Predictability of the mid‐summer surface air temperature over the Yangtze River valley in the National Centers for Environmental Prediction Climate Forecast System

Abstract. We compare herein polar processing diagnostics derived from the four most recent “full-input” reanalysis datasets: the National Centers for Environmental Prediction Climate Forecast System Reanalysis/Climate Forecast System, version 2 (CFSR/CFSv2), the European Centre for Medium-Range Weather Forecasts Interim (ERA-Interim) reanalysis, the Japanese Meteorological Agency's 55-year (JRA-55) reanalysis, and the National Aeronautics and Space Administration (NASA) Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA-2). We focus on diagnostics based on temperatures and potential vorticity (PV) in the lower-to-middle stratosphere that are related to formation of polar stratospheric clouds (PSCs), chlorine activation, and the strength, size, and longevity of the stratospheric polar vortex. Polar minimum temperatures (Tmin) and the area of regions having temperatures below PSC formation thresholds (APSC) show large persistent differences between the reanalyses, especially in the Southern Hemisphere (SH), for years prior to 1999. Average absolute differences of the reanalyses from the reanalysis ensemble mean (REM) in Tmin are as large as 3 K at some levels in the SH (1.5 K in the Northern Hemisphere – NH), and absolute differences of reanalysis APSC from the REM up to 1.5 % of a hemisphere (0.75 % of a hemisphere in the NH). After 1999, the reanalyses converge toward better agreement in both hemispheres, dramatically so in the SH: average Tmin differences from the REM are generally less than 1 K in both hemispheres, and average APSC differences less than 0.3 % of a hemisphere. The comparisons of diagnostics based on isentropic PV for assessing polar vortex characteristics, including maximum PV gradients (MPVGs) and the area of the vortex in sunlight (or sunlit vortex area, SVA), show more complex behavior: SH MPVGs showed convergence toward better agreement with the REM after 1999, while NH MPVGs differences remained largely constant over time; differences in SVA remained relatively constant in both hemispheres. While the average differences from the REM are generally small for these vortex diagnostics, understanding such differences among the reanalyses is complicated by the need to use different methods to obtain vertically resolved PV for the different reanalyses. We also evaluated other winter season summary diagnostics, including the winter mean volume of air below PSC thresholds, and vortex decay dates. For the volume of air below PSC thresholds, the reanalyses generally agree best in the SH, where relatively small interannual variability has led to many winter seasons with similar polar processing potential and duration, and thus low sensitivity to differences in meteorological conditions among the reanalyses. In contrast, the large interannual variability of NH winters has given rise to many seasons with marginal conditions that are more sensitive to reanalysis differences. For vortex decay dates, larger differences are seen in the SH than in the NH; in general, the differences in decay dates among the reanalyses follow from persistent differences in their vortex areas. Our results indicate that the transition from the reanalyses assimilating Tiros Operational Vertical Sounder (TOVS) data to advanced TOVS and other data around 1998–2000 resulted in a profound improvement in the agreement of the temperature diagnostics presented (especially in the SH) and to a lesser extent the agreement of the vortex diagnostics. We present several recommendations for using reanalyses in polar processing studies, particularly related to the sensitivity to changes in data inputs and assimilation. Because of these sensitivities, we urge great caution for studies aiming to assess trends derived from reanalysis temperatures. We also argue that one of the best ways to assess the sensitivity of scientific results on polar processing is to use multiple reanalysis datasets.

Download Full-text

Seasonal Prediction of Summer Precipitation in the Middle and Lower Reaches of the Yangtze River Valley: Comparison of Machine Learning and Climate Model Predictions

Water ◽

10.3390/w13223294 ◽

2021 ◽

Vol 13 (22) ◽

pp. 3294

Author(s):

Chentao He ◽

Jiangfeng Wei ◽

Yuanyuan Song ◽

Jing-Jia Luo

Keyword(s):

Neural Network ◽

Machine Learning ◽

Yangtze River ◽

Climate Models ◽

Summer Precipitation ◽

Yangtze River Valley ◽

River Valley ◽

Training Data ◽

The Yangtze River ◽

Learning Methods

The middle and lower reaches of the Yangtze River valley (YRV), which are among the most densely populated regions in China, are subject to frequent flooding. In this study, the predictor importance analysis model was used to sort and select predictors, and five methods (multiple linear regression (MLR), decision tree (DT), random forest (RF), backpropagation neural network (BPNN), and convolutional neural network (CNN)) were used to predict the interannual variation of summer precipitation over the middle and lower reaches of the YRV. Predictions from eight climate models were used for comparison. Of the five tested methods, RF demonstrated the best predictive skill. Starting the RF prediction in December, when its prediction skill was highest, the 70-year correlation coefficient from cross validation of average predictions was 0.473. Using the same five predictors in December 2019, the RF model successfully predicted the YRV wet anomaly in summer 2020, although it had weaker amplitude. It was found that the enhanced warm pool area in the Indian Ocean was the most important causal factor. The BPNN and CNN methods demonstrated the poorest performance. The RF, DT, and climate models all showed higher prediction skills when the predictions start in winter than in early spring, and the RF, DT, and MLR methods all showed better prediction skills than the numerical climate models. Lack of training data was a factor that limited the performance of the machine learning methods. Future studies should use deep learning methods to take full advantage of the potential of ocean, land, sea ice, and other factors for more accurate climate predictions.

Download Full-text