1 Introduction

Study of Indian Summer Monsoon (ISM) rainfall is essential in understanding the hydrological cycle of the country (Guhathakurta and Rajeevan 2008). Northeast (NE) Indian region receives some of the heaviest rainfall during ISM (around 1400 mm) as compared to other parts of Indian subcontinent (Prabhu et al. 2017). This area is dominated by dynamic weather arising due to its diverse orographic conditions (Mahanta et al. 2013). Both local and remote forces (Prabhu et al. 2017) govern ISM rainfall in the eastern-most part of the country. In the last few decades, this region has experienced a decline in ISM rainfall (Preethi et al. 2017) leading to drought like conditions in different states of NE India. Western disturbance and Nor’westers are major phenomenon governing moisture transport over NE India during winter and pre-monsoon months, respectively (Laskar et al. 2015; Jeelani et al. 2018).

Studying isotopic composition of water vapor provides information about the history of mixing and phase change associated with a given air parcel (Dansgaard 1964; Gat 1996; Mook 2000). Isotopic analysis of precipitation has been widely used to understand moisture sources over specific regions (Deshpande et al. 2010; Srivastava et al. 2014; Lekshmy et al. 2015). Source of moisture during ISM over Northeastern states has been studied using stable water isotopes in rainwater samples (Breitenbach et al. 2010). Isotopic composition of water vapor differs with that of precipitation due to different sources and hence, remote sensing gives new information about the hydrological cycle, which is unattainable from surface (Worden et al. 2007). Rainfall depletes the resulting vapor, as heavier isotopes tend to condense easily. On the other hand, isotope ratios can indicate how the atmosphere is conditioned by various surface processes like evaporation and transpiration, which can in turn influence precipitation. Limited work has been carried out connecting monsoon precipitation with satellite-based observations of stable water isotopes over India.

Satellite-based remote sensing of water isotopes in atmosphere has been used to understand various hydrological processes (Worden et al. 2007; Frankenberg et al. 2009; Lee et al. 2012). Satellite-based sensors like SCanning Imaging Absorption Spectrometer for Atmospheric CHartographY (SCIAMACHY), Tropospheric Emission Spectrometer (TES), Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) and Atmospheric Chemistry Experiment-Fourier Transform Spectrometer (ACE-FTS) have provided isotopic ratio observations with global coverage (Worden et al. 2006; Nassar et al. 2007; Payne et al. 2007; Schneider et al. 2018). SCIAMACHY, on board ENVISAT-1 launched in March 2002, was a spectrometer designed to measure the composition of trace gases in the troposphere and stratosphere. SCIAMACHY retrieves columnar water isotope ratios of HDO/H2O using the Shortwave-Infrared Carbon Monoxide Retrieval (SICOR) algorithm (Schneider et al. 2018) and performs joint retrieval of CH4, H2O, HDO and CO (Scheepmaker et al. 2013) to reduce uncertainty. Limitations in sensitivity of current instruments restrict the retrieval of H 182 O; hence, observations of D/H ratio are only available from space-based platforms.

Several global climate phenomenon like El-Niño and La-Niña accompanied by atmospheric circulations, such as El-Niño-Southern Oscillation (ENSO) and Indian Ocean Dipole (IOD), also affect the ISM rainfall variability (Krishna Kumar et al. 1995) resulting in drought and flood-like situations in different parts of the country (Philander 1983). Out of these, ENSO is considered an important driving factor that affects the inter-annual variations in ISM (Walker 1923). To monitor ENSO, various indices like Southern Oscillation Index (SOI), Multivariate ENSO Index (MEI), NOAA Oceanic Niño Index (ONI), etc. have been used to quantify the nature of atmosphere–ocean circulations. MEI is well suited to monitor ENSO events as compared to others, because it takes into account six variables of meteorological and oceanographic components (Wolter and Timlin 1998) over the Pacific Ocean. Studies have been carried out to establish relationship between stable water isotopes and ENSO variability in different parts of the globe. Vuille and Werner (2005) studied the relationship between South American Summer Monsoon and ENSO analyzing the stable isotopes in precipitation. Similar studies have been carried out by Okazaki et al. (2015) over Western Africa connecting ENSO with inter-annual variability of isotopic composition in water vapor using satellite-based observation and model simulations. However, there exists a gap area in understanding the effect of these phenomenon on the distribution of water isotope ratios in the atmosphere over Indian landmass. In this paper, we attempt to understand the relationship between water isotope ratios and monsoon precipitation over NE India during pre-monsoon and monsoon seasons. We also studied the response of atmosphere over Indian mainland in connection to ENSO events.

