1. Introduction
The Indian summer monsoon (ISM), which accounts for 70% of the annual rainfall over the country, is one of the most awaited weather transitions over the region (Parthasarathy et al. 1994) because of its tremendous socioeconomic and hydrological impacts. The southwesterly winds bringing the monsoon over the Indian peninsula first arrive over the southern Indian state of Kerala, widely known as the “gateway” of the Indian monsoon. The monsoon onset over Kerala (MOK) marks the end of scorching early summer and the commencement of the rainy season. It is well known that the normal date of MOK is 1 June; however, it varies by a few days from year to year with a standard deviation of 8–9 days (Ananthakrishnan and Soman 1991; Joseph et al. 1994). Although the date of MOK does not have considerable correlations with the subsequent progression of the ISM to northern India and the seasonal mean rainfall (Bansod et al. 1991), the timing of MOK can have significant impact on agricultural productivity. A delay in the onset of the monsoon by a few weeks affects the initiation of agricultural activity while an early onset might not be utilized to its full advantage without an advanced forecast. Thus, the monitoring and forecasting of MOK is essential for the economy of the country.
The MOK is associated with changes in the large-scale dynamical parameters as well as local moisture parameters (Ananthakrishnan et al. 1983; Ananthakrishnan and Soman 1988; Joseph et al. 1994, 2006; Pai and Rajeevan 2009; Pearce and Mohanty 1984; Soman and Krishna Kumar 1993, among others). The differential heating of land and sea is the primary cause that drives the ISM. In the boreal summer, the Tibetan Plateau acts as an elevated heat source and the increased diabatic heating over the region favors the seasonal reversal of tropospheric temperature and pressure gradients (Flohn 1960; Xavier et al. 2007), and thus the occurrence of onset of monsoon (Murakami and Ding 1982; Yanai et al. 1992). Krishnamurti et al. (1981) showed that the kinetic energy over the southeast Arabian Sea (AS) increases by an order of magnitude just prior to MOK. In conjunction with the monsoon onset, a rapid, substantial, and sustained increase in rainfall occurs along with the establishment of a cross-equatorial low-level jet along the Somali coast. The southwesterlies associated with this low-level jet deepen and can reach even up to 600-hPa level (Pearce and Mohanty 1984; Pai and Rajeevan 2009). In the upper troposphere, the midlatitude westerly jet shifts northward to the north of the Tibetan Plateau (Yin 1949) and a tropical easterly jet stream (TEJ) is prevalent over the Indian region (Koteswaram 1960), defining the southern flank of the Tibetan high . Thus, monsoon onset is marked by strong vertical wind shear over the Indian subcontinent with westerlies prevailing in the lower levels and easterlies in the upper troposphere. MOK is also associated with an increase in the height of increased water vapor and midtropospheric moisture (700–500 hPa) about 8–10 days earlier (Rao 1976; Soman and Krishna Kumar 1993; Simon and Joshi 1994). A band of deep convection [termed the maximum cloud zone by Sikka and Gadgil (1980)] in the east–west direction propagates northward through the southern tip of India (Joseph et al. 1994).
Based on the abovementioned changes in the large-scale and local meteorological conditions associated with MOK, several researchers have formulated criteria for objectively defining the start of monsoon season (Ananthakrishnan and Soman 1988; Fasullo and Webster 2003; Joseph et al. 2006; Lu and Chan 1999; Lu et al. 2009; Pai and Rajeevan 2009; Soman and Krishna Kumar 1993; Taniguchi and Koike 2006; Wang et al. 2009; Zeng and Lu 2004, among others). The specification of a criterion for the onset is generally a subjective decision based on an overall judgment that takes into account of the changes in the circulation features, seasonal reversal of winds, and a sustained increase in rainfall over Kerala (Flatau et al. 2003). Ananthakrishnan and Soman (1988) and Soman and Krishna Kumar (1993) considered the transition from light to heavy rainfall category as the date of onset, provided that average daily rainfall during the first 5 days after the transition should not be less than 10 mm. Fasullo and Webster (2003) developed a Hydrological Onset and Withdrawal Index (HOWI) using the vertically integrated moisture transport, whereas Joseph et al. (2006) utilized the depth of westerlies and widespread convection around Kerala. Wang et al. (2009) determined MOK by the establishment of rapid and sustained 850-hPa zonal wind averaged over the southern AS (5°–15°N, 40°–80°E). Zeng and Lu (2004) and Lu et al. (2009) proposed use of precipitable water data for defining onset. The India Meteorological Department (IMD), the authorized weather and climate forecasting agency under the government of India, follows the criteria defined by Pai and Rajeevan (2009), which combines the rainfall over various stations over the state of Kerala, depth of westerlies up to 600 hPa, and satellite-observed outgoing longwave radiation values north of the equator. The conditions to be satisfied are the following: 1) If after 10 May, 60% of the available 14 stations in Kerala report rainfall of 2.5 mm or more for two consecutive days, the MOK may be declared on the second day, provided the following criteria are also satisfied in concurrence. 2) The depth of westerlies should be maintained up to 600 hPa in the area 0°–10°N, 55°–80°E. The zonal wind speed over the area bounded by 5°–10°N, 70°–80°E should be on the order of 15–20 kt at 925 hPa. The source of data can be RSMC New Delhi wind analysis and/or satellite-derived winds. 3) The INSAT-derived OLR value should be below 200 W m−2 in the area 5°–10°N, 70°–75°E. Special emphasis is given to the sharp increase in rainfall over Kerala, although other factors are also confirmed before declaring the onset.
