Stable Isotopes as Validation Tools for Global Climate ... - AMS Journals

3 downloads 0 Views 2MB Size Report
Sep 15, 2002 - For example, the wet season deuterium excess differences between .... depleted or enriched water vapor (Gat and Matsui 1991). Differing ...
2664

JOURNAL OF CLIMATE

VOLUME 15

Stable Isotopes as Validation Tools for Global Climate Model Predictions of the Impact of Amazonian Deforestation A. HENDERSON-SELLERS Environment, Australian Nuclear Science and Technology Organisation, Lucas Heights, New South Wales, Australia

K. MCGUFFIE Department of Applied Physics, University of Technology, Sydney, Broadway, New South Wales, Australia

H. ZHANG Bureau of Meteorology Research Centre, Melbourne, Victoria, Australia (Manuscript received 7 May 2001, in final form 12 November 2001) ABSTRACT This paper examines changes in isotopic abundances for 18O and deuterium in precipitation over the Amazon basin based on data in the Global Network on Isotopes in Precipitation (GNIP) database from the International Atomic Energy Agency (IAEA)/WMO. The analysis is conducted in the context of recent changes (and anticipated future changes) to the land surface hydrology as a result of tropical deforestation. Statistically significant temporal changes (1965–90) in selected stable isotopic signatures in the Amazon have been compared with global climate model (GCM) predictions revealing notable differences. For example, the wet season deuterium excess differences between Belem and Manaus, Brazil, are consistent with recent GCM simulations only if there has been a relative increase in evaporation from nonfractionating water sources over this period. No significant change in dry season isotopic characteristics is found despite earlier predictions that land-use change signals would be found. Results of GCM simulations of Amazonian deforestation suggest that the recent stable isotope record is more consistent with the predicted effects of greenhouse warming possibly combined with forest removal than with the predicted effects of deforestation alone.

1. Isotopic analysis as a tool in Amazonian research a. History of isotopes in the Amazon Isotopic data have been used to determine important characteristics of Amazonian hydrologic cycling by means of testing and tuning parameters of simple models for about 30 years. Despite a long history of isotopic records from the Amazon, there has been, to date, no attempt to use isotopic data to evaluate global climate models’ (GCMs’) predictions of the possible impacts of Amazonian deforestation. Salati et al. (1979) used 1-yr isotope data from precipitation and river samples and results from a sector box model to reinforce Molion’s (1975) conclusion that about half the Amazon basin’s water is recycled. On the basis of 13 months of data (October 1972–October Corresponding author address: Professor A. Henderson-Sellers, Australian Nuclear Science and Technology Organisation, Lucas Heights Science and Technology Centre, Private Mail Bag 1, Menai, NSW, 2234 Australia. E-mail: [email protected]

q 2002 American Meteorological Society

1973), they were able to identify that the Amazonian inland gradient of depletion of the heavy isotope of oxygen is surprisingly weak compared to other continental areas. This showed that a proportion of the Amazon’s hydrologic recycling is from nonfractionating sources, that is, transpiration and full canopy reevaporation. This moisture recycling within the Amazon basin leads to a seasonally averaged gradient of only 1.5‰ (1000 km) 21 in d18O going inland on an east to west transect (Fig. 1) as compared with 2.0‰ (1000 km) 21 in Europe and elsewhere (Rozanski et al. 1993). Despite the apparent simplicity in water movement and cycling in the Amazon, some observations seem to point to isotope heterogeneity in originating air masses. For example, in 1981, Leopoldo (1981) reported values of the stable isotopes of oxygen and hydrogen as measured in samples of stemflow and throughflow at the Duke Reserve, near Manaus, Brazil. Although his results were somewhat contradictory, the most likely source of observed differences are believed to be different initial sources. Matsui et al. (1983) also note the role of source regions on isotopic signature. The fate of water intercepted by the canopy is crucially important

15 SEPTEMBER 2002

HENDERSON-SELLERS ET AL.

FIG. 1. Isotopic signature (d18O) across a longitudinal transect of the Amazon from Belem on the coast and at the river mouth through Manaus and Porto Velho to Sao Gabriel, Brazil, in the west. The open diamonds are the 1960s values (similar to those analyzed by Salati et al. 1979) while the solid squares are for the 1980s. Trend lines are shown that illustrate the east–west gradient of 18O over the two periods.

to a complete understanding of the Amazon’s forest hydrology. Any realistic model of deforestation must capture the partitioning of moisture fluxes between evaporation and transpiration, and isotopic data present an independent data source that can be used to gain an insight into these hydrological processes. An important review of Amazonian isotopic and other data was published by Salati and Vose in 1984 (Salati and Vose 1984). This paper was influential because its publication coincided with the first reports of a simulation using a GCM to assess the possible impact of deforestation of the Amazon on the local and global climate. Indeed, Salati and Vose (1984) quote preliminary (1983) results from the work of Henderson-Sellers and Gornitz (1984). The Salati and Vose (1984) paper, primarily a collection of their, and others’, previous work with isotopic analyses, reinforced the conclusion that the Amazon recycles about half its water within its basin. Specifically, these authors calculated that the central Amazon has a water recycling time of about 5.5 days. In this period about half of all rainfall is reevaporated or transpired and of this around 50% refalls as precipitation (Fig. 2). Matsui et al. (1983) examined the isotopic signature on a daily basis and concluded that, on the whole, simple models did not capture the complexity of the system. They suggested that any isotopic signal of deforestation would be most detectable in the winter (dry) season. In 1991, the results of two isotopic models of Amazonian precipitation and its implications for regional hydrology and climate were published. Gat and Matsui (1991) employed a simple box model of the central Amazon basin to demonstrate that some of the water

2665

FIG. 2. Schematic of the Amazonian basin water cycle. Note that of the 100% precipitation only about one-quarter runs off while the rest is recycled either through reevaporation or transpiration.