2 Study area and data used

SCIAMACHY Level-2 point retrievals of HDO and H2O for each orbit (available at ftp://ftp.sron.nl/pub/pub/DataProducts/SCIAMACHY_HDO/ (Schneider et al. 2018)) were selected for the period 2003–2011 over Northeast India with bounding box 22°–28°N and 88°–98°E as shown in figure 1. This region comprises over 60% forest cover (Jain et al. 2013) along with the flood plains of Brahmaputra river in Assam and Bangladesh. Forests of NE India are regarded as the northern-most limit of tropical rainforests in the world (Procter et al. 1998). This region is an important biodiversity hotspot of India and a wide variety of flora and fauna are found here (Dikshit and Dikshit 2014). The onset of ISM in the region is usually between 1 and 10 June every year (IMD Report 2009).

Figure 1
figure 1

Study area (white bounding box) overlaid on MODIS LULC for 2009 (Friedl et al. 2010).

Full size image

Precipitation data was obtained from Global System for Mapping of Precipitation (GSMaP), provided by JAXA’s Global Rainfall Watch System. Gauge calibrated GSMaP data is available at multiple spatial (0.1° × 0.1° and 0.25° × 0.25° grids) as well as temporal scales (hourly and daily). Daily GSMaP data at spatial resolution of 0.1° × 0.1° was used in our study for the year 2003–2011. Multivariate El-Niño Index (MEI) data was obtained from NOAA (https://www.esrl.noaa.gov/psd/enso/mei/).

3 Methodology

Orbit-wise SCIAMACHY Level-2 retrievals of columnar HDO and H2O concentrations for the period 2003–2011 were used in this study. Individual point retrievals were filtered using the quality control criteria suggested by Schneider et al. (2018). Methane and water vapor being the major gases, constraints are applied on their concentration values based on a priori model estimates. This allows removal of spurious outliers from the data. Based on this, only those data points were considered which meet the following conditions:

$$ 0.9 < \frac{{{\text{C}}_{{{\text{CH}}_{4} }} }}{{{\text{C}}_{{{\text{CH}}_{{4 \cdot {\text{TM}}_{5} }} }} }} < 1.1, $$
(1)
$$ 0.7 < \frac{{{\text{C}}_{{{\text{H}}_{2} {\text{O}}}} }}{{{\text{C}}_{{{\text{H}}_{2} {\text{O}} \cdot {\text{ECMF}}}} }}, $$
(2)

where, \( {\text{C}}_{{{\text{CH}}_{4} }} \) and \( {\text{C}}_{{{\text{H}}_{2} {\text{O}}}} \) are columnar concentrations of CH4 and H2O and \( {\text{C}}_{{{\text{CH}}_{4} \cdot {\text{TM}}5}} \) and \( {\text{C}}_{{{\text{H}}_{2} {\text{O}} \cdot {\text{ECMF}}}} \) are corresponding model predictions of columnar concentrations. These two criteria act as simple cloud filter. P15.9 and P84.1 (15.9th and 84.1th percentile) of logarithmic HDO and H2O data distribution (Schneider et al. 2018) are used to define σ as follows:

$$ \sigma = \frac{1}{2}\left( {{\text{P}}_{84.1} - {\text{P}}_{15.9} } \right). $$
(3)

Data points with absolute error in H2O and HDO concentrations within 5σ around median (μe) are used. Points are also filtered on the basis of root mean square of the spectral fit residual crms keeping data only within 6σ around median (μrms).

$$ \mu_{{{\text{e}}\left( {{\text{H}}_{2} {\text{O}}} \right)}} - 5\sigma_{{{\text{e}}\left( {{\text{H}}_{2} {\text{O}}} \right)}} < log {\text{ e}}_{{{\text{H}}_{2} {\text{O}}}} < \mu_{{{\text{e}}\left( {{\text{H}}_{2} {\text{O}}} \right)}} + 5\sigma_{{{\text{e}}\left( {{\text{H}}_{2} {\text{O}}} \right)}} , $$
(4)
$$ \mu_{{{\text{e}}\left( {\text{HDO}} \right)}} - 5\sigma_{{{\text{e}}\left( {\text{HDO}} \right)}} < log {\text{ e}}_{\text{HDO}} < \mu_{{{\text{e}}\left( {\text{HDO}} \right)}} + 5\sigma_{{{\text{e}}\left( {\text{HDO}} \right)}} , $$
(5)
$$ \mu_{\text{rms}} - 6\sigma_{\text{rms}} < log {\text{ c}}_{\text{rms}} < \mu_{\text{rms}} + 6\sigma_{\text{rms}} . $$
(6)