A successful multi-index definition of MOK needs to satisfy two conditions. First, the local indices should capture the time when the large-scale meteorological conditions start influencing the local scale. Second, from a forecast perspective, the large-scale indices should be skillfully forecastable quantities. Kerala frequently witnesses widespread rainfall activity due to premonsoon thundershowers, which must be differentiated from the monsoon rains (Flatau et al. 2003). The prediction of MOK should thus be viewed as the prediction of the transition from premonsoon localized synoptic activity to large-scale sustained rainfall activity, so as to circumvent the “false” or “bogus” onsets that are unrelated to the large-scale monsoonal system (Goswami and Gouda 2010). As the bogus onsets can be followed by extended periods of dry weather, the false prediction can have considerable agricultural and economic impacts.
It has been proposed by various researchers that the onset date could be predicted in terms of antecedent circulation (Kung and Sharif 1980, 1982), moisture (Kumar 2004; Simon et al. 2006), thermal features (Joshi et al. 1990; Luo and Yanai 1984), and premonsoon thunderstorm activity (Ghanekar et al. 2003). Several studies have reported on the prediction of the onset date of the summer monsoon over India using statistical methods (Ghanekar et al. 2003; Pai and Rajeevan 2009; Raju et al. 2007; Ramesh et al. 1996). However, none of these algorithms explicitly focuses on using an index that could draw on dynamical model forecasts as well as observational quantities to define MOK. Although the prediction of MOK has been explored using dynamical models (Goswami and Gouda 2010), explicit use of the large-scale conditions along with rainfall (e.g., the kinetic energy or strength of the low-level jet trajectories over oceanic region, upper air circulation conditions, etc.) has not been attempted. Furthermore, efforts to predict MOK in real time using a dynamical frame work is still emergent and need to be explored in detail.
Therefore, the main objectives of the present study are 1) to develop an objective criterion for the real-time operational dynamical prediction of MOK in an extended range (~2–3 weeks in advance) time scale using an ensemble prediction system (EPS), 2) to formulate the criterion in such a way that it successfully eliminates the possibility of bogus onsets and that the predicted MOK dates should largely match the IMD-declared MOK dates, and 3) to assess the ability of the EPS in reproducing the observed space–time evolution of various meteorological parameters during MOK.
The EPS used in this study is constructed indigenously at the Indian Institute of Tropical Meteorology (IITM), India, under the National Monsoon Mission (http://www.tropmet.res.in/monsoon/) based on the state-of-the-art coupled general circulation model Climate Forecast System version 2 (CFSv2; Saha et al. 2014) developed at the National Centers for Environmental Prediction (NCEP; United States). The extended-range prediction group at IITM has been providing the extended-range forecasts of active/break spells of ISM since 2011 up to four pentad lead using the EPS based on CFSv2 [see Abhilash et al. (2012, 2014b) and Sahai et al. (2013) for details on the generation of EPS] from 16 May to 28 September at 5-day intervals (16 May, 21 May, 26 May, …, 23 September, 28 September), and the forecasts have been updated at the IITM website (http://www.tropmet.res.in/erpas/) on a real-time basis since 2013. Given that the normal date of MOK is 1 June and our objective is to make extended-range forecasts (~15–20 days), we have utilized the forecasts from 16 May initial conditions (IC) of each year during 2001–14, in concurrence with the date on which IMD issues the onset forecast every year. Additionally, the choice of 16 May IC is suitable for MOK forecast as it can predict the onsets that happen after mid-May and before the first week of June considering the fact that the climatological MOK date is 1 June with a standard deviation of 8–9 days.
2. Model, data, and methodology
a. Model and observational datasets
We use the NCEP CFSv2 (Saha et al. 2014), with the Global Forecast System (GFS) as its atmospheric component, coupled to the GFDL Modular Ocean Model version 4p0d (MOM4; Griffies et al. 2004), a four-layer Noah land surface model (Ek et al. 2003), and a two-layer sea ice model (Wu et al. 2005). In this study, output from the CFSv2-based Grand Ensemble Prediction System (CGEPS; Abhilash et al. 2015) is used. The CGEPS includes three subensembles, namely from CFSv2 run at T126 (~100 km; hereafter termed CFS126) and T382 (~38 km; hereafter termed CFS382) horizontal resolutions, plus the GFS forced with bias-corrected SST from CFS126 (hereafter termed GFSbc). CGEPS has 21 ensemble members of GFSbc and 11 members each of CFS126 and CFS382. Every year, each of these ensembles are run for 45 days lead time every 5 days, starting from 16 May to 28 September. As mentioned earlier, the forecasts initialized on 16 May of every year during 2001–14 are utilized here. For the experimental details and skills of GFSbc, CFS126, and CFS382, see Abhilash et al. (2012), Abhilash et al. (2014a,b), and Sahai et al. (2015), respectively. Each of the ensemble members of the CGEPS is generated by slightly perturbing the initial atmospheric conditions with a random matrix (random number at each grid point) generated from a random seed. To center the ensemble seeding around an unperturbed analysis, Gaussian random numbers with mean 0 and unit standard deviation are used to rescale wind, temperature, and moisture differences between a short-term forecast and its corresponding analysis. It is ensured that amplitude of perturbation varies in accordance with the uncertainty in the analysis. Details on the formulation of EPS can be found in Abhilash et al. (2012, 2014b) and Borah et al. (2013).
