1. Introduction

Stable isotopes (¹⁸O, D) are among the most powerful tools for investigating the hydrological cycle. The International Atomic Energy Agency/Global Network of Isotopes in Precipitation (IAEA/GNIP) database (IAEA 2001) has been widely used for this purpose. In South America, the tropical rainforests in the Amazon basin transpirate the enriched water, which results in the smallest longitudinal gradient of δ¹⁸O in the meteoric water among all the continents (see Henderson-Sellers et al. 2002, and references therein). Here, transpiration is the nonfractioning process of the stable isotope of water lost through the leaf stomata. Contrarily, evaporation is the fractioning process of the stable isotope of water lost from soil, leaf surfaces, lakes, rivers, and so on. This study follows these definitions, and only when the amount of water vapor lost from the surface is important, will we use the term “evapotranspiration,” which includes both fractioning and nonfractioning processes.

In the other regions of South America such as in the central Andes, precipitable water comes from tropical South America, tropical North Atlantic, subtropical South America, and tropical South Atlantic (Vuille et al. 2003). This is clarified by the atmospheric general circulation models (AGCMs) and the IAEA/GNIP database. However, the hydrological cycle of Pantanal, located in the center of South America (Fig. 1), has not been fully investigated from the isotopic point of view.

Cuiabá (15.60°S, 56.10°W, 165 m ASL), one of the IAEA/GNIP stations, is located on the northern fringe of Pantanal, which is the largest wetland in the world (Fig. 1). Here, Dansgaard (1964) found that the amount effect of precipitation was important for the isotopic variations, by the preliminary analysis of the IAEA/GNIP database from 1961 to 1962. Dansgaard (1964) also mentioned that the temperature effect was insignificant for the isotopic variations at Cuiabá, since it is located in the Tropics.

Gat and Matsui (1991) tabulated that δ¹⁸O in precipitation at Cuiabá was positive from June to October, meaning that more enriched water than the Standard Mean Ocean Water was observed in the dry season in Pantanal. They also showed that the seasonal variation of δ¹⁸O at Cuiabá was the largest among 10 IAEA/GNIP stations located in the tropical South America. Also, Rozanski and Araguás-Araguás (1995) confirmed the findings of Dansgaard (1964) by extending the period of the IAEA/GNIP database. However, all of these studies lack the viewpoint of the year-to-year variations.

According to the authors’ preliminary analysis of the IAEA/GNIP database, the year-to-year variations of δ¹⁸O and δD in precipitation at Cuiabá were the largest during February within a year (Table 1). In comparison with δ¹⁸O in precipitation during February at Manaus (Fig. 2), one of the IAEA/GNIP stations in the Amazon basin (3.12°S, 60.02°W, 72 m ASL; Fig. 1), meteoric water at Cuiabá is more depleted and variable, although the mean precipitation is larger in Manaus and the standard deviations of Manaus and Cuiabá are equal (Fig. 2). If the variation of δ¹⁸O were determined by the amount effect alone, meteoric water at Manaus would be more depleted. In fact, meteoric water at Cuiabá is more depleted and variable, so that the amount effect does not fully account for the variation of the depleted water at Cuiabá (Fig. 2).

The enriched water observed at Cuiabá is partly explained by the transpiration of the tropical rainforests in the Amazon basin (Salati et al. 1979). This is because the vapor flux from the Amazon basin contributes to the precipitation at Cuiabá (Fig. 1). Also, the continentality effect (Dansgaard 1964) is partly responsible for the depleted water at Cuiabá (Fig. 2). Vuille et al. (2003) estimated that this effect in the Andes, along with the altitude effect, was as large as −14 ± 2 per mil of δ¹⁸O in the rainy season around 0°–10°S of the South American continent. Since this finding was based on the IAEA/GNIP database at Izobamba in the Andes (0.37°S, 78.55°W, 3058 m ASL; Fig. 1), the year-to-year variations of δ¹⁸O at Cuiabá appear fairly large (Fig. 2), even if the difference in the altitudes of Izobamba and Cuiabá is considered.

