Effects of large-scale deforestation on precipitation in the monsoon regions: Remote versus local effects

Significance

Biogeophysical effects such as albedo and evapotranspiration changes due to deforestation were shown by several studies in the 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 of the world. In contrast to the dominant role of local effects on temperature changes, we find that the remote effects have a larger influence than local effects on shifting the location of the Intertropical Convergence Zone and hence precipitation in all the monsoon regions. This result has important implications for assessing the net benefits of climate change mitigation strategies such as afforestation/reforestation and for understanding changes in monsoon rainfall in past climates.

Abstract

In this paper, using idealized climate model simulations, we investigate the biogeophysical effects of large-scale deforestation on monsoon regions. We find that the remote forcing from large-scale deforestation in the northern middle and high latitudes shifts the Intertropical Convergence Zone southward. This results in a significant decrease in precipitation in the Northern Hemisphere monsoon regions (East Asia, North America, North Africa, and South Asia) and moderate precipitation increases in the Southern Hemisphere monsoon regions (South Africa, South America, and Australia). The magnitude of the monsoonal precipitation changes depends on the location of deforestation, with remote effects showing a larger influence than local effects. The South Asian Monsoon region is affected the most, with 18% decline in precipitation over India. Our results indicate that any comprehensive assessment of afforestation/reforestation as climate change mitigation strategies should carefully evaluate the remote effects on monsoonal precipitation alongside the large local impacts on temperatures.

Historical land cover change has been one of the major drivers of climate change. By the 1750s, ∼6–7% of the global land surface area had been deforested for agriculture. Today, croplands and pasture lands make up approximately one third of the global land surface (1–4). In terms of area, croplands and pasture lands increased globally from 620 million ha in 1700 to 4,960 million ha by 2000 (1). This large-scale conversion of forests to croplands or grasslands can impact climate through biogeochemical (changes in atmospheric composition) and biogeophysical (changes in physical land 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 investigated by numerous studies (5–10). These studies find that the biogeochemical process primarily causes global effects while biogeophysical processes cause strong local effects. The combined biogeochemical and biogeophysical effects from land cover change in the Holocene before 1850 were modeled as a global mean warming of 0.73 K (9).

During the historical period (1750 to present day), deforestation-associated CO₂ emissions have contributed ∼180 ± 80 PgC to the cumulative anthropogenic CO₂ emissions (11) and a warming of ∼0.16–0.30 K (biogeochemical effect) to anthropogenic climate change (5, 6). This warming is probably partly offset by the biogeophysical effect of albedo increase, which may have caused a global mean cooling by ∼0.03–0.27 K (5, 7, 8). However, other major biogeophysical processes, such as reduction in evapotranspiration and roughness length due to deforestation, could result in warming (12).

Several studies have investigated the link between land cover change and local climate change (13–16). For example, deforestation (16) in the tropics (18.75°S−15°N) reduces precipitation over Amazon by 138 mm/y (9.2%) and increases the temperature by 1.6 K. Another study (17) simulates a 266 mm/y reduction in precipitation over tropics due to tropical deforestation. The biogeophysical effects can also have remote effects via changes in atmospheric circulation (13, 18–20). For instance, recent studies (13, 21) find a shift in Intertropical Convergence Zone (ITCZ) due to afforestation in entire midlatitudes or over Eurasia. These studies suggest that the ITCZ shifts can have consequences for precipitation in the monsoon 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 remote effects can affect the monsoon regions. To our knowledge, no study 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 has a larger influence on precipitation in monsoon regions than the local effect, although the local effect has a larger impact on surface temperature changes as shown in several previous studies (13–15, 21). The remote effect can be quantified through a relationship between the ITCZ location and the atmospheric heat transport at the equator. Our investigation has direct relevance to changes in precipitation in monsoon regions in the past [for instance, during the Last Glacial Maximum (LGM) and at the Cretaceous–Tertiary boundary when large areas of forests were completely removed], to make improved assessment of risks to agriculture from changes to rainfall in the tropics (22) and to integrated assessments of afforestation/reforestation as climate change mitigation strategies.