For the observations, the IMD-TRMM merged rainfall dataset (Mitra et al. 2009) and the reanalysis dataset from the Modern-Era Retrospective Analysis for Research and Applications (MERRA; Rienecker et al. 2011) have been used. To check the reliability of the results and to confirm that the results are not an artifact of the observational dataset used, we also used reanalysis datasets from NCEP and the National Center for Atmospheric Research (NCAR; Kalnay et al. 1996) and ERA-Interim (Dee et al. 2011). The meteorological parameters utilized from reanalysis datasets are zonal, meridional, and vertical components of wind, specific humidity, air temperature, and geopotential height at pressure levels, surface pressure, and mean sea level pressure (MSLP). Additionally, the tropospheric temperature (TT) and kinetic energy (KE) of rotational wind and diabatic heating (Q1) are computed for further diagnostics. Anomalies of various observational parameters except rainfall are calculated based on the daily mean climatology of 30 yr (1981–2010). The rainfall anomalies are calculated based on the daily climatology of 14 yr (1998–2011). In the case of model, the climatologies are based on 12-yr (2001–12) hindcasts.
b. Criteria for predicting onset
The prerequisite to predict MOK is to devise an objective criterion that could successfully avoid bogus onsets. Goswami and Gouda (2010) indicated that rainfall is the most suitable variable for defining MOK, as it is characterized by persistence, significance, and coverage. Taniguchi and Koike (2006) suggested that onset criterion based on the wind speed west of Kerala is easier, clearer, and better screens out bogus onset cases. According to Pai and Rajeevan (2009), the depth of low-level westerlies up to 600 hPa is elemental for MOK.
As the main characteristic that differentiates the monsoon rains from the premonsoon thundershowers is the seasonal reversal of the low-level winds, persistence in both rainfall and wind is to be considered to eliminate the likelihood of bogus onsets. Considering all these factors, we developed a criterion using low-level wind as well as rainfall from all 43 members of the CGEPS for the real-time extended-range prediction of MOK. The first index has been defined from rainfall over Kerala (hereafter termed the ROK index), and the second is based on the strength of the low-level jet over the AS (hereafter termed the UARAB index). As Pai and Rajeevan (2009) suggested that the depth of low-level westerlies up to 600 hPa is elemental for MOK, a third index has been defined from the depth of westerlies up to 600 hPa (hereafter termed the Udepth index). In the criteria used by the IMD, rainfall over 14 stations in Kerala are considered. However, since we are formulating the criteria in the dynamical framework, it is difficult to use the station-wise data. Therefore, ROK is defined as the rainfall averaged over 8°–12°N, 74°–78°E (region R1 in Fig. 1). UARAB is defined as the zonal wind at 850 hPa averaged over 5°–12°N, 55°–75°E (region R2 in Fig. 1), while Udepth is defined as the zonal wind at 600 hPa over the same (R2) region. Thirty-day mean values of forecasted ROK and UARAB (hereafter termed as ROKM and UARABM respectively) starting from 17 May are computed, for each ensemble member and for each year. MOK for each member is then defined individually as the date on which both ROK and UARAB exceed 50% of their mean (ROKM and UARABM respectively), and one of them surmounts 70% of its mean, for 5 consecutive days, provided the value of Udepth during the period exceeds zero. The ensemble mean MOK date (of all 43 members) is treated as the final predicted MOK date. In this way, the uncertainties arising from the differences in the evolution of monsoon in individual ensemble members are taken care of. While choosing the threshold values, special care has been taken in selecting the one with which the predicted MOK dates have minimum root-mean-square error (RMSE) and maximum correlation coefficient (CC), with respect to IMD-declared MOK dates.
Regions selected for defining the ROK, UARAB, and Udepth indices that have been used for predicting MOK. The area labelled R1 (R2) denotes the region used for defining ROK (UARAB/Udepth) index.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
The previous study that attempted the dynamical prediction of MOK is by Goswami and Gouda (2010). However, their study used only rainfall; and before formulating the criterion, they applied statistical techniques such as debiasing of the rainfall with observed rainfall. On the other hand, as our objective is to formulate a criterion for real-time MOK prediction, the present study uses the forecast outputs from the EPS. Also, the criterion has been devised by considering the inherent biases and uncertainties within the system.
The main advantage of the criterion proposed in the present study is that it considers the persistence in the rainfall and low-level cross-equatorial flow post-MOK date, thereby avoiding the possibility of false or bogus onsets. However, this criterion can only be utilized in the prediction mode, as it requires the future evolution of the parameters used.
