Science of the Total Environment 468469 (2013) S85S92

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Land atmosphere feedbacks and their role in the water resources

of the Ganges basin

R.J. Harding a,, E.M. Blyth a, O.A. Tuinenburg b, A. Wiltshire c

a
Centre of Ecology and Hydrology, Wallingford, Oxon., UK, OX10 8BB
b
Wageningen University and Research Centre, Wageningen, Netherlands
c
Met Ofce, FitzRoy Road, Exeter, UK, EX1 3PB

H I G H L I G H T S

The Ganges and Indus basins contain one of the largest contiguous areas of irrigation in the world.
The processes and magnitude of feedbacks between the land surface and the atmosphere depend on spatial scale.
Feedbacks into the Planetary Boundary Layer from irrigated land will reduce evaporation by as much as 31%.
Increased cloud amount due to evaporation by irrigated areas may reduce evaporation by between 5 and 10%.
Approximately 40% of the precipitation falling in the Himalayas originates from the irrigated areas in northern India and Pakistan.

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

Article history: The northern Indian subcontinent has frequently been identied as a hotspot for land atmosphere interac-
Received 21 June 2012 tions. It is also a region with the highest concentration of irrigated land and highest (and increasing) popu-
Received in revised form 28 February 2013 lation density in the world. The available water in the region with which to grow food depends on the
Accepted 1 March 2013
Asian monsoon, groundwater and melt from Himalayan snows. Any changes or disruptions to these sources
Available online 1 April 2013
of water could threaten the food supply. It is therefore essential to understand how the land surface, and in
Keywords:
particular irrigated land, interacts with the atmosphere. It is anticipated that the interactions will occur on
Irrigation many scales. To an extent the magnitude and form of these will depend on the depth of the atmosphere
Ganges which is affected. Thus at the local, or micro, scale it is the surface layer (some 10s m deep) which is cooled
Feedbacks and moistened by the evaporation of irrigated water, at the meso-scale the Planetary boundary layer (up to 1
Rainfall or 2 km) will be modied with possible atmospheric moistening, increased cloud and rain formation and at
very large scales the whole dynamics of the south Asian Monsoon will be affected. This illustrates a strong
interaction between the Asian monsoon and the regional topography. Of considerable signicance is the nd-
ing in this paper that up to 60% of the evaporation from irrigated areas in the summer months is ultimately
recycled to Himalayan rainfall and so feedbacks to river ows in the Ganges.
Crown Copyright 2013 Published by Elsevier B.V. All rights reserved.

1. Introduction
The Ganges Basin is home to almost one half a billion people and is
one of the most densely populated regions of the world. The majority
of the population rely on agricultural production in the region and
75% are rural populations. Currently agricultural production in the
basin broadly balances consumption, although increasing population,
increasing wealth and climate change are likely to push this balance
into decit in the coming decades. About one half of the total area
of the Ganges basin is under cultivation and of this one third (approx-
imately 35 m ha) is irrigated. The largest contiguous areas of irriga-
tion (with over 75% of the land area irrigated) are found in north
India and Pakistan along the Ganges and Indus rivers (see e.g. Siebert
Corresponding author. Tel.: +44 1491 692240.
E-mail address: rjh@ceh.ac.uk (R.J. Harding).
et al., 2005). In the states of Haryana and Uttar Pradesh, for example,
over 50% of the total geographical areas of the states are under irrigation,
(http://www.fao.org/nr/water/aquastat/irrigationmap/in/index.stm). A
recent study Biemans et al. (2013this issue) suggests only 44% of the ir-
rigation demand in the Ganges can be met by surface waters and the re-
mainder is extracted from other sources, mainly deep groundwater. This
has led to a widely observed decrease in regional ground water levels
(e.g. Rodell et al., 2009; Wada et al., 2010). Only 10% of the ow of the
Ganges is supplied by the melt from the snow and ice areas of the
Himalayas (Immerzeel et al., 2010), the rest coming from the rainfall
within the basin, primarily during the monsoon season.
In the agricultural regions of the Ganges the average rainfall is be-
tween 600 and 800 mm, concentrated during the monsoon period, July
to September. Outside the monsoon period and during monsoon breaks
irrigation is necessary to maintain the high agricultural production. This
leads to evaporation rates much higher than for rainfed agriculture
(or natural forest cover).