recycling is from isotopically fractionating sources. Using data from the International Atomic Energy Agency/ World Meteorological Organization (IAEA/WMO) global station network up to 1981 (IAEA/WMO 1999), they interpreted a 13‰ deviation from the meteoric water line as suggesting that 20%–40% of the recycled moisture within the basin is derived from sources such as lakes, the river itself, or standing water, which fractionate isotopes of oxygen and hydrogen. Victoria et al. (1991) combined IAEA/WMO isotopic results from Belem and Manaus, Brazil (1972–86), and a box/sector model of Dall’Olio (1976) and Salati et al. (1979) to show that wet season recycling is primarily by transpiration, while dry season recycling is mostly accomplished by reevaporation of precipitation intercepted on the canopy. Gat (2000) reviews this work and reports that an updated model of the Amazon’s water balance, which uses isotopic input, improves earlier predictions. b. Physical setting of Amazonia Seasonal variation in mean monthly temperature for the Amazon basin is less than 28C. Combined with the simple topography of the inner basin, the trade wind flow provides essentially a single water source. The Amazon basin is essentially a large flat plain with topographic variations of less than 120 m across its 3400km extent (Salati et al. 1979). The basin is open to the Atlantic on the eastern side, but natural topographic boundaries to the north, west, and south prevent other regions from significantly influencing the inner basin (Lettau et al. 1979; Salati et al. 1979). Annual average rainfall estimates for the basin as a whole vary from 2000 to 2400 mm and the river is responsible for pro-

2666

JOURNAL OF CLIMATE

FIG. 3. The Amazon basin showing locations of six IAEA/WMO isotope sampling stations.

ducing 15% of the global freshwater flow (Salati and Vose 1984), and thus plays a significant role in the global water cycle. Crucial to this is the tropical rainforest that covers 75% of the basin (e.g., Lettau et al. 1979; Fearnside 1993), which constitutes approximately half of the earth’s tropical forests (e.g., Dickinson 1987). The shape and topography of the basin as well as the typical direction of winds cause isolated air masses to travel long distances over the forest, with increasing contributions of recycled water from basin sources along the way. Recycled moisture has been estimated to contribute 48% to rainfall between Belem and Manaus (Salati et al. 1979), whereas in the western Amazon as much as 88% of the rainfall is thought to be derived from recycled moisture (Lettau et al. 1979). 2. Isotopic processes and data The stable isotopic composition of the atmospheric moisture and precipitation is intimately related to the origin of the water and its fate in the atmosphere. Within the atmospheric component of the hydrologic cycle, the only significant form of isotopic fractionation of water molecules arises from the physical processes of evaporation and condensation (e.g., Dansgaard 1964). By simultaneously investigating the two primary isotopes

VOLUME 15

in water (HDO and H 218O), it is possible to track the history of evaporative and condensation processes (e.g., Ingraham and Craig 1986). When water phase changes take place slowly, the relative concentrations of heavy water isotopes in each phase vary according to welldefined functions of temperature. The isotopic content of a moist air mass that has been moisture depleted through equilibrium condensation processes changes according to a well-established relationship dD 5 8d18O 1 10 (Craig 1961) known as the meteoric water line (MWL; Dansgaard 1964). Deviations from the MWL are usually expressed in terms of the ‘‘deuterium excess’’ parameter defined as d 5 dD 2 8d18O (Dansgaard 1964). Departures from a value of 10 for the deuterium excess indicate a secondary source of either isotopically depleted or enriched water vapor (Gat and Matsui 1991). Differing circumstances controlling evaporation and condensation and airmass mixing can all lead to complex isotopic signals. Stable isotopic data in rainfall have been collected by the IAEA/WMO in the Amazon basin since 1961 (e.g., Rozanski et al. 1993) as part of a global monitoring program over 550 stations. Of the five stations that have been established within the low-lying regions of the Amazon basin (Fig. 3), the most extensive datasets are available from Belem and Manaus (Table 1). In addition to isotopic information, temperature, humidity, precipitation, and precipitation type were also recorded at these sites. The main focus of analysis here is on data from Belem, situated near the coast at the entrance of the Amazon basin, and Manaus, which is located centrally within the basin (Fig. 3). There have been several isotopic studies of the Amazon (Salati et al. 1979; Victoria et al. 1991; Gat and Matsui 1991; Moreira et al. 1997; Gat 2000) using relatively simple models (Gat and Matsui 1991); however, no one has compared observed isotopic signatures with results of GCM simulations. In this paper, we examine the changes in isotopic abundances over the Amazon basin as characterized by the Global Network on Isotopes in Precipitation (GNIP) dataset. In view of the changes in the hydrological cycle predicted to occur as a result of deforestation in this region, it seems likely that these changes will affect the relative importance of evaporative processes and, hence, the isotopic signature of the precipitation. In the next section, trends in Ama-

TABLE 1. IAEA/WMO Amazon basin isotope collection station location details and availability. The last column is the percentage of the total months of observation for which both deuterium (D) and oxygen 18 ( 18O) observations are available. Station

IAEA id no.

Belem Cayenne Manaus Porto Velho Sao Gabriel Izobamba

8219100 8140500 8233100 8282500 8210600 8404400

Location 1.438S, 48.488W 4.838N, 52.378W 3.128S, 60.028W 8.778S, 63.928W 13.08S, 67.088W 0.378S, 78.558W

Alt (mASL)

Operational period

No. months

% time for D and 18O

24 8 72 105 87 3058

1965–87 1962–75 1965–90 1965–83 1961–90 1968–97

264 156 300 216 348 324

87.5% 38.5% 52.0% 33.3% 26.7% 93.1%

15 SEPTEMBER 2002

HENDERSON-SELLERS ET AL.

