Summer high temperature extremes over Northeastern China predicted by spring soil moisture

Current seasonal climate predictions mainly reside in the ocean anomalies. However, the prediction skills are generally limited over many extra-tropical land areas where the oceanic effects are relatively weak. In this study, we address the potential of preceding spring soil moisture condition to predict summer hot days over Northeastern China, a typical Northern Hemisphere mid-latitude region. The results show that spring soil moisture condition over Central-Eastern China is closely related with following summer hot days over Northeastern China for the period of 1979–2017. The statistical model based on the preceding spring soil moisture condition yields temporal cross-validated correlation skill of 0.57 for summer hot days over Northeastern China. The spatial pattern correlation skills of independent hindcast experiments for 2009–2017 are also high, ranging from 0.87 to 0.94. Our results can be easily applied to practical prediction of summer hot days over Northeastern China, and help to provide better climate services and reduce the detrimental effects of extreme heat over this extra-tropical region.

www.nature.com/scientificreports www.nature.com/scientificreports/ spring soil moisture anomalies over Central-Eastern China to following summer geopotential height anomalies over downstream Northeastern China. The positive summer geopotential height anomalies associated with dry spring soil moisture anomalies tend to lead to more downward solar radiation (Fig. 5a) and stronger sinking motion (Fig. 5b), which may subsequently enhance surface heat fluxes and subsidence warming in summer over Northeastern China, respectively. In addition, decreased precipitation associated with increased geopotential heights may result in significant soil dryness over Northeastern China (Fig. 5c), increasing local sensible heat and the entrainment of heat into planetary boundary layer (Fig. 5d). These changed processes possibly induced by dry spring soil moisture condition over upstream Central-Eastern China together may cause more summer high temperature extremes over Northeastern China. Due to complex dynamical and thermodynamic processes involved, numerical experiments need to be further performed to better understand the underlying mechanisms explaining the close association of spring soil moisture condition over Central-Eastern China with summer hot extremes over Northeastern China.
Prediction of summer hot days over Northeastern China. As described above, the summer hot days over Northeastern China have strong correlations with preceding spring soil moisture condition over Central-Eastern China. Here, we use soil moisture from GLDAS-Noah which has the highest correlation with summer hot days among three GLDAS soil moisture data and SPEI, as the predictor, and further establish linear regression models for summer hot days averaged over Northeastern China. Figure 6 shows that the correlation coefficient between the observed summer hot days and leave-one-out cross-validation estimates over Northeastern China is 0.57 for 1979-2017. And the result indicates that about one third (32.5%) of the total variance of summer hot days over Northeastern China can be predicted by the statistical models based on the spring soil moisture condition.
Then, we predict spatial patterns of summer hot days over Northeastern China with independent hindcast experiments for 2009 to 2017 by using 30-year-moving prediction models for each of 136 stations. Figure 7 presents examples of 2016 and 2017 for the observed and predicted summer hot days at 136 stations over Northeastern China. The observed and predicted summer hot days generally show the consistent spatial pattern with a obvious south-to-north gradient. We also calculate the spatial pattern correlation coefficients between the observed and predicted summer hot days over Northeastern China for 2009-2017 (Fig. 8). The correlation coefficients range from 0.87 to 0.94, indicating that the statistic models we developed have high prediction skills for spatial patterns of summer hot days over Northeastern China.