Results

Global-Scale Temperature and Precipitation Responses.

We investigate the effects of Global, Boreal, Temperate, and Tropical deforestation on global climate (i.e., surface temperature and precipitation) relative to the CTL (control) case. Global mean surface air temperature decreases by 1.50 K, 0.90 K, 0.47 K, and 0.04 K in the Global, Boreal, Temperate, and Tropical cases, respectively (Table 1). Correspondingly, the global mean precipitation decreases 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 mean cooling of 1.5 K in the Global case is in close agreement with previous modeling studies that find 1.7 K (14) or 1.6 K (15) cooling due to global deforestation. Compared with the preindustrial climate, all simulations except the Tropical case show significant cooling in the Northern Hemisphere (NH; Fig.1 A−C). In the Global case, cooling averaged over NH is about 3 K. In the Boreal case, the NH mean cooling is ∼2.20 K, and it is ∼1.50 K in the Temperate case (Fig. 1 B and C). The simulated NH cooling is consistent with an observational study (23) that finds strong evidence for cooling over northern latitudes (>45°N) and weak evidence for warming below 35°N as a result of the biogeophysical responses to deforestation.

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 ± 0.03
Precipitation, 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/y	574.11 ± 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 ± 0.07
Precipitation, 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 mean differences. Values within the parentheses show the percentage changes relative to control.

^*Albedo changes given in parentheses are absolute changes in percentage.

Fig. 1.

Changes in annual mean surface temperature between the deforestation experiments and the control simulation over the last 50 y of the 80-y simulations. (A) Global, (B) Boreal, (C) Temperate, and (D) Tropical. Hatched areas are regions where changes are statistically significant at the 95% confidence level. Significance level is estimated using a Student’s t test with a sample of 50 annual mean differences and SE corrected for temporal serial correlation (51). Line plots show the zonal mean surface temperature changes. Shading in line plots represents the one SD estimated from the control simulation. Temperature changes in all panels indicate a larger local effect of deforestation.

The NH cooling in the Global, Boreal, and Temperate cases is due to the increased land surface albedo and reduced absorption of solar radiation at the surface (15). The albedo changes in the northern middle and high latitudes are more than 25–30% (SI Appendix, Fig. S1 A–C). The large albedo changes are because of (i) the conversion of forests into grasslands, which have relatively larger albedo, and (ii) the seasonal presence of snow in these areas—deforestation exposes the surface snow cover, which has a larger albedo. This albedo-related cooling can be also reinforced by snow and sea ice−albedo feedbacks (24). The importance of the seasonal snow cover for albedo effect in the Global, Boreal, and Temperate cases can be inferred by noting that the changes in albedo in the tropical regions in the Global case do not exceed 0–5% (SI Appendix, Fig. S1A). In the Tropical deforestation simulation, the global mean cooling is of only 0.04 K (Table 1) because the influence of albedo and evapotranspiration almost counterbalance each other. However, we find strong local 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 evapotranspiration may have compensating effects (15). The decrease in evapotranspiration from deforestation leads to a decrease in clouds over tropical land that allows more downward solar radiation at the surface [SI Appendix, Fig. S2 and Table S1 (15)]. Thus, while deforestation brightens the surface in the tropics, it also tends to decrease cloudiness and darkens the planet. These effects nearly cancel each other, and hence the magnitude of cooling and the changes in planetary albedo and net flux at the top of the atmosphere (TOA) are much smaller in low latitudes compared 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 warming of 0.2 K (20°S−20°N, Fig. 1D) in the Tropical deforestation case, and strong localized cooling in the midlatitudes (−0.8 K in 20°N−50°N) and high latitudes (−4 K in 50°N−90°N) in Temperate and Boreal deforestation cases, respectively. The strong cooling in NH (in the Global and Boreal cases) is similar to the climate that prevailed during LGM when temperatures were much cooler than today [by ∼3.6 K to 5.7 K (25–27)].The annual mean precipitation declines 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 in tropical regions in association with a southward shift of the ITCZ (Fig. 2 A−C).