0048-9697/$ see front matter. Crown Copyright 2013 Published by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.scitotenv.2013.03.016
S86 R.J. Harding et al. / Science of the Total Environment 468469 (2013) S85S92
In many studies the Ganges basin appears as a hotspot particularly
susceptible to land/atmosphere interactions. For example Koster et al.
(2004) and Seneviratne et al. (2006) identify, within an ensemble of at-
mospheric models, some regions of high land-atmosphere coupling
through the soil moisture, i.e. the Great Plains of north America, north-
ern India and the Sahel. High evaporation rates associated with irrigated
areas will lead locally to lower temperatures and higher humidity close
to the surface, leading to reduced potential and actual evaporation (see
e.g. Douglas et al., 2006). At a regional scale higher humidity in the
lower atmospheric (or planetary) boundary layer may increase cloud
cover and rainfall. The reduced temperature and surface turbulent
heat ux may, however, reduce the depth of the planetary boundary
layer and the generation of convection, and hence convective rainfall.
The balance of these various effects will be determined by the magni-
tude of the surface uxes and the stratication (temperature and hu-
midity) of the lower atmosphere.
Classically the strength of the monsoon systems is regarded as a
consequence of the land-sea temperature contrast however the situ-
ation is more complex, particularly for the South Asian monsoon
which is a fully coupled oceanlandatmosphere system (Turner
and Annamalai, 2012). The land/sea warming contrast occurs over a
signicant depth of the troposphere and is initiated in the spring by
surface sensible heating over the Tibetan Plateau, brought about by
the increasing solar forcing, while latent heat release during the mon-
soon helps maintain the contrast as the solar forcing declines later in
the year (Li and Yanai, 1996; Levine and Turner, 2012). Reduced sur-
face temperatures due to irrigation would therefore be expected to
reduce the intensity of the monsoon systems (Lee et al., 2009) how-
ever increased evaporation and latent heat released at height may
ameliorate this effect. Douville et al. (2001) contrast the land surface
inuence in India and the Sahel and conclude that although precipita-
tion does increase as a consequence of increasing evaporation this is
counterbalanced, in the case of the Indian peninsula, by a reduced
moisture convergence. Saeed et al. (2009) looked at these inuences
in more detail using a regional climate model, with and without irri-
gation. They found increased rainfall over the irrigated areas due to
increased local moisture recycling and also an increase of the pene-
tration of rain bearing depressions travelling inland from the Bay of
Bengal, caused by a reduction in the westerly ows from the Arabian
Sea. The overall impact of irrigation in the Indian sub-continent is
complex and will depend on the thermodynamics of the atmosphere,
the regional circulations and spatial scales of the irrigation. In paper
we will discuss the strength of these different feedbacks from the
Fig. 1. Chain of feedbacks between the land surface and atmosphere as a function of spatial scale.
extensive irrigated land and the meteorology at different scales:
micro-scale (less than 10 km), meso-scale (or regional scale) (10 km
to 100 km) and climate (or continental) scale (above 100 km) and
their respective impacts on the water resources of the region (see
Fig. 1).
2. Microscale
2.1. Different drivers of evaporation in the Indian subcontinent
Evaporation at the land surface depends on the energy supply, the
drying power of the air and water supply. The equation that most
neatly summarises these controls on evaporation is the Penman-
Monteith (Monteith, 1965) equation:
 