conclusions and Discussion
Northeastern China which is located in middle latitudes of Northern Hemisphere, is increasingly influenced by extreme high temperature events during the summer. However, current limited skills of seasonal climate prediction impede our ability to deal with these heat events. In this study, we identify that preceding spring soil moisture conditions over Central-Eastern China have robust and negative associations with summertime hot days over Northeastern China for 1979-2017. These close relationships point to that spring soil moisture condition can be used as the potentially important predictor for summer hot days over Northeastern China.
Dry soil moisture condition in spring over upstream Central-Eastern China corresponds well to summer geopotential heights over Northeastern China. The increased geopotential height may subsequently rise downward solar radiation reaching land surface and enhance subsidence warming, leading to warmer surface air temperature and more hot days over Northeastern China during the summer. In addition, the increased geopotential height tends to result in less precipitation and drier soil over Northeastern China, which may subsequently cause stronger local surface sensible heat fluxes and enhance the entrainment of warm air into the planetary boundary layer. These possible physical processes directly and indirectly resulted from dry soil moisture condition favor the occurrence of summer hot days over Northeastern China.
Spring soil moisture condition over Central-Eastern China is further used to predict hot days over Northeastern China during the following summer. The statistical models developed with spring soil moisture  www.nature.com/scientificreports www.nature.com/scientificreports/  This study highlights the importance of spring soil moisture condition over upstream Central-Eastern China for predicting summer high temperature extremes over Northeastern China. Meanwhile, some limitations exist. These proposed physical mechanisms underlying the close association of spring soil moisture condition with following summer high temperature extremes over Northeastern China need to be further tested by model simulations in the future. In addition to soil moisture condition, sea surface temperature, soil temperature, vegetation, sea ice among others have the potential for summer high temperature extreme prediction 47,48 . For example, Zhang et al. made the relative accurate prediction of summer high temperature extremes over the source region of the ancient Silk Road in China by taking spring soil temperature, sea surface temperature and large-scale climate indices as preceding precusors 49 . In particular, the ENSO is well known to largely control the interannual climate variability on the global scale. However, our preliminary analysis does not identify significant correlations of summer high temperature extremes over Northeastern China with preceding ENSO indices, which are subject to additional investigation. Current dynamical models generally show very low skills in forecasting high temperature extremes, and realistic soil moisture initialization and improved representation of soil processes have the potential to enhance the monthly-to-seasonal forecast skills in dynamical models 17,50 . Our findings are expected to facilitate the improvement of practical seasonal forecasting for summer hot days over Northeastern China.  www.nature.com/scientificreports www.nature.com/scientificreports/ Methods Data preparation. The daily surface air temperature records for 699 stations in China were obtained from the China Meteorological Data Service Center. In this study, we chose 136 stations over Northeastern China with complete records of daily maximum surface air temperature in summer (June-July-August) for 1979-2017 (Fig. 1a). Soil moisture data were obtained from the Global Land Data Assimilation System (GLDAS) version 1.0 at a resolution of 1.0° × 1.0°, which were produced by driving offline land surface models with integrated observation based data 51 . Spring (March-April-May) soil moisture data from GLDAS-Noah, GLDAS-Mosaic, and GLDAS-CLM for 1979-2017 were applied with soil layers of 0.1-1 m, 0.02-1.5 m, 0.09-1.38 m, respectively. We took the standardized precipitation evapotranspiration index (SPEI) data at a 3-month time scale as a proxy of soil moisture 52 to test if our results depend on the choice of soil moisture datasets.
The monthly geopotential heights at 200 hPa, 500 hPa and 850 hPa, surface downward solar radiation, 500 hPa vertical velocity and planetary boundary layer height were taken from the European Center for Medium-range Weather Forecasts (ECMWF)'s ERA-interim reanalysis dataset 53 .
Hot day definition and EOF analysis. A day in summer (June-July-August) was defined as a hot day in which daily maximum temperature was equal to or above 30 °C. EOF analysis, which can extract the principal spatial and temporal characteristics from the variable, is applied to check the spatial homogeneity of the summer hot days at 136 stations over Northeastern China. correlation with spring soil moisture. The correlation analysis was employed to examine the relationship between summer hot days averaged over Northeastern China and spring soil moisture over East Asia for 1979-2017. We also conducted the correlation analysis after removing the linear trends of data. Student's t-test was used to identify the key region of statistical significance for both original and detrended data. As shown in Fig. 2, the identified key region of spring soil moisture is located in Central-Eastern China. In addition, we also tested the robustness of the key region by the correlation analysis of the data for 1979-2008 (Figs S1 and S2). physical process analysis. Regression analysis was applied to explored the possible physical processes associated with the link between summer hot days over Northeastern China and spring soil moisture conditions over Central-Eastern China. We calculated the regression coefficients of 200 hPa, 500 hPa and 850 hPa geopotential heights, surface downward solar radiation, 500 hPa vertical velocity, soil moisture, and planetary boundary layer height in summer against spring soil moisture averaged over Central-Eastern China for 1979-2017. prediction model development and evaluation. We developed linear regression prediction model for summer hot days over Northeastern China based on the identified spring soil moisture over Central-Eastern China. We used leave-one-out cross-validation to evaluate how well the statistical models perform. Each time, one year observation from 1979-2017 was chosen as the validation set, and the remaining observations were used as the training set. Both spring soil moisture and summer hot days for each prediction year are dropped out when statistical model is developed. This cross-validation process was repeated 39 times until all predictions were conducted. We used the correlation coefficient between the observed and the corresponding cross-validation estimates to measure the prediction skill. Spatial pattern prediction. Finally, we conducted independent hindcast experiments of summer hot days at 136 stations over Northeastern China from 2009 to 2017. The 30-year-moving prediction models at each station were established based on the identified spring soil moisture over Central-Eastern China. For example, we used the data for 1979-2008, 1980-2009 and 1981-2010 to establish the prediction models for summer hot days at each station in 2009, 2010 and 2011, respectively. The observed and predicted summer hot days at 136 stations in 2016 and 2017 were shown in Fig. 7. The correlation coefficients between the observed and predicted spatial patterns of summer hot days for 2009-2017 were computed to evaluate the prediction skills (Fig. 8).

Data Availability
The data sets within the article are available from authors upon reasonable request.