2667

FIG. 4. Monthly mean values of d18O and deuterium excess (left-hand scales) at Belem and Manaus, Brazil; and Izobamba, Ecuador, for the 1960s (solid diamonds) and the 1980s (open squares) running from Oct to Oct. Mean monthly precipitation (mm) for the whole period is also shown as a histogram (right-hand scales). The error bars are 61 standard error.

zonia’s stable isotope record are discussed and, in the subsequent section, we introduce some possible means of evaluating GCMs using isotopic data. Section 5 attempts a comparison between these new results and current GCM simulations of the Amazon and, finally, section 6 draws some conclusions. 3. Trends in recent stable isotopic data in the Amazon For each Amazon station listed in Table 1, we have examined the records for temporal trends. We find noticeable changes in isotopic characteristics in the wet season, which extends from about December to May (e.g., Fig. 4). The Amazonian continental gradient of d18O, already the weakest in the world, has been further weakened over the last three decades in the wet months from December to May (Fig. 1). Interestingly, Matsui et al. (1983) and Moreira et al. (1997) suggested that dry season data would provide the strongest signal of deforestation effects on the hydrological cycle; that is, our results regarding wet season changes appear to be at odds with these earlier predictions of the detectability of future isotopic signals. In the 1960s, when the collection of isotopic data began in the Amazon basin, monthly average values of

the deuterium excess at Manaus were significantly greater than at Belem for both the wet and dry seasons (Fig. 5). Large changes are observed in the seasonal deuterium excess inside the basin at Manaus. Although the annual mean deuterium excess at Manaus in the 1980s (11.83 6 0.44‰) is not significantly different to that in the 1960s (12.62 6 0.48‰), results now show a significant difference between the wet and dry season values (Fig. 5). In particular, the deuterium excess has decreased significantly in the wet season and increased in the dry season. The gradient in deuterium excess between Belem and Manaus is also much reduced in the wet season. Plausible explanations of the wet season deuterium excess decrease involve either more nonfractionating (e.g., canopy evaporation) or less fractionating (e.g., lake evaporation) recycling, or both. Thus the observed temporal shift in isotope data (1960s–80s) requires a change in the water recycling behavior in the Amazon. These new stable isotopic results therefore provide a novel means of testing the performance of GCMs used to predict the climate of the Amazon basin. Global climate models, which are employed to predict changes in the hydrology and climate of the basin, should be capable of simulating the observed stable isotope trends. If the GCMs can reproduce the observed temporal iso-

2668

JOURNAL OF CLIMATE

VOLUME 15

1) a further temporal weakening of the continental isotopic depletion (Figs. 1 and 4); and 2) a temporal change in the wet season water recycling in terms of the relative importance of fractionating as compared to nonfractionating sources (Fig. 5). 4. Can GCMs’ simulation of Amazonia be evaluated using isotopic data?

FIG. 5. (a) Dry (Jun–Nov) and (b) wet (Dec–May) half-year deuterium excess comparisons for Manaus and Belem for the earliest decade of IAEA/WMO data (1965–75) and the most recent decade (1980–90). Error bars show 61 standard error.

topic trends, then they may also permit an explanation of their cause. If they cannot, then it is reasonable to reevaluate the merit of other aspects of their predictions pertaining to changes in the Amazon forest. The stable isotopic data analyzed here reveal interesting characteristics that may be indicative of largescale environmental changes and are worthy of investigation as tools by means of which GCMs, and other models of the Amazon, can be evaluated. The Amazonian stable isotopes show the following:

These isotope–GCM ‘‘missing links’’ warrant further detailed study. One possibility is that the temporal isotopic records are illustrative of the impact of Amazonian deforestation. Specifically, the isotopic observations are consistent with the GCM deforestation predictions, which show less overall transpiration (Fig. 6) only if there has been a relative decrease in the evaporation of water from lakes and other fractionating sources over this period. At present, the available GCM studies are unable to demonstrate or deny the validity of this conclusion. Another possibility that deserves some consideration is that the disturbances in the isotopic record over time have not been caused (or not solely caused) by forest removal. Although the regional extent of deforestation in the Amazon is great (e.g., Fearnside 1993), there are other effects that may also be contributing to the observed temporal shifts in the isotopic signatures (e.g., Houghton et al. 2001). The first GCM simulation of the impact of Amazonian deforestation was published by Henderson-Sellers and Gornitz in 1984. Since then, there have been a large number of simulations. McGuffie et al. (1998) reviewed the problems associated with correctly specifying climate model parameters in both control (present day) and deforested simulations. Some of the differences in the outcomes of predictions are due to the imposed differences in surface albedo, surface roughness, density and mix of original and replacing vegetation, soil type and state, and so on (Fig. 7). There is general model

FIG. 6. Changes simulated by CCM1-Oz to the Amazon water budget following deforestation. Differences (mm month 21 ) and percentages are all from the forested case in Fig. 9 for the same periods.

15 SEPTEMBER 2002

HENDERSON-SELLERS ET AL.