Along with this, data points are also filtered on the basis of number of iterations performed to reach an optimum solution, Niter ≤ 12 and observations with solar zenith angle \( v_{\text{sz}} \) < 70°. These quality control criteria prevent spurious concentration observations from being used in analysis. Once the valid data points along with the day of observation are obtained, these are binned onto 2° × 2° grids for every 5-day period (pentad). Stable isotope ratio of HDO/H2O is usually defined in parts per thousand (per mil—‰) relative to a standard (Vienna Standard Mean Ocean Water, VSMOW) and given by delta (δ—notation) expressed by equation 7.

$$ {\updelta }{\text{D}} = \left[ {\frac{{\left( {{\text{HDO}}/{\text{H}}_{2} {\text{O}}} \right)_{\text{sample}} }}{{\left( {{\text{HDO}}/{\text{H}}_{2} {\text{O}}} \right)_{\text{VSMOW}} }} - 1} \right] \times 1000\left( {\permille} \right), $$
(7)

where the value of (HDO/H2O)VSMOW is 3.11 × 10–4. δD was computed for each valid retrieval and binned across entire study area for every pentad. These pentad-averaged total columnar δD were plotted to depict the temporal variability over study area and understand its seasonal cycle.

Valid observations over NE India for the year 2003–2011 across all the individual orbits were binned to calculate pentad-averaged columnar δD for the entire region. Averaged columnar δD for pre-monsoon (April and May) and monsoon (June–August) months was correlated with the rainfall data. Pearson correlation coefficient maps were generated between area-averaged columnar δD for pre-monsoon (April–May) and monsoon (June–August) months and grid-level total precipitation for the corresponding months. Contours were generated for those regions where correlation was significant (p < 0.05). Correlation coefficient maps were also generated between grid-level (2° × 2°) mean-monthly columnar δD values and monthly Multivariate El-Niño Index (MEI). Monthly MEI was computed by taking mean of ith and (i − 1)th bi-monthly MEI values. Values were represented for significant correlations only, where p < 0.05. Methodology used in this paper has been summarized in the flow chart shown in figure 2.

Figure 2
figure 2

Flowchart of methodology used.

Full size image

4 Results

4.1 Temporal variability of columnar δD

Figure 3(a) represents the temporal variability of mean columnar δD over the study area (black line) along with standard deviation of all valid observations within a pentad (grey shaded region). From the figure, it is evident that columnar δD has a cyclic behavior. Fitting a simple second order Fourier series helps us visualize the two distinct frequency components: high δD values during summer season (owing to enriched vapor and reduction in fractionation factor) and sudden reduction in δD values with onset of monsoon during June. The values are low throughout winter months owing to lack of available moisture and increased fractionation factor at lower temperatures. From figure 3(b), the mean values range between − 210‰ during winter months (December–January) and − 105‰ during summer months (April–May). Although the inter-annual variability follows a distinct cyclic pattern, there are evident deviations from this trend during some years. These fluctuations or departures from mean can provide vital clues about the ongoing processes occurring in the region that affect the stable water isotope ratios in the atmosphere.

Figure 3
figure 3

(a) Temporal variability of pentad averaged columnar δD over study area. Black line represents mean of all valid observations and shaded area represents one standard deviation. Yellow line is a simple second-order Fourier series fit. (b) Inter-annual mean over all pentads for 9 years (2003–2011).

Full size image

4.2 Relationship between δD and precipitation

Precipitation occurs when vapor is removed from the atmosphere and since phase change influences water isotope ratios, precipitation plays a direct role in governing columnar δD. Heavier isotopes of water (like HDO) tend to condense rather easily and hence, the remaining vapor after precipitation is mostly depleted in heavy isotopes. δD values, along with specific humidity and temperature, provide valuable information about the source of vapor. To understand this linkage, we plot inter-annual δD during pre-monsoon (April–May) and monsoon (June–August) seasons with the mean seasonal precipitation during monsoon months (figure 4). From figure 4, it was observed that variability of δD during pre-monsoon season closely follows the trend in monsoon precipitation. During 2006, we see a reduction in water isotope ratios in both pre-monsoon and monsoon seasons and correspondingly lowest precipitation was observed in the region during this time. This trend, however, breaks in year 2011. Relationship between these two variables can be understood in detail using a grid-level analysis. Nevertheless, this figure does provide us with evidence of possible correlation between pre-monsoon isotope ratios and monsoon precipitation.

Figure 4
figure 4

Inter-annual variability of water isotope ratios during pre-monsoon and monsoon season compared with average JJA precipitation for the corresponding year.