E A cp q=ra = 1 rs =ra 1
Where E is the latent heat of evaporation (W m 2), is the rate
of change of the saturated humidity with temperature (K 1), A is the
available energy (W m 2), is the density of the air (kg m 3), cp is
the heat capacity of the air (J kg 1 K 1), q is the humidity decit
of the air (kg kg 1), ra is the aerodynamic resistance (s m 1), rs is
the surface resistance (s m 1) and is the psychometric constant
(cp/, K 1).
From this expression, we can split the potential evaporation
(Weedon et al., 2011) into the two main drivers of evaporation: the
energy coming from the sun (expressed in the numerator Eq. (1) by
the A, referred to here as Radiation Evaporative Demand or RED)
and the drying of the air (expressed in the numerator of Eq. (1) by
cpq/ra, referred to here as Wind Evaporative Demand or WED).
These two factors change with region and with season. For instance,
Shuttleworth et al. (2009) dened the ratio of these two drivers of
evaporation as a climatological resistance which give a theoretical
explanation for differences in trends between open water and actual
evaporation rates, depending on the aridity of the climate.
In the India region, the two drivers have quite different values. Using
meteorological data derived from observations (the WATCH forcing
data, see Weedon et al., 2011) the values of the RED and WED vary
over the season for a typical area within the Ganges river basin, as
shown in Fig. 2.
From Fig. 2, it can be seen that for the Ganges river basin, evapora-
tion is mainly driven by radiation. It reaches a peak in the pre-
 R.J. Harding et al. / Science of the Total Environment 468469 (2013) S85S92
Fig. 2. Mean-monthly values of RED (open circles) and WED (closed squares) for a region
in the Ganges River basin.
monsoon months (May and June) and a minimum in the winter when
the sun is low (November to March) and in the middle of the mon-
soon (July and August) when it is cloudy.
A key period for the evaporation is the pre-monsoon. To illustrate
the spatial patterns of evaporative demand for this period, Fig. 3
shows the mean values of RED and WED for June.
It is clear from these two contrasting maps that the Himalayas sig-
nicantly cool the air such that the wind-driven evaporative demand
is low (WED) compared to the radiation driven demand (RED) for
these high-altitude regions. For the rest of the subcontinent, the
drivers are of similar magnitude. The irrigated areas (not shown) lie
along the southern edge of the Himalayas region, and a comparison
of the maps shows that the irrigated area is half covered by the low
WED region (in purple). This means that the feedbacks due to the
changes in the wind driven demand are only half as important as pos-
sible changes to the net radiation due to changes in clouds, which is
also evident from Fig. 2.
It is also possible to study the contrasting role of water demand
and water supply on actual evaporation. Teuling et al. (2009) (using
FLUXNET data) and Miralles et al. (2011) (using an earth observation
product, combined with a simple land-water accounting model that
assimilates observed surface soil moisture and rainfall) have pub-
lished maps of the extent to which the actual evaporation is dominat-
ed by the evaporative demand compared to the supply of water. The
Teuling et al analysis does not include Asia, but in the Miralles et al
analysis (see Fig. 8 in Miralles et al., 2011) for the Ganges river
basin, the actual evaporation is dominated by the evaporative de-
mand. This could reect the fact that the region is irrigated.
2.2. Scale and feedbacks
In theory it is possible to study the impact of an increase in water sup-
ply without considering feedbacks. If the area where the increase is very
small (for instance an Evaporation Pan of the order of a few metres)
then the impact of the evaporation on the overlying temperature and hu-
midity is mixed into the lower atmosphere downwind before it can im-
pact on the evaporation itself. An isolated patch will therefore not have
an effect on the atmosphere and can be considered as a one-way (atmo-
sphere to land) process. Larger patches (up to a few hundred metres)
may moisten and cool the surface layers (the lower few 10s metres) of
the atmosphere. Studies of the effect of a heterogeneous mixture of
such dry and wet patches have shown that there is an impact of the
size of the patches since the evaporation from the wet patch is enhanced
by warm dry air owing off the dry patch more than the evaporation
from the dry patch is suppressed by the cool moist air owing off the
wet patch (Blyth, 1995). The modelling studies showed that there can
S87
be an increase of up to 30% when the size of patch decreases from
1 km down to 1 m. This is an extreme case with well watered vegetation
mixed in with bare, dry soils. If there are several of these patches and
cumulatively they amount to a signicant area, then the evaporation
from these patches will affect the overlying atmospheric conditions.
If the anomalous area is spatially large (greater than 10 km), the
evaporative uxes will alter the overlying atmosphere (at least as
high as the Planetary Boundary Layer, 1 to 2 km) to a much greater
extent, resulting in a change in the evaporation. In other words,
there is a feedback. This is discussed in the next section.
3. Mesoscale
3.1. Planetary boundary layer (PBL) feedbacks
The planetary boundary layer is the part of the atmosphere that is
impacted by the surface during the course of a day. During the day,
the sun warms the surface and the surface warms and moistens this
layer of the atmosphere through the turbulent uxes of sensible
heat and moisture. The second part of the feedback chain is the
impact of evaporation (and also sensible heat uxes) on the PBL
moisture and temperature and the subsequent effect that has on
evaporation itself. This straightforward feedback loop is logical, but
the state (temperature and humidity) of the PBL is not only depen-
dent on the surface uxes of heat and moisture; also the interaction
with the overlying free atmosphere via entrainment plays an impor-
tant role. During the course of a day PBL drying can occur due to
mixing in of dry air, in spite of an upward surface moisture ux.
This feedback needs to be considered when trying to estimate surface
evaporation (McNaughton and Spriggs, 1986; Jacobs and de Bruin,
1992). Heerwaarden et al. (2010) developed an analytical model to
quantify the effects of this feedback while Blyth and Jacobs (2011)
used a numerical model to show how the PBL structure would affect
the water resources of a region. The simple numerical model, de-
scribed below, was used with theoretical boundary layer structures.
The PBL can be envisaged as a box of air of height h (m) with a
semi-permeable lid at the top (Huntingford and Monteith, 1998).
This box grows as the day progresses with heat coming into the box
from below and with air being entrained into the box from above.
The evolution of the temperature (and humidity) within the bound-
ary layer is governed by the latent heat (E) and the sensible heat,
(available energy, A, minus latent heat: AE) entering the PBL
open-box from the surface and the temperature and humidity of the
air being entrained into the box from above as the box grows:
m dh
cp h AE cp s hm 2
t dt
qm dh
h E qs hqm 3
t dt
Where s (K) is the temperature of the original morning prole
outside the PBL (it depends on h) and qs (kg kg 1) is the specic hu-
midity of the original morning prole. A simple model for the rate of
growth (dh/dt) is to assume that it is determined by the energy going
into the PBL from the surface and the thermodynamic stability of the
original morning prole, given by the two lapse rates of temperature
and humidity, and q respectively:
  
dh=dt AE 0:07E= cp 0:61 q 4
The lapse rate describes the atmospheric stability: a low value is
similar to a lid being placed on the PBL, with a strong negative feed-
back keeping the evaporation limited whereas a high value of the
lapse rate results in a quickly growing PBL, with warm dry air being
entrained into the PBL enhancing the evaporation. According to this
S88 R.J. Harding et al. / Science of the Total Environment 468469 (2013) S85S92

