• No results found

Prediction of monsoon using a seamless coupled modelling system

N/A
N/A
Protected

Academic year: 2023

Share "Prediction of monsoon using a seamless coupled modelling system"

Copied!
11
0
0

Loading.... (view fulltext now)

Full text

(1)

*For correspondence. (e-mail: ashis.mitra@nic.in)

Prediction of monsoon using a seamless coupled modelling system

Ashis K. Mitra

1,

*, E. N. Rajagopal

1

, G. R. Iyengar

1

, D. K. Mahapatra

1

,

I. M. Momin

1

, A. Gera

1

, K. Sharma

1

, J. P. George

1

, R. Ashrit

1

, M. Dasgupta

1

, S. Mohandas

1

, V. S. Prasad

1

, Swati Basu

1

, A. Arribas

2

, S. F. Milton

2

,

G. M. Martin

2

, D. Barker

2

and M. Martin

2

1National Centre for Medium Range Weather Forecasting, A-50, Institutional Area, Sector 62, Noida 201 309, India

2UK Met Office, FitzRoy Road, Exeter, EX1 3PB, UK

Rainfall for India and South Asian region mainly comes during the summer monsoon period. Realistic model representation of monsoon rainfall variability is a scientific challenge and is of great societal conse- quences. Due to scale interactions, study of monsoon variability and predictability from hours to a season using a realistic model is of vital importance. Under the Ministry of Earth Sciences, National Monsoon Mission, National Centre for Medium Range Weather Forecasting (NCMRWF) is focussing on the week-2 forecast of monsoon rainfall using a coupled model and its initialization. Improving the skill of models within the timescale of the first two weeks period is vital because of its seamless connection to the mon- soon intraseasonal and seasonal variability and pre- diction. Model development for monsoon rainfall has to be dealt in a holistic way, including the scales from days to a season. NCMRWF on its effort to undertake advanced research in tropical/monsoon simulation/

prediction across a range of timescales from hours to a season, has made one incremental step forward by implementing a state-of-the-art coupled model. In this study, we report results from monsoon simulations from the high-resolution global assimilation–forecast set-up in NWP medium range timescale and also from the simulation results from a relatively coarser coupled model. The monsoon lower tropospheric flow and associated rainfall patterns are quite realistic. The existing model biases in flow, rainfall and surface parameters are also discussed. To remain at the fore- front of monsoon prediction, a much higher resolution version of the coupled model and its initialization component will be implemented in the coming years on a more powerful supercomputer. The focus will be to improve simulation of coupled monsoon variability in timescale of hours to a season with an aim to improve forecast skill in week-2 time-period over the Indian region.

Keywords: Coupled model, data assimilation, monsoon modelling, rainfall variability.

Introduction

WEATHER and climate of the South Asian summer mon- soon region are a complex coupled atmosphere–ocean–

land–ice interactive phenomenon. To understand and pre- dict the monsoon and tropical weather systems, better numerical simulation models and data assimilation systems have been developed by leading modelling centres. How- ever, simulation of Indian summer monsoon rainfall vari- ability at various space and timescales by numerical modelling is still a challenging task as the model skill is lagging behind those for the mid-latitude weather sys- tems. In tropical regions, due to slowly varying boundary ocean/land conditions, potentially it is possible to simu- late/predict the variability beyond days up to a season1. However, current models have biases for monsoon/tropics, which probably are limiting the skill of rainfall prediction

beyond a few days. For India and many neighbouring countries of South Asia, the monsoon rainfall variability has strong socio-economic consequences. Therefore, there is an urgent requirement of improved understanding of the interconnections between the Indian regional mon- soon rainfall variability and other modes of weather/climate variability through enhanced observations and modelling activities2. Recently, studies suggest that human activities through industrialization and increased population have started affecting the weather/climate regimes through anthropogenic warming and are a cause of global con- cern. Due to existing uncertainties in model simulations of monsoon at shorter scales of days to a season, it is not possible to use present-day models with confidence to assess the future projections related to monsoon climate change. Seeing the challenges and utility of monsoon modelling, Ministry of Earth Sciences (MoES), New Delhi has launched the ‘National Monsoon Mission’ pro- ject during its XII Plan period to boost monsoon research in an accelerated manner by collaborating with multiple international agencies and the world scientific community3.

(2)

The whole issue of modelling the complex monsoon system from hours to a season has to be dealt with as a single entity. The evolution of weather and climate is linked by the same physical processes in the atmosphere–

ocean–land–cryosphere system operating across multiple space and timescales. It is well documented that the model biases seen at extended range (month to a season) are in fact generated within the first few days of model integration and complex feedback processes4–7. For mon- soon the errors develop within the medium-range time- scale itself, and it is more serious when the monsoon is either in active or break state8. Therefore, from model development viewpoint, it is better to use the same model across a range of temporal and spatial scales. The benefit of using a seamless prediction system across a wide range of scales is highly recommended. The World Climate Research Programme’s (WCRP’s) world modelling summit in 2008 also adopted the notion of weather and climate as a seamless problem from model development angle9–11. In a recent review article12, the practical experi- ences at UKMO in seamless prediction approach was summarized indicating real benefit for further model developments. Moving ahead with the seamless approach for monsoon forecasts is the right direction in bridging the gap between weather and climate research13.