Full size image

Since δD and precipitation were found to be inter-related, studying the spatial correlation between the two variables can reveal information about surface and lower atmospheric processes governing monsoonal precipitation. The aim of this analysis was to identify the areas where monsoon precipitation is affected by water isotope ratios. This could also imply that the conditioning of atmosphere by surface processes (like evaporation, transpiration and moisture transport) during pre-monsoon season plays a vital role in governing the monsoon precipitation in those regions. With this basis, we computed Pearson correlation coefficients between area-averaged columnar δD (during pre-monsoon and monsoon) for the study region and the corresponding grid-level seasonal precipitation. Four maps representing the spatial distribution of correlation between these two variables were generated (figure 5). The regions where significant correlation (p < 0.05) was observed have been represented with contours. Artifacts in GSMaP precipitation data resulted in spurious lines on the right side of these images.

Figure 5
figure 5

Pearson correlation coefficient (r) between area-averaged columnar δD and grid-level seasonal precipitation for the study period. (a) pre-monsoon δD with pre-monsoon precipitation, (b) pre-monsoon δD with monsoon precipitation, (c) monsoon δD with pre-monsoon precipitation, and (d) monsoon δD with monsoon precipitation. Contours represent regions with significant correlation (p < 0.05).

Full size image

From figure 5(a), we observe significant negative correlation between pre-monsoon δD and pre-monsoon precipitation over parts of central Bangladesh and West Bengal. This could be due to the fact that precipitation causes the remaining vapor to be depleted in heavier isotopes. Hence, these regions show a direct negative correlation between the two variables. From figure 5(b), we observe a positive correlation between pre-monsoon δD and monsoon precipitation over Meghalaya and parts of Assam. This region receives some of the highest rainfall in India and a positive correlation implies that the conditioning of atmosphere occurring during pre-monsoon season is linked with the amount of precipitation during monsoon season. This figure provides evidence that surface and atmospheric processes affecting pre-monsoon water-isotope ratios in this region also govern monsoon rainfall. Evapotranspiration plays a crucial role in determining isotope ratios in the overlying atmosphere. On one hand, evaporation results in depleted water vapor and on the other, transpiration does not cause fractionation at steady state resulting in similar isotope ratios in vapor and liquid phase. An increase in pre-monsoon δD could signify enhanced evapotranspiration from this region and this could positively influence the monsoon precipitation.

4.3 Relationship between δD and ENSO

Columnar isotope ratios are not only governed by local processes but also by global climate phenomenon such as ENSO, IOD etc. ENSO plays an important role in influencing global climate and different regions respond differently to these oscillations. Multivariate El-Niño Index (MEI) is a comprehensive quantification of ENSO phenomenon, which considers into account six different variables. Figure 6(a) represents regions showing significant Pearson correlation coefficient (p < 0.05) between monthly MEI and monthly columnar δD at 2° × 2° grids with zero-lag. Maximum correlation of 0.33 was observed between the two variables. Effect of ENSO on water isotope ratio was highest over central and southern India including Maharashtra, Chhattisgarh, Orissa, Telangana and Andhra Pradesh. Parts of Mizoram, Tripura, Bangladesh and West Bengal also showed weak correlation between the two variables indicating that the atmosphere in these regions responds to ENSO phenomenon within a month. Parts of Myanmar, Laos and central Thailand also show significant positive correlation with MEI due to their vicinity to the Pacific Ocean where ENSO occurs. Figure 6(b) shows the correlation coefficient where MEI lags δD by one month. The effect of ENSO was observed even after 1 month in same regions where peak correlation was seen with zero lag, however, these effects were greatly reduced. From figure 6(a, b), we can conclude that the atmosphere over these regions responds to ENSO within a month and it plays a role in influencing water isotope ratios.

Figure 6
figure 6

Pearson correlation coefficient between MEI and mean monthly columnar δD with (a) no lag and (b) 1 month lag. Only significant values (p < 0.05) are represented.

Full size image

5 Conclusion

In this paper, we discussed the relationship between monsoon precipitation and SCIAMACHY observed columnar water isotope ratios during pre-monsoon and monsoon seasons for Northeast India. Significant positive correlation (r > 0.7, p < 0.05) between δD and grid-level GSMaP precipitation was observed over parts of Meghalaya and central Assam, indicating strong correlation between pre-monsoon surface/atmospheric processes and amount of monsoon rainfall over the region. This finding may have implications in inclusion of water isotope ratios as an input parameter for various weather forecasting models. Effect of ENSO on water isotope ratios in atmosphere was studied and a positive correlation was observed between δD and MEI over parts of Central India, Myanmar, Laos and Thailand. Further study needs to be carried out to understand the variations in atmospheric water isotopes in response to declining monsoon precipitation over Northeast India.