Fig. 3. Map over India of RED (left) and WED (right) in W m2 for month of June.

simple theory, the relevant lapse rate is that given by the morning ra-
diosonde. A study of the atmospheric proles in the region from ra-
diosondes was made by Tuinenburg et al. (2011). As an example of
the range of atmospheric stabilities in the region, the proles of tem-
perature for June are analysed here. The range of temperature lapse
rates from these proles is summarised as a Histogram in the Fig. 4.
The mean temperature lapse is 0.004 K km 1 and the range is
0.003 K km 1.
To illustrate how important this PBL feedback is on the surface
evaporative demand, a numerical example is drawn up. The morning
temperature is taken to be 20 C and the humidity is 0.016 kg kg 1
(these are taken from the WATCH forcing data for the region) and
the mid-day maximum available energy of 500 Wm 2 is used. In
the rst instance the effect of dry surface conditions on the evapora-
tive demand in the presence of feedback is calculated and then it is
shown how sensitive the results are to the stability of the atmo-
sphere. Table 1 shows the evaporative demand of a large area that
has a well watered surface compared to the evaporative demand of
a dry surface. The results are given for the several stabilities show in
Fig. 4. The simple model described above (Eqs. (1) to (4)) is used
with representative values of surface and aerodynamic resistances
and air temperatures and humidities for the region as indicated in
the table.
The model reproduces similar temperatures with irrigated and
non-irrigated surfaces (not shown) to the model results reported in
Section 3.2. This example shows that the existence of irrigation can
alter the temperature and humidity to decrease the evaporative de-
mand on a cloud-free day by an average of 26%, while the sensitivity
to the existing proles in the region show that this can vary from 12%
to 31%. The weakness of this analysis is that it does not consider the
likelihood of advection and the possibility that the presence, or ab-
sence, of irrigated land on the regional scale is likely to inuence
the initial stratication of the atmosphere. These limitations can
only be addressed using more complete three dimensional models.
3.2. Feedback on cloud formation
It is possible to study cloud development using a one-dimensional
framework similar to that used in the previous section for the PBL de-
velopment, however, it is much less robust. Ek and Holtslag (2004)
used a model to explore the role of soil moisture on cloud develop-
ment in the Netherlands. They performed Single Column Model
(SCM) studies for a few special cases where PBL cloud formation
appeared to be highly sensitive to both surface evaporation and the
atmospheric stability above the PBL. The idea that cloud formation
is a result of simple one-dimensional thermodynamic processes has
also been explored by Findell and Eltahir (2003) who studied atmo-
spheric proles over the USA to create maps of areas and conditions
where soil moisture values could affect the formation of convection.
Tuinenburg et al. (2011) used this framework to explore the possibil-
ity of feedbacks between the Indian irrigated land and rainfall in the
region. They found that the strongest potential for feedbacks occurred
in the pre-monsoon and the post-monsoon periods.
These one-dimensional analyses capture only some of the process-
es involved in cloud and rainfall production and three dimensional
circulations and advection may be important. This might be especially
true in this region where the monsoon system dominates the meteo-
rology of the important wet season. To explore the interaction be-
tween the land and the atmosphere more fully a regional climate
model can be used. This incorporates many more impacts of the
land surface on the atmosphere including the mesoscale dynamics
that may form as a result of the pattern of altered energy balance

Table 1
Evaporative demand (ED) under different conditions, using A = 500 hour since sunrise/
12 W m2, =293 K, qm = 0.01 kg kg1, ra = 20 sm1, rs = 30 sm1 for irrigated,
120 sm1 for dry.