3. Results and discussion
As mentioned earlier, MOK is closely linked with changes and seasonal reversal in large-scale as well as local meteorological conditions. In this section, a detailed diagnosis on the evolution of the various meteorological parameters associated with the predicted MOK date is made and compared with observations to examine whether the EPS could reproduce the observed transitions contemporaneous with MOK, allowing us to scrutinize the robustness of the devised MOK criteria.
a. Predicted MOK dates and the associated evolution of ROK and UARAB indices
Based on the devised criteria, the MOK dates predicted during 2001–14 are shown in Table 1 along with the actual MOK dates declared by IMD. The table also presents the standard deviation among the 43 ensemble members used. It can be seen from the table that out of 14 years, the MOK dates of 11 years are well predicted (error is within 3 days). This illustrates the robustness of the proposed criterion. The biggest difference in the actual and predicted MOK dates is in 2002 and 2003 (8 and 9 days respectively).
Forecasted and observed MOK dates during 2001–14.
Figure 2 shows the temporal evolution of the ROK and UARAB indices in both observation and MME. The figure suggests that both rainfall and low-level wind increases sharply across the MOK date, in observation as well as MME. It is seen that both the ROK and UARAB indices show persistence in the magnitude subsequent to the forecasted MOK, whereas in the case of observations, both the indices show fluctuations after the IMD-declared MOK date. It is interesting to note that during some years (e.g., 2001, 2011, and 2013), the magnitude of the ROK (UARAB) index is consistently above 10 mm day−1 (10 m s−1) after the forecasted MOK date. On the other hand, the magnitude of the ROK (UARAB) index is persistently below 10 mm day−1 (10 m s−1) respectively for more than a week during 2007 and 2014. This could be attributed to the consideration of the mean of fixed 30 days starting from 17 May of every year, in the formulation of criterion. Still, both the indices show persistence after MOK, indicating the success of the devised criterion in avoiding bogus onsets, as suggested by Goswami and Gouda (2010). This persistence is noticed in the case of individual ensemble members also (not shown).
Evolution of ROK and UARAB indices in (a) OBS and (b) MME. For each year, the ROK index is shown as solid bars, the UARAB index as a thick solid line, and the MOK date as an open circle.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
The maximum discrepancy between the actual and forecasted MOK was noted in 2002 and 2003 (Table 1). During both years, the forecasted MOK was earlier by 8–9 days, compared to the IMD-declared MOK. However, it is noteworthy that during these years, both the indices were persistent after the predicted MOK date, indicating that MOK dates forecasted by the MME are reasonable measures of a genuine onset (albeit in an inaccurate set of model forecasts).
b. Spatiotemporal characteristics during MOK
For any dynamical framework and the criterion for predicting MOK to be successful, it should reproduce the transitions in the large scale as well as local meteorological parameters like rainfall, lower- and upper-level wind, tropospheric temperature, vertical instability, diabatic heating, and so on. This subsection examines the composite spatiotemporal evolution of these parameters in MME with the inception of MOK, and compared with observations to check the robustness of the devised criterion.
1) Composite spatial structure
As mentioned earlier, MOK is characterized by sustained rainfall over Kerala along with the intensification of the cross-equatorial flow. Figure 3 shows rainfall (shaded) and 850-hPa wind (streamlines) composited for the MOK dates during 2001–14. The CGEPS MME could predict the observed spatial characteristics of rainfall and wind realistically, including the clockwise gyre over eastern equatorial Indian Ocean (IO), the cross-equatorial flow along the Somali coast, and the increased rainfall along the Kerala coast. However, MME failed to reproduce the Southern Hemispheric Mascarene high, as observed, at least in the ensemble mean. The increased rainfall observed over the Head Bay region is also missing in MME.
Composite rainfall (shaded; mm day−1) and 850-hPa wind circulation (streamlines) during MOK for the period 2001–14. Observational rainfall values are from IMD–TRMM merged data and wind from MERRA data.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
The development of a strong north–south gradient of tropospheric temperature [defined as the air temperature averaged between 600 and 200 hPa by Xavier et al. (2007)] is another measure of monsoon onset (Murakami and Ding 1982; Xavier et al. 2007) and is a key factor in sustaining the monsoon current. Figure 4 shows the composite TT and MSLP during MOK. The patterns of TT and MSLP are consistent among all the three reanalysis datasets used. It is interesting to note from the figure that although the TT values are ~1°C lower in the MME compared to the observations, its north–south gradient, which is vital in the onset and maintenance of monsoon, is better reproduced. Additionally, the pressure pattern along peninsular India associated with the observed MOK is agreeably reproduced by the MME. Thus, it is confirmed that even though we have considered only rainfall over Kerala and the strength and depth of cross-equatorial westerlies in formulating the MOK criteria, the predicted MOK dates are robust and represent the seasonal reversal in the TT gradients associated with large-scale monsoonal flow.
