Effects of large-scale deforestation on precipitation inthe monsoon regions: Remote versus local effectsN. Devaraju1, Govindasamy Bala, and Angshuman Modak

Divecha Center for Climate Change & Center for Atmospheric and Oceanic Sciences, Indian Institute of Science, Bangalore 560012, India

Edited by Robert E. Dickinson, The University of Texas at Austin, Austin, TX, and approved February 6, 2015 (received for review December 8, 2014)

In this paper, using idealized climate model simulations, we in-vestigate the biogeophysical effects of large-scale deforestationon monsoon regions. We find that the remote forcing from large-scale deforestation in the northern middle and high latitudesshifts the Intertropical Convergence Zone southward. This resultsin a significant decrease in precipitation in the Northern Hemispheremonsoon regions (East Asia, North America, North Africa, and SouthAsia) and moderate precipitation increases in the Southern Hemi-sphere monsoon regions (South Africa, South America, and Australia).The magnitude of the monsoonal precipitation changes depends onthe location of deforestation, with remote effects showing a largerinfluence than local effects. The South Asian Monsoon region isaffected the most, with 18% decline in precipitation over India. Ourresults indicate that any comprehensive assessment of afforestation/reforestation as climate change mitigation strategies should carefullyevaluate the remote effects onmonsoonal precipitation alongside thelarge local impacts on temperatures.

deforestation | biogeophysical effects | Hadley Cell movement | ITCZ shift |monsoon regions

Historical land cover change has been one of the major driversof climate change. By the 1750s, ∼6–7% of the global land

surface area had been deforested for agriculture. Today, croplandsand pasture lands make up approximately one third of the globalland surface (1–4). In terms of area, croplands and pasture landsincreased globally from 620 million ha in 1700 to 4,960 million haby 2000 (1). This large-scale conversion of forests to croplands orgrasslands can impact climate through biogeochemical (changes inatmospheric composition) and biogeophysical (changes in physicalland surface characteristics such as albedo, evapotranspiration,and roughness length) processes.The impacts of past, present, and future biogeochemical and

biogeophysical effects from land use change have been inves-tigated by numerous studies (5–10). These studies find that thebiogeochemical process primarily causes global effects while bio-geophysical processes cause strong local effects. The combinedbiogeochemical and biogeophysical effects from land cover changein the Holocene before 1850 were modeled as a global meanwarming of 0.73 K (9).During the historical period (1750 to present day), deforestation-

associated CO2 emissions have contributed ∼180 ± 80 PgC to thecumulative anthropogenic CO2 emissions (11) and a warming of∼0.16–0.30 K (biogeochemical effect) to anthropogenic climatechange (5, 6). This warming is probably partly offset by the bio-geophysical effect of albedo increase, which may have caused aglobal mean cooling by ∼0.03–0.27 K (5, 7, 8). However, othermajor biogeophysical processes, such as reduction in evapotrans-piration and roughness length due to deforestation, could result inwarming (12).Several studies have investigated the link between land cover

change and local climate change (13–16). For example, defores-tation (16) in the tropics (18.75°S−15°N) reduces precipitation overAmazon by 138 mm/y (9.2%) and increases the temperature by1.6 K. Another study (17) simulates a 266 mm/y reduction in precipi-tation over tropics due to tropical deforestation. The biogeophysicaleffects can also have remote effects via changes in atmospheric

circulation (13, 18–20). For instance, recent studies (13, 21) finda shift in Intertropical Convergence Zone (ITCZ) due to affores-tation in entire midlatitudes or over Eurasia. These studies suggestthat the ITCZ shifts can have consequences for precipitation in themonsoon regions of northeast Asia and South Asia.Most of the monsoon regions are located within the vicinity of

ITCZ. Thus, the ITCZ shift due to land cover change via remoteeffects can affect the monsoon regions. To our knowledge, nostudy has quantified the ITCZ shift and its effect, due to large-scale deforestation, on all of the monsoon regions. In this paper,we show that the remote effect of large-scale deforestation hasa larger influence on precipitation in monsoon regions than thelocal effect, although the local effect has a larger impact onsurface temperature changes as shown in several previous studies(13–15, 21). The remote effect can be quantified through a re-lationship between the ITCZ location and the atmospheric heattransport at the equator. Our investigation has direct relevanceto changes in precipitation in monsoon regions in the past [forinstance, during the Last Glacial Maximum (LGM) and at theCretaceous–Tertiary boundary when large areas of forests werecompletely removed], to make improved assessment of risks toagriculture from changes to rainfall in the tropics (22) and tointegrated assessments of afforestation/reforestation as climatechange mitigation strategies.

ResultsGlobal-Scale Temperature and Precipitation Responses.We investigatethe effects of Global, Boreal, Temperate, and Tropical deforesta-tion on global climate (i.e., surface temperature and precipita-tion) relative to the CTL (control) case. Global mean surface airtemperature decreases by 1.50 K, 0.90 K, 0.47 K, and 0.04 K in

Significance

Biogeophysical effects such as albedo and evapotranspirationchanges due to deforestation were shown by several studies inthe past to exert strong influence on local surface temperatures.In this study, we assess the remote versus local effects of large-scale deforestation on precipitation in the monsoon regions ofthe world. In contrast to the dominant role of local effects ontemperature changes, we find that the remote effects havea larger influence than local effects on shifting the location ofthe Intertropical Convergence Zone and hence precipitation inall the monsoon regions. This result has important implicationsfor assessing the net benefits of climate change mitigationstrategies such as afforestation/reforestation and for under-standing changes in monsoon rainfall in past climates.