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 represents the ±1 SD estimated from the control simulation. Comparison of B with D clearly indicates that the remote effect has a larger influence on tropical precipitation than the local effect. The location of the precipitation centroid in the ITCZ region in the CTL case and the shifts in the experiments are shown above the panels.

ITCZ Shift.

Following refs. 28 and 29, we use the precipitation centroid (P_CENT) as a metric for locating the ITCZ precipitation maximum. The precipitation centroid is defined as the median of the zonal average precipitation from 20°S to 20°N. The zonal mean precipitation (from the average of the last 50 y) interpolated to a 0.01° grid in the tropics (20°S−20°N) allows us to locate the precipitation centroid to a higher precision than the grid resolution. We find a southward shift of annual mean ITCZ (P_CENT) in all of the deforestation simulations: ∼1.70° in Global, ∼1.02° in Boreal, ∼1.11° in Temperate, and ∼0.03° in Tropical case from its original position of 0.50°S in CTL case (Table 2). During boreal summer [June−August (JJA)] the location of ITCZ is inside the NH (5.60°N) in the CTL and it shifts southward by about 1.30° in Global, 0.80° in Boreal, 0.62° in Temperate, and 0.23° in Tropical deforestation simulations. In the austral summer [December−February (DJF)], the ITCZ is located inside the SH (5.16°S) and the southward shifts have slightly reduced magnitude (Table 2).

Table 2.

The global mean and annual mean ITCZ location as represented by the precipitation centroid (P_CENT) in the control simulation and its southward shift in the four deforestation experiments

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 segments of the last 50 y of the control experiment.

The shifts in ITCZ in our deforestation simulations are associated with changes in meridional heat transports (Fig. 3A), which is in agreement with earlier modeling studies that investigated the impact of various climate forcings such as imposed ice cover in the high latitudes, artificially enhanced albedo, etc. (30–32). In the case of Global and Boreal deforestation, for example, the NH high latitudes absorb less solar radiation because of the increase in albedo, which leads to a larger deficit in energy in the NH. For the Global case, the TOA NH energy deficit is −0.72 W/m², −0.54 W/m² in the Boreal case, and −0.29 W/m² in the Temperate case (SI Appendix, Table S2). This NH energy deficit necessitates an increase in heat transport (Fig. 3A) into the NH from the SH: Since the sign of the vertically averaged meridional heat transport in the tropical region is dominated by the upper branch of the Hadley cell, a southward shift of the NH Hadley cell (and hence the ITCZ) would facilitate more heat transport into the NH. The southward shift of ITCZ in association with NH climate cooling (∼2.45 K, Fig. 1 A and B) in our Global and Boreal deforestation experiments agree qualitatively with simulated ITCZ shifts [∼1°S (33)] and the NH cooling during LGM [∼3 K (30)]. The increased heat transport in the midlatitudes (Fig. 3A) is likely associated with enhanced baroclinic eddy activity because of increased meridional temperature gradient (Fig. 1) and the corresponding increase in vertical shear in NH midlatitude zonal mean westerly winds (SI Appendix, Figs. S3 and S4).

Fig. 3.