The objective of this study is to investigate why such large year-to-year variations of δ¹⁸O were observed in the precipitation during February at Cuiabá (Fig. 2). The role of the seasonal emergence of wetland in Pantanal in the hydrological cycle during the rainy season will be discussed as well. The understanding of what controls the modern temporal variability of δ¹⁸O in precipitation will certainly give some implications as to the paleoclimate, although this study does not directly treat the paleoclimatic data.

2. Amount effect and the occurrences of storm precipitation

Monthly data of δ¹⁸O, δD, and the amount of precipitation were taken from the IAEA/GNIP database. Observations at Cuiabá were available from 1961 through 1987. For February, there were seven missing readings for δ¹⁸O and eight for δD, while there were no missing readings for the amount of precipitation. From a visual inspection of Fig. 2, the most depleted years (1963, 1978, and 1968) and the most enriched years (1966, 1984, and 1983) were selected for the analysis.

Figure 3 depicts the relationship between δ¹⁸O and precipitation for February. Although not statistically significant at the 5% level, precipitation amount increased in the most depleted years (1963, 1978, and 1968; Fig. 2). This fixes the amount effect as shown by the previous studies (Dansgaard 1964; Rozanski and Araguás-Araguás 1995). However, in some years, such as in 1966 (the most enriched year), precipitation got enriched even though precipitation exceeding the long-term mean was observed (Fig. 3). The latter result is believed to be caused by differences in the way the precipitation events occur (Dansgaard 1964). Specifically, even if the expected amount of precipitation is observed during a given month, δ¹⁸O will be depleted if heavy precipitation (i.e., storm precipitation) is concentrated in a few days out of the month. In contrast, δ¹⁸O will be enriched if light precipitation is observed for many days out of the month.

The occurrences of storm precipitation were counted with these considerations in mind. Here, storm precipitation was defined as the continuous precipitation event separated from other such events by the absence of precipitation for at least one day (JMA Statistical Division 1960). Namely, the time period of storm precipitation was variable and depended on how long a precipitation event continued. For this analysis, we used the daily precipitation data at Cuiabá from 1978 to 1987 archived in the Global Daily Summary (NOAA/NCDC 1994), along with the data from 1961 to 1977 archived in the Center for Weather Forecasting and Climate Studies/National Institute of Space Studies (CPTEC/INPE) in Brazil.

Figure 4 displays the statistically significant relationship (at the 5% level) between δ¹⁸O and the occurrences of storm precipitation, with the threshold of the storm precipitation set to 10 mm. This threshold was determined by trial and error, based on the criteria for daily precipitation defined by JMA (1991). In February 1966, the number of occurrences of storm precipitation equal to or exceeding 10 mm was smaller than in any other year, although the amount of precipitation was the next to the largest (Fig. 3). Evidently, the accumulation of light precipitation brought about the enriched meteoric water in February 1966. Namely, what was implied by Dansgaard (1964) was proved by the actual meteorological data.

The large year-to-year variations of δ¹⁸O in precipitation during February at Cuiabá are explained by the amount effect as well as by the occurrences of storm precipitation. Next, the origin of precipitation will be investigated in relation to the vapor flux field.

3. Vapor flux field and d-excess

We used the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996) to calculate the vertically integrated vapor flux for both the most depleted years (1963, 1978, and 1968) and the most enriched years (1966, 1984, and 1983). Since the composite fields for these years were qualitatively similar to those shown in Fig. 1, the anomalies from Fig. 1 were investigated.

Figure 5a illustrates the anomaly field for 1963 (the most depleted year). It is seen that the northward anomaly prevails around Pantanal and Cuiabá. As it happens, the vapor flux from the Amazon basin weakened in 1963. These northward anomalies were also observed in 1978 and 1968 around Pantanal and Cuiabá (Figs. 5b and 5c).

