Introduction

1. The Seasonal Footprinting Mechanism in CFSv2: Simulation and Impact on ENSO Prediction 1,2Kathy Pegion(Kathy.Pegion@noaa.gov) & 2Michael Alexander 1University of Colorado/CIRES & 2Physical Sciences Division/ESRL/NOAA Introduction Does the CFSv2 Simulate the SFM and its Relationship with ENSO? The seasonal footprinting mechanism (SFM) is thought to be a pre-cursor to the El Nino Southern Oscillation (ENSO). Fluctuations in the North Pacific Oscillation (NPO), a north-south dipole in sea level pressure, impact the ocean via surface heat fluxes during winter, leaving a sea-surface temperature (SST) “footprint” in the subtropics. This footprint persists through the spring, impacting the tropical Pacific atmosphere-ocean circulation throughout the following year (Vimont et al. 2001, GRL; Vimont et al. 2003, J. Climate). There is also evidence that this mechanism is more effective depending on the state of the equatorial western Pacific thermocline (Anderson 2007, J. Climate). The simulation of the SFM in the NCEP/Climate Forecast System, version 2 (CFSv2) is likely to have an impact on operational predictions of ENSO. We explore the simulation of the SFM in the COLA/CFSv2 decadal simulations and compare it with reanalysis products. Additionally, we investigate the relationship between the SFM and ENSO prediction in the NCEP/CFSv2 retrospective forecasts. CFSv2 Composites SLPI Neg(N=59)-Pos(N=54) Observed SLPI vs. ENSO Indices CFSv2 Conditional Probabilities P(ENSO|SLPI) P(ENSO|Z15I) Model Data COLA/CFSv2 Decadal Experiments The CFSv2 also shows a shift in probabilities for El Nino (La Nina) events when the SLPI is <-1σ (>1σ) and for El Nino (La Nina) when Z15I >1σ (< -1σ). However, the conditional probabilities are much smaller than in the observations. NCEP CFSv2 Retrospective Forecasts The weaker SFM in the CFSv2 is also apparent in the relationship between ENSO indices and SLPI. The difference in the relationship between SLPI and central Pacific vs. eastern Pacific ENSO indices is not well represented by the model. (left panels) The relationship between SLPI and Nino3.4 is stronger when the Z15I and SLPI have opposite sign. This indicates that a deeper (shallower) thermocline in the equatorial west Pacific, together with a negative (positive) NPO is more effective in producing El Nino (La Nina) events. (right panels) CFSv2 is able to simulate the SFM, although it is weaker in the model than observations. The southern lobe of the NPO is weak in the model, leading to weaker warming in the sub-tropics, weaker air-sea feedbacks, and a weaker El Nino. The same is true for the positive sign of the NPO and subsequent La Nina events. What is the SFM & how is it related to ENSO? Observed Composites (1958-2007) SLPI Neg(N=10)-Pos(N=9) Observed SLPI vs. ENSO Indices What is the Impact of the SFM on ENSO Prediction? NCEPR1 SLP (hPa) & Hadley SST(°C) NCEPR1 10m Winds (m/x) & SODA Z15 (m) DJF Negative Phase of North Pacific Oscillation CFSv2 Forecast RMSE CFSv2 Forecast Composites CFSv2 6-month Forecast Reliability Forecasts initialized in March Forecasts Initialized in March, valid Sep-Oct-Nov Forecasts Initialized in Feb, Mar, Apr Westerly wind anoms in subtropics oppose the trades & reduce heat flux MAM Warm Subtropical SST “footprint” JJA Enhanced westerly wind anomalies in the deep tropics reduce heat flux along the EQ SON El Nino Conditions SLPI & Z15I have opposite sign SLPI & Z15I have same sign SLPI (Nov-Mar SLP anomalies [10°N-25°N;175°W-145°W]) In a composite sense, the CFSv2 re-forecasts initialized in March capture the correct relationship between the SLPI and forecasts of tropical Pacific SSTs in fall. Z15I (Dec-Jan depth of 15°C isotherm anomalies [5°N-5°N;160°E-180] ) CFSv2 forecasts of ENSO at 6-month lead times have some reliability for SLP(0) cases, but are generally not reliable for SLPI (+/-) cases. This indicates that long lead probabilistic ENSO forecasts are poor when ENSO is triggered by the SFM. Relatively large errors in forecast SSTs in the subtropics and eastern equatorial Pacific are already present at <1month lead-time. Errors in Z15 are also present in the western and eastern equatorial Pacific at <1month. These errors become very large at 3-months lead, particularly in the ENSO region. The relationship between SLPI and ENSO indices is stronger in the central Pacific (Nino4 &Nino3.4) than in the eastern Pacific (Nino3 and Nino1+2), indicating that the SFM may have an impact on the location of maximum SST anomalies for ENSO events. (left panels) The observed relationship between SLPI and Nino3.4 does not appear to be impacted by the sign of Z15I, in contrast to other studies (e.g. Anderson 2007). The relationship between Z15I and Nino3.4 indicate a stronger relationship when Z15I and SLPI have opposite sign, than when they have the same sign. (right panels) The SFM provides a link between north Pacific SLP anomalies (i.e. NPO) and Tropical Pacific SST anomalies. The negative (positive) phase of the NPO in winter leads to El Nino (La Nina) conditions by the following fall, via this mechanism. Conclusions Observed Conditional Probabilities P(ENSO|SLPI) P(ENSO|Z15I) The CFSv2 is able to simulate some basic aspects of the SFM. However, the southern lobe of the NPO is weak and does not impact the Tropics as strongly as observed. The difference in the relationship between SLPI and central Pacific vs. eastern Pacific ENSO indices is not well represented by the model. This may have a negative impact on ENSO prediction and seasonal forecasts that depend on ENSO teleconnections. The CFSv2 is able to capture the correct direction of shifted probabilities for ENSO events conditional upon the SFM (indicated by SLPI) and the current state of the tropical Pacific (indicated by Z15I). However, the shift in probabilities is more pronounced in the observations than in the model. Although CFSv2 long-lead forecasts indicate the correct phase of ENSO based on SLPI in a composite sense, ENSO forecasts triggered by the SFM have poor reliability. This indicates an area for potential improvement in long-lead ENSO forecasting. Large forecast errors in SST and Z15 are already present in forecasts at less than 1-month lead time. This indicates it is possible that these errors are caused by spin-up as the model adjusts from the initial analysis. % Conditional probabilities indicate that there is a strong preference for El Nino (La Nina) events when SLPI < -1σ (>1σ). There is also a preference for La Nina (El Nino) events when Z15I < -1σ (>1σ).