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

We also quantify the relationship between the ITCZ location (P_CENT) and atmospheric heat transport at the equator (AHT_eq) as shown in SI Appendix, Fig. S5. We adopt the procedure discussed in ref. 29 for the estimation of annual mean AHT_eq and its seasonal cycle (Fig. 3 A and C). In the CTL case, there is a linear relationship between P_CENT and AHT_eq in the seasonal time scale (SI Appendix, Fig. S5). The location of P_CENT in the control simulation varies from 7.50°S in February to 7.27°N in August. We estimate that the southward movement of P_CENT is ∼2.65° ±0.5° per petawatts (PW) of AHT_eq on the seasonal time scale. This linear relationship is in close agreement with a recent study (29) that finds a shift of −2.4° per PW for CMIP3 models and −2.7° per PW in observations. The link between changes in AHT_eq and ITCZ shifts in our deforestation experiments are consistent with the rate determined for seasonal scale (SI Appendix, Fig. S5). We find that the southward shift of ITCZ and the corresponding increase in AHT_eq in the deforestation simulations occur almost throughout the year (Fig. 3 B and C). The changes in net TOA radiative flux are larger in NH high latitudes in the Global, Boreal, and Temperate cases compared with the tropical latitudes (Fig. 3D and SI Appendix, Table S1): While large surface albedo changes lead to large changes in TOA fluxes in the high latitudes, reductions in clouds in the tropical regions because of reduced evapotranspiration nearly offset the surface albedo-induced TOA flux changes (15) (SI Appendix, Fig. S2 and Table S1).

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

Effects on Monsoon Regions.

To understand the changes in NH and SH monsoon precipitation, we chose the monsoonal regions based on the criteria developed in ref. 34 that relies on the annual range of precipitation. The regional monsoons as shown by boxes 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 American (SA), and (vii) Australian (AUS) monsoons. As stated before, the changes in circulation in our deforestation simulations (except Tropical case) decrease the mean monsoon precipitation in NH monsoon regions and slightly increase the mean precipitation in SH monsoon regions in association with the southward shift in ITCZ (Fig. 4). For example, during JJA, the NH monsoonal regions receive less precipitation in the Global case: EA monsoon precipitation decreases by 10.2% (0.7 mm/d), monsoon precipitation over SAs decreases by 11.8% [0.64 mm/d, during June−September (JJAS)] and NA monsoon precipitation decreases by 4% (0.15 mm/d) (Fig. 4 and SI Appendix, Table S4). In contrast, during DJF (summer in SH), the SH monsoonal regions receive more precipitation in the Global case: SAf monsoon precipitation increases by 2.2% (0.13 mm/d) and AUS monsoon precipitation increases by 2.1% (0.18 mm/d). The SA monsoon shows a slight increase but a much larger increase in the dry season (8%; JJA) and 2.9% increase in the entire wet season (November−May).

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 monsoon domains as defined in ref. 34 in our deforestation simulations relative to the control simulation. Changes are shown for the annual, JJA (JJAS for South Asia), and DJF means. Error bar represents the SE from the 50 annual and seasonal mean differences.

We also quantify the location of P_CENT for each of the defined monsoon regions (SI Appendix, Table S5). We find that the location of P_CENT in both NH and SH monsoonal regions shifts southward (hence precipitation decreases in NH and increases in SH as shown in Fig. 4) although the shifts are too small in SH monsoonal domains. Overall, the changes in monsoon regions are larger in the Boreal and Temperate deforestation cases than in the Tropical case, indicating the dominance of remote effects over the local effects. The exception is SA monsoon, which shows ∼9% decline during JJA contributing to an overall annual mean reduction (Fig. 4D) that is larger than the increase in the Temperate case and comparable to the Boreal case. This decrease is also consistent with several Amazon deforestation studies (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 on the location of deforestation. In our deforestation experiments, the SAs monsoon region is affected the most, with 12% decline in precipitation (Fig. 4G).

Effects over India.