Composite tropospheric temperature (shaded; K) and mean sea level pressure (contour; hPa) during MOK for the period 2001–14, for MME as well as three reanalysis datasets.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
2) Spatiotemporal evolution
The composite spatial structure of rainfall, 850-hPa wind, TT, and MSLP associated with the MOK date has been discussed in the previous subsection. Lagged spatial composites around MOK, using 3-day running means centered from lag −6 to lag +3 days, which will elucidate the spatiotemporal evolution of various dynamic and thermodynamic meteorological parameters in conjunction with MOK, are examined next. Here, lag 0 is the mean of day −1, day 0, and day +1 where day 0 is the MOK date. Similarly, lag −6 is the mean of day −7, day −6, and day −5; lag −3 that of day −4, day −3, and day −2; and so on.
Figure 5 depicts the evolution of precipitation and KE of the rotational component of low-level (850 hPa) wind, from lag −6 to lag +3. The MME performs well in reflecting the observed evolution of precipitation. It is seen that the rainfall band slowly move eastward from central AS to Kerala coast from lag −6 to lag 0, in both OBS and MME. Although the KE of rotational wind is stronger over Northern Hemisphere, particularly over the AS, in the MME than observation, its evolution is realistic. The core of the low-level wind over the Somali coast is also captured in the MME reasonably. The Somali jet speed index [defined by Boos and Emanuel (2009) as the square root of twice the spatial mean kinetic energy of 850-hPa horizontal wind over the Arabian Sea, in the spatial domain 5°S–20°N, 50°–70°E] is plotted in Fig. 6 to assess the strength of cross-equatorial flow. It is found that the winds are stronger in MME, compared to the observations. While the winds show a sharp increase almost 3–4 days prior to MOK (except in MERRA), they are sustained only after the MOK date, further illustrating the reliability of the developed criterion.
Composite evolution of rainfall (shaded; mm day−1) and kinetic energy of rotational component of 850-hPa wind (contour; m2 s−2) during MOK. Observational wind data are from MERRA.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
Temporal evolution of Somali jet speed index (m s−1) during MOK. Day 0 corresponds to the MOK date. The index is shown for all three reanalysis datasets and MME.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
With the onset of monsoon, the upper-level westerlies shift to the north of the Tibetan Plateau and the Tibetan anticyclone becomes active in concurrence with the increased heating over the Tibetan region (Flohn 1960). Further, the associated upper-level easterlies preponderate over the Indian region (Koteswaram 1960). Figure 7 portrays the composite evolution of TT and 200-hPa wind. The MME predicts these features well, albeit with a cold bias in the absolute temperatures. Xavier et al. (2007) noted that with the inception of MOK, the north–south TT gradient slowly builds up. Figure 8 displays the meridional profile of TT averaged over the Indian region 65°–95°E and the temporal evolution of its meridional gradient [defined by Xavier et al. (2007) as the difference of TT values between a northern area (5°–35°N, 40°–100°E) and a southern area (15°S–5°N, 40°–100°E)]. Although the CFS model has an intrinsic cold temperature bias (Fig. 8a), its TT gradient is comparable to (or even stronger with respect to NCEP) that of observations (Fig. 8b). Thus, it is established that the CGEPS MME is good enough to reproduce the transitions in the large-scale dynamic/thermodynamic variables, even with inherent biases.
As in Fig. 5, but for tropospheric temperature (shaded; K) and 200-hPa wind (streamlines). Observations are from MERRA.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
(a) Meridional profile of tropospheric temperature (TT; K) averaged over 65°–95°E during MOK and (b) temporal evolution of the TT gradient. Day 0 corresponds to the MOK date in Fig. 8b. Here, TT is shown for all three reanalysis datasets and MME.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
As in Fig. 5, but for Q1 that is vertically integrated from 850 to 200 hPa (shaded; W m−2) and 850-hPa wind (streamlines). Observations are from MERRA.
Citation: Journal of Climate 28, 15; 10.1175/JCLI-D-14-00842.1
The above analyses indicate that the criterion devised from MME using rainfall and low-level wind indices alone is quite successful in capturing the composite spatiotemporal evolution of various dynamic/thermodynamic parameters associated with MOK, thus supporting the robustness of the developed criterion and suitability of the CGEPS MME for MOK prediction.
4. Conclusions
MOK marks the beginning of rainy season in India and hence the extended-range prediction of the exact date of MOK is desirable for agricultural planning, dam and water management, tourism, etc. However, Kerala receives premonsoon thundershowers prior to MOK and one has to clearly distinguish the premonsoon synoptic activity from the large-scale sustained monsoon activity. The criterion devised for predicting MOK should thus be objective and must avoid “bogus” onsets.
Although there are several previous studies that proposed criteria for defining MOK, the only previous study that attempted the dynamical prediction of MOK is by Goswami and Gouda (2010). The efforts to predict MOK in real time using a dynamical frame work are still emergent. The present study is a first-of-its-kind attempt in this direction. In this study, an objective criterion was formulated using three indices defined from the large-scale wind circulation as well as rainfall, for the real-time extended-range prediction of MOK. The feasibility of the real-time prediction of MOK was explored using the CGEPS MME initialized on 16 May of every year during 2001–14. The MOK date has been identified for all individual ensemble members and the ensemble mean date is treated as the final predicted MOK date from the CGEPS MME. The MOK dates forecasted by the MME matched well with the MOK dates declared by the IMD, signifying the real-time operational effectiveness of the proposed criterion. Subsequent to the forecasted MOK, model rainfall and low-level wind are persistent in their magnitude for more than 10 days, confirming that the devised criterion effectively avoid the likelihood of bogus onsets.
