Reconstruction and Downscaling of the South Asian Monsoon Extended Range forecast on regional scale from NCEP-CFSv2 operational forecast output through detection of monsoon intraseasonal oscillations using Self Organizing Map

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 7 January 2015: 10:45 AM
123 (Phoenix Convention Center - West and North Buildings)
Rajib Chattopadhyay, Indian Institute of Tropical Meteorology, Pune, Maharashtra, India; and A. Sahai, S. Joseph, and S. Abhilash

A recent study by Sahai et al.,(2014) has developed a method to identify and track the Monsoon Intraseasonal oscillations (MISOs) using Self Organizing Maps or SOM algorithm (Kohenen 1997). This method of post-processing is shown to be equivalent to other standard techniques based on empirical orthogonal functions (EOF analysis) routinely used to monitor or track MISOs or Madden-Julian Oscillations (MJO) in real-time applications (Suhas et al. 2012; Wheeler and Hendon 2004). We show in this study that the regional biases in the extended range forecasts (i.e. forecast up to 15-20 days lead time) from Climate Forecast System (CFS) model, version 2 developed at National Centre for Environmental Prediction (NCEP) US, can be corrected up to a significant extent based on the identification of MISO phase propagation from the model forecast output which has less biases on the large scale. The NCEP CFSv2 coupled model shows realistic northward phase propagation of MISO as well as realistic forecast of the large scale MISO defined through the formulation of large scale indices to capture MISO from precipitation as shown in Sahai et al. (2014). However, the spatial pattern of the forecasted rainfall at large lead time is subjected to biases due to errors in smaller scales at large lead time (i.e. beyond weather scales). It is seen that most of the precipitation bias in the NCEP-CFSv2 forecast is associated with regional scales where the large scale low frequency signal is affected due to small scale errors. We employ a SOM based technique to correct this bias with an aim to improve the phase correlation between the observed and predicted pattern. The SOM based post processing and bias correction has two steps: (a) to identify the MISO from forecast and observation in a standard [MISO1, MISO2] phase space, where MISO1 and MISO2 are the first two principal components of the large scale patterns derived through SOM technique [Sahai et al., 2014] and (b) correct the bias through a statistical mapping of rainfall for each phases of MISO as defined in a [MISO1, MISO2] phase space. The statistical mapping technique is developed based on the position coordinate of the forecasted MISO indices in the [MISO1, MISO2] phase space. If we assume that the [MISO1, MISO2] phase space is divided into 8 equal octants and each octant has an angular span of 45, we can reconstruct the composite lifecycle of MISO based on the history of points clustered in each phase that is derived from observation. The life cycle of MISO is assumed to be represented by the eight spatial maps defined through the composite of rainfall for the days clustered in the each of the 8 octants in the phase space. Based on these composite spatial maps from observation and the angular location of the forecasted MISO indices, the spatial pattern of forecasted rainfall is derived through an angular interpolation technique. The forecasted rainfall is shown to be retaining the large scale signal at every grid point and the biases due to unphysical errors are significantly reduced. The most important advantage of this method is that the derivation of forecasted rainfall is dependent on observed composite pattern (in addition to the forecasted MISO indices). Thus if the observation has very high resolution data, the large scale MISO indices obtained from a low resolution forecast can be used to reconstruct a high resolution and bias corrected forecast representing the intraseasonal patterns over large scale. Thus we achieve both reconstruction of large scale forecasted pattern as well as downscaling using SOM.

References: Kohenen, T., 1997: Self Organizing Maps. Springer Berlin / Heidelberg. Sahai, A. K., R. Chattopadhyay, S. Joseph, S. Abhilash, N. Borah, and B. N. Goswami, 2014: A new method to compute the principal components from self-organizing maps: an application to monsoon intraseasonal oscillations. Int. J. Climatol., 34, 29252939. (Accessed July 31, 2014). Suhas, E., J. Neena, and B. Goswami, 2012: An Indian monsoon intraseasonal oscillations (MISO) index for real time monitoring and forecast verification. Climate Dynamics, 112. (Accessed August 31, 2012). Wheeler, M. C., and H. H. Hendon, 2004: An All-Season Real-Time Multivariate MJO Index: Development of an Index for Monitoring and Prediction. Mon. Wea. Rev., 132, 19171932.