Lapse rate, ED with dry ED with irrigated surface Role of irrigation: %

(K m1) surface (W m2) change in evaporative
(W m2) demand

177 156 12
0.0015 221 179 19
0.0025 245 189 22
0.0035 261 193 26
0.0045 273 196 28
0.0055 282 198 30
0.0063 289 199 31
Fig. 4. Histogram of temperature lapse rates June over India.
 R.J. Harding et al. / Science of the Total Environment 468469 (2013) S85S92
at the surface. A version of the Met Ofce Hadley Centre RCM
(HadRM3) was developed to include the MOSES II land surface
scheme (Essery et al., 2003) and a simple irrigation scheme incorpo-
rated. In the MOSES II land surface scheme plants become water
stressed when soil moisture drops below a critical point. At this
point plants start to limit their stomatal conductance and control
the loss of moisture, the effect of this is to limit their uptake of carbon
dioxide and therefore photosynthesis. In the irrigation scheme water
is added to the soil to keep soil moisture at or above the critical point.
Soil moisture is updated every 5 min to be consistent with the atmo-
spheric timestep in the RCM. Implicit within this approach is a very
efcient method of irrigation, particularly compared to spray or
ood irrigation. Furthermore, the assumption is made that water for
irrigation is not limited by supply. The area equipped for irrigation
is derived from Siebert et al. (2005) global map of irrigated area.
Two model simulations using the RCM were made, one a control
and the other including the irrigation scheme. Boundary conditions
for driving the RCM runs were taken from a HadCM3 run using an
A1B scenario. The RCMs were allowed to reach a quasi-equilibrium
before simulations were completed over a ten year period between
1990 and 2000. Fig. 5 shows the resulting impact of irrigation to air
temperature, humidity and downward short-wave radiation that
were simulated by the model. This modelling study demonstrated
that, although changes in modelled rainfall were not signicant,
there was an important impact on reduced air temperatures which
affect the evaporative demand (this can be seen in the simple
one-dimensional analysis) and in an increased cloudiness which
also reduces evaporation.
The size of this decrease in shortwave radiation is of the order of
10 W m 2. Compared to the potential evaporation of the irrigated
area (see Fig. 2) this is about 5 to 10% of the potential evaporation.
The changes in the temperature and humidity however reduce the
potential evaporation by 15%. So, although the radiation is a strong
driver of the evaporation in the irrigated region, these feedbacks on
the humidity and temperature are up to three times stronger.
4. Climate scale
4.1. Monsoon system
The monsoon system directly controls the transport of water into
the region and is an atmospheric response to the contrast between
the heating of the land surface and the oceans. The land warms quick-
ly in the spring and the ocean slowly, retaining its heat for longer. The
subsequent impact on the dynamics of the global climate is that there
is a strong seasonality to the weather in the Indian region.
The impact can be altered by the strength of the contrast between
the land and the ocean. Clearly with an increase in water on the land,
Fig. 5. Maps of changes in air temperature, humidity and downward shortwave radiation due to presence of irrigation in the RCM model during March, April and May averaged over
a ten year period between 1990 and 2000.
S89
the contrast will be reduced. Modelling studies (Saeed et al., 2009)
showed that the reduced land-sea contrast caused by the presence
of the irrigation is a factor in the monsoon. As expected, when the
RCM model was run with and without irrigation, the irrigation weak-
ened the monsoon ows but there was an interaction with the topog-
raphy of the region so the weakened westerly ows from the Arabian
sea allowed increased penetration of rainfall systems along the
Ganges Basin from the Bay of Bengal.
Tuinenburg et al. (2011)'s study of the observed (from Radio-
sondes) atmospheric structures in the region show a potential alter-
ation of the timing of the monsoon due to changes in PBL moisture
from irrigated land. During the pre-monsoon, monsoon and post-
monsoon season, a positive inuence of soil moisture to convective
precipitation is expected. When the land surface is wet due to irriga-
tion, convective showers could be initiated during the pre-monsoon
and post-monsoon periods (MarchMay and SeptemberOctober).
From the water availability perspective, this could lengthen the mon-
soon period. Positive feedbacks are also expected during the monsoon
season. Although most precipitation falls during the main period
(JuneAugust), monsoon break periods, during which the dominant
monsoon ow stops for up to three weeks, are a common feature
and are linked to the interaction between the monsoon convection
and the larger scale circulations (Krishnan et al., 2009). The effect of
a moist land surface might trigger some additional convection thus
ameliorating somewhat the persistence of breaks.
4.2. Feedbacks from irrigation to Himalayan glaciers
In addition to these alterations of the monsoon system there is
another potential large-scale feedback that occurs due to the simple
addition of moisture from the land into the atmosphere: recycling of
the moisture from the irrigated region into the precipitation in the