Author contributions: G.B. designed research; N.D. performed research; N.D. and G.B.contributed new reagents/analytic tools; N.D. and A.M. analyzed data; and N.D., G.B.,and A.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: dev@caos.iisc.ernet.in.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1423439112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1423439112 PNAS | March 17, 2015 | vol. 112 | no. 11 | 3257–3262

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

5, 2

020

the Global, Boreal, Temperate, and Tropical cases, respectively(Table 1). Correspondingly, the global mean precipitation de-creases by 33.40 mm/y (3.21%), 17.70 mm/y (1.70%), 10.50 mm/y(1.01%), and 4.88 mm/y (0.50%). Our simulated global meancooling of 1.5 K in the Global case is in close agreement withprevious modeling studies that find 1.7 K (14) or 1.6 K (15)cooling due to global deforestation. Compared with the pre-industrial climate, all simulations except the Tropical case showsignificant cooling in the Northern Hemisphere (NH; Fig.1 A−C).In the Global case, cooling averaged over NH is about 3 K. In theBoreal case, the NH mean cooling is ∼2.20 K, and it is ∼1.50 K inthe Temperate case (Fig. 1 B and C). The simulated NH cooling isconsistent with an observational study (23) that finds strong evi-dence for cooling over northern latitudes (>45°N) and weak evi-dence for warming below 35°N as a result of the biogeophysicalresponses to deforestation.The NH cooling in the Global, Boreal, and Temperate cases is

due to the increased land surface albedo and reduced absorptionof solar radiation at the surface (15). The albedo changes in thenorthern middle and high latitudes are more than 25–30% (SIAppendix, Fig. S1 A–C). The large albedo changes are because of(i) the conversion of forests into grasslands, which have relativelylarger albedo, and (ii) the seasonal presence of snow in theseareas—deforestation exposes the surface snow cover, which hasa larger albedo. This albedo-related cooling can be also rein-forced by snow and sea ice−albedo feedbacks (24). The impor-tance of the seasonal snow cover for albedo effect in the Global,Boreal, and Temperate cases can be inferred by noting that thechanges in albedo in the tropical regions in the Global case donot exceed 0–5% (SI Appendix, Fig. S1A). In the Tropical de-forestation simulation, the global mean cooling is of only 0.04 K(Table 1) because the influence of albedo and evapotranspira-tion almost counterbalance each other. However, we find stronglocal warming of more than 1 K in the forested regions of tropics:Amazon, Central Africa, and South Asia (Fig. 1D).In the Global case, increases in surface albedo in the tropical

regions (0–5%) do not produce much cooling (SI Appendix, Fig.S1A and Table S1), indicating that the changes in evapotrans-piration may have compensating effects (15). The decrease inevapotranspiration from deforestation leads to a decrease inclouds over tropical land that allows more downward solar ra-diation at the surface [SI Appendix, Fig. S2 and Table S1 (15)].Thus, while deforestation brightens the surface in the tropics, italso tends to decrease cloudiness and darkens the planet. Theseeffects nearly cancel each other, and hence the magnitude ofcooling and the changes in planetary albedo and net flux at the

top of the atmosphere (TOA) are much smaller in low latitudescompared with high latitudes (SI Appendix, Fig. S2 and Table S1).The strong local temperature responses from biogeophysical

processes can be inferred from a strong tropical mean warmingof 0.2 K (20°S−20°N, Fig. 1D) in the Tropical deforestationcase, and strong localized cooling in the midlatitudes (−0.8 K in20°N−50°N) and high latitudes (−4 K in 50°N−90°N) in Temperateand Boreal deforestation cases, respectively. The strong cooling inNH (in the Global and Boreal cases) is similar to the climate thatprevailed during LGMwhen temperatures were much cooler thantoday [by ∼3.6 K to 5.7 K (25–27)].The annual mean precipitationdeclines in the NH and increases in the Southern Hemisphere(SH) in Global, Boreal, and Temperate deforestation simulations(Fig. 2 A−C). Large changes in precipitation are prominent intropical regions in association with a southward shift of the ITCZ(Fig. 2 A−C).

ITCZ Shift. Following refs. 28 and 29, we use the precipitationcentroid (PCENT) as a metric for locating the ITCZ precipitationmaximum. The precipitation centroid is defined as the median ofthe zonal average precipitation from 20°S to 20°N. The zonalmean precipitation (from the average of the last 50 y) in-terpolated to a 0.01° grid in the tropics (20°S−20°N) allows us tolocate the precipitation centroid to a higher precision than thegrid resolution. We find a southward shift of annual mean ITCZ(PCENT) in all of the deforestation simulations: ∼1.70° in Global,∼1.02° in Boreal, ∼1.11° in Temperate, and ∼0.03° in Tropicalcase from its original position of 0.50°S in CTL case (Table 2).During boreal summer [June−August (JJA)] the location ofITCZ is inside the NH (5.60°N) in the CTL and it shifts south-ward by about 1.30° in Global, 0.80° in Boreal, 0.62° in Tem-perate, and 0.23° in Tropical deforestation simulations. In theaustral summer [December−February (DJF)], the ITCZ is lo-cated inside the SH (5.16°S) and the southward shifts haveslightly reduced magnitude (Table 2).The shifts in ITCZ in our deforestation simulations are asso-

ciated with changes in meridional heat transports (Fig. 3A),which is in agreement with earlier modeling studies that in-vestigated the impact of various climate forcings such as imposedice cover in the high latitudes, artificially enhanced albedo, etc.(30–32). In the case of Global and Boreal deforestation, forexample, the NH high latitudes absorb less solar radiation be-cause of the increase in albedo, which leads to a larger deficit inenergy in the NH. For the Global case, the TOA NH energy deficitis −0.72 W/m2, −0.54 W/m2 in the Boreal case, and −0.29 W/m2 inthe Temperate case (SI Appendix, Table S2). This NH energy deficitnecessitates an increase in heat transport (Fig. 3A) into the NH

