation
from continentally recycled transpiration are less resilient to land-use
changes in their source region, especially if a region's precipitation
depends on transpiration in the dry season. This is because, when vegetation
is removed, the mechanism to retain and draw moisture from the root zone is
lost as well, and total evaporation will be significantly reduced.
Our results suggest that the effect of land-use change on moisture
recycling is very different during wet and dry seasons, and also
during summer and winter, indicating that seasonality is important to
consider when analysing effects of land-use change. During the wet
season, increased or decreased interception could amplify or attenuate
the local moisture recycling signal. Still, we conclude that land-use
change needs to be drastic to influence the evaporative fluxes in
a way that this signal would have continental-scale influence. During
the dry season, land-use change (in particular deforestation) could
lead to reduced transpiration, which reduces moisture recycling, and
as such could have a domino effect on precipitation downwind. Such
potential effects of forest-to-agriculture conversion make the already
challenging task of sustainably producing enough food for a growing
population even more challenging.
On the other hand, moisture-recycling-dependent regions such as western Africa
could potentially benefit from increased rainfall due to large-scale
implementations of water harvesting, small reservoirs, and agroforestry
not only in western Africa itself but also in
central Africa (upwind).
For future studies, we expect that coupled land–biosphere–atmosphere
models will be increasingly used for predicting climate impacts due to
land-use changes. However, we must not forget the tremendous
uncertainty in the process understanding and parameterisation
underlying these models e.g.. It is not uncommon
for different models to predict different outputs for temperature
e.g., and especially precipitation
e.g., and fundamental issues are still debated,
such as the partitioning of evaporation
. Another issue requiring attention is
that recent studies have shown that increased atmospheric carbon
dioxide reduces transpiration . Our paper
shows that this will likely reduce moisture recycling and precipitation in
some regions (see Figs. 1c, 2c and d and 6c and d), making them more
vulnerable to droughts, but this clearly needs more quantification.
This paper stresses the fact that the land surface has a large
potential to influence the hydrological cycle. Quantification of exact
regional and planetary boundaries of tolerable
land-use changes before drastic precipitation changes are expected
is, however, difficult to provide. This is because our results only
allow for a first-order estimate of land-use change impacts, whereas
very drastic land-use change affects the energy balance and wind
patterns as well
e.g.. Nonetheless,
we anticipate that our results may help future coupled
land–atmosphere research to interpret whether the findings are
the result of moisture recycling or other climatic processes. As such,
we hope that this paper is useful for providing a larger context to
future regional studies examining the impact of land-use changes on
the hydrological cycle.