Research ArticleCLIMATOLOGY

Oceanic and radiative forcing of medieval megadroughts in the American Southwest

See allHide authors and affiliations

Science Advances  24 Jul 2019:
Vol. 5, no. 7, eaax0087
DOI: 10.1126/sciadv.aax0087
  • Fig. 1 Time series and megadrought state changes.

    (A) The monthly NINO3.4 index from PHYDA (27) in black, with dashed gray lines showing the 5th and 95th percentiles of the PHYDA reconstruction ensemble and the yellow line showing a 20-year locally weighted linear regression smoothing of the NINO3.4 time series. (B) Annual AMO [or North Atlantic SST (NASST)] index from PHYDA in black, with dashed gray lines and yellow line as in (A). (C) The mean NINO3.4 and AMO [corresponding to black lines in (A) and (B)] along with a global mean forcing estimate (see Materials and Methods) and highlighted megadrought periods. (D to F) Probability density functions (PDFs) of temperature for the NINO3.4, AMO, and forcing estimate during the reference years of 1601–1925 (teal), all the states during each of the megadroughts (orange), and the remaining nondrought segments from 800 to 1600 (black). NINO3.4 and AMO PDFs are for the full reconstruction ensemble from PHYDA, while the forcing PDF is for the full ensemble forcing estimate (see Materials and Methods). (G to I) Change in percentiles of the PDFs shown in (D) to (F) relative to the corresponding 1601–1925 reference PDF. Percentile changes show which parts of the megadrought and nondrought distributions change relative to the reference distribution.

  • Fig. 2 Hydroclimate and megadrought regression analysis.

    (A) The bootstrap distributions of adjusted r2 for the decadal linear regression of the predictand North American Southwest PDSI (NASW PDSI) and the predictors of NINO3.4 (or N), AMO (or A), and total forcing (Forc or F) (see Materials and Methods). For this and the other panels in this figure, the central box mark indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, while the whiskers extend to the most extreme data points not considered outliers (outliers not shown for clarity). (B) Bootstrap distributions of adjusted r2 for the decadal linear regression with the predictands of NINO3.4, AMO, and NASW averaged 2-m temperature (T2m) and the predictor of total forcing (see Materials and Methods). (C) Bootstrap regression β1 parameter distributions for the same regressions as shown in (B) (see Materials and Methods). Here and in (D), the numbers at the top indicate the whole percentage of data points below and above zero (below:above). (D) Bootstrap distributions of β1 regression parameter values for a logistic regression binned over megadroughts (see Materials and Methods); here, the predictand is the binary variable of either being in a megadrought or not, while the predictors are the NINO3.4, AMO, and forcing indices. The suffix “All” indicates regressions with all data points, and the suffix “NLV” (or no large volcanoes) indicates regressions with outlier volcanic eruption bins removed for all variables (see Materials and Methods). (E) Percentage of bootstrap regressions for which a particular logistic regression model has a minimum information criterion, suggesting that it is the best of the tested models (see Materials and Methods). “AIC” is the Akaike information criterion corrected for the sample size, and “BIC” is the Bayesian information criterion. The regression models all use the megadrought predictand with different combinations of the predictors NINO3.4, AMO, and total forcing (see Materials and Methods).

  • Fig. 3 SOM analysis of the change in the incidence of modes of SST variability and consequent hydrological change.

    (A) SOM SST patterns (April to next calendar year March annual mean) with PDSI composites (JJA) over the best matching units for the years 800–1925 in PHYDA. The boxes in the lower left corner of each panel indicate the percent change in the frequency of occurrence of that pattern during the megadroughts relative to the 1601–1925 reference period. std units, standard units. (B) Frequency of occurrence of the SOM SST patterns during the full reference period 1601–1925, during megadroughts within the period 800–1600, and the remainder nonmegadrought years within the period 800–1600. The gray error bars around the reference value are the 2σ range of a bootstrap resampling (see Materials and Methods).

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/7/eaax0087/DC1

    Fig. S1. Ensemble drought agreement and comparison of NASW PDSI indices.

    Fig. S2. Global mean temperature from PHYDA.

    Fig. S3. Energy balance model estimates of the historical, global mean forced climate response.

    Fig. S4. SOM analysis using six SOM nodes.

    Fig. S5. SOM analysis using 12 SOM nodes.

    Fig. S6. Temperature and PDSI composite over the megadroughts.

    Fig. S7. Power spectra comparisons.

    Fig. S8. Cross-correlation comparisons.

    Fig. S9. NINO3.4 correlation maps.

    References (53, 54)

  • Supplementary Materials

    This PDF file includes:

    • Fig. S1. Ensemble drought agreement and comparison of NASW PDSI indices.
    • Fig. S2. Global mean temperature from PHYDA.
    • Fig. S3. Energy balance model estimates of the historical, global mean forced climate response.
    • Fig. S4. SOM analysis using six SOM nodes.
    • Fig. S5. SOM analysis using 12 SOM nodes.
    • Fig. S6. Temperature and PDSI composite over the megadroughts.
    • Fig. S7. Power spectra comparisons.
    • Fig. S8. Cross-correlation comparisons.
    • Fig. S9. NINO3.4 correlation maps.
    • References (53, 54)

    Download PDF

    Files in this Data Supplement:

Navigate This Article