Table 1. Global and annual mean changes of key climatic variables averaged over the last 50 y of the 80-y simulations

Variable CTL Global−CTL Boreal−CTL Temperate−CTL Tropical−CTL

Surface temperature, K 287.16 ± 0.03 −1.50 ± 0.03 −0.90 ± 0.02 −0.47 ± 0.04 −0.04 ± 0.03Precipitation, mm/y 1039.33 ± 0.82 −33.39 ± 1.15 (−3.21) −17.69 ± 1.08 (−1.70) −10.50 ± 0.95 (−1.01) −4.88 ± 0.98 (−0.47)Evapotranspiration over land,

mm/y574.11 ± 1.05 −61.01 ± 1.05 (−10.63) −28.05 ± 0.93 (−4.88) −16.53 ± 1.65 (−2.88) −20.51 ± 0.80 (−3.57)

Land surface albedo* 0.266 ± 0.000 0.056 ± 0.000 (5.6) 0.029 ± 0.000 (2.9) 0.013 ± 0.000 (1.3) 0.004 ± 0.000 (0.4)TOA albedo* 0.304 ± 0.000 0.007 ± 0.000 (0.7) 0.004 ± 0.000 (0.4) 0.002 ± 0.000 (0.2) 0.000 ± 0.000 (0.0)Sensible heat flux, W/m2 17.62 ± 0.02 0.1 ± 0.01 (0.57) 0.02 ± 0.01 (0.11) −0.06 ± 0.02 (0.34) 0.13 ± 0.02 (0.74)Latent heat flux, W/m2 82.27 ± 0.05 −2.64 ± 0.05 (−3.21) −1.43 ± 0.04 (−1.74) −0.82 ± 0.03 (−0.99) −0.35 ± 0.04 (−0.42)Changes over India in JJAS

Surface temperature, K 275.45 ± 0.07 0.32 ± 0.07 −0.16 ± 0.06 0.28 ± 0.05 0.33 ± 0.07Precipitation, mm/d 6.43 ± 0.12 −1.14 ± 0.09 (−18.00) −0.35 ± 0.08 (−5.60) −0.55 ± 0.05 (−8.60) −0.25 ± 0.04 (−4.00)Surface albedo* 0.124 ± 0.000 0.011 ± 0.000 (1.1) 0.002 ± 0.000 (0.2) 0.005 ± 0.000 (0.5) 0.005 ± 0.000 (0.5)

Changes over India in June−September (JJAS) for selected variables are also shown. Uncertainty is given by the SE computed from 50 annual meandifferences. Values within the parentheses show the percentage changes relative to control.*Albedo changes given in parentheses are absolute changes in percentage.

3258 | www.pnas.org/cgi/doi/10.1073/pnas.1423439112 Devaraju et al.

Dow

nloa

ded

by g

uest

on

May

5, 2

020

from the SH: Since the sign of the vertically averaged meridionalheat transport in the tropical region is dominated by the upperbranch of the Hadley cell, a southward shift of the NH Hadley cell(and hence the ITCZ) would facilitate more heat transport into theNH. The southward shift of ITCZ in association with NH climatecooling (∼2.45 K, Fig. 1 A and B) in our Global and Boreal de-forestation experiments agree qualitatively with simulated ITCZshifts [∼1°S (33)] and the NH cooling during LGM [∼3 K (30)]. Theincreased heat transport in the midlatitudes (Fig. 3A) is likely as-sociated with enhanced baroclinic eddy activity because of increasedmeridional temperature gradient (Fig. 1) and the correspondingincrease in vertical shear in NH midlatitude zonal mean westerlywinds (SI Appendix, Figs. S3 and S4).We also quantify the relationship between the ITCZ location

(PCENT) and atmospheric heat transport at the equator (AHTeq)as shown in SI Appendix, Fig. S5. We adopt the procedure dis-cussed in ref. 29 for the estimation of annual mean AHTeq andits seasonal cycle (Fig. 3 A and C). In the CTL case, there isa linear relationship between PCENT and AHTeq in the seasonaltime scale (SI Appendix, Fig. S5). The location of PCENT in thecontrol simulation varies from 7.50°S in February to 7.27°N inAugust. We estimate that the southward movement of PCENT is∼2.65° ±0.5° per petawatts (PW) of AHTeq on the seasonal timescale. This linear relationship is in close agreement with a recentstudy (29) that finds a shift of −2.4° per PW for CMIP3 modelsand −2.7° per PW in observations. The link between changes inAHTeq and ITCZ shifts in our deforestation experiments areconsistent with the rate determined for seasonal scale (SI Ap-pendix, Fig. S5). We find that the southward shift of ITCZ andthe corresponding increase in AHTeq in the deforestation sim-ulations occur almost throughout the year (Fig. 3 B and C). Thechanges in net TOA radiative flux are larger in NH high latitudesin the Global, Boreal, and Temperate cases compared with thetropical latitudes (Fig. 3D and SI Appendix, Table S1): While largesurface albedo changes lead to large changes in TOA fluxes in thehigh latitudes, reductions in clouds in the tropical regions because ofreduced evapotranspiration nearly offset the surface albedo-inducedTOA flux changes (15) (SI Appendix, Fig. S2 and Table S1).