Since MOK is associated with changes in the large-scale dynamical parameters as well as local moisture parameters, a model-based measure of onset makes sense only if the model reproduces the observed spatial structures and the temporal evolution of various meteorological parameters associated with the incidence of the MOK. Therefore, to examine the sturdiness of the devised criteria and to assess the competency of the CGEPS MME in MOK prediction, the composite spatial structure and the spatiotemporal evolution of various meteorological parameters associated with the forecasted MOK were compared with the observational analysis. Although the MME forecasts have realistic spatiotemporal patterns of rainfall, circulation, and temperature (Figs. 3 and 4), they exhibit a cold TT bias (Figs. 7 and 8) and stronger-than-observed winds (see Figs. 5 and 6). Despite an inherent cold TT bias, the MME predicted a realistic evolution of north–south TT gradient (Fig. 8b). The temporal evolution of various meteorological parameters (including rainfall, KE of low-level rotational wind, Q1, etc.) also appear reasonable in MME. Thus, it could be affirmed that the criterion devised from the CGEPS MME is robust and can be used for the real-time extended range prediction of MOK. It is also confirmed the CGEPS is a good choice for MOK prediction and monsoon studies, despite inherent biases. Yet, it is not to be forgotten that the RMSE in the prediction of MOK date is about 3.6 days, which is about half of the standard deviation of the onset. Also, the model prediction is within a day in only about 57% of the 14 years considered. It is anticipated that with the developmental works ongoing at IITM, under the National Monsoon Mission Project, to improve the physical processes in the CFS model, we could overcome the limitations in the current CGEPS and this would lead to a better model for monsoon prediction studies.
The criterion proposed in the present study can be used only in the dynamical prediction framework, as it necessitates input data on the future evolution of rainfall and low-level wind. Although this limits the usage of the criterion in the observational framework in real time, the future persistence is crucial for distinguishing the true onsets from bogus onsets.
Acknowledgments
IITM is fully supported by Ministry of Earth Sciences, Government of India. BEM gratefully acknowledges the financial support given by the Earth System Science Organization, Ministry of Earth Sciences, Government of India (Grant/Project MM/SERP/ Univ_Miami_USA/2013/INT-1/002) to conduct this research under Monsoon Mission. The authors gratefully acknowledge the three reviewers for their constructive and insightful comments that helped considerably in improving the manuscript. The model runs were carried out on Prithvi IBM High Performance Computing System installed at IITM, Pune. The figures in this paper were generated using GrADS and Xmgrace.
REFERENCES
Abhilash, S., S. Joseph, R. Chattopadhyay, S. Pattnaik, R. P. M. Krishna, S. De, A. K. Sahai, and B. N. Goswami, 2012: Performance of an ensemble prediction system based on CFSV2 for the extended range prediction of active-break spells of Indian summer monsoon rainfall during 2011. IITM Research Rep. RR-128, 56 pp.
Abhilash, S., A. K. Sahai, N. Borah, R. Chattopadhyay, S. Joseph, S. Sharmila, S. De, and B. N. Goswami, 2014a: Does bias correction in the forecasted SST improve the extended range prediction skill of active-break spells of Indian summer monsoon rainfall? Atmos. Sci. Lett., 15, 114–119, doi:10.1002/asl2.477.
Abhilash, S., A. K. Sahai, S. Pattnaik, B. N. Goswami, and A. Kumar, 2014b: Extended range prediction of active-break spells of Indian summer monsoon rainfall using an ensemble prediction system in NCEP Climate Forecast System. Int. J. Climatol.,34, 98–113, doi:10.1002/joc.3668.
Abhilash, S., and Coauthors, 2015: Improved spread–error relationship and probabilistic prediction from CFS-based Grand Ensemble Prediction System. J. Appl. Meteor. Climatol., 54, 1569–1578, doi:10.1175/JAMC-D-14-0200.1.
Ananthakrishnan, R., and M. K. Soman, 1988: The onset of the southwest monsoon over Kerala: 1901–1980. Int. J. Climatol., 8, 283–296, doi:10.1002/joc.3370080305.
Ananthakrishnan, R., and M. K. Soman, 1991: The onset of the southwest monsoon in 1990. Curr. Sci., 61, 447–453.
Ananthakrishnan, R., J. M. Pathan, and S. S. Aralikatti, 1983: The onset phase of the southwest monsoon. Curr. Sci., 52, 155–164.
Bansod, S. D., S. V. Singh, and R. H. Kripalani, 1991: The relationship of monsoon onset with subsequent rainfall over India. Int. J. Climatol., 11, 809–817, doi:10.1002/joc.3370110707.
Boos, W. R., and K. A. Emanuel, 2009: Annual intensification of the Somali jet in a quasi-equilibrium framework: Observational composites. Quart. J. Roy. Meteor. Soc., 135, 319–335, doi:10.1002/qj.388.
Borah, N., S. Abhilash, S. Joseph, R. Chattopadhyay, S. Sharmila, and A. K. Sahai, 2013: Development of extended range prediction system using CFSv2 and its verification. IITM Research Rep. RR-130, 62 pp.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828.