In spite of all the latest developments in modelling and data assimilation, the skill of medium range rainfall forecast for large-scale monsoon is only up to day three to five14. Even state-of-the-art global models show large errors for longer lead times, which could be related to unrealistic rep- resentation of diurnal cycle of monsoon rain over land and ocean15. The limiting skill of medium-range monsoon rain- fall forecasts for the India (South Asia) region could be re- lated to the lack of model capability to capture the processes of the coupled monsoon system fully resolving the range of scale-interactions. Other things such as poor ocean–atmosphere coupling, missing processes, poor model resolution, etc. could equally play a role. It has been documented that during active phase of MJO, the medium-range forecast skill drops further due to lack of proper representation of the dominant convective vari- ability in the atmosphere-only model16. The background MJO scale needs to be resolved in the medium-range forecast model. Study of features related to intraseasonal oscillations of tropics and monsoons are important to un- derstand, simulate and predict the dynamical and physical processes related to tropical South Asian monsoons17,18. To better represent the air–sea interaction, we have to consider the ocean thermo-haline circulation and large- scale changes in ocean heat content to interact with the atmosphere. Additionally realistic representation of evo- lution of El Niño and Indian Ocean Dipole is required for monsoon modulation. For tropical weather systems, there is a requirement of using coupled models19.

All these studies point towards the use of coupled ocean–atmosphere model for resolving monsoon/tropics

even in medium range and up to a season in a seamless manner. From Atmospheric Model Inter-Comparison Pro- ject (AMIP) model diagnostic studies and subsequent use of coupled models, it is clear that for monsoon simulation beyond medium range (monthly/seasonal), atmosphere- only models cannot be used20–22. Coarse-resolution cou- pled models (used in AR4) also cannot simulate the nonlinear SST–rainfall relationship23. However, recent studies with high-resolution coupled models are provid- ing enough hints on the degree of sophistication we need for monsoon prediction. Current coupled models can po- tentially produce skillful prediction of Indian summer monsoon rainfall and its interannual variability24,25. The role of coupled processes to capture MJO over the Indian and Pacific Ocean regions in model simulations has been studied extensively. Improved representation of the diur- nal cycle coming from an increased upper ocean vertical resolution is a key factor in improved MJO skill26. There are other important studies suggesting improvement of in- traseasonal variability in the model due to coupling27,28. Due to better representation of the Intra-seasonal Vari- ability (ISV) in coupled models, NCEP is planning to use its latest coupled model CFSV2 for its forecasts up to week-6 having several ensemble members. ECMWF has very high-resolution coupled model for its monthly fore- casts, which can capture the ISV better than the atmo- sphere only model. In fact current operational coupled models can capture the summer intraseasonal oscillations and their northward propagation with some skill, 1–3 weeks in advance29. Latest coupled medium-range experi- ment studies at UK Met Office (UKMO) with a high resolution coupled model also suggest tropical skill improvement over the atmosphere only model30. These coupled models should have the capability to take care of the natural variability of the earth system, including the externally forced changes. In this context realistic repre- sentation of current anthropogenic source of greenhouse gases, aerosol concentrations (including volcanic) and solar irradiance are important parameters to be taken care of in the seamless model to be used from days to a season.

Many recent studies show that the representation of seasonal monsoon rainfall in coupled models is improv- ing gradually as the models become more sophisticated in terms of resolution and physics. For example, the EU ENSEMBLES models were better than the DEMETER models31. Current operational models from lead centres have further improved. The predictive nature of the sea- sonal monsoon through its teleconnection with ENSO can be captured well by improving the mean state of biases in the tropical Indo-Pacific Ocean. However, current cou- pled models have biases in their mean monsoon simula- tions. Kim et al.32 compared seasonal skill from two state-of-the-art latest global coupled models from Euro- pean Centre for Medium Range Weather Forecasting (ECMWF) and National Centre for Environmental Pre-

(3)

diction (NCEP), USA. It is noted that the current models are superior compared to their earlier versions, where they capture well the large-scale monsoon wind variabi- lity and have better skill than persistence for rainfall for Asian monsoon. The models have best skill over the tro- pics compared to higher latitudes, which is related to bet- ter simulation of ENSO and its amplitude in the earth system. Monsoon–ENSO variability is also captured well in the newer models. However, for monsoon, these cou- pled models have biases seen in SST, low-level wind and rainfall, and also it is seen that the precipitation skill is still poor. It is believed that the source of predictability in extended range (month to a season) for the South Asia region comes from the intraseasonal variability (ISO). As the ISO plays a dominant role, if they are not simulated realistically, the seasonal average rainfall coming from the simulated ISO will naturally have biases33,34.

Understanding the monsoon variability from days to a season is vital for model development and realistic simu- lation of the present-day scenario and future changes of climate state. The international scientific community un- der the coupled model inter-comparison project CMIP5 is studying the simulation skill of monsoon rainfall variabil- ity35. Simulation of intraseasonal variability remains problematic. Due to prevailing model biases for the Indian Ocean region and discrepancies among model simu- lations, the projected changes in monsoon precipitation climate still remain uncertain36. Realistic simulation of monsoon ISO, the monsoon–ENSO variability in these coupled models could be the key to simulated future changes in monsoon rainfall under climate change sce- nario. These ISO simulations and removal of model biases are closely related to the model errors originating during the first few days to weeks. Therefore, it is more relevant to study the model biases in a seamless frame- work, and realistic simulation of monsoon ISO is the key and remains the most challenging task for the modelling community.

In this study, we report recent results from the mon- soon simulations at both medium-range and seasonal timescales and discuss the nature of its biases. These results are from the unified seamless modelling system that has been recently implemented at NCMRWF. The medium-range forecast biases are from a high-resolution atmosphere-only model, whereas the coupled model simulations are from a relatively coarser resolution.