In association with the ITCZ shifts, we simulate large changesin vertical motion in the tropics (30°N−30°S, more than 2 hPa/d)in the case of Global, Boreal, and Temperate cases but relativelysmaller changes (less than 0.5 hPa/d) in the case of Tropi-cal experiment (SI Appendix, Fig. S6). Thus, the Hadley cell isaffected significantly in Boreal and Temperate cases whileleast affected in the Tropical case, indicating the dominance ofremote effects over the local effects. Further, in the Global,Boreal, and Temperate simulations, there is a large reductionin atmospheric water vapor in association with NH cooling(SI Appendix, Fig. S7). The ITCZ shift simulated in the de-forestation experiments could alter the precipitation over NHand SH monsoon regions.

Effects on Monsoon Regions. To understand the changes in NHand SH monsoon precipitation, we chose the monsoonal regionsbased on the criteria developed in ref. 34 that relies on the an-nual range of precipitation. The regional monsoons as shown byboxes in Fig. 2D and SI Appendix, Table S3 are: (i) East Asian(EA), (ii) North American (NA), (iii) North African (NAf),(iv) South Asian (SAs), (v) South African (SAf), (vi) South Ameri-can (SA), and (vii) Australian (AUS) monsoons. As stated be-fore, the changes in circulation in our deforestation simulations(except Tropical case) decrease the mean monsoon precipitationin NH monsoon regions and slightly increase the mean precipi-tation in SH monsoon regions in association with the southwardshift in ITCZ (Fig. 4). For example, during JJA, the NH mon-soonal regions receive less precipitation in the Global case: EAmonsoon precipitation decreases by 10.2% (0.7 mm/d), monsoonprecipitation over SAs decreases by 11.8% [0.64 mm/d, during June−September (JJAS)] and NA monsoon precipitation decreases by4% (0.15 mm/d) (Fig. 4 and SI Appendix, Table S4). In contrast,during DJF (summer in SH), the SH monsoonal regions receivemore precipitation in the Global case: SAf monsoon precipitationincreases by 2.2% (0.13 mm/d) and AUS monsoon precipitationincreases by 2.1% (0.18 mm/d). The SA monsoon shows a slight in-crease but a much larger increase in the dry season (8%; JJA) and2.9% increase in the entire wet season (November−May).

Fig. 1. Changes in annual mean surface temperature between the de-forestation experiments and the control simulation over the last 50 y of the80-y simulations. (A) Global, (B) Boreal, (C) Temperate, and (D) Tropical.Hatched areas are regions where changes are statistically significant at the95% confidence level. Significance level is estimated using a Student’s t testwith a sample of 50 annual mean differences and SE corrected for temporalserial correlation (51). Line plots show the zonal mean surface temperaturechanges. Shading in line plots represents the one SD estimated from thecontrol simulation. Temperature changes in all panels indicate a larger localeffect of deforestation.

Fig. 2. Same as Fig. 1 but for changes in precipitation (mm/d). (A) Global,(B) Boreal, (C) Temperate, and (D) Tropical. Shading in line plots representsthe ±1 SD estimated from the control simulation. Comparison of B with Dclearly indicates that the remote effect has a larger influence on tropicalprecipitation than the local effect. The location of the precipitation centroidin the ITCZ region in the CTL case and the shifts in the experiments areshown above the panels.

Devaraju et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3259

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

5, 2

020

We also quantify the location of PCENT for each of the definedmonsoon regions (SI Appendix, Table S5). We find that the locationof PCENT in both NH and SH monsoonal regions shifts southward(hence precipitation decreases in NH and increases in SH asshown in Fig. 4) although the shifts are too small in SH mon-soonal domains. Overall, the changes in monsoon regions arelarger in the Boreal and Temperate deforestation cases than inthe Tropical case, indicating the dominance of remote effectsover the local effects. The exception is SA monsoon, whichshows ∼9% decline during JJA contributing to an overall annualmean reduction (Fig. 4D) that is larger than the increase in theTemperate case and comparable to the Boreal case. This de-crease is also consistent with several Amazon deforestationstudies (17, 35, 36) that showed reduced annual mean rainfall.We conclude from Fig. 4 and SI Appendix, Tables S4 and S5,that the decline/increase in monsoon precipitation depends onthe location of deforestation. In our deforestation experiments,the SAs monsoon region is affected the most, with 12% declinein precipitation (Fig. 4G).

Effects over India. The SAs region covers 0°N to 40°N and 42°E to110°E in NH, and it has been suggested that the SAs monsoon isan integral part of the Indian Ocean ITCZ (37). Since a largepart of India receives most of the rainfall during the summermonths JJAS, we restrict our discussion only to these months.Further, our analysis of SAs monsoon here is confined to India.In India, except in the Boreal case, trees are locally converted tograsslands over part of the domain (Tropical and Temperate) orover the full domain (Global; SI Appendix, Fig. S8). The changein mean surface temperature over India in the Global case is0.32 K, −0.16 K in the Boreal case, 0.28 K in the Temperatecase, and 0.33 K in the Tropical case (Table 1), indicating thatthe local deforestation has larger influence on temperaturechange. However, the regional cooling in the seasonally snow-covered northern Himalayas is likely due to the dominance ofthe albedo effect in the Global and Temperate cases and due toremote effects amplified by albedo feedback in the Boreal case(SI Appendix, Fig. S8).In contrast to the local cooling effect due to increased surface

albedo in middle and high latitudes, deforestation leads to localwarming in the tropics (i.e., central and southern India) becauseof decreased (increased) partitioning of the surface radiation tolatent (sensible) heating (15, 38–42). Reduced latent heat fluxesdue to local deforestation can also lead to warming by affectingcloud formation: Drying of the boundary layer as a result ofdeforestation could lead to reduced clouds that could in turnallow more downward solar radiation at the surface and hencewarming (15). The increase in surface albedo over India is small(∼0–5%) in all of the experiments (SI Appendix, Fig. S1). Overall,in the tropics, the changes in evapotranspiration and roughness