Himalayas onto the glaciers that feed the Ganges river that supplies
the water for the irrigation.
To determine the Himalaya glacier moisture sources, an atmo-
spheric moisture tracking scheme is used that is based on the quasi-
isentropic back-trajectory tracking (QIBT) scheme (Dirmeyer and
Brubaker, 2007). This scheme uses forcing from atmospheric vari-
ables (three dimensional wind speeds and specic humidity) from at-
mospheric model output or reanalysis data. For each location and
time in the analysis, the precipitation is split into parcels of water
that are back-tracked through the atmosphere, along the path that
the parcel has travelled. Along this path, the evaporation is attributed
to the precipitation that is tracked back. Thus the sources of precipita-
tion can be located.
In this study, the data used to force the water-tracking scheme is
the ERA-interim dataset (Dee et al., 2011) and the method used is de-
scribed in Tuinenburg et al. (2012). This reanalysis data describes the
S90 R.J. Harding et al. / Science of the Total Environment 468469 (2013) S85S92
atmospheric moisture ow well. However, the surface moisture
uxes (evaporation and precipitation) are less certain, especially in
the Himalayas where measurements are sparse. Moreover, although
Tuinenburg et al. (2012) showed that the soil moisture assimilation
follows the annual irrigation cycle in India, no explicit irrigation is in-
cluded in ERA-interim. In this study, the moisture parcels are tracked
for a period of 30 days, any moisture still left in the parcel after this
period is regarded as unclassied. The evaporation sources are classi-
ed according to the amount of irrigation present at that location, the
river basin of the location and whether the evaporation originates
from land or sea. The temporal variability of these sources is shown
in Fig. 6.
The amount of moisture that is classied in the tracking scheme is
always at about 80% of the moisture released. The remaining 20% of
the moisture cannot be classied within the 30 days of tracking, this
means that this moisture denitely does not originate from local
sources. The other two curves in the gure are a subset of the land
surface sources: sources from irrigated areas and sources from the
Ganges river basin.
This analysis shows that the largest part (about 80%) of the precip-
itation in the Himalayas originates as evaporation from land surfaces.
The fraction of moisture that originates from local sources (here as-
sumed to be equivalent to the Ganges basin) shows a peak of 30%
during spring and a broader peak of 50% during autumn with a total
of 32%. These peaks correspond to the transition periods between
the northerly winter monsoon and southerly summer monsoon
ows. During the main monsoon season (JJA) the Himalaya precipita-
tion is dominated by moisture that has evaporated from the Indian
Ocean, while during the winter period (DJF), the Himalaya precipita-
tion is limited and is supplied by evaporation from far upwind areas.
However, during the periods between the (wet) summer monsoon
and the (dry) winter monsoon, the contribution of local evaporation
to Himalaya precipitation is greatest because these periods are not
dominated by strong monsoon ows and evaporation is not
transported out of the Ganges/Himalaya domain as quickly as during
the monsoon periods.
The source of the Himalaya glacier precipitation is shown in the
three-month footprint (Fig. 7). These gures show the spatial distri-
bution of the evaporation that falls again as precipitation in area
enclosed in the black lines. The scales show mm of evaporation over
Fig. 6. Annual cycle of classied sources of precipitation in the Himalayas (10 day
smoothing).
the three month period, which correspond to the classied precipita-
tion in the three month period in the black enclosed area (1.00, 1.47,
4.51 and 2.00 mm/day for DJF, MAM, JJA and SON, respectively). As
suggested by Keys et al. (2012), the scale also shows the correspond-
ing cumulative percentage of the evaporation.
During the dry winter season DJF, the moisture comes from the
west and apart from some evaporation from nearby the precipitation
location, most of the precipitation originates from the oceanic areas.
During MAM, more precipitation is tracked backwards, and the
winds have shifted slightly, so more evaporation sources from the
land surface in central Asia and the Middle East are present. June
JulyAugust is the wettest season. The moisture in the Northern
part of the enclosed area originates from the land masses in Europe
and Asia, which have a higher evaporation now than in other seasons,
whereas the moisture in the Eastern part of the enclosed area origi-
nates from oceanic areas, and is transported northwards by the mon-
soon ows. During JJA, the land surface in India provides a signicant
share of the precipitation in the focus area. During SON, the majority
of the precipitation in the focus area originates as evaporation from
south Asia's land surface. Oceanic sources contribute some moisture
to the precipitation as well, but the transport of evaporation from
land surface further away is negligible.
5. Discussion and conclusions
Interactions between the atmosphere and the land surface, partic-
ularly irrigation, will occur on many scales. To a large extent the mag-
nitude and form of these will depend on the depth of the atmosphere