The SAs region covers 0°N to 40°N and 42°E to 110°E in NH, and it has been suggested that the SAs monsoon is an integral part of the Indian Ocean ITCZ (37). Since a large part of India receives most of the rainfall during the summer months 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 to grasslands over part of the domain (Tropical and Temperate) or over the full domain (Global; SI Appendix, Fig. S8). The change in mean surface temperature over India in the Global case is 0.32 K, −0.16 K in the Boreal case, 0.28 K in the Temperate case, and 0.33 K in the Tropical case (Table 1), indicating that the local deforestation has larger influence on temperature change. However, the regional cooling in the seasonally snow-covered northern Himalayas is likely due to the dominance of the albedo effect in the Global and Temperate cases and due to remote 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 local warming in the tropics (i.e., central and southern India) because of decreased (increased) partitioning of the surface radiation to latent (sensible) heating (15, 38–42). Reduced latent heat fluxes due to local deforestation can also lead to warming by affecting cloud formation: Drying of the boundary layer as a result of deforestation could lead to reduced clouds that could in turn allow more downward solar radiation at the surface and hence warming (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 contributions from changes in evapotranspiration and roughness length appear to overwhelm the cooling effect due to increase in surface albedo, and hence we simulate a significant warming over central and 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 Appendix, Fig. S9) but in the Tropical case, the precipitation reduces significantly only over central India. This suggests the dominance of remotely induced effect over local effect when precipitation changes are considered. The decline in Indian mean precipitation 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 reduction in precipitation in the Tropical case is likely dominated by local effect of decreased evapotranspiration (SI Appendix, Fig. S10), but in Global and Temperate cases, it is likely associated with both the local and remote effects. In the Boreal case, the reduction in 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 the centroid 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, to 7.02°N in the Boreal case, to 6.80°N in the Temperate case, and to 7.59°N in the Tropical case (SI Appendix, Fig. S9). We find a strong correlation between NH cooling-induced ITCZ shift and precipitation decline over India: The Global case shows the largest shift of 2.17° southward with largest decline in precipitation (18%, Table 1). In the Tropical case, there is little cooling in the NH and hence smaller shift in ITCZ (∼0.21°) and smaller decline in precipitation over India (4%, Table 1).

We simulate anomalous easterly winds over India (SI Appendix, Fig. S11) that are associated with a weakened SAs monsoon circulation and reduced rainfall. This circulation change is associated with high-pressure anomalies in Eurasia and South Asia (SI Appendix, Fig. S12). High-pressure anomalies are often related to descending motion and thus indicate less favorable conditions for precipitation. As a result, monsoon-related convection, precipitation, and also evaporation over India are reduced (SI Appendix, Figs. S9 and S10). Total cloud cover decreases up to 9.5% (SI Appendix, Fig. S13), and evapotranspiration decreases up to 1 mm d⁻¹ (SI Appendix, Fig. S10). All of these changes indicate a weaker local moisture recycling and weaker moisture convergence over India.

There are several studies (ref. 43 and references therein) that link the Eurasian snow cover and its teleconnection to Indian monsoon rainfall: Increased snow cover is associated with reduced rainfall over India. In our Global, Boreal, and Temperate deforestation simulations, there is an increase in Eurasian snow depth due to NH cooling (SI Appendix, Fig. S14): The Eurasian snow depth (0°−180°E, 40°N−78°N) increases, respectively, by 20 cm, 15 cm, and 5 cm during spring in these cases. The corresponding increases in snow cover area are 2 million km², 1.2 million km², and 0.24 million km², respectively. Anomalous high pressures (SI Appendix, Fig. S12) are associated with these increases in snow cover. This in turn increases the strength of the easterly winds over India (SI Appendix, Fig. S11) and also extends the surface cooling southward (Fig. 1).

Discussion and Conclusions

In this study, we have investigated the effects of large-scale deforestation on ITCZ and its implications for the NH and SH monsoon regions. The removal of forests in the NH high latitudes results in less solar radiation absorption because of the increase in albedo which leads to an energy deficit (Fig. 3D) in the NH high latitudes. This energy deficit in the NH necessitates an increase in northward heat transport across the equator (Fig. 3A). As the sign of the vertically integrated meridional heat transport is determined by the upper branch of the Hadley cell, a shift in NH Hadley cell and ITCZ southward is implied. This leads 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%) and least decline in the Tropical case (4%). Hence, the effect on precipitation is location dependent, with maximum impact for high latitude (Boreal) and least impact for tropical (Tropical) case (Table 2). The other NH monsoon regions are also affected significantly, i.e., NA, NAf, and EA monsoon precipitation have also declined because of the ITCZ shift. However, precipitation increases in SH monsoon regions where the shift in ITCZ southward favors the precipitation increase there. Our results indicate that the remote effects (from biogeophysical changes due to deforestation) on precipitation in the monsoonal regions have larger influence than the local effects, although local effects dominate in case of temperature changes.