length could overwhelm the surface albedo effect: The contri-butions from changes in evapotranspiration and roughness lengthappear to overwhelm the cooling effect due to increase in surfacealbedo, and hence we simulate a significant warming over centraland southern India (SI Appendix, Fig. S8).In the Global, Boreal, and Temperate deforestation cases, we

find a larger decrease in precipitation all over India (SI Appen-dix, Fig. S9) but in the Tropical case, the precipitation reducessignificantly only over central India. This suggests the dominanceof remotely induced effect over local effect when precipitationchanges are considered. The decline in Indian mean precipita-tion is 1.14 ± 0.09 mm/d (18%), 0.35 ± 0.08 mm/d (5.50%), 0.55 ±0.05 mm/d (∼8.60%), and 0.25 ± 0.04 mm/d (4%) in the Global,Boreal, Temperate, and Tropical cases, respectively. The reduc-tion in precipitation in the Tropical case is likely dominated bylocal effect of decreased evapotranspiration (SI Appendix, Fig.S10), but in Global and Temperate cases, it is likely associated withboth the local and remote effects. In the Boreal case, the reductionin India mean precipitation is entirely due to the remote effects.As discussed earlier, SAs monsoon is an integral part of ITCZ.

During JJAS, we locate the ITCZ over the Indian Ocean as thecentroid of precipitation (SI Appendix, Fig. S9) in the CTL case,at 7.80°N. It shifts southward to 5.88°N in the Global case, to7.02°N in the Boreal case, to 6.80°N in the Temperate case, andto 7.59°N in the Tropical case (SI Appendix, Fig. S9). We finda strong correlation between NH cooling-induced ITCZ shift andprecipitation decline over India: The Global case shows the largestshift of 2.17° southward with largest decline in precipitation (18%,Table 1). In the Tropical case, there is little cooling in the NH andhence smaller shift in ITCZ (∼0.21°) and smaller decline in pre-cipitation over India (4%, Table 1).We simulate anomalous easterly winds over India (SI Appen-

dix, Fig. S11) that are associated with a weakened SAs monsooncirculation and reduced rainfall. This circulation change is as-sociated with high-pressure anomalies in Eurasia and SouthAsia (SI Appendix, Fig. S12). High-pressure anomalies are often

Table 2. The global mean and annual mean ITCZ location asrepresented by the precipitation centroid (PCENT) in the controlsimulation and its southward shift in the four deforestationexperiments

Experiments

Global ITCZ location in CTL and its shift (PCENT)toward south relative to CTL

Annual JJA DJF

CTL 0.50°S ± 0.20° 5.60°N ± 0.35° 5.16°S ± 0.17°Global−CTL 1.70° 1.30° 0.80°Boreal–CTL 1.02° 0.80° 0.47°Temperate–CTL 1.11° 0.62° 0.47°Tropical−CTL 0.03° 0.23° 0.03°

The uncertainty is given by ±1 SD of the ITCZ position in 10 5-y segmentsof the last 50 y of the control experiment.

Fig. 3. Changes in (A) annual mean meridional heat transport by the at-mosphere (PW), (B) latitudinal location of the precipitation centroid (PCENT)as a function of season, and (C) atmospheric heat transport (AHT) at theequator as a function of season and (D) annual mean top of atmosphere (TOA)net energy fluxes in the Global (red), Boreal (green), Temperate (blue), andTropical (cyan) deforestation experiments relative to CTL. Shading in A and Dshows the ±1 SD estimated from the control.

3260 | www.pnas.org/cgi/doi/10.1073/pnas.1423439112 Devaraju et al.

Dow

nloa

ded

by g

uest

on

May

5, 2

020

related to descending motion and thus indicate less favorableconditions for precipitation. As a result, monsoon-related con-vection, precipitation, and also evaporation over India are re-duced (SI Appendix, Figs. S9 and S10). Total cloud cover decreasesup to 9.5% (SI Appendix, Fig. S13), and evapotranspiration de-creases up to 1 mm d−1 (SI Appendix, Fig. S10). All of thesechanges indicate a weaker local moisture recycling and weakermoisture convergence over India.There are several studies (ref. 43 and references therein) that

link the Eurasian snow cover and its teleconnection to Indianmonsoon rainfall: Increased snow cover is associated with re-duced rainfall over India. In our Global, Boreal, and Temperatedeforestation simulations, there is an increase in Eurasian snowdepth due to NH cooling (SI Appendix, Fig. S14): The Eurasiansnow depth (0°−180°E, 40°N−78°N) increases, respectively, by20 cm, 15 cm, and 5 cm during spring in these cases. The cor-responding increases in snow cover area are 2 million km2,1.2 million km2, and 0.24 million km2, respectively. Anomaloushigh pressures (SI Appendix, Fig. S12) are associated with theseincreases in snow cover. This in turn increases the strength of theeasterly winds over India (SI Appendix, Fig. S11) and alsoextends the surface cooling southward (Fig. 1).

