Research ArticleCLIMATOLOGY

Skillful prediction of summer rainfall in the Tibetan Plateau on multiyear time scales

See allHide authors and affiliations

Science Advances  09 Jun 2021:
Vol. 7, no. 24, eabf9395
DOI: 10.1126/sciadv.abf9395
  • Fig. 1 The TP and the CMIP6 multimodel ensemble projections of the TP summer precipitation.

    (A) The surface elevation of the TP (shaded; unit: m). Distribution of natural lakes (blue shades) and major rivers (light blue lines; including Yellow, Yangtze, Mekong, Salween, Brahmaputra, Ganges, Indus, and Amu Darya) on the TP are also shown. The TP is divided into 12 large river basins (26). (B) Time series of observations and CMIP6 multimodel ensemble projections of the summer precipitation anomalies (unit: mm/day) over the TP relative to 1986–2005. The historical simulations (gray lines) during 1960–2014 and projections under SSP1-2.6 (green lines), SSP2-4.5 (blue lines), SSP3-7.0 (purple lines), and SSP5-8.5 (red lines) emission scenarios during 2015–2100 from 19 CMIP6 models (table S1) are used. The thick (dashed) lines represent the ensemble mean (individual ensemble) simulations, respectively. The observations were derived from Asian Precipitation-Highly Resolved Observational Data Integration Towards Evaluation of Water Resources (APHRODITE) (black line), Global Precipitation Climatology Project (GPCP) (orange line), and Global Precipitation Climatology Centre (GPCC) (brown line). Both the observations and simulations are smoothed with a 9-year running average. The inset figure in (B) emphasizes the projections for the period of 2014–2040.

  • Fig. 2 Decadal prediction skill for 2- to 9-year forecast periods of the averaged summer rainfall over the TP.

    (A) ACC skill for 138-member ensemble mean prediction. (B) ACC skill for 10-member ensemble mean prediction (averaged over all possible combinations). (C) MSSS skill for 138-member ensemble mean prediction. Only the values higher than 0 are shown. (D) RPC calculated from 138-member predictions. The dots denote values passing the 95% confidence level (see Materials and Methods). The elevation of 2500 m is used as the outline of TP. The outline of the ITP is also shown.

  • Fig. 3 Decadal prediction for the summer ITP precipitation.

    (A) The 2- to 9-year ensemble mean predictions of the summer ITP precipitation (unit: mm/day) from decadal hindcasts (raw hindcast ensembles, red line), the quasi–real-time predictions (raw prediction ensembles, blue line), and the corresponding observations (black line). (B) Same as in (A), but after applying variance adjustment to the raw predictions. Cor noted in parentheses is the correlation coefficient between the observation and the predictions. Shading in (A and B) show the 5 to 95% uncertainty range. (C) Correlation of the predicted summer ITP rainfall with the observational reference along the forecast time for 4-year averages. The black dashed line represents the persistence predictions. The vertical segments denote 5 to 95% uncertainty range of all possible combinations. (D) Correlation as a function of ensemble size for model ensemble mean predictions of an independent ensemble member (blue line; averaged over all possible combinations) and for model ensemble mean predictions of the observations (red line). Shading shows the 5 to 95% confidence interval. Time series of the predictions and the observations in (C and D) are detrended. (E) Relationship between the 8-year running mean of the summer ITP rainfall and the summer subpolar gyre (SPG) SSTA in observations. (F) Same as in (E), but for the predicted 2- to 9-year summer SPG SSTA and the summer ITP rainfall from the decadal hindcasts. The COR noted in the top left in (E and F) is the correlation coefficient for the observations and ensemble mean predictions, respectively.

  • Fig. 4 Mechanism for the decadal variability of ITP summer precipitation and sources of decadal predictability in DCPP.

    Regressions onto the 8-year running averaged summer ITP rainfall index for (A) the observed 200-hPa eddy geopotential height (shaded; unit: m) and WAF (vector; unit: m2·s−2) and (B) the observed SSTA (shaded; unit: °C) and land precipitation (shaded; unit: mm/day). (C and D) Same as in (A and B), but for the regressions of the 2- to 9-year averaged ensemble mean predictions onto the 2- to 9-year averaged ensemble mean predicted ITP summer precipitation index. The dots denote values passing the 95% confidence level (see Materials and Methods). (E and F). ACC skill for 2- to 9-year forecast periods of the averaged (E) 200-hPa eddy geopotential height and (F) SSTA and land precipitation. The dots denote values passing the 95% confidence level (see Materials and Methods).

  • Fig. 5 The 2020–2027 averaged summer TP rainfall anomalies in real-time forecasts.

    (A and B) Spatial distributions of the 2020–2027 averaged summer precipitation anomalies (mm/day) from (A) decadal forecast from 2018 (averaged over raw forecast ensembles) and (B) decadal forecast from 2018 (averaged over variance-adjusted forecast ensembles). Regions with ACC < 0 are masked. (C and D) Probability distributions (%) of the ensembles for ITP summer precipitation anomalies from (C) raw forecast ensembles and (D) variance-adjusted forecast ensembles. The blue lines and numbers in (C and D) show the ensemble mean values, and the member sizes are also noted for each scenario. The precipitation anomalies are relative to the period of 1986–2005.

Supplementary Materials

  • Supplementary Materials

    Skillful prediction of summer rainfall in the Tibetan Plateau on multiyear time scales

    Shuai Hu and Tianjun Zhou

    Download Supplement

    This PDF file includes:

    • Supplementary Text S1
    • Tables S1 and S2
    • Figs. S1 to S7

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article