Our results are qualitatively in agreement with other studies (13, 30, 33) that use different forcing, but similar hemispheric asymmetries (e.g., cooling in NH and slight warming in SH) as simulated in our deforestation experiments. Paleoclimate data on a variety of timescales also suggest similar atmospheric circulation changes during periods when NH is colder (44, 45), which causes the ITCZ to displace southward, which in turn changes the precipitation pattern (46). The ITCZ shifts simulated in this study are also consistent with a number of modeling studies that use various other forcings (30, 33, 47, 48).

The United Nations Framework Convention on Climate Change and its Kyoto protocol came into effect with an objective to stabilize atmospheric concentrations of greenhouse gases (49). Avoidance of deforestation, restoration of degraded forests, and afforestation/reforestation are some terrestrial carbon sequestration strategies suggested by Kyoto Protocol for mitigating climate change. It is the biogeochemical effect that is accounted for in these strategies. The biogeophysical effects, such as albedo and evapotranspiration changes that could offset or enhance the biochemical effects, are not included. The importance of these biogeophysical effects for local and global temperature change was highlighted by several modeling studies (e.g., refs. 15 and 16). The results presented in this paper further indicate that the assessment of remote effects from biogeophysical changes could be more important than local effects when impacts on precipitation in the monsoon regions are considered.

There are some limitations to our study. First, the results obtained in this study are from a single model; the magnitude of ITCZ shifts may vary from model to model. Therefore, a multimodel analysis will be required to provide uncertainty estimates on the magnitude of ITCZ shifts. Second, our model lacks dynamic ocean and dynamic sea ice components and biogeochemical cycles, and hence feedbacks related to deep ocean circulation and biogeochemical interactions are not modeled here. However, we believe that our results are fundamental, and the inclusion of other physical processes and feedbacks in the model would not alter the qualitative result that the high-latitude cooling associated with deforestation will shift the ITCZ southward and reduce rainfall in NH monsoonal regions.

Model and Experiments

In this study, we use the latest version of the National Center for Atmospheric Research Community Atmosphere Model 5.0 coupled to the land surface model Community Land Model 4 and a slab ocean model (SOM). The SOM configuration uses a thermodynamic sea ice model to represent the interactions between the ocean and sea ice components of the climate system (50). A horizontal resolution of 1.9° latitude × 2.5° longitude, 27 levels in the vertical, and a time step of 30 min are used here. All our simulations use preindustrial (1850) levels of atmospheric CO₂ concentration (285 ppm) and N deposition (20.3 TgN/y). We perform a control and four large-scale deforestation simulations: (i) CTL: the control simulation with vegetation corresponding to the preindustrial period; (ii) Global: same as CTL but all of the tree plant function types (PFTs) across the globe are replaced by grasses; (iii) Boreal: same as CTL but all of the tree PFTs in the boreal region (50°N−90°N) are replaced by grasses; (iv) Temperate: same as CTL but all of the tree PFTs in the midlatitude region (20°S−50°S and 20°N−50°N) are replaced by grasses; and (v) Tropical: same as CTL but all of the tree PFTs in the tropical region (20°S−20°N) are replaced by grasses. We impose the changes in vegetation as a step function change at the start 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 surface temperature is only 0.01 K in the control simulation. The first 30 y are discarded as model spin-up, and the remaining 50 y are analyzed for the results. Since the carbon emissions from deforestation are not used to increase the atmospheric CO₂ in our deforestation experiments, our analysis investigates only effects of biogeophysical changes from deforestation. In this paper, we mainly focus on the impacts of biogeophysical effects on the ITCZ shift and the associated changes in monsoon precipitation. A comparison of changes in precipitation in the monsoon regions in the Boreal case against the Tropical case should reveal relative magnitudes of remote and local effects on monsoons.