FIG. 7. Schematic illustration of processes simulated in a GCM predicting changes following tropical deforestation. Note that there are many feedbacks within even this restricted subset of the simulated environment. At present, there is no consensus on the direction of the sign change in atmospheric moisture convergence.

agreement that both precipitation and evaporation decrease following deforestation. Almost all models predict increased surface temperatures (DT), but there is less consensus on the sign of the change in atmospheric moisture convergence (e.g., McGuffie and HendersonSellers 2001, their Table 5). One means of assessing GCM performance in this area is to utilize the results of Gat and Matsui (1991) regarding the relative amounts of water recycled in the Amazon from fractionating and nonfractionating sources. By comparing deuterium and oxygen isotopic observations with results from their box model of the central Amazon basin, Gat and Matsui (1991) deduced that 10%–20% of the input precipitation is reevaporated from fractionating sources (i.e., sources where water remains after evaporation such as lakes and rivers), 30%–40% from nonfractionating sources (e.g., transpiring plants and complete reevaporation of canopyintercepted water), with about half of the total hydrological budget going to runoff (Fig. 8). Note that these values differ from those shown in Fig. 2 especially in terms of the relative proportions of water recycling and running off. These values could be used to evaluate the hydrologic budget components of GCMs especially if agreement can be achieved between the Gat and Matsui (1991) proportions and those of earlier researchers. Canopy interception is both very difficult to measure and monitor and equally challenging to treat in numerical models. Complete reevaporation of canopy moisture is, of course, nonfractionating. However, the frequency of rainfall events will have a bearing on the prevalence of such complete evaporation. If new rain falls before all the previously intercepted water can be evaporated, then some wash through of isotopically enriched water occurs. This means that some degree of fractionation is

2669

FIG. 8. Schematic of the components of the Amazon water budget modeled by Gat and Matsui (1991): Fp is the precipitation (or rain flux), tFp is the fraction of the precipitation that is transpired, xFp is the fraction from lakes, and yFp is the fraction of the precipitation that runs off.

possible from canopy evaporation but that the degree of fractionation is unknown. McGuffie et al. (1998) report on a series of GCM experiments conducted using the National Center for Atmospheric Research’s (NCAR’s) Community Climate Model (CCM1-Oz; see also McGuffie et al. 1995; Zhang et al. 1996a). Figure 9 illustrates the components of the Amazonian water budget derived from these CCM1-Oz simulations for the annual and two 3-month seasonal means. While the amount of water recycled via transpiration does not vary all that greatly through the year in the GCM simulations, its percentage contribution to the total ranges from 38% in the wet season to 73% in the dry season. Although the land surface scheme [the Biosphere Atmosphere Transfer Scheme (BATS)] used in the CCM1Oz simulations does permit inclusion of lakes, this option was not used in these GCM experiments (McGuffie et al. 1995). It is therefore not possible to examine the Gat and Matsui (1991) conclusion regarding the fraction of recycled moisture from lakes directly in terms of these GCM results. However, the simulations in Fig. 9 do appear to be in contrast with those of Victoria et al. (1991). They claimed on the basis of isotope analysis that transpiration is the major source of recycled water in the wet season, while Fig. 9c shows this GCM simulates transpiration as being much more significant in the Amazon forest’s dry season budget of recycled water. The representation of low absolute seasonality in transpiration, but with a large percentage range in Fig. 9, appears plausible on first inspection. It seems reasonable that the tropical forest will photosynthesize, and hence transpire, roughly homogeneously through the year. Nonetheless, the isotopic data analyzed by Victoria et al. (1991) show a deuterium excess of 14% in the dry season (June–November) requiring significant input of recycled water from one or more fractionating sources such as lakes and rivers. Interestingly, the CCM1-Oz model does not capture this component of the Amazon’s

2670

JOURNAL OF CLIMATE

VOLUME 15

FIG. 9. One GCM (CCM1-Oz) simulation of the water budget of the Amazon for (a) annual means, (b) wet season (DJF), and (c) dry season (JJA) all given in mm month 21 and as percentages of the total precipitation. N.B., There is no requirement for the seasonal percentages to add up to 100% because plants, e.g., can tap soil moisture from previous seasons.

hydrological cycle. These results, while not absolutely contradicting those of Victoria et al. (1991) who found from isotope data that transpiration is dominant in the wet season, seem to be in contention with them. This GCM (CCM1-Oz) has the vast majority (73%) of the recycling being by transpiration in the dry season, while the isotopic data suggest this dominance actually occurs in the wet season. A more thorough examination of the components of the Amazonian water cycle using isotopic data might both reveal inadequacies in current climate model simulations and, hopefully, indicate how simulations by GCMs could be improved to more completely and correctly capture the moisture exchanges. This is worthwhile because it offers the possibility of improving the simulation of important aspects of water recycling in Amazonia. It is also important because it has been shown that tropical deforestation has the potential to excite large-scale Rossby waves in the atmosphere (Zhang et al. 1996b). These waves can propagate from the source of their initiating disturbance into the midand high latitudes of both hemispheres and, hence, prompt impacts far distant from deforestation in the Amazon. 5. Evaluating GCM predictions for the Amazon with stable isotopic data The record of deuterium and 18O depletions in the Amazon has been shown to change over time. Specifically, the isotopic depletion in the central Amazon basin, which had already the weakest continental gradient in the world (Rozanski et al. 1993), has weakened further in the wet season since the 1960s (Fig. 1). This suggests that the Amazonian hydrological recycling is intensifying at least in this season. Furthermore, the statistically significant decrease in deuterium excess at

Manaus (Fig. 4) shows that fractionating water sources (i.e., those that leave water behind such as lakes and rivers) have become relatively less important as compared with nonfractionating sources, which include transpiration and complete reevaporation of intercepted canopy water. The picture derived from the temporal shifts in the Amazonian stable isotopic record suggests the following: 1) more intense water recycling in the wet season; 2) increases in transpiration and/or canopy evaporation in the wet season; and 3) decreases in runoff in the wet season. A variety of hypotheses can be erected to try to explain these observed isotopic trends. These include natural variability and large-scale atmospheric circulation changes including interdecadal changes in the nature of the Walker circulation over the period in question. Although the isotopic results from the wet season are statistically significant, the number of observations is small compared to that typically associated with meteorological data. Large-scale circulation anomalies, for example, associated with El Nin˜o–Southern Oscillation (ENSO) changes could have an influence on these timescales (e.g., Ting and Hoerling 1993). Interannual variations in the timing and position of rainfall associated with the intertropical convergence zone (ITCZ) may also have an influence as has been noted in previous analyses of isotopic records for this region (Salati et al. 1979). The choice of only two ‘‘seasons’’ here has been made to try to minimize the effects of changes in timing of ITCZ rainfall but it is possible that it has not been fully removed. In addition, the effects of large ENSO events may bias the analysis reported here. Further investigations are being pursued to consider ways of removing this possible bias. Table 2 summarizes the observed signals in the stable

15 SEPTEMBER 2002

HENDERSON-SELLERS ET AL.