Ek, M., K. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. Tarpley, 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 108, 8851, doi:10.1029/2002JD003296.
Fasullo, J., and P. J. Webster, 2003: A hydrological definition of Indian monsoon onset and withdrawal. J. Climate, 16, 3200–3211, doi:10.1175/1520-0442(2003)016<3200a:AHDOIM>2.0.CO;2.
Flatau, M., P. Flatau, J. Schmidt, and G. Kiladis, 2003: Delayed onset of the 2002 Indian monsoon. Geophys. Res. Lett., 30, 1768, doi:10.1029/2003GL017434.
Flohn, H., 1960: Recent investigations on the mechanisms of the “summer monsoon” of southern and eastern India. Monsoons of the World, India Meteorological Department, 75–88.
Ghanekar, S. P., P. V. Puranik, and U. V. Bhide, 2003: Forecasting the onset of monsoon over Kerala using the peak in pre-monsoon convective activity over south peninsular India. Mausam, 54, 645–652.
Goswami, P., and K. C. Gouda, 2010: Evaluation of a dynamical basis for advance forecasting of the date of onset of monsoon rainfall over India. Mon. Wea. Rev., 138, 3120–3141, doi:10.1175/2010MWR2978.1.
Griffies, S. M., M. J. Harrison, R. C. Pacanowski, and A. Rosati, 2004: A technical guide to MOM4. GFDL Ocean Group Tech. Rep. 5, 342 pp.
Joseph, P. V., J. K. Eischeid, and R. J. Pyle, 1994: Interannual variability of the onset of the Indian summer monsoon and its association with atmospheric features, El Niño and sea surface temperature anomalies. J. Climate, 7, 81–105, doi:10.1175/1520-0442(1994)007<0081:IVOTOO>2.0.CO;2.
Joseph, P. V., K. P. Sooraj, and C. K. Rajan, 2006: The summer monsoon onset process over South Asia and an objective method for the date of monsoon onset over Kerala. Int. J. Climatol., 26, 1871–1893, doi:10.1002/joc.1340.
Joseph, S., and Coauthors, 2015: North Indian heavy rainfall event during June 2013: Diagnostics and extended range prediction. Climate Dyn.,44, 2049–2065, doi:10.1007/s00382-014-2291-5.
Joshi, P. C., B. Simon, and P. S. Desai, 1990: Atmospheric thermal changes over the Indian region prior to the monsoon onset as observed by satellite sounding data. Int. J. Climatol., 10, 49–56, doi:10.1002/joc.3370100106.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, doi:10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Koteswaram, P., 1960: The Asian summer monsoon and the general circulation over the tropics. Monsoons of the World, India Meteorological Department, 105–110.
Krishnamurti, T. N., P. Ardanuy, Y. Ramanathan, and R. Pasch, 1981: The onset-vortex of the summer monsoon. Mon. Wea. Rev., 109, 344–363, doi:10.1175/1520-0493(1981)109<0344:OTOVOT>2.0.CO;2.
Kumar, M. R. R., 2004: Forecasting of onset of southwest monsoon over Kerala coast using satellite data. IEEE Geosci. Remote Sens. Lett., 1, 265–267, doi:10.1109/LGRS.2004.832226.
Kung, E. C., and T. A. Sharif, 1980: Regression forecasting of the onset of the India summer monsoon with antecedent upper air conditions. J. Appl. Meteor., 19, 370–380, doi:10.1175/1520-0450(1980)019<0370:RFOTOO>2.0.CO;2.
Kung, E. C., and T. A. Sharif, 1982: Long-range forecasting of the Indian summer monsoon onset and rainfall with upper air conditions. J. Meteor. Soc. Japan, 60, 672–681.
Lu, E., and J. C. L. Chan, 1999: A unified monsoon index for south China. J. Climate, 12, 2375–2385, doi:10.1175/1520-0442(1999)012<2375:AUMIFS>2.0.CO;2.
Lu, E., X. Zeng, Z. Jiang, Y. Wang, and Q. Zhang, 2009: Precipitation and precipitable water: Their temporal-spatial behaviors and use in determining monsoon onset/retreat and monsoon regions. J. Geophys. Res., 114, D23105, doi:10.1029/2009JD012146.
Luo, H., and M. Yanai, 1984: The large-scale circulation and heat sources over the Tibetan Plateau and surrounding areas during the early summer of 1979. Part II: Heat and moisture budgets. Mon. Wea. Rev., 112, 966–989, doi:10.1175/1520-0493(1984)112<0966:TLSCAH>2.0.CO;2.
Mitra, A. K., A. K. Bohra, M. Rajeevan, and T. N. Krishnamurti, 2009: Daily Indian precipitation analysis formed from a merge of rain-gauge data with the TRMM TMPA satellite-derived rainfall estimates. J. Meteor. Soc. Japan, 87A, 265–279, doi:10.2151/jmsj.87A.265.
Murakami, T., and Y. H. Ding, 1982: Wind and temperature changes over Eurasia during the early summer of 1979. J. Meteor. Soc. Japan, 60, 183–196.