which is affected. Thus at the local, or micro, scale it is the surface
layer some 10s m deep which is cooled and moistened by the evapo-
ration of irrigated water reducing the potential evaporation. Calcula-
tions presented here show that over the Ganges region the
evaporation demand is primarily driven by radiation and so the
micro-scale feedbacks of moistening the surface layers may have a
smaller effect than in more arid regions. At the meso-scale the Plane-
tary Boundary Layer will be modied through moistening and in-
creased cloud formation. The degree to which this happens will
depend on both the surface uxes and the stratication of the atmo-
sphere into which the PBL is growing. Using typical measured atmo-
spheric proles and calculated evaporation rates into a boundary
layer model shows the evaporative demand could be reduced by
26% on a cloud free day due to high surface evaporation. More de-
tailed calculations using a regional climate model which include
cloudy days suggest smaller impacts with a 15% decrease in evapo-
rative demand due to the cooling and moistening of the atmosphere
and a further 2 to 10% due to increases in cloud amount.
At larger scales rainfall will be impacted. Tuinenburg et al. (2011)
show recycling ratios of up to 60% for the whole of the Ganges basin
during June, July and August, but again this is a very scale dependent
process and the majority of the evapo(transpi)ration is recycled a
large distance (500 to 1000 km) away from source. Interestingly a
considerable amount of water is transferred from the irrigated re-
gions to the Himalayas, thus providing a vital hydrological feedback.
The majority of studies of feedbacks rely on models. A recent study
of the impact of soil moisture on precipitation from observed satellite
products (Taylor et al., 2012) shows that convective triggering often
takes place at the interface between wet and dry soil patches,
where both a sufcient surface heating and further fuelling of the
convective system with moisture occur. The process involved is
small scale, but the effect is large scale. The process is not captured
by either the PBL modelling shown above nor the RCM models,
which always calculate an increase in rainfall with an increase in
soil moisture, while this analysis of the observations show that a de-
crease in soil moisture triggers rainfall events. Taylor et al. (2012) il-
lustrates there are many important interactions between the land
surface and the atmosphere which occur at sub-grid scale which are
 R.J. Harding et al. / Science of the Total Environment 468469 (2013) S85S92
Fig. 7. Evaporation sources of Himalaya precipitation during the four seasons, based on ERA-interim (2004). The areal sum of evaporation equals the total precipitation in the
enclosed Himalaya area.
not included in our current global (and often regional) climate
models. Further feedbacks are likely to occur through changes in the
chemical and aerosol loading of the lower atmosphere through
man-made causes, for example, Bollasina et al. (2011) suggest that
increasing aerosols may be the cause of the observed drying of the
south-Asian monsoon in the last 50 years.
Irrigation enhances the regional evaporation in the Ganges basin
considerably. This reduces the regional evaporative demand some-
what but, more importantly, enhances the regional rainfall (although
these effects are far from local). Unfortunately these effects depend
critically on the season, on the stratication of the atmosphere and
on the dynamics of the monsoon (modulated considerably by the
presence of the complex topography of the region) so are difcult
to generalise. However, considerations of the water resources of the
region and possible impact and adaptation of global and regional
changes (climate as well as increasing consumption) will need to
take these feedbacks into account. This can only be done by consistent
multi-scale modelling of the combined land surface, ocean and land
surface system. For example, caution needs to be taken in using Re-
gional Climate Models which do not include irrigation to drive ofine
hydrological impacts models to inform adaptation to climate change.
The missing processes and feedbacks may signicantly alter the
projected regional climate change as well as the projected impacts
on water resources.
S91
Conict of interest
The Authors know of no conicts of interest with respect to this paper.
Acknowledgements
A. Wiltshire was partly supported by the Joint DECC/Defra Met Of-
ce Hadley Centre Climate Programme (GA01101) and the HighNoon
project funded by the European Commission Framework Programme
7 under Grant Nr. 227087. Obb Tuinenberg was also part funded by
the EU HighNoon Project. Richard Harding and Eleanor Blyth were
supported by the NERC Changing Water Cycle Programme and core
funding.
References
Biemans H, Speelman L, Ludwig F, Moors E, Wiltshire AJ, Kumar P, Gerten D, Kabat P.
Future water resources for food production in ve South Asian river basins and po-
tential for adaptation A modeling study. Sci Total Environ 2013;486
469(Suppl.):S11731(this issue).
Blyth EM. Using a simple SVAT scheme to describe the effect of scale on aggregation.
Boundary Layer Meteorol 1995;72:26785.
Blyth EM, Jacobs CMJ. Developing a feedbacks toolkit for regional water resource as-
sessments. Hydro-climatology: Variability and Change (Proceedings of symposium