Discussion and ConclusionsIn this study, we have investigated the effects of large-scale de-forestation on ITCZ and its implications for the NH and SHmonsoon regions. The removal of forests in the NH high lat-itudes results in less solar radiation absorption because of theincrease in albedo which leads to an energy deficit (Fig. 3D) inthe NH high latitudes. This energy deficit in the NH necessitatesan increase in northward heat transport across the equator (Fig.3A). As the sign of the vertically integrated meridional heattransport is determined by the upper branch of the Hadley cell,a shift in NH Hadley cell and ITCZ southward is implied. Thisleads to reduced monsoon precipitation in NH (EA, NA, NAf,and SAs) and increased precipitation over SH monsoon regions(AUS, SAf, and SA).Summer monsoon precipitation over India during JJAS shows

a maximum decline in the Global deforestation case (18%) andleast decline in the Tropical case (4%). Hence, the effect on

precipitation is location dependent, with maximum impact forhigh latitude (Boreal) and least impact for tropical (Tropical)case (Table 2). The other NH monsoon regions are also affectedsignificantly, i.e., NA, NAf, and EA monsoon precipitation havealso declined because of the ITCZ shift. However, precipitationincreases in SH monsoon regions where the shift in ITCZ south-ward favors the precipitation increase there. Our results indicatethat the remote effects (from biogeophysical changes due to de-forestation) on precipitation in the monsoonal regions have largerinfluence than the local effects, although local effects dominatein case of temperature changes.Our results are qualitatively in agreement with other studies (13,

30, 33) that use different forcing, but similar hemispheric asym-metries (e.g., cooling in NH and slight warming in SH) as simu-lated in our deforestation experiments. Paleoclimate data on avariety of timescales also suggest similar atmospheric circulationchanges during periods when NH is colder (44, 45), which causesthe ITCZ to displace southward, which in turn changes the pre-cipitation pattern (46). The ITCZ shifts simulated in this studyare also consistent with a number of modeling studies that usevarious other forcings (30, 33, 47, 48).The United Nations Framework Convention on Climate

Change and its Kyoto protocol came into effect with an objectiveto stabilize atmospheric concentrations of greenhouse gases (49).Avoidance of deforestation, restoration of degraded forests, andafforestation/reforestation are some terrestrial carbon seques-tration strategies suggested by Kyoto Protocol for mitigatingclimate change. It is the biogeochemical effect that is accountedfor in these strategies. The biogeophysical effects, such as albedoand evapotranspiration changes that could offset or enhance thebiochemical effects, are not included. The importance of thesebiogeophysical effects for local and global temperature changewas highlighted by several modeling studies (e.g., refs. 15 and16). The results presented in this paper further indicate that theassessment of remote effects from biogeophysical changes couldbe more important than local effects when impacts on pre-cipitation in the monsoon regions are considered.There are some limitations to our study. First, the results ob-

tained in this study are from a single model; the magnitude ofITCZ shifts may vary from model to model. Therefore, a multi-model analysis will be required to provide uncertainty estimateson the magnitude of ITCZ shifts. Second, our model lacks dy-namic ocean and dynamic sea ice components and biogeochemicalcycles, and hence feedbacks related to deep ocean circulation andbiogeochemical interactions are not modeled here. However, webelieve that our results are fundamental, and the inclusion ofother physical processes and feedbacks in the model would notalter the qualitative result that the high-latitude cooling asso-ciated with deforestation will shift the ITCZ southward andreduce rainfall in NH monsoonal regions.

Model and ExperimentsIn this study, we use the latest version of the National Center for AtmosphericResearch Community Atmosphere Model 5.0 coupled to the land surfacemodel Community Land Model 4 and a slab ocean model (SOM). The SOMconfiguration uses a thermodynamic sea ice model to represent the inter-actions between the ocean and sea ice components of the climate system(50). A horizontal resolution of 1.9° latitude × 2.5° longitude, 27 levels inthe vertical, and a time step of 30 min are used here. All our simulationsuse preindustrial (1850) levels of atmospheric CO2 concentration (285ppm) and N deposition (20.3 TgN/y). We perform a control and four large-scale deforestation simulations: (i ) CTL: the control simulation with veg-etation corresponding to the preindustrial period; (ii ) Global: same as CTLbut all of the tree plant function types (PFTs) across the globe arereplaced by grasses; (iii ) Boreal: same as CTL but all of the tree PFTs in theboreal region (50°N−90°N) are replaced by grasses; (iv) Temperate: sameas CTL but all of the tree PFTs in the midlatitude region (20°S−50°S and20°N−50°N) are replaced by grasses; and (v) Tropical: same as CTL but allof the tree PFTs in the tropical region (20°S−20°N) are replaced by grasses.

Annual JJA DJF

-12

-8

-4

0

ChangeinPrecipitation(%) EAST ASIA (10oN-40oN, 90oE-150oE)

Annual JJA DJF

03691215

ChangeinPrecipitation(%) AUSTRALIA (20oS-0,110oE-150oE)

Annual JJA DJF-12

-9

-6

-3

0

ChangeinPrecipitation(%) NORTH AMERICA (0oN-30oN, 230oE-300oE)

Annual JJA DJF-10

-5

0

5

10

ChangeinPrecipitation(%) SOUTH AMERICA (35oS-0, 290oE-320oE)

Annual JJA DJF-12-9-6-303

ChangeinPrecipitation(%) NORTH AFRICA (10oS-30oN,30oW-40oE)

Annual JJA DJF-20

246

ChangeinPrecipitation(%) SOUTH AFRICA (30oS-0, 10oE-60oE)

Annual JJAS DJF

-15

-10

-5

0

ChangeinPrecipitation(%)

TROPICAL-CTLTEMPERATE-CTLBOREAL - CTLGLOBAL -CTL

SOUTH ASIA (0-40oN,42oE-110oE)

A B

C D

E F

G

Fig. 4. Percentage change in precipitation over NH monsoon regions (A, C,E, and G) and SH monsoon regions (B, D, and F) averaged over the monsoondomains as defined in ref. 34 in our deforestation simulations relative to thecontrol simulation. Changes are shown for the annual, JJA (JJAS for SouthAsia), and DJF means. Error bar represents the SE from the 50 annual andseasonal mean differences.