Precipitation greater than the long-term mean was observed in 1963, 1978, and 1968 (Fig. 3), while the vapor flux from the Amazon basin weakened in those years (Fig. 5). In terms of the atmospheric water balance, local evapotranspiration must contribute to the greater precipitation in these depleted years. The variations in d-excess (Dansgaard 1964) support this idea. The rank of d-excess values for these years are third (1963), first (1978), and fifth (1968), as shown in Fig. 6, although depleted precipitation was not observed in the second- (1976) and fourth-ranked (1977) years, which are not specifically designated in Fig. 6. This is presumably due to the moderate amount of precipitation (Fig. 3) and the moderate number of the occurrences of storm precipitation (Fig. 4). From Fig. 6, it is evident that fast evaporation (not transpiration) must contribute to precipitation greater than the long-term mean in 1963, 1978, and 1968.

In the enriched years (1966, 1984, and 1983), the observed precipitation does not have lower values of d-excess (Fig. 6). Rather, d-excesses in these years are nearly equal to 10, a value that is widely observed in meteoric water (Craig 1961). In the enriched years, the anomaly fields of the vapor flux do not have common features, unlike the depleted years (Fig. 5). In 1966, the eastward anomaly prevailed around Cuiabá (Fig. 7a), while the westward anomaly prevailed in 1984 (Fig. 7b). Only in 1983 did the southward anomaly prevail at Cuiabá, which transported vapor from the Amazon basin (Fig. 7c). We, therefore, conclude that in the enriched years, the occurrences of storm precipitation are important in 1966 (Fig. 4), while the amount effect is responsible for 1984 (Fig. 3). In 1983, enriched meteoric water is attributed to both the occurrences of storm precipitation and vapor flux field (Figs. 3 and 7).

4. Conclusions and discussion

The present study produced the following important findings:

1. The year-to-year variations of δ¹⁸O in precipitation are the largest in February at Cuiabá. In both the depleted years (1963, 1978, and 1968) and the enriched years (1966, 1984, and 1983), the variations of δ¹⁸O can primarily be explained by the amount effect and by the occurrences of storm precipitation.

2. In all of the depleted years, a weakening of the vapor flux from the Amazon basin is found. In these years, precipitation greater than the long-term mean is observed along with fairly large values of d-excess.

3. In the enriched years, common features are not found in the anomaly fields of the vapor flux as opposed to those of the depleted years.

Vuille et al. (2003) mentioned that due to the inadequate data coverage of the IAEA/GNIP database, an analysis of the δ¹⁸O–climate relationship in a given region will always be incomplete. They, therefore, used AGCMs to help understand the δ¹⁸O–climate relationship in South America. Contrarily, we can come to the above conclusions, by combining the IAEA/GNIP database with other available data. This implies the importance of the data analysis itself. Especially, by analyzing the daily precipitation data, we can point out the importance of the occurrences of storm precipitation for the year-to-year variations of δ¹⁸O, which was implied by Dansgaard (1964).

It is evident that fast evaporation must contribute to precipitation more than the long-term mean in the depleted years at Cuiabá (Fig. 6). Considering the atmospheric water balance, where does this much evaporation come from? It is suggested that this vapor comes from the seasonal emergence of wetland in Pantanal, which becomes grassland in the dry season (e.g., Prance and Schaller 1982). This seasonal change in the surface conditions has already been captured by passive microwave remote sensing studies (e.g., Giddings and Choudhury 1989). The seasonal emergence of wetland in Pantanal seems to play an important role in the variations of stable isotopes during the rainy season.

In most of the GCMs, Pantanal is designated as grassland [e.g., the Simple Biosphere model (SiB2) of Sellers et al. (1996)] where surface conditions do not change drastically from grassland to wetland within a year. Certainly, evaporation in the rainy season and transpiration in the dry season affect the variations in the stable isotopes. Henderson-Sellers et al. (2002) mentioned the importance of accurately handling terrestrial open waters such as lakes and rivers in GCMs from an isotopic point of view. In this sense, it is also necessary to consider drastic seasonal changes from grassland to wetland within a year, as seen in Pantanal, when GCMs are used to simulate the hydrological cycle.