Are Long-Term Changes in Mixed Layer Depth Influencing North Pacific Marine Heatwaves?

AFFILIATIONS: Amaya—Cooperative Institute for Research in Environmental Sciences (CIRES), University of Colorado Boulder, Boulder, Colorado; Alexander—NOAA Physical Sciences Laboratory, Boulder, Colorado; Capotondi—CIRES, University of Colorado Boulder, and NOAA Physical Sciences Laboratory, Boulder, Colorado; Deser—National Center for Atmospheric Research, Boulder, Colorado; Karnauskas—CIRES and Department of Atmospheric and Oceanic Sciences, University of Colorado Boulder, Boulder, Colorado; Miller—Scripps Institution of Oceanography, University of California San Diego; Mantua—NOAA Southwest Fisheries Science Center, Fisheries Ecology Division


I n boreal summer
, the northeast Paci c Ocean (NE-Pac) experienced a resurgence of extremely warm upper ocean temperatures (Fig. a).The strength and pattern of the sea surface temperature anomalies (SSTAs) earned this event the moniker "Blob ." (Amaya et al.
; herea er A ), a reference to the original warm "Blob" that initiated a multi-year marine heatwave (MHW) that devastated regional ecosystems over - (Bond et al. ; Cavole et al. ; Amaya et al. ; Piatt et al. ).In particular, the intraseasonal persistence of the Blob .generated similar widespread concern among shery and wildlife managers for sensitive marine ecosystems along the west coast of North America (NOAA ).Blob .primarily resulted from a record minimum mixed layer depth (MLD; Fig. a shading), which formed due to weaker than normal wind speeds and strong surface heating from reduced cloud cover (A ).Equation ( ) illustrates how shallow mixed layer depths a ect mixed layer temperature changes, T m / t, when consider- ing only local heat sources and sinks (i.e., neglecting advection) and separating each budget term into mean and perturbation components: where Q is the net surface heat ux into the ocean, h is the MLD, ρ is seawater density, and c p is the speci c heat of seawater.Primes denote time anomalies and overbars

|
represent time mean values.For the full derivation see Alexander and Penland ( ).For Blob ., strong downward Q anomalies (i.e., positive Q ) led to excess heat being distributed over a thin climatological mixed layer, since h ¯ is small in summer (term I).More importantly, anomalous MLD shoaling (i.e., negative h ) contributed to upper ocean warming through term II (A ).As discussed in A , the MLD anomalies (MLDAs) superpose on a MLD shoaling trend from to the present, which they suggest may indicate a role for anthropogenic forcing.Upper ocean warming in response to future climate change is expected to reduce mixing and shoal the mixed layer (Capotondi et al. ; Alexander et al. ).A long-term trend in the mean MLD would have signi cant implications for SSTAs since, according to Eq. ( ), decreasing the mean MLD (h ¯) results in a stronger temperature response for the same heat ux and MLD anomalies.Therefore, if the observed MLD shoaling rst reported by A is robust across di erent datasets and consistent with the projected response to anthropogenic climate change, then Blob .may have been exacerbated by anthropogenic forcing.Here, we investigate the presence of NEPac MLD trends in a suite of observational analyses.We then compare these results to coupled model simulations to assess the potential in uence of anthropogenic climate change on NEPac MLD trends, and by extension, on the likelihood and intensity of the MHW.

Data and methods.
For observed MLD, we use monthly mean data from the NOAA Global Ocean Data Assimilation System (GODAS; Behringer and Xue ), ECMWF Ocean Reanalysis System (ORAS ; Balmaseda et al.
), Simple Ocean Data Assimilation version (SODA ; Carton et al.
), and gridded Argo pro les (Hosoda et al. ).See Table 1 for more details.We estimate the externally forced MLD trends using the Community Earth System Model version Large Ensemble (CESM -LE; Kay et al.
).Additionally, we use models from phase of the Coupled Model Intercomparison Project (CMIP ; Taylor et al. ) with the same radiative forcing protocol.Model details are provided in Table 1 and also Table ES in the supplemental material.We use the ensemble mean of each model ensemble (CESM -LE and CMIP ) as two estimates of the forced response.
To compare trends across datasets, we calculate MLD in each observational analysis and coupled model simulation as the interpolated depth at which potential density rst exceeds .kg m -greater than the surface value (Suga et al. ).For datasets that do not include potential density, we calculate it from monthly mean potential temperature and salinity pro les.To compare to Blob ., we only analyze MLD values averaged over boreal summer [June-August (JJA)].Unless otherwise speci ed, all anomalies are relative to the period -, which is the longest overlapping period for the data used in this study.
Our results are not sensitive to the choice of MLD de nition.Additionally, while it is preferred to calculate long-term MLD trends based on daily mean values, many of the datasets only provided monthly means (e.g., ORAS , ORAS , CESM -LE, and CMIP ).However, we do not expect our results or conclusions to be in uenced by this choice, since the temperature and density gradients are very strong at the base of the mixed layer in summer.Finally, we de ne the term "NEPac" to represent the region bounded by °-°N, °-°W (black box, Fig. ), the same area used in A .

Results.
MLD trends in observations.We begin by assessing MLD trends in observations.Interannual MLD variability in the NEPac is quite consistent across the various observational analyses (Fig. b, Table ), particularly during the Argo era ( -present).However, earlier in the instrumental record, two groupings emerge, with GODAS and ORAS exhibiting a more pronounced shoaling trend than ORAS and SODA .While the magnitude of the observed MLD shoaling varies among datasets, the average NEPac trend (− .m decade − ) is significant at the % confidence level (Table ).Given that the climatological JJA MLD in the NEPac region is ~ m, such a trend would correspond ~ % decrease in the mean MLD from to .Creating two groupings of observational analyses (GODAS, ORAS , and Argo vs ORAS and SODA ), we produce two observational MLD trend maps from to  -; Figs. g,h).For the NEPac, the JJA MLD time series show significant forced trends in both the CESM -LE ).
. The average of GODAS, ORAS , and Argo shows widespread MLD shoaling trends with two main centers of action, one around the Aleutian Islands and one o the California coast (Figs.c,d; see also Fig. ES ).While the average of ORAS and SODA shows weaker MLD trends overall, the two centers of action are also generally present in these data (Fig. ES ).Additionally, the close spatial correspondence of the MLDAs near California (Fig. a, shading) with some of the observed trends (Fig. c and Fig. ES ) suggests that this extreme event was likely exacerbated by these longer-term features.MLD trends in climate models.Are these and other North Pacific MLD trends attributable to anthropogenic forcing?To address this question, we show maps of JJA MLD trends from to for the ensemble means of CESM -LE and CMIP to estimate the forced component (Figs.e,f).There is some spatial correspondence with the observational analyses (Fig. c and Fig. ES ), especially with CMIP .The spatial similarities between the historical trends in observations and the forced trends in models are even more apparent when the latter are extended into the future (

Table 1 .
Observational and coupled model data used in this study and their JJA-averaged MLD trends in the NEPac (black box; Fig.1a).For ensemble datasets, the mean MLD trend is reported with minimum and maximum ensemble trends in parentheses.For Argo data, the trend is reported for 2001-19 are bolded and are based on a 95% Mann-Kendall test.Datasets marked with an asterisk (*) did not provide potential density as a variable.Therefore, the potential density used to calculate MLD for these datasets is | that modulate the mixed layer heat budget when assessing the in uence of climate change on future MHWs, which complements recent studies focusing primarily on the in uence of climate change on the SST itself(Frölicher et al.  ; Jacox et al.  ;  Walsh et al.