2671

TABLE 2. Stable isotopic and global climate model characterization of the Amazon. Recent isotopic record: (derived here)

Amazonian deforestation simulations (e.g., McGuffie et al. 1998) Greenhouse impacts in Amazon (e.g., Houghton et al. 2001)

—More water recycling in the wet season —Greater relative importance of transpiration and canopy evaporation in the wet season —Decrease in runoff ratio (equivalent to a decrease in atmospheric moisture convergence) —Decrease in precipitation —Decrease in evaporation —Less water recycling —Possible decrease in atmospheric moisture convergence —Increase in precipitation —Increase in evaporation —Intensification of hydrological recycling —Sign of atmospheric moisture convergence change unknown

isotope record from the Amazon and the changes predicted from GCM experiments. It is not clear that the observed isotopic changes are a simple and predicted result of large-scale deforestation. On the contrary, it is plausible that part of the isotopic signal could owe its origins to the direct and indirect effects of greenhouse gas increases or other concurrent disturbances in climate or human systems. A further use is made here of the box model of Gat and Matsui (1991) to reconsider their conclusions about the best fit of relative components of the central Amazon’s water budget (Fig. 8). Among the many disagreements among GCM representations of the impact of Amazonian deforestation, the sign of the change in moisture convergence is important and outstanding (Fig. 7). The challenges associated with predicting this are illustrated in Zhang et al. (1996a), who show the changes in the vertically integrated water flux across the north, south, east, and west boundaries of the Amazon basin derived from one set of GCM simulations of deforestation (Fig. 10). While the inflow of atmospheric moisture from the east (the main Amazonian water source of the Atlantic Ocean) changes very little as a result of deforestation, the inand outflows in the other four quadrants do change. The impact of deforestation is particularly clear in the west where all four seasons exhibit outflows and all values are much larger than in the control simulations. To the north and south, the signs of the seasonal in/outflows are not affected by deforestation but the magnitudes are. In- and outflows are decreased in the north but increased in the south. The overall effect is a decrease in atmospheric moisture convergence following deforestation for CCM1-Oz. The removal of the dense rainforest canopy in the Amazon and its replacement with grass and scrub has been hypothesized to have a significant effect on the basin hydrology. The sign of the net effect of these important changes can be deduced from the basin model of Gat and Matsui (1991). Considering the temporal changes in isotopic composition over the last 30 years at Manaus, it is possible to use their model and to match new parameter values (cf. Fig. 8). The changed isotopic data can be readily represented in the Gat and Matsui (1991) model if a reduction of around 0.1 is made to

the runoff ratio (Fig. 11). Using this simple model, it is possible to match the changed wet season isotopic signature at Manaus between the periods 1965–75 and 1980–90 by decreasing the runoff fraction, y, by approximately 0.1. This result is consistent with the (larger number of ) deforestation experiments that find a decrease in moisture convergence (see McGuffie and Henderson-Sellers 2001, their Table 5). Thus there is consistency between the observed temporal trends in the stable isotope data and the GCMs that predict decreased atmospheric moisture convergence over the Amazon basin. However, it is important to realize that decreased moisture convergence is not necessarily the result of deforestation alone. The other changes in the Amazonian hydrology suggested by the observed temporal trends in the stable isotopes indicate an intensification of hydrological recycling, which seems very unlikely to be the result of replacing tropical forest by scrub or grassland. Here we review other possible climate changes and their predicted impacts on the Amazon. There have been very few GCM studies so far that have attempted to assess the joint impact of deforestation and greenhouse gas increases in the Amazon. Table 3 lists the imposed changes and predicted outcomes for the three available published studies. The paper by Henderson-Sellers et al. (1995a), was not focused on tropical deforestation but did consider plant physiological responses to increased atmospheric CO 2 levels including stomatal closure. Costa and Foley’s (2000) study is a much more sophisticated evaluation of the independent and combined effects of stomatal closure in response to an enriched CO 2 atmosphere, deforestation, and greenhouse warming. Zhang et al. (2001) consider the latter two effects but not the plant physiological responses. The challenge for the future use of isotopic signatures for GCM evaluation is to know which of these representations most closely fit ‘‘present-day’’ isotopic measurements. Table 4 lists the changes predicted by Zhang et al. (2001) and Costa and Foley (2000) in response to the combined effects of deforestation of the Amazon and global greenhouse gas increases equivalent to doubled atmospheric CO 2 . Figure 12 depicts the sets of predic-

2672

JOURNAL OF CLIMATE

VOLUME 15

FIG. 10. CCM1-Oz predicted seasonal moisture convergence fluxes from the four canonical directions for the Amazon basin before and following deforestation. Predictions are in millions of kg s 21 (see Zhang et al. 1996a,b).