Devaraju et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3261

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

5, 2

020

We impose the changes in vegetation as a step function change at thestart of the simulations. Our simulations typically reach equilibrium in∼30 y, and all simulations last for 80 y. The drift in the last 50 y in surfacetemperature is only 0.01 K in the control simulation. The first 30 y arediscarded as model spin-up, and the remaining 50 y are analyzed for theresults. Since the carbon emissions from deforestation are not used toincrease the atmospheric CO2 in our deforestation experiments, ouranalysis investigates only effects of biogeophysical changes from de-forestation. In this paper, we mainly focus on the impacts of bio-geophysical effects on the ITCZ shift and the associated changes in

monsoon precipitation. A comparison of changes in precipitation in themonsoon regions in the Boreal case against the Tropical case shouldreveal relative magnitudes of remote and local effects on monsoons.

ACKNOWLEDGMENTS. We thank Prof. J. Srinivasan, Indian Institute ofScience, and two anonymous reviewers for their helpful comments on theoriginal manuscript. We also thank Dr. A. Donohoe for providing clarificationon the calculation of precipitation centroid. Computations were carried out atCentre for Atmospheric and Oceanic Sciences High Performance Computingfacility funded by Fund for Improvement of S & T Infrastructure, Departmentof Science and Technology.

1. Goldewijk KK, Beusen A, van Drecht G, de Vos M (2011) The HYDE 3.1 spatially explicitdatabase of human-induced global land-use change over the past 12,000 years. GlobEcol Biogeogr 20(1):73–86.

2. Ramankutty N, Foley JA (1999) Estimating historical changes in global land cover:Croplands from 1700 to 1992. Glob Biogeochem Cycles 13(4):997–1027.

3. Williams M (1990) Forests. The Earth as Transformed by Human Action: Global andRegional Changes in the Biosphere over the Past 300 Years, eds Turner BL, et al.(Cambridge Univ Press, Cambridge, UK), pp 179–201.

4. Richards JF (1990) Land transformation. The Earth as Transformed by Human Action:Global and Regional Changes in the Biosphere over the Past 300 Years, eds Turner BL,et al. (Cambridge Univ Press, Cambridge, UK), pp 163–178.

5. Mathews HD, Weaver AJ, Meissner KJ, Gillet NP, Eby M (2004) Natural and anthro-pogenic climate change: Incorporating historical land cover change, vegetation dy-namics and the global carbon cycle. Clim Dyn 22(5):461–479.

6. Pongratz J, Reick CH, Raddatz T, Claussen M (2010) Bio-geophysical versus bio-geochemical climate response to historical anthropogenic land cover change. Geo-phys Res Lett 37(8):L08702.

7. Lawrence PJ, et al. (2012) Simulating the biogeochemical and biogeophysical impactsof transient land cover change and wood harvest in the Community Climate SystemModel (CCSM4) from 1850 to 2100. J Clim 25(9):3071–3095.

8. Brovkin V, et al. (2013) Effect of anthropogenic land-use and land-cover changes onclimate and land carbon storage in CMIP5 projections for the twenty-first century.J Clim 26(18):6859–6881.

9. He F, et al. (2014) Simulating global and local surface temperature changes due toHolocene anthropogenic land cover change. Geophys Res Lett 41(2):623–631.

10. Boysen LR, et al. (2014) Global and regional effects of land-use change on climate in21st century simulations with interactive carbon cycle. Earth Syst Dyn 5(2):309–319.

11. Ciais P, et al. (2013) Carbon and other biogeochemical cycles. Climate Change 2013:The Physical Science Basis. Contribution of Working Group I to the Fifth AssessmentReport of the Intergovernmental Panel on Climate Chang, eds Stocker TF, et al.(Cambridge Univ Press, Cambridge, UK), pp 465–570.

12. Davin EL, de Noblet-Ducoudré N (2010) Climatic impact of global-scale deforestation:Radiative versus non-radiative processes. J Clim 23(1):97–112.

13. Swann ALS, Fung IY, Liu Y, Chiang JCH (2014) Remote vegetation feedbacks and themid-Holocene green Sahara. J Clim 27(13):4857–4870.

14. Devaraju N, Cao L, Bala G, Caldeira K, Nemani R (2011) A model investigationof vegetation-atmosphere interactions on a millennial timescale. Biogeosciences8(12):3677–3686.

15. Bala G, et al. (2007) Combined climate and carbon-cycle effects of large-scale de-forestation. Proc Natl Acad Sci USA 104(16):6550–6555.

16. Bathiany S, Claussen M, Brovkin V, Raddatz T, Gayler V (2010) Combined biogeo-physical and biogeochemical effects of large scale forest cover changes in the MPIEarth system model. Biogeosciences 7(5):1383–1399.

17. Sud YC, et al. (1996) Biogeophysical consequences of a tropical deforestation sce-nario: A GCM simulation study. J Clim 9(12):3225–3247.

18. Chase TN, Pielke RA, Kittel TGF, Nemani RR, Running SW (2000) Simulated impacts ofhistorical land cover changes on global climate in northern winter. Clim Dyn 16(2-3):93–105.

19. Dallmeyer A, Claussen M (2011) The influence of land cover change in the Asianmonsoon region on present-day andmid-Holocene climate. Biogeosciences 8(6):1499–1519.

20. Mahmood R, et al. (2013) Land cover changes and their biogeophysical effects onclimate. Int J Climatol 34(4):929–953.