J-H02 held during IUGG2011 in Melbourne, Australia, July 2011), 344. IAHS
Publ.; 2011.
S92 R.J. Harding et al. / Science of the Total Environment 468469 (2013) S85S92
Bollasina MA, Ming Y, Ramaswamy V. Anthropogenic aerosols and the weakening of
the south Asian summer monsoon. Science 2011;334:5025.
Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S. The ERA-Interim
reanalysis: conguration and performance of the data assimilation system. Q J R
Meteorol Soc 2011;137:55397.
Dirmeyer PA, Brubaker KL. Characterization of the global hydrologic cycle from a
back-trajectory analysis of atmospheric water vapor. J Hydrometeorol 2007;8:
2037.
Douglas EM, Niyogi D, Frolking S, Yeluripati JB, Pielke Sr RA, Niyogi N, et al. Changes in
moisture and energy uxes due to agricultural land use and irrigation in the Indian
Monsoon Belt. Geophys Res Lett 2006;33:L14403.
Douville H, Chauvin F, Broqua H. Inuence of soil moisture on the Asian and African
monsoons. Part1: mean monsoon and daily precipitation. J Climate 2001;14:
2381403.
Ek MB, Holtslag AAM. Inuence of soil moisture on boundary layer cloud development.
J Hydrometeorol 2004;5:8699.
Essery RLH, Best MJ, Betts RA, Cox PM, Taylor CM. Explicit representation of subgrid
heterogeneity in a GCM land surface scheme. J Hydrometeorol 2003;4(3):
53043.
Findell KL, Eltahir EAB. Atmospheric controls on soil moistureboundary layer interac-
tions. Part II: Feedbacks within the continental United States. J Hydrometeorol
2003;4:57083.
Heerwaarden CC, Villa-Guerau de Arellano J, Gounou A, Guichard F, Couvreux F. Under-
standing the daily cycle of evapotranspiration: a method to quantify the inuence
of forcings and feedbacks. J. Hydrometeorology 2010;11:140522.
Huntingford C, Monteith JL. The behaviour of a mixed-layer model of the convective
boundary layer coupled to a big leaf model of surface energy partitioning. Bound-
ary Layer Meteorol 1998;88:87-101.
Immerzeel WW, Beek van LPH, Bierkens MFP. Climate change will affect the Asian
water towers. Science 2010;328:13825.
Jacobs CMJ, De Bruin HAR. The sensitivity of regional transpiration to land-surface
characteristics: signicance of feedback. J Clim 1992;5:68398.
Keys PW, van der Ent RJ, Gordon LJ, Hoff H, Nikoli R, Savenije HHG. Analyzing
precipitationsheds to understand the vulnerability of rainfall dependent regions.
Biogeosciences 2012;9:73346. http://dx.doi.org/10.5194/bg-9-733-2012.
Koster RD, Dirmeyer PA, Guo ZC, Bonan G, Chan E, Cox P, et al. Regions of strong
coupling between soil moisture and precipitation. Science 2004;305:113840.
Krishnan R, Kumar V, Sugi M, Yoshimura J. Internal feedbacks from monsoonmidlatitude
interactions during droughts in the Indian summer monsoon. J Atmos Sci 2009;66(3):
55378.
Lee E, Chase TN, Rajagopalan B, Barry RG, Biggs TW, Lawrence PJ. Effects of irrigation
and vegetation activity on early Indian summer monsoon variability. Int J Climatol
2009;29:57381.
Levine RC, Turner AG. Dependence of Indian monsoon rainfall on moisture uxes
across the Arabian Sea and the impact of coupled model sea surface temperature
biases. Clim. Dyn. 2012;38(1112):216790.
Li C, Yanai M. The onset and interannual variability of the Asian summer monsoon in
relation to landsea thermal contrast. J Climate 1996;9:35875.
McNaughton KG, Spriggs TW. A mixed layer model for regional evaporation. Boundary
Layer Meteorol 1986;34:24362.
Miralles DG, De Jeu RAM, Gash JH, Holmes TRH, Dolman AJ. Magnitude and variability
of land evaporation and its components at the global scale. Hydrol Earth Syst Sci
2011;15:96781.
Monteith JL. Evaporation and environment. Symp Soc Exp Biol 1965:20534.
Rodell M, Velicogna I, Famiglietti JS. Satellite-based estimates of groundwater deple-
tion in India. Nature 2009;460:999-1002.
Saeed F, Hagemann S, Jacob D. Impact of irrigation on the South Asian summer
monsoon. Geophys. Res. Lett. 2009;36. L20711.
Seneviratne SI, Luthi D, Litschi M, Schar C. Landatmosphere coupling and climate
change in Europe. Nature 2006;443:2059.
Shuttleworth WJ, Serrat-Capdevila A, Roderick ML, Scott RL. On the theory relating
changes in area-average and pan evaporation. Q J R Meteorol Soc 2009;135:123047.
Siebert S, Doll P, Hoogeveen J, Faures J-M, Frenken K, Feick S. Development and valida-
tion of the global map of irrigation areas. Hydrol Earth Syst Sci 2005;9:53547.
Taylor CM, de Jeu RAM, Guichard F, Harris PP, Dorigo WA. Afternoon rain more likely
over drier soils. Nature 2012;489:4236.
Teuling AJ, Hirschi M, Ohmura A, Wild M, Reichstein M, Ciais P, et al. A regional per-
spective on trends in continental evaporation. Geophys Res Lett 2009;36:l02404.
http://dx.doi.org/10.1029/2008gl036584.
Tuinenburg OA, Hutjes RWA, Jacobs CMJ, Kabat P. Diagnosis of local landatmosphere
feedbacks in India. J Climate 2011;24:25166.
Tuinenburg OA, Hutjes RWA, Kabat P. The fate of evaporated water from the Ganges
basin. J Geophys Res 2012;117(D01107):17.
Turner AG, Annamalai H. Climate change and the South Asian summer monsoon. Na-
ture 2012;2:58795.
Wada Y, van Beek LPH, van Kempen CM, Reckman JWTM, Vasak S, Bierkens MFP. Global
depletion of groundwater resources. Geophys Res Lett 2010;37:L20402.
Weedon GP, Gomes S, Viterbo P, Shuttleworth WJ, Blyth E, sterle H, et al. Creation of the
WATCH forcing data and its use to assess global and regional reference crop evapo-
ration over land during the twentieth century. J Hydrometeorol 2011;12:82348.