tions made by these two groups annually and for the wet season [December–January–February (DJF)] from the CCM1-Oz simulations. All four GCM simulation combinations in Table 4 predict decreases in atmospheric moisture convergence. These are therefore also consistent with the temporal trend in stable isotopes determined using the Gat and Matsui box model (see Fig. 11). The impact of deforestation alone, however, tends to decrease the intensity of hydrological recycling. This is also clear in Fig. 12 in which comparisons between the ‘‘control’’ cases of both available GCM simulations and the perturbations suggest that, in order to match the newly derived isotope results (i.e., more intense wet season water recycling and increases in either or both transpiration or complete canopy reevaporation), a greenhouse gas increase case is required. In particular, increasing hydrological intensity is delivered to Amazonia in all simulations of greenhouse gas increases. When both greenhouse gas increases and deforestation impacts are considered, it is possible to see aspects of

hydrological recycling that are congruent with the isotopically derived implications regarding changed relativities of nonfractionating and fractionating input. Finally, the CCM1-Oz simulations also allow investigation of the atmospheric moisture convergence changes in all four simulation scenarios (viz., control, deforested 2 3 CO 2 , and combined). Figure 13 represents these flux convergences for the wet season. Interestingly, there is a monotonic increase in the inflow of moisture from the Atlantic (east) in these scenarios and an increase in outflow to the west and south in the combined (CO 2 doubling plus deforestation) case. Despite the overall basin change being consistent with the isotopic temporal trend, nothing further distinguishes the greenhouse gas deforestation impacts here. Thus the wet season (DJF) atmospheric moisture convergence predictions from this GCM are not able to determine clearly the more important factors likely to be affecting the stable isotopic trends detected in the Amazon. Our analysis reveals temporal trends in stable isotopes

15 SEPTEMBER 2002

2673

HENDERSON-SELLERS ET AL.

FIG. 11. Diagram of the results of the sets of equations used by Gat and Matsui (1991). Imposed are recent observations from Manaus showing that a plausible explanation for the displacement in the observed stable isotopes is a decrease in runoff of 10%.

in the Amazon which suggest the following: (i) more intense water recycling in the wet season; (ii) a relative increase in nonfractionating sources (i.e., transpiration or complete reevaporation of water intercepted on the canopy); and (iii) less runoff in the wet season. The first two of these results are not in agreement with GCM deforestation simulations and the third only agrees with about half the published predictions, for example, McGuffie et al. (1998). Previous studies (Matsui et al. 1983; Moreira et al. 1997) have suggested that the isotopic signal of deforestation would be most detectable in the dry season, but no such signal is significant in this analysis. Other plausible reasons for the observed changes include air mass (i.e., circulation) changes and greenhouse gas increases. These may themselves be connected since both large-scale deforestation and global warming are predicted to alter circulation patterns (e.g., Fig. 10). Both are occurring concurrently. Results from the two GCM simulations that consider both greenhouse gas increases and Amazonian deforestation (Table

4 and Figs. 12 and 13) suggest that the combined effects of greenhouse and deforestation may be somewhat coherent with the observed isotope-derived trends. Specifically, Zhang et al. (2001) find that the reduction in precipitation and evaporation in the wet season is much smaller in the combined experiment even though their annual totals are rather similar (Table 4). Costa and Foley (2000), who incorporate plant physiology, have annual reductions in both precipitation and evaporation that are much smaller for the combined impact case than for deforestation alone (Table 4). Figure 12 shows that the Zhang et al. (2001) wet season recycling is equal for transpiration and only reduced by ;10 mm month 21 for reevaporated interception. These figures, while not agreeing with the isotope results, are closer than for deforestation alone. The published results of Costa and Foley (2000) do not give wet season totals but their annual case shows a vast increase in reevaporated interception for the annual case for doubled CO 2 and deforestation. The atmospheric moisture convergence differences in the wet season (Fig. 13) are not greatly different among the four cases except that for greenhouse plus deforestation there is a larger atmospheric moisture flux flowing out of the basin to the west. Overall, we cannot find any GCM simulation that agrees well with the isotope-derived changes in wet season Amazonian water cycling. However, the most plausible hypothesis is that these trends may be due to the combined impact of greenhouse gas increases and largescale forest removal. 6. Use of stable isotopes in evaluating models of the Amazon’s climate and hydrology and the possible impacts of deforestation and global warming Isotopic analysis and modeling studies relating to the Amazon date back to the early 1970s and estimates of the likely impacts of deforestation begin around 1984. Results from isotopic research have influenced climate and hydrological modeling but the two communities have rarely worked closely. The potential importance of the impact of Amazonian deforestation to distant lo-

TABLE 3. Annually averaged regional response to Amazon tropical deforestation for surface temperature T, precipitation P, evaporation E, and moisture convergence from recent GCM studies that have included the effects of greenhouse gas increases. (N/A means the information is not available; a sign change is given only when precise values are unavailable).

Study Henderson-Sellers et al. (1995a) Doubled stomatal resistance and warming no deforestation Costa and Foley (2000) Doubled CO 2 and deforestation with plant physiological response Zhang et al. (2001) Doubled CO 2 and deforestation

Roughness change

DT (8C)

DP (mm)

DE (mm)

Moisture convergence change

No change

No change

1

N/A

2

1

0.135/0.173

0.151/0.05

13.5

2153

2146

2

0.12/0.19

2.0/0.2

12.95

2317

2179

2

Albedo change

2674

JOURNAL OF CLIMATE

VOLUME 15

TABLE 4. Combined impact of greenhouse gas increase (doubled CO 2 ) and deforestation on the predicted Amazon climate. In all cases, the difference is from the present-day situation (simulated). Annual averages are given for the whole Amazon basin. The annual averages are given for both models and the wet season (DJF) means are shown for the CCM1-Oz results.