Using this advanced assimilation-modelling system, the future plans are also discussed.

Modelling system

The monsoon system involves multi-scale interaction among atmosphere, land surface, ocean, cryosphere and biosphere. Ideally, the model should be able to represent these processes and the model resolution should be good

enough to allow realistic interaction among the said com- ponents. At NCMRWF the unified modelling system of UKMO is further being developed jointly for monsoon over different space and timescales (days to a season).

Currently, for medium-range forecasts, a global high resolution (25 km at mid-latitudes) having 70 layers in the vertical with 4D variational atmospheric data assimi- lation has been set up. A detailed description of the implementation of the ‘NCMRWF Unified Model’

(NCUM) system can be found in (ref. 37). The grid point atmospheric model has non-hydrostatic dynamics which can handle all scales of motion. It has all the latest physi- cal parameterization schemes taking care of surface exchanges, non-local boundary layer mixing, mass flux- type convection, interactive cloud-radiation, orography based on high-resolution elevation data and gravity wave drag. The 4D variational data assimilation system uses all global conventional and satellite data. A vast amount of satellite radiance data from different sensors is assimi- lated as direct radiance. This unified atmospheric model can easily be configured to a higher resolution regional nested model over the area of interest. At NCMRWF, a 12 km version of regional Unified Model (UM) has been set up nested within the global model. A further very high resolution model of 4 km was also tested for Delhi region nested within the Indian domain 12 km regional UM (Figure 1). These very high resolution models are suitable for realistic representation of local-scale heavy rainfall38. It has the potential to be configured as a convection- permitting model at 1.5 km model resolution for mon- soon.

Figure 1. Domain of configurations of NCMRWF Unified Model.

(4)

The unified modelling framework software can easily incorporate other modules like ocean, sea-ice, land- surface, etc. into the full modelling system desired. The current NCMRWF coupled system configuration consists of a global coupled ocean–atmosphere general circulation model to calculate the evolution of ocean and atmosphere simultaneously. The NCMRWF coupled modelling system is being jointly developed for monsoon prediction with UKMO. The core model is the latest coupled UKMO HadGEM3AO system39,40. It comprises of a range of spe- cific model configurations incorporating different levels of complexity, but with a common physical framework.

The HadGEM3AO family includes a coupled configura- tion, with or without a vertical extension in the atmos- phere to include a well-resolved stratosphere, and an earth-system configuration which includes dynamic vege- tation, ocean biology and atmospheric chemistry. The atmosphere, ocean and sea-ice are coupled using a version of the OASIS coupler developed at CERFACS in France. This coupler is used to interpolate between oce- anic and atmospheric grids with a coupling interval of 3 h, which allows some resolution of diurnal cycle. A Gaussian method is used for interpolation in both direc- tions, which automatically accounts for the coastline of the two models. In the NCMRWF coupled configuration, the atmosphere horizontal resolution is roughly 1.875° × 1.25° (roughly 140 km at mid-latitudes) and 85 layers in the vertical (50 levels are below 18 km). In the NCMRWF coupled configuration, the ocean model used is Nucleus for European Modelling of the Ocean (NEMO) in the ORCA tripolar grid configuration, which has 1° × 1° horizontal resolution in mid-latitudes and en- hanced meridional resolution near the equator (0.33° at the equator). In the vertical it has 75 layers resolving the upper ocean in a very fine way (1 m resolution near the ocean surface). NEMO is a community ocean modelling framework owned and maintained by a consortium of in- stitutions41,42. NEMO has a large user community and is used for research and operational applications. For sea- ice simulation in the coupled system, CICE, the Los Alamos National Laboratory (LLNL, USA) sea-ice model is used43. A realistic sea-ice model and its proper initiali- zation will help in capturing the recently noted variability of Arctic sea-ice as seen in 2012 summer. In the set-up for simulating up to a season, all climate forcings (aero- sols, methane, CO2 concentration, etc.) are set to observed values for the period 1960–2000 and follow the emissions scenario A1B afterwards. Ozone is fixed to observed cli- matological values and includes a seasonal cycle. In the coupled configuration the atmospheric model component has 85 levels in the vertical, with the model top height is around 85 km called as hi-top system. This hi-top model feature is useful to capture the sudden stratospheric warming (SSW) and quasi-biennial (QBO) features and their teleconnections with other major climate signals44. Due to better representation of weather conditions in the

stratosphere, the higher parts of atmosphere, the SSW is captured more realistically, including its propagation to the troposphere, and hence the extreme cold surface con- ditions in parts of Europe are simulated better. Realistic simulation of QBO helps reduce tropical biases in the model. In addition to improved stratosphere representa- tion, this model enables better troposphere–stratosphere interaction, polar jet, which in turn reduces spin-up prob- lems in the stratosphere and tropopause. This hi-top model will simulate the linkage (teleconnections) between monsoon and mid/high-latitude weather systems.

Model skill for monsoon

First we see the mean monsoon simulation and biases during the 2012 season in the medium-range NWP set-up.

Figure 2 shows the 850 hPa wind for Indian monsoon re- gion, and its bias compared to the analysis. The analysis is from the 4D variational data assimilation system of the NCUM implemented at NCMRWF. The features of the low-level monsoon circulation like cross-equatorial flow, low-level westerly Somali jet and monsoon trough are brought out well by the analysis–forecast system. How- ever, around the equator and eastern equatorial Indian

Figure 2. Mean monsoon 850 hPa winds (mps) during 2012 season (JJAS) and its bias from NCMRWF Unified Model Numerical Weather Prediction global set-up.

