Predictable skill and its association to sea‐surface temperature variations in an ensemble climate simulation

Abstract

[1] Simulated near-surface air temperature (T a ) and precipitation in an ensemble climate simulation is assessed. The ensemble climate-simulation was constructed with the Center for Ocean-Land-Atmosphere (COLA) atmospheric general circulation model (AGCM) in conjunction with the Atmospheric Model Intercomparison Project Phase II (AMIP II). To diagnose the ensemble simulation, a measure of ''predictable skill'' is formalized. This diagnostic is based upon the statistical significance of spatial correlation over any given region (for this analysis, North America) between the ensemble mean and observed anomalies and the inter-member scatter of the spatial anomaly correlation. Using this measure as a function of time, periods of predictable skill within the COLA AMIP II ensemble simulation are identified, and through a point-wise, multiple correlation technique, spatial patterns of contemporaneous sea-surface temperature (SST) variability are also constructed. This diagnosis can essentially define which skillfully simulated climate signals over North America (given by predictable skill) can be attributed to specific SST anomalies. Spatially coherent patterns of SST variability are found to be associated with predictable skill. These patterns are, not surprisingly, primarily related to the El Niño Southern Oscillation (ENSO). One of the more prevalent ENSO associations to predictable skill (for the COLA AGCM) is found in sub-tropical regions of the western Pacific. Moreover, a skillful episode of simulated North American precipitation is seen to be associated with SST variability over the east Indian Ocean. In addition, a strong contemporaneous association of T a predictable skill with tropical Atlantic SST variability is found. Complementary simulations are performed with the COLA AGCM to further assess the degree of attribution of predictable skill to these regions of SST variability. The results of the experimental simulations suggest that, for the AMIP II simulation period, the COLA AGCM can be skillfully attributed to SST variability primarily associated with ENSO, warm tropical Atlantic anomalies, and an abrupt, cooling event over the east Indian Ocean.Citation: Schlosser, C. A., and B. P. Kirtman (2005), Predictable skill and its association to sea-surface temperature variations in an ensemble climate simulation,

Non-stationarity of the signal and noise characteristics of seasonal precipitation anomalies

“…Other studies have shown similar results for both potential predictability and seasonal forecast and hindcast skill (e.g. Brankovic et al 1994;Kumar and Hoerling 1998;Zwiers et al 2000;Hoerling and Kumar 2003;Schlosser and Kirtman 2005;Phillips 2006). Here we evaluate correlations between signal, noise, and potential predictability, and four dominant modes (EOFs) of twentieth century global SST variability.…”

“…Interannual variability of signal and potential predictability have been largely attributed ENSO, with both exhibiting increases over much of the globe during ENSO events compared to ENSO-neutral periods (e.g. Brankovic et al 1994;Kumar and Hoerling 1998;Pegion et al 2000;Peng et al 2000;Zwiers et al 2000;Phillips 2006;Schlosser and Kirtman 2005;Wu and Kirtman 2006). More recently, Nakaegawa et al (2004) showed a widespread positive trend in the potential predictability of 500 hPa geopotential heights during boreal winter from 1950-2000, and Kang et al (2006 showed positive trends in the leading principal components of the signal variance and potential predictability of boreal winter precipitation and temperature over the twentieth century.…”

“…Similarly, analysis of precipitation and streamflow forecast and hindcast skill has shown interannual and interdecadal variability in skill associated with ENSO, PDO, and other ocean-atmosphere teleconnections (e.g. Hamlet and Lettenmaier 1999;Schlosser and Kirtman 2005;Grimm et al 2006). These results suggest that the physical mechanisms governing seasonal precipitation anomalies vary in time.…”

On the Relationship Between Probabilistic and Deterministic Skills in Dynamical Seasonal Climate Prediction

“…In many of these works, the assessments are mainly from the deterministic forecast angle. Several important aspects, such as physical sources of seasonal forecast skill (e.g., Butler et al, ; Chowdary et al, ; Kumar et al, ; Lee, Lee, et al, ; Lee, Wang, et al, ; Lee et al, ; Li et al, ; Manzanas et al, ; Quan et al, ; Rodwell et al, ; Scaife et al, ; Schlosser & Kirtman, ; Stockdale et al, ; Yang et al, , ), the relation between model fidelity in simulating climatology and skill in predicting seasonal anomaly (e.g., DelSole & Shukla, ; Jia et al, ; Lee et al, ; Sperber & Palmer, ), one‐ versus two‐tier modeling strategy (e.g., Beraki et al, ; Graham et al, ; Guérémy et al, ; Kug et al, ; Landman et al, ; Wang et al, ; Zhu & Shukla, ), and MME versus SME (e.g., Kang & Shukla, ; Krishnamurti, ; Krishnamurti et al, ; Pavan & Doblas‐Reyes, ; Peng et al, ; Yoo & Kang, ) were explored and discussed, in general terms or in the context of predicting specific climate phenomena. In view of the importance of probabilistic forecast, quite a few recent studies have started to pay due attention to or even concentrate on assessing the forecast skill from the probabilistic angle (e.g., Alessandri et al, ; Becker & van den Dool, ; Hagedorn et al, ; Kharin et al, ; Kirtman et al, ; Min et al, ; Palmer et al, ; Sohn et al, ; Tippett et al, ; Wang et al, ; Weisheimer et al, ; Yan & Tang, ; Yang et al, ).…”

Potential Predictability of North American Surface Temperature. Part I: Information-Based versus Signal-To-Noise-Based Metrics

“…This is often employed using signal-to-noise ratio (SNR)-based metrics (e.g., Kumar et al 2003;Peng and Kumar 2005;Kang and Shukla 2006;Quan et al 2006;Peng et al 2011). With the analysis of SNR, it was found that the seasonal climate predictability in NA relies mostly on the state of the tropical SST anomalies, and that the extratropical atmospheric seasonal variability over the Pacific-North American region has a pronounced response to the interannual variability of the tropical SST, especially in winter (e.g., Shabbar and Barnston 1996; Barnston and Smith 1996;Kumar and Hoerling 1998;Shukla 1998;Shukla et al 2000;Schlosser and Kirtman 2005;Peng et al 2011). The SNR-based predictability measures assume that the signal is always predictable and noise is certainly unpredictable.…”