Zhang et al. (2001) Deforestation Both

DT(8C) Annual/wet season

DP (mm) Annual/wet season

10.3/10.38 12.95/12.7

2402/2110 2317/257

2222/256 2179/238

Negative/negative Negative/negative

2266 2153

2223 2146

Negative Negative

Costa and Foley (2000) (annual case only) Deforestation 11.4 Both 13.5

cations has been suggested by, for example, Zhang et al. (1996b). The possibility of teleconnections as well as local impacts increases the importance of using all available data in the evaluation and, hopefully, validation of global climate models used to predict the consequences of deforestation (cf. Gates et al. 1996). In this paper, we have shown that results derived from isotopic data from the IAEA/WMO network can be compared with outputs from GCMs. We find that water recycling in the central Amazon has changed over the last 30 years, significantly so in the wet season. The statistically significant wet season changes we report may be related to, or even exaggerated by, ENSO events that are known to modify the position of the Walker circulation and, hence, the moisture climatology of the Amazon. While some GCM results may be consistent with this conclusion, selected model sets analyzed here have been found to be failing to correctly simulate the relative components of transpiration and reevaporated canopy interception for the complementary dry season. The failure by the CCM1-Oz simulations to correctly represent the relative seasonal importance of transpiration (nonfractionating) as compared to the fractionating evaporation seems likely to be traced to the land surface parameterization scheme. Although the BATS land surface scheme does have an option to incorporate open water bodies, this was not used in any published Amazon simulations. It is very likely that no other GCM simulation of the Amazon has involved open lakes and river surfaces as a source of recycling water in the basin. This may be a significant feature and may need to be included in land surface parameterization, at least for the Amazon, in the future. Dansgaard (1964) and others examined subtle mechanisms responsible for local changes in deuterium excess in the Mediterranean and in the lee of mountains, but we believe that none of these are likely to pertain to the quasi-homogeneous flat Amazon Basin. These findings warrant further detailed analysis and extension to the very large number of GCMs already claiming to predict the impacts of Amazonian deforestation. Care and caution are demanded in all interpretations because isotopic signals are not straightforward. They are a function of temperature (not important in Tropics); of airmass sources; of precipitation type,

DE (mm) Annual/wet season

Moisture convergence change annual/wet season

that is convection versus frontal; and of other factors (e.g., Dansgaard 1964). At a minimum, it is recommended that all future GCM predictions of the climatic and hydrological impacts of Amazonian deforestation be evaluated against all the available isotopic data. Additionally, it might be valuable for current assessments of the performance of land surface schemes in atmospheric models to include an evaluation of small landlocked water bodies as sources of atmospheric moisture (e.g., Irannejad et al. 2000). A new intercomparison of climate model simulations of Amazonian hydrology might also be possible, perhaps under the auspices of the Atmospheric Model Intercomparison Project (AMIP II; e.g., Phillips et al. 2000). There is potential to explore isotopic modification by Amazonian deforestation by utilizing state-ofthe-art land surface schemes combined with one of the current ‘‘isotope’’ GCMs (e.g., Hoffman et al. 2000). Such an examination would fit into the suite of land surface intercomparisons organized by the Project for Intercomparison of Land-surface Parameterization Schemes (PILPS; e.g., Henderson-Sellers et al. 1995b; Schlosser et al. 2000). Data from the isotopic archive for Amazonia could be offered through the Global Land Atmosphere System Study (GLASS) part of the Global Energy and Water Cycle Experiment (GEWEX) to form the basis of a new stand-alone and coupled (to an isotope-tracking GCM) pair of intercomparison and validation experimental simulations. Finally, the great need for new validation data for GCMs, and the obvious and beneficial synergy derivable, seems to demand that new observational programs, such as the Large Scale Biosphere Atmosphere Experiment in Amazonia (LBA), embrace isotopic studies as a potentially valuable tool for climate and hydrological model validation (e.g., Vo¨ro¨smarty et al. 2001). Simulation validation and model improvement opportunities include the following: (i) regional to basin-scale moisture convergence estimates; (ii) evaluation of model partitioning among transpiration, free evaporation, and canopy evaporation; and (iii) detection and attribution of the impacts of forest change and greenhouse gas increases. We strongly recommend joint investigation of the Amazon by isotope experts and climate modelers.

15 SEPTEMBER 2002

HENDERSON-SELLERS ET AL.

2675

FIG. 12. Annual mean simulations of the hydrologic budget of the Amazon basin derived from the only two studies to date that include both the impacts of deforestation and of greenhouse gas increases (i.e., Costa and Foley 2000; Zhang et al. 2001). The main difference between the GCMs is that the land surface scheme used by Costa and Foley (2000) includes physiological plant responses to increases in atmospheric CO 2 . Annual means are shown for both GCM simulation sets and also the wet season (DJF) for Zhang et al. (2001).

2676

JOURNAL OF CLIMATE

VOLUME 15

FIG. 13. Impact of doubling CO 2 and Amazonian deforestation and their combination on the moisture convergence fluxes in Dec–Feb. The same four directions of flow as in Fig. 10. Units are millions of kg s 21 .

Acknowledgments. We would like to thank Dr. Scott Chambers of ANSTO who performed some of the initial dataset development. REFERENCES Costa, M. H., and J. A. Foley, 2000: Combined effects of deforestation and doubled atmospheric CO 2 concentrations on the climate of Amazonia. J. Climate, 13, 18–34. Craig, H., 1961: Isotopic variations in meteoric water. Science, 133, 1702–1703. Dall’Olio, A., 1976: A composicao isotopica das precipitacoes do Brasil: Modelos isotermicos e a influencia de evapo-transpiracao na Bacia amazonica. M.S. thesis, Dept. of Meteorology, University of Sao Paolo, Piracicaba, Brazil, 180 pp. Dansgaard, W., 1964: Stable isotopes in precipitation. Tellus, 16, 436– 468. Dickinson, R. E., Ed., 1987: Geophysiology of Amazonia: Vegetation and Climate Interactions. J. Wiley, 526 pp. Fearnside, P. M., 1993: Deforestation in Brazilian Amazonia: The effect of population and land tenure. Ambio, 22, 537–545. Gat, J. R., 2000: Atmospheric water balance—The isotopic perspective. Hydrol. Processes, 14, 1357–1369.