(5)

Figure 3. Mean monsoon rainfall (cm/day) during 2012 season (JJA) from NCMRWF Unified Model Numerical Weather Prediction global set-up; observations, day-1, day-3 and day-5 forecasts.

Ocean, the model has a tendency to strengthen the east- erly flow feeding into the cross-equatorial winds in the western equatorial Indian Ocean. Part of this easterly bias causes to weaken the monsoon flow in the western equa- torial Indian Ocean (north of the equator). Another west- erly bias is seen in the northern Arabian Sea, which ultimately strengthens the monsoon over northwest India and might supply more moisture to the monsoon trough region and the Himalayan foothills. The monsoon trough flow is also seen to be stronger compared to the analysis.

This will have an impact on the systematic errors in mon- soon trough region rainfall. Another noticeable wind bias is over the Bay of Bengal region, trying to weaken the monsoon flow, which will affect the rainfall bias in the northern part of the Bay of Bengal. Figure 3 shows the NCUM model rainfall simulations at day-1, day-3 and day-5, taken from real-time runs at NCMRWF. These model rainfalls are compared with the satellite and gauge- merged daily gridded rainfall data jointly prepared by IMD and NCMRWF45. Compared to the observed rainfall features, the model simulations up to day-5 look quite good. The heavy rain zone of the west coast and eastern

part of the Bay of Bengal are comparing well. The mon- soon trough and the rain-shadow zone in southeast part of India are also simulated well. However, by day-5 we start noticing some biases. One excessive rainfall band near the equatorial India Ocean (70°E, 5°S) is quite notice- able. Figure 4 shows the bias (forecast minus observa- tion) of NCUM rainfall from day-1 through day-5. The said equatorial Indian Ocean positive bias is the most dominant one. This positive rain bias is also seen in the southern part of the Bay of Bengal. Most of the peninsu- lar Indian land region (south of the monsoon trough) has a dry bias. The monsoon trough and the region north of the monsoon trough (Himalayan foothills) have wet bias.

These are the model biases in very early days of the model forecast (up to day-5), which represent the case of monsoon 2012. Though the model biases are similar for other recent years, we have to analyse more years in the context of initialized NWP runs.

Mean monsoon from model simulations, also known as model climate, shows the capability of the model to cap- ture the complex processes of monsoon. In the model the simulated variability is also related to the mean monsoon.

(6)

In a recent study of seven coupled model hindcast data, it was seen that the models that more closely replicate the observed climatological mean tend to have better skill in seasonal forecasting25. In a recent study46 the importance of model climate for issuing a real-time forecast to users has been shown. The skill of capturing seasonal monsoon in two state-of-the-art coupled models from ECMWF and NCEP is discussed in the detailed diagnostic work by Kim et al.32. Next, we examine the model output from

Figure 4. Mean rainfall bias (forecast–observation) of NCMRWF Unified Model during monsoon 2012 season (JJA; in mm/day).

coupled NCUM runs. Here the start date is 9 May, and we did hindcast runs in NCMRWF HPC system from 1996 to 2009. The coupled NCUM was integrated for six months stretching up to 31 October for each of the above- mentioned years. The JJA mean features were computed from these hindcast runs, and were compared with Euro- pean Reanalysis (ERA) reanalyses data. ERA and Xie–

Arkin monthly rainfall were used as observed gridded data for model comparison. The seasonal mean monsoon features of the atmosphere and oceans as simulated in the coupled model are discussed now. The first-order ocean diagnostics like the sea surface temperature (SST) and sea surface salinity (SSS) are examined. Figures 5 and 6 show the global coupled model simulated SST and SSS for a monsoon representative month of July for the whole hindcast period, and their comparison with mean ob- served features respectively. Reynolds SST and WOCE data for SSS are used here as observations. In a broader sense, the SST simulation for July looks good, with small biases. The SSS simulations also look good compared to observations. However, in the Indian Ocean the gradients of the SSS are sharper compared to the observations.

Now we examine the mean monsoon atmospheric con- ditions simulated by coupled NCUM from the said hind- cast dataset. Figure 7 shows the coupled model simulated mean sea-level pressure (MSLP) for monsoon period along with observations from ERA. While plotting, to avoid values near high mountain regions, a data mask is used to remove values at heights higher than 925 hPa of pressure level. The monsoon trough region (lower values of MSLP) over central India extending to the northern Bay of Bengal region is simulated well in the coupled model. However, over the north Indian region the model produces a stronger monsoon trough, as seen in the bias plot. From this we note that the model simulates a stronger monsoon compared to ERA observations. This will have an impact on model rainfall bias. A wet bias over northwest India, including the Pakistan region is expected, which was seen in the NWP medium-range model rainfall simulation bias in Figure 4. Similar wet bias in rainfall is also expected from the coupled model, which exhibits deep monsoon trough.

Figure 8 shows the coupled model-simulated near- surface temperature (T2m) and its comparison with ERA observations, and the model bias. The mean surface tem- perature is simulated quite well capturing the warmer northwest India region (desert area), matching well with observations. Another warm, dry region near Chennai (east coast of India) is also captured well. The tempera- ture biases are very less. Figures 9 and 10 show the coupled model-simulated low-level (850 hPa) and upper- level (200 hPa) winds along with ERA observations and the respective biases. The low-level wind flow is the heart of the monsoon system, strongly linked to the moisture convergence and rainfall distribution. The cross- equatorial flow, low-level westerly Somali jet and monsoon

(7)

Figure 5. Mean July sea surface temperature (°C) from 14 seasons’ hindcast of coupled NCMRWF Unified Model and comparison with observations.

