Predictability of the anomaly pattern of summer extreme high temperature days over northern China

Destructive extreme high-temperature events have attacked northern China (NC) more frequently in recent decades, with regional discrepancies. However, to what extent the distribution of extreme high temperature days (EHDs) over NC is predictable remains unexplored. To address this issue, this study investigates the predictability sources and builds physics-based empirical (P-E) models to predict the spatial distribution of the June–July mean number of EHDs over NC and estimate its predictability via Predictable Mode Analysis. Two observed major empirical orthogonal function (EOF) modes of EHDs over NC are extracted that explain 64% of the total variance during the period 1980–2009. The first mode is a homogeneous mode of EHDs dominated by a barotropic high-pressure anomaly over NC. The second mode is a southwest–northeast dipole pattern of EHDs modulated by a dipole pattern of geopotential height anomalies centered over Northwest China and Northeast Asia. Tracking back the origins of boundary anomalies, physically meaningful predictors are selected to establish a set of P-E models to predict the first two principal components (PCs). The causative links between the predictors and corresponding PCs are discussed using lead–lag correlation analysis. The P-E models can predict the first two modes well, with significant skill, and thus the first two modes can be regarded as the predictable modes. If the predictable modes can be perfectly predicted, 64% of the total observed variability of EHDs over NC is potentially predictable. By reconstructing the pattern of the predictable modes and corresponding predicted PCs, the pattern of EHD anomalies over NC can be predicted. The temporal mean pattern correlation coefficient (areal mean temporal correlation coefficient) for the prediction is 0.26 (0.27) during the independent forecast period (2010–2019), providing a reference for the actual predictability.