The Spring Onset forecasting

The map above shows the spring onset forecasting (ΔSI-x; spring index anomaly) in units of days. Negative values mean early onset from a climatology defined for 1982-2010 and positive values for late-onset. The black solid lines indicate the forecasting skill (local skill), e.g., validated regions where the spring onset has correlation thresholds of r > 0.45. Also, the gray shaded zones are regions where the spring onset (SI-x) has occurred.

The actual forecasted day of the spring onset (SI-x) is indicated in the table below in calendar days (Day-Of-the-Year units, DOY). This table describes four products related to the SI-x forecast for different initializations (e.g., Jan 15, Jan 31, etc.). Four fields in the table are updated monthly: (1) the Spring Index anomaly (ΔSI-x); (2) the Spring Index (SI-x); (3) a Global Skill defined as CDF of centered anomaly correlation for the entire domain (Wilks, 2011); with two thresholds (0.5 and 0.8) identify skill for different initializations; and (4) a Local Skill defined using Pearson correlation.

Carlos M. Carrillo and Toby R. Ault
Emergent Climate Risk Lab, @ECRL_Cornell
Dept. of Earth & Atmos. Sciences, Cornell University

The SI-x was computed as in Ault et al. (2015) with current climatic conditions, which data sets are multiple simulations of daily maximum and minimum temperature of the CFSv2 model. The SI-x product was post-processing using an ensemble model statistic output (EMOS), Non-homogeneous Gaussian Regression (Gneiting et al., 2005). The different ensemble members are defined with four daily runs at different initializations: 00Z, 06Z, 12Z, and 18Z. Thus, a total of 120 model realizations were used in each issued SI-x forecast. The EMOS product was calibrated with a re-forecasting dataset from 1982-2010 (see table below historical validation) and further details can be found in Carrillo et al. (2018).