——, and E. Matsui, 1991: Atmospheric water balance in the Amazon Basin: An isotopic evapotranspiration model. J. Geophys. Res., 96 (D7), 13 179–13 188. Gates, W. L., and Coauthors, 1996: Climate models—Evaluation. Climate Change 1995: The Science of Climate Change, J. T. Houghton et al., Eds., Cambridge University Press, 229–284. Henderson-Sellers, A., and V. Gornitz, 1984: Possible climatic impacts of land cover transformations, with particular emphasis on tropical deforestation. Climatic Change, 6, 231–258. ——, K. McGuffie, and C. Gross, 1995a: Sensitivity of global climate model simulations to increased stomatal resistance and CO 2 increases. J. Climate, 8, 1738–1756. ——, A. J. Pitman, P. K. Love, P. Irannejad, and T. H. Chen, 1995b: The Project for Intercomparison of Land Surface Parameterization Schemes (PILPS): Phases 2 and 3. Bull. Amer. Meteor. Soc., 76, 489–503. Hoffman, G., J. Jouzel, and V. Masson, 2000: Stable isotopes in atmospheric general circulation models. Hydrol. Processes, 14, 1385–1406. Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson, Eds., 2001: Climate Change 2001: The Scientific Basis. Cambridge University Press, 881 pp. IAEA/WMO, cited 1999: Global network for isotopes in precipitation.

15 SEPTEMBER 2002

HENDERSON-SELLERS ET AL.

The GNIP Database, Release 3, October 1999. [Available online at http://www.iaea.org/programs/ri/gnip/gnipmain.htm.] Ingraham, N. L., and R. C. Craig, 1986: Hydrogen isotope study of large-scale meteoric water transport in northern California and Nevada. J. Hydrol., 85, 183–197. Irannejad, P., A. Henderson-Sellers, T. Phillips, and K. McGuffie, 2000: Analysis of AMIP II models’ simulations of land surface climates. GEWEX News, 10, 5–8. Leopoldo, R. R., 1981: Aspetos hidrologicos at florista amazonica denga na regiao de Manaus. Ph.D. thesis, University National Estado San Paolo, Botucato, SP, Brazil, 147 pp. Lettau, H., K. Lettau, and L. C. B. Molion, 1979: Amazonia’s hydrologic cycle and the role of atmospheric recycling in assessing deforestation effects. Mon. Wea. Rev., 107, 227–238. Matsui, E., E. Salati, M. N. G. Ribeiro, C. M. Reis, A. C. S. N. F. Tancredi, and J. R. Gat, 1983: Precipitation in the Central Amazon Basin: The isotopic composition of rain and atmospheric moisture at Belem and Manaus. Acta Amazon., 13, 307–369. McGuffie, K., and A. Henderson-Sellers, 2001: Forty years of numerical climate modelling. Int. J. Climatol., 21, 1067–1109. ——, ——, H. Zhang, T. Durbidge, and A. J. Pitman, 1995: Global climate sensitivity to tropical deforestation. Global Planet. Change, 10, 97–128. ——, ——, and ——, 1998: Modelling climatic impacts of future rainforest destruction, Human Activities and the Tropical Rainforest, B. K. Maloney, Ed., Kluwer, 169–193. Molion, L. C. B., 1975: A climatonomic study of the energy and moisture fluxes of the Amazonas basin with considerations of deforestation effects. Ph.D. thesis, University of Wisconsin— Madison, 133 pp. Moreira, M. Z., L. D. L. Sternberg, L. A. Martinelli, R. L. Victoria, E. M. Barbosa, L. C. Bonates, and D. C. Nepstad, 1997: Contribution of transpiration to forest ambient vapour based on isotopic measurements. Global Change Biol., 3, 439–450. Phillips, T. J., A. Henderson-Sellers, P. Irannejad, K. McGuffie, and

2677

H. Zhang, 2000: On validation and diagnosis of landsurface climate simulations. Climate Change Newsl., 12, 3–5. [Available online at http://www.brs.gov.au/publications/ ccn/ccn12v1/research.html.] Rozanski, K., L. Araguas-Araguas, and R. Gonfiantini, 1993: Isotopic patterns in modern global precipitation. Climate Change in Continental Isotopic Records, Geophys. Monogr., No. 78, Amer. Geophys. Union, 1–36. Salati, E., and P. B. Vose, 1984: Amazon Basin: A system in equilibrium. Science, 225, 129–137. ——, A. D. Olio, E. Matsui, and J. R. Gat, 1979: Recycling of water in the Amazon Basin: An isotopic study. Water Resour. Res., 15, 1250–1258. Schlosser, C. A., and Coauthors, 2000: Simulations of a boreal grassland hydrology at Valdai, Russia: PILPS Phase 2(d). Mon. Wea. Rev., 128, 301–321. Ting, M., and M. P. Hoerling, 1993: Dynamics of stationary wave anomalous during the 1986/87 El Nin˜o. Climate Dyn., 9, 147– 164. Victoria, R. L., L. A. Martinelli, J. Mortatti, and J. Richey, 1991: Mechanisms of water recycling in the Amazon Basin: Isotopic insights. Ambio, 20, 384–387. Vo¨ro¨smarty, C. J., C. J. Willmott, B. Fekete, S. Webber, J. A. Marengo, C. Fernandez-Jauregui, and M. Concepcio´n Donoso, 2001: Large-scale hydrometric integration studies over the LBA domain. BAHC News, 8, 9–10. Zhang, H., A. Henderson-Sellers, and K. McGuffie, 1996a: Impacts of tropical deforestation. Part I: Process analysis of local climatic change. J. Climate, 9, 1497–1517. ——, K. McGuffie, and A. Henderson-Sellers, 1996b: Impacts of tropical deforestation. Part II: The role of large-scale dynamics. J. Climate, 9, 2498–2521. ——, A. Henderson-Sellers, and K. McGuffie, 2001: The compounding effects of tropical deforestation and greenhouse warming on climate. Climatic Change, 49, 309–338.