Figure 6. Mean July sea surface salinity (psu units) from 14 seasons’ hindcast of coupled NCMRWF Unified Model and comparison with observations.

trough are captured well in the coupled model, as seen in the 850 hPa flow pattern. At 200 hPa level (Figure 10), the mid-latitude westerly system, the Tibetan anti- cyclone and the tropical easterly jet are matching well with ERA observation patterns. The core of the tropical easterly jet is seen to extend across the African continent.

However, the low-level flow in the coupled model is stronger compared to observations. The bias plot brings out clearly the intensification of low-level flow over the Arabian Sea and convergence around the west coast of India. This Arabian Sea flow intensification gets a feed also from the easterly winds in the east equatorial Indian

(8)

Figure 7. Mean monsoon (JJA) MSLP (hPa) from 14 seasons’ hindcast of coupled NCMRWF Unified Model and com- parison with observations.

Figure 8. Mean monsoon (JJA) near-surface temperature T2m (°C) from 14 seasons’ hindcast of coupled NCMRWF Unified Model and comparison with observations.

Figure 9. Mean monsoon (JJA) 850 hPa winds (mps) from 14 seasons’ hindcast of coupled NCMRWF Unified Model and comparison with observations.

Ocean region. Over the Bay of Bengal there is a north- westerly wind bias, weakening the monsoon there. These low-level wind biases are similar to the model errors discussed in the NWP context in the medium range. The stronger westerly over the Bay of Bengal region forms a cyclonic wind bias around the eastern part of the Bay (Arrakan coast), which will result in excess moisture convergence and associated rainfall. The coupled model- simulated mean monsoon rainfall, corresponding to observations from CMAP rain47 and the model bias are shown in Figure 11. The distribution of monsoon rainfall

from the coupled model looks quite realistic. The observed rainfall maxima over the west coast, monsoon trough, northeastern Bay of Bengal and the rainshadow region of southeast India are brought out well. The rain- fall distribution shows that the coupled model has a good representation of monsoon, which is consistent with model-simulated MSLP and low-level flow patterns. The biases in the mean monsoon coupled model simulations are consistent with biases in low-level flow and MSLP distribution. The equatorial rainfall bias might be related to the equatorial low-level wind bias. The wet bias over

(9)

Figure 10. Mean monsoon (JJA) 200 hPa winds (mps) from 14 seasons’ hindcast of coupled NCMRWF Unified Model and comparison with observations.

Figure 11. Mean monsoon (JJA) rainfall (mm/day) from 14 seasons’ hindcast of coupled NCMRWF Unified Model and comparison with observations.

the west coast and monsoon trough region is related to the stronger monsoon flow in the model. Another wet- bias region is the northeast part of the Bay of Bengal (Myanmar coast region), which is again related to the cyclonic low-level wind bias there. Over peninsular India (south of the monsoon trough), there is a slight dry bias region. Interestingly, these rainfall mean biases in the coupled model are mostly similar to the day-5 forecast rainfall bias in the medium range coming from high- resolution global NWP set-up for monsoon 2012. Of course, we have to examine the mean rainfall bias in the NWP context by taking model output from more mon- soon seasons. From this analysis of model output, it is not possible to say more about the location and intensity of the mean monsoon rain biases. In a recent study it is shown that in the coupled system, the cold bias over the Arabian Sea region could be a possible reason for less monsoon rainfall over India region48. A more in-depth study and much more experimentation with this coupled model will be required to understand the source of error and its growth from days to season.

Way ahead

To further refine the coupled models, improving the MJO/MISO simulations will be the main focus of research in the coming years at least for monsoon and the Indian Ocean region. During 2013, it is planned to increase the horizontal resolution of atmospheric component to around 85 km and the ocean component to 25 km. Mov-

ing towards higher model resolution, particularly for monsoon/tropics is important, as it will help resolve sub- synoptic and mesoscale features49. For capturing the South Asia land precipitation during monsoon, very high- resolution coupled model is required with high-capacity computing machines50,51. With coupled model, realistic representation of MJO variations especially over the Indian Ocean region also need further research52. Depict- ing its propagation and traversing the Maritime Continent into the western Pacific is a challenging task. Improved initial conditions of ocean–atmosphere system are of great importance in this context. At NCMRWF, the ocean initialization scheme, NEMOVar, compatible to the ocean model is being implemented. This scheme will use all available in situ temperature and salinity data, surface forcing from UM short-range atmospheric forecasts, SST analyses, sea-ice concentration/extent estimates and satel- lite altimetry measurements to create the best estimate of ocean state. NEMOvar is an advanced ocean data assimi- lation system, which was made operational recently at ECMWF for its coupled forecast system. This will make use of the upcoming altimeter data from SARAL/Altika.

MoES is also enhancing Indian Ocean observations gradually to better define the initial ocean state in cou- pled models.

The global NCUM will be upgraded to 15 km horizon- tal resolution. The nested regional UM for India domain will also be upgraded to 8 km. NCMRWF has completed experiments with surface wind data from Indian satellite Oceansat-II, and the same is being put into operational

(10)

UM runs now. Experiments are on to include Megha- Tropiques data into the system. The upcoming INSAT- 3D satellite radiance data will also be assimilated to the UM 4D variational system. With mesoscale observing network from India Meteorological Department (IMD), a regional UM-based 4D variational data assimilation scheme will be made operational at NCMRWF. To verify the model simulations there is a requirement of high- resolution gridded daily and even hourly rainfall datasets.