Pai, D. S., and M. Rajeevan, 2009: Summer monsoon onset over Kerala: New definition and prediction. J. Earth Syst. Sci., 118, 123–135, doi:10.1007/s12040-009-0020-y.
Parthasarathy, B., A. A. Munot, and D. R. Kothawale, 1994: Droughts over homogeneous regions of India: 1871–1990. Drought Network News, No. 6, National Drought Mitigation Center, Lincoln, NE, 17–19.
Pearce, R. P., and U. C. Mohanty, 1984: Onsets of the Asian summer monsoon 1979–82. J. Atmos. Sci., 41, 1620–1639, doi:10.1175/1520-0469(1984)041<1620:OOTASM>2.0.CO;2.
Raju, P. V. S., U. C. Mohanty, and R. Bhatla, 2007: Interannual variability of onset of the summer monsoon over India and its prediction. Nat. Hazards, 42, 287–300, doi:10.1007/s11069-006-9089-7.
Ramesh, K. J., S. Basu, and Z. N. Begum, 1996: Objective determination of onset, advancement and withdrawal of the summer monsoon using large-scale forecast fields of a global spectral model over India. Meteor. Atmos. Phys., 61, 137–151, doi:10.1007/BF01025702.
Rao, Y. P., 1976: Southwest Monsoon. India Meteorological Department, 367 pp.
Rienecker, M. M., and Coauthors, 2011: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 3624–3648, doi:10.1175/JCLI-D-11-00015.1.
Saha, S., and Coauthors, 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 2185–2208, doi:10.1175/JCLI-D-12-00823.1.
Sahai, A. K., and Coauthors, 2013: Simulation and extended range prediction of monsoon intraseasonal oscillations in NCEP CFS/GFS version 2 framework. Curr. Sci., 104, 1394–1400.
Sahai, A. K., S. Abhilash, R. Chattopadhyay, N. Borah, S. Joseph, S. Sharmila, and M. Rajeevan, 2015: High-resolution operational monsoon forecasts: An objective assessment. Climate. Dyn.,44, 3129–3140, doi:10.1007/s00382-014-2210-9.
Sikka, D. R., and S. Gadgil, 1980: On the maximum cloud zone and the ITCZ over Indian longitude during southwest monsoon. Mon. Wea. Rev., 108, 1840–1853, doi:10.1175/1520-0493(1980)108<1840:OTMCZA>2.0.CO;2.
Simon, B., and P. C. Joshi, 1994: Determination of moisture changes prior to the onset of south-west monsoon over Kerala using NOAA/TOVS satellite data. Meteor. Atmos. Phys., 53, 223–231, doi:10.1007/BF01029613.
Simon, B., S. H. Rahman, and P. C. Joshi, 2006: Conditions leading to the onset of the Indian monsoon: A satellite perspective. Meteor. Atmos. Phys.,93, 201–210, doi:10.1007/s00703-005-0155-6.
Soman, M. K., and K. Krishna Kumar, 1993: Space–time evolution of meteorological features associated with the onset of the Indian summer monsoon. Mon. Wea. Rev., 121, 1177–1194, doi:10.1175/1520-0493(1993)121<1177:STEOMF>2.0.CO;2.
Taniguchi, K., and T. Koike, 2006: Comparison of definitions of Indian summer monsoon onset: Better representation of rapid transitions of atmospheric condition. Geophys. Res. Lett., 33, L02709, doi:10.1029/2005GL024526.
Wang, B., Q. Ding, and P. V. Joseph, 2009: Objective definition of the Indian summer monsoon onset. J. Climate, 22, 3303–3316, doi:10.1175/2008JCLI2675.1.
Wu, X., K. S. Moorthi, K. Okomoto, and H.-L. Pan, 2005: Sea ice impacts on GFS forecasts at high latitudes. Eighth Conf. on Polar Meteorology and Oceanography, San Diego, CA, Amer. Meteor. Soc., 7.4. [Available online at http://ams.confex.com/ams/pdfpapers/84292.pdf.]
Xavier, P. K., C. Marzin, and B. N. Goswami, 2007: An objective definition of the Indian summer monsoon season and a new perspective on the ENSO–monsoon relationship. Quart. J. Roy. Meteor. Soc., 133, 749–764, doi:10.1002/qj.45.
Yanai, M., and T. Tomita, 1998: Seasonal and interannual variability of atmospheric heat sources and moisture sinks as determined from NCEP–NCAR reanalysis. J. Climate, 11, 463–482, doi:10.1175/1520-0442(1998)011<0463:SAIVOA>2.0.CO;2.
Yanai, M., C. Li, and Z. Song, 1992: Seasonal heating of the Tibetan Plateau and its effects on the evolution of the Asian summer monsoon. J. Meteor. Soc. Japan, 70, 319–351.
Yin, M. T., 1949: A synoptic-systematic aerological study of onset of the summer monsoon over India and Burma. J. Meteor., 6, 393–400, doi:10.1175/1520-0469(1949)006<0393:SASOTO>2.0.CO;2.
Zeng, X., and E. Lu, 2004: Globally unified monsoon onset and retreat indexes. J. Climate, 17, 2241–2248, doi:10.1175/1520-0442(2004)017<2241:GUMOAR>2.0.CO;2.