21. Swann ALS, Fung IY, Chiang JCH (2012) Mid-latitude afforestation shifts generalcirculation and tropical precipitation. Proc Natl Acad Sci USA 109(3):712–716.

22. Lawrence D, Vandecar K (2014) Effects of tropical deforestation on climate and ag-riculture. Nat Clim Change 5:27–36.

23. Lee X, et al. (2011) Observed increase in local cooling effect of deforestation at higherlatitudes. Nature 479(7373):384–387.

24. Betts RA (2001) Biogeophysical impacts of land use on present-day climate: Near-surface temperature change and radiative forcing. Atmos Sci Lett 2(1-4):39–51.

25. Braconnot P, et al. (2007) Results of PMIP2 coupled simulations of the mid-Holoceneand Last Glacial Maximum Part 1: Experiments and large-scale features. Clim Past 3(2):261–277.

26. Feldberg MJ, Mix AC (2002) Sea-surface temperature estimates in the Southeast Pa-cific based on planktonic foraminiferal species; modern calibration and Last GlacialMaximum. Mar Micropaleontol 44(1-2):1–29.

27. Baker PA, et al. (2001) Tropical climate changes at millennial and orbital timescales onthe Bolivian Altiplano. Nature 409(6821):698–701.

28. Frierson DMW, Hwang Y-T (2012) Extratropical influence on ITCZ shifts in slab oceansimulations of global warming. J Clim 25(2):720–733.

29. Donohoe A, Marshall J, Ferreira D, Mcgee D (2013) The relationship between ITCZlocation and cross-equatorial atmospheric heat transport: From the seasonal cycle tothe Last Glacial Maximum. J Clim 26(11):3597–3618.

30. Chiang J, Bitz C (2005) The influence of high latitude ice on the position of the marineintertropical convergence zone. Clim Dyn 25(5):477–496.

31. Voigt A, Stevens B, Bader J, Mauritsen T (2014) Compensation of hemispheric albedoasymmetries by shifts of the ITCZ and tropical clouds. J Clim 27(3):1029–1045.

32. Broccoli AJ, Dahl KA, Stouffer RJ (2006) Response of the ITCZ to Northern Hemispherecooling. Geophys Res Lett 33(1):L01702.

33. McGee D, Donohoe A, Marshall J, Ferreira D (2013) Changes in ITCZ location andcross-equatorial heat transport at the Last Glacial Maximum, Heinrich Stadial 1, andthe mid-Holocene. Earth Planet Sci Lett 390:69–79.

34. Wang B, Ding Q (2006) Changes in global monsoon precipitation over the past 56years. Geophys Res Lett 33(6):L06711.

35. Moore N, Arima E, Walker R, Da Silva R (2007) Uncertainty and changing hydro-climatology of the Amazon. Geophys Res Lett 34(14):L14707.

36. Sampaio G, et al. (2007) Regional climate change over eastern Amazonia caused bypasture and soybean cropland expansion. Geophys Res Lett 34(17):L17709.

37. Gadgil S (2003) The Indian monsoon and its variability. Annu Rev Earth Planet Sci31:429–467.

38. Feddema JJ, et al. (2005) The importance of land-cover change in simulating futureclimates. Science 310(5754):1674–1678.

39. DeFries RS, Foley JA, Asner GP (2004) Land-use choices: Balancing human needs andecosystem function. Front Ecol Environ 2(5):249–257.

40. Henderson-Sellers A, et al. (1993) Tropical deforestation: Modeling local- to regional-scale climate change. J Geophys Res 98(D4):7289–7315.

41. Nobre CA, Sellers PJ, Shukla J (1991) Amazonian deforestation and regional climatechange. J Clim 4(10):957–988.

42. Dickinson RE, Henderson-Sellers A (1988) Modeling tropical deforestation—A study ofGCM land surface parameterizations. Q J R Meteorol Soc 114(480):439–462.

43. Saha SK, Pokhrel S, Chaudhari HS (2013) Influence of Eurasian snow on Indian summermonsoon in NCEP CFSv2 freerun. Clim Dyn 41(7-8):1801–1815.

44. Pandey A, Mishra R (2013) Variability of the Southwest Monsoon since the last 25 000Years and their possible causes: Role of Northern Hemisphere versus Tibetan Plateau.J Earth Sci 24(3):428–436.

45. Yanase W, Abe-Ouchi A (2007) The LGM surface climate and atmospheric circulationover East Asia and the North Pacific in the PMIP2 Coupled Model Simulations. ClimPast 3(3):439–451.

46. Swingedouw D, et al. (2009) Impact of freshwater release in the North Atlantic underdifferent climate conditions in an OAGCM. J Clim 22(23):6377–6403.

47. Haywood JM, Jones A, Bellouinand N, Stephenson D (2013) Asymmetric forcing fromstratospheric aerosols impacts Sahelian rainfall. Nat Clim Chang 3:660–665.

48. Schneider T, Bischoff T, Haug GH (2014) Migrations and dynamics of the intertropicalconvergence zone. Nature 513(7516):45–53.

49. United Nations Framework Convention on Climate Change (2010) The CancunAgreements: Outcome of the Work of the Ad Hoc Working Group on Long-TermCooperative Action Under the Convention (United Nations, Bonn).

50. Neale RB, et al. (2012) Description of the NCAR Community Atmosphere Model (CAM5.0) (Natl Cent Atmos Res, Boulder, CO), NCAR/TN-486+STR.

51. Zwiers FW, von Storch H (1995) Taking serial correlation into account in tests ofthe mean. J Clim 8(2):336–351.

3262 | www.pnas.org/cgi/doi/10.1073/pnas.1423439112 Devaraju et al.

Dow

nloa

ded

by g

uest

on

May

5, 2

020