High-resolution realistic observed gridded daily dataset from reanalysis products is required for model validation and development. Daily rainfall dataset will be useful to study the space and timescales of the northward- propagating convective rainfall bands during summer53. Uncertainty of representing tropical intraseasonal rainfall variability among various reanalysis datasets is a cause of worry54,55. It is desirable that a high-resolution 4D varia- tional data assimilation system be used where observed rainfall, moisture and wind information could be assimi- lated specially for the monsoon region to create a reanalysis dataset. NCMRWF has initiated this joint work.

In the tropics/monsoon, there are some uncertainties in forecasts because the observations with which the models are initialized are incomplete, and the models are also not perfect. These uncertainties will be taken into account by making an ensemble of forecasts, intended to cover the range of possible outcomes consistent with imperfections in the initialization and modelling systems. In addition to estimating uncertainties, these ensembles can also be used to provide probabilistic information that can be extremely valuable in many areas of weather/climate applications. One way to cater to the model uncertainties will be by representing them through the use of stochastic physics schemes, effective as seen from current research.

Both for the medium-range NWP and coupled system ensemble runs of several members will be set-up next year with the availability of more computing resources.

Land-surface processes provide significant feedback to monsoon rainfall variability, and hence a realistic detailed land-surface model along with its proper initialization is an import factor in simulation/prediction of monsoon rainfall in different timescales, including implication on monsoon climate change studies56. In future, JULES advanced land-surface model and the corresponding land- surface data assimilation scheme will be used in the NCMRWF coupled modelling system for the seamless prediction system development57. This will enable dynamic evolution of land-surface and vegetation parameters interacting and modulating the monsoon flow.

There is still much more research work to be done in this developing area of tropical/monsoon meteorology. After the onset of monsoon near the southern tip of India, northwestward propagation of rainfall band in June is sensitive to soil moisture and associated feedback proc- esses, for which a realistic land-surface scheme (and cor- responding initialization) has to be integrated to the

modelling system58. Due to human activities the land use, greenhouse gas emission and aerosols have to be consid- ered seriously in the modelling systems. Seeing the com- plexities of the earth system modelling, where one has to introduce the impact of living organisms, human responses and chemical reactions, it is possible that a new genera- tion of models will be needed to handle the multiscale nature of the climate system59.

1. Charney, J. G. and Shukla, J., Predictability of monsoons. In Monsoon Dynamics (eds Lighthill, J. and Perace, R. P.), Cam- bridge University Press, Cambridge, 1981, pp. 99–109.

2. Ghelli, A., Cloke, H. and Kulkarni, A., Monsoons: prediction, variability and impact. Meteorol. Appl. (Spec. Issue), 2012, 19.

3. MoES National Monsoon Mission, 2012; http://moes.gov.in/

monsoon_mission.pdf

4. Klein, S. A., Jiang, X., Boyle, J., Malyshev, S. and Xie, S., Dia- gnosis of the summertime warm and dry bias over the U.S. Southern Great Plains in the GFDL climate model using a weather forecast- ing approach. Geophys. Res. Lett., 2006, 33, L18805; doi:

10.1029/2006GL027567.

5. Rodwell, M. J. and Palmer, T. N., Using NWP to assess climate models. QJRMS, 2007, 133, 129–146.

6. Song, S. and Mapes, B., Interpretations of systematic errors in the NCEP CFS at lead times of 2, 4, 8,..., 256 days. J. Adv. Modell.

Earth Syst., 2012, 4, 11, M09002.

7. Martin, G. M., Milton, S. F., Senior, C. A., Brooks, M. E., Ineson, S., Reichler, T. and Kim, J., Analysis and reduction of systematic errors through a seamless approach to modelling weather and climate. J. Climate, 2010, 23, 5933–5957.

8. Rai, S. and Krishnamurthy, V., Error growth in climate forecast system daily retrospective forecasts of South Asian monsoon.

JGR, 2011, 116, D03108; doi: 10.1029/2010JD014840.

9. Brunet, G. et al., Collaboration of the weather and climate com- munities to advance subseasonal-to-seasonal prediction. Bull. Am.

Meteorol. Soc., 2010, 91, 1397–1406.

10. Shukla, J. et al., Revolution in climate prediction is both necessary and possible: a declaration at the world modelling summit. Bull.

Am. Meteorol. Soc. (AMS), 2009, 90, 175–178.

11. Hurrell, J. et al., A unified modelling approach to climate system prediction. Bull. Am. Meteorol. Soc., 2009, 90, 1819–1831.

12. Brown, A., Milton, S., Cullen, M., Golding, B., Mitchell, J. and Shelly, A., Unified modelling and prediction of weather and climate: a 25-year journey. Bull. Am. Meteorol. Soc., 2012, 93, 1865–1877.

13. Chang, Chih-Pei, Lau, Ngar-Cheung, Johnson, R. H. and Jiao, M., Bridging weather and climate in research and forecasts of the global monsoon system. Bull. Am. Meteorol. Soc., 2011, 92, 369–373.

14. Mitra, A. K., Iyengar, G. R., Durai, V. R., Sanjay, J., Krishnamurti, T. N., Mishra, A. and Sikka, D. R., Experimental real-time multi- model ensemble (MME) prediction of rainfall during monsoon 2008:

large-scale medium-range aspects. JESS, 2011, 120, 27–52.

15. Chakraborty, A., The skill of ECMWF MRF during YOTC 2008.

Mon. Weather Rev., 2010, 138, 3787–3805.

16. Hendon, H. H., Liebmann, B., Newman, M. and Glick, J. D., MRF errors associated with active episodes of the MJO. Mon. Weather Rev., 2000, 128, 69–86.

17. Goswami, B. N., South Asian monsoon. In Intra-Seasonal Vari- ability in the Atmosphere–Ocean Climate System (eds Lau, W. K.

M. and Waliser, D. E.), Springer-Praxis, 2005, chap. 2, pp. 19–61.

18. Xavier, P., Intraseasonal convective moistening in CMIP3 models.

J. Climate, 2012, 25, 2569–2577.

19. Sandery, P. A., Brassington, G. B., Craig, A. and Pugh, T., Impacts of ocean–atmosphere coupling on TC intensity change and ocean prediction in the Australia region. Mon. Weather Rev., 2010, 138, 2074–2091.

(11)

20. Gadgil, S. and Sajani, S., Monsoon precipitation in the AMIP runs. Climate Dyn., 1998, 14, 659–689.

21. Nanjundiah, R. S., Vidyunmala, V. and Srinivasan, J., On the difference in the seasonal cycle of rainfall over India in the Com- munity Climate System Model (CCSM2) and Community Atmos- pheric Model (CAM2). GRL, 2005, 32, L20720, doi:10.1029/

2005GL024278.

22. Mitra, A. K., Stefanova, L., Vijaya Kumar, T. S. V. and Krishna- murti, T. N., Seasonal prediction for the Indian monsoon region with FSU ocean–atmosphere coupled model: model mean and 2002 anomalous drought. Pure Appl. Geophys., 2005, 162, 1431–1454.

23. Rajendran, K., Nanjundiah, R. S., Gadgil, S. and Srinivasan, J., How good are the simulations of tropical SST-rainfall relationship by IPCC AR4 atmospheric and coupled model? JESS, 2012, 121, 595–610.

24. DelSole, T. and Shukla, J., Model fidelity versus skill in a sea- sonal forecasting. J. Climate, 2010, 23, 4794–4806.

25. Gadgil, S. and Srinivasan, J., Monsoon prediction: are dynamical models getting better than statistical models? Curr. Sci., 2012, 103, 257–259.

26. Woolnough, S. J., Vitart, F. and Balmaseda, M. A., The role of the ocean in the MJO: implications for MJO prediction. QJRMS, 2007, 133, 117–128.

27. Lin, J. L. et al., Sub-seasonal variability associated with the Asian monsoon simulation by 14 IPCC AR4 coupled GCMs. J. Climate, 2008, 21, 4541–4567.

28. Sperber, K. R. and Annamalai, H., Coupled model simulation of boreal summer intraseasonal (30–50 day) variability: Part 1, Sys- tematic error and caution on use of metrics. Climate Dyn., 2008, 31, 345–372.

29. Fu, J., Lee, J., Wang, B. and Vitart, F., Intraseasonal forecasting of Asian summer monsoon in four operational and research models.

J. Climate, 2012 (in press); doi: 10.1175/JCLI-D-12-00252.1.

30. Johns, T. et al., Report on extensive coupled ocean–atmosphere trials on NWP (1–15 day) timescales. UKMO PWS Key Delivery Report, UKMO, Exeter, UK, 2012, p. 54.

31. Rajeevan, M., Unikrishnan, C. K. and Preethi, B., Evaluation of the ENSEMBLES multi-model seasonal forecasts of Indian sum- mer monsoon variability. Climate Dyn., 2012, 38, 2257–2274.

32. Kim, H.-M., Webster, P. J., Curry, J. A. and Toma, V. E., Asian summer monsoon prediction in ECMWF system 4 and NCEP CFSV2 retrospective seasonal forecasts. Climate Dyn., 2012, 39, 2975–2991.

33. Krishnamurty, V. and Shukla, J., Intraseaosnal and interannual variability of rainfall over India. J. Climate, 2000, 13, 4366–4377.

34. Krishnamurthy, V. and Shukla, J., Intraseasonal and seasonally persisting patterns of Indian monsoon rainfall. J. Climate, 2007, 20, 3–20.

35. Sperber, K. R. et al., The Asian summer monsoon: an intercom- parison of CMIP5 vs CMIP3 simulations of the late 20th century.

Climate Dyn., 2012 (accepted).

36. Turner, A. and Annamalai, H., Climate change and the South Asian summer monsoon. Nature Climate Change, 24 June 2012, 2012, p. 9; doi: 10.1038/NCLIMATE1495.

37. Rajagopal, E. N. et al., Implementation of the UM model based analysis–forecast system at NCMRWF. NCMRWF Technical Report No. NMRF/TR/2012, 2012, p. 45, available from the NCMRWF website.

38. Kendon, E. J., Roberts, N. M., Senior, C. A. and Roberts, M. J., Realism of rainfall in a very high-resolution RCM. J. Climate, 2012, 25, 5791–5806.

39. Hewitt, H. T. et al., Design and implementation of the infrastruc- ture of HadGEM3: the next-generation Met Office climate model- ling system. Geosci. Model Dev., 2011, 4, 223–253.

40. Arribas, A. et al., The GloSea4 ensemble prediction system for seasonal forecasting. Mon. Weather Rev., 2011, 139, 1891–1910.

41. Madec, G. NEMO ocean engine, Note du Pole de modelisation de l’Institut Pierre-Simon Laplace No. 27; 2008; http://www.nemo- ocean.eu/About-NEMO/Reference-manuals

42. Madec, G. and Imbard, M., A global ocean mesh to overcome the North Pole singularity. Climate Dyn., 1996, 12, 381–388.

43. Hunke, E. C. and Lipscomb, W. H., CICE: the Los Alamos Sea Ice Model Documentation and Software User’s Manual, Version 4.1, No.LA-CC-06-012, 2010, p. 76, Available from Los Alamos National Laboratory, USA (http://oceans11.lanl.gov/trac/CICE).

44. Fereday, D. R., Maidens, A., Arribas, A., Scaife, A. A. and Knight, J. R., Seasonal forecasts of northern hemisphere winter 2009/10.

Environ. Res. Lett., 2012, 7, doi:10.1088/1748-9326/7/3/034031.

45. Mitra, A. K., Bohra, A. K., Rajeevan, M. N. and Krishnamurti, T. N., Daily Indian precipitation analyses formed from a merged of rain-gauge with TRMM TMPA satellite-derived rainfall esti- mates. J. Meteorol. Soc. Jpn. A, 2009, 87, 265–279.

46. Wallace, E. and Arribas, A., Forecasting with reference to a specific climatology. MWR, 2012, 140, 3795–3802.

47. Xie, P. and Arkin, P. A., Global precipitation: a 17-year monthly analysis based on observations, satellite estimates, and numerical model outputs. BAMS, 1997, 78, 2539–2558.

48. Levine, R. C. and Turner, A. G., Dependence of Indian monsoon rainfall on moisture fluxes across the Arabian Sea and the impact of coupled model SST biases. Climate Dyn., 2012, 38, 2167–2190.

49. Jung, T. et al., High-resolution global climate simulations with ECMWF model in Project Athena: experimental design, model climate and seasonal forecast skill. J. Climate, 2012, 25, 3155–

3172.

50. Wen, M. et al., Impacts of model resolutions and initial conditions on prediction of Asian summer monsoon by NCEP CFS. WAF (AMS), 2012, 27, 629–646.

51. Shukla, J. et al., Toward a new generation of world climate research and computing facilities. BAMS (AMS), 2010, 91, 1407–

1412.

52. Weaver, S. J., Wang, W., Chen, M. and Kumar, A., Representation of MJO variability in the NECP CFS. J. Climate, 2011, 24, 4676–

4694.

53. Chakraborty, A. and Nanjundiah, R. S., Space-time scales of northward propagation of convection during boreal summer. Mon.

Weather Rev., 2012, 140, 3857–3866.

54. Mishra, V. et al., A comparative study of the Indian summer mon- soon hydroclimate and its variations in three reanalyses. Climate Dyn., 2012, 20, doi:10.1007/s00382-012-1319-y.

55. Wang, J., Wang, W., Fu, X. and Seo, K. H., Tropical intraseasonal rainfall variability in the CFSR. Climate Dyn., 2012 (in press), doi: 10.1007/s00382-011-1087-0.

56. Martin, G. M. and Levine, R. C., The influence of dynamic vege- tation on the present-day simulation and future projections of the South Asian summer monsoon in the HadGEM2 family. Earth Syst. Dyn., 2012, 3, 245–261.

57. Best, M. J. et al., The Joint UK Land Environment Simulator (JULES), model description – Part 1: Energy and water fluxes.

Geosci. Model Dev., 2011, 4, 595–640.

58. Krishnamurti, T. N. et al., Modelling of forecast sensitivity on the march of monsoon isochrones from Kerala to New Delhi: the first 25 days. JAS, 2012, 69, 2465–2487.

59. Slingo, J. et al., Developing the next-generation climate system models: challenges and achievements. Philos. Trans. R. Soc. Lon- don, Ser. A, 2009, 367, 815–831.

ACKNOWLEDGEMENTS. We thank S. D. R. Sikka, Chairman, RAC NCMRWF for useful discussions and guidance on coupled modelling project at NCMRWF. We also thank Dr Shailesh Nayak, Secretary MoES, for his encouragement for the NCMRWF coupled modelling and ocean initialization project.

References

Related documents

1.2 North-East Monsoon: Normally, in Telangana State only 14% of annual rainfall is through North- East Monsoon and the average normal rainfall of North-East Monsoon is 129

While Greenpeace Southeast Asia welcomes the company’s commitment to return to 100% FAD free by the end 2020, we recommend that the company put in place a strong procurement

Active and break events are identified from the average rainfall data over a critical area, called the core monsoon zone within which the monsoon trough/ Continental

The seasons of 1997 and 1983 for which all models generated false alarms and also the excess monsoon season of 1994, for which only two models predicted a substantive posi-

In this study, we investigate the nature of the SST–rainfall relationship simulated by AGCMs and the coupled versions (CGCMs) of nine global climate models used in the Fourth

3.6., which is a Smith Predictor based NCS (SPNCS). The plant model is considered in the minor feedback loop with a virtual time delay to compensate for networked induced

The expansion of flood forecasting activity was done using the rainfall runoff model by developing MIKE 11 model using the daily rainfall received through India Meteorological

The matter has been reviewed by Pension Division and keeping in line with RBI instructions, it has been decided that all field offices may send the monthly BRS to banks in such a