Asymmetric Emergence of Low-to-No Snow in the American Cordillera


 Societies and ecosystems within and downstream of mountains rely on seasonal snowmelt to satisfy their water demands. Anthropogenic climate change has reduced mountain snowpacks worldwide, altering snowmelt magnitude and timing. Here, the global warming level leading to widespread and persistent mountain snowpack decline, termed low-to-no snow, is estimated for the world's most latitudinally contiguous mountain range, the American Cordillera. We show a combination of dynamical, thermodynamical, and hypsometric factors results in an asymmetric emergence of low-to-no snow conditions within the midlatitudes of the American Cordillera. Low-to-no snow emergence occurs approximately 20 years earlier in the Southern Hemisphere, at a third of the local warming level, and coincides with runoff efficiency declines in both dry and wet years. Prevention of a low-to-no snow future in either hemisphere requires the level of global warming to be held to, at most, +2.5 °C.

. Therefore, it is imperative to understand and estimate how these cross-scale 84 interactions influence the emergence of a low-to-no snow future across the American Cordillera, 85 particularly for instilling resilience into water resource management. 86 Identifying Low-to-No Snow Emergence 87 The inherent challenges in modeling the cross-scale interactions that shape the moun- no snow emergence. The latter definition is based on annual peak snow water equiva-102 lent (SWE) percentiles (see Methods). Next, we assess low-to-no snow persistence, its 103 connection to warming, and identify if there is hemispheric asymmetry. Last, we detail 104 how the mountainous hydrologic cycle is fundamentally altered following the emergence 105 of low-to-no snow through alterations in how water is deposited, stored and transferred 106 downstream. 107 Over the historical period  inter-annual variability in peak SWE is high 108 along the American Cordillera, representative of a weak signal-to-noise of snow loss (Fig-109 ure 1 and S1). The signal of change refers to the trend in peak SWE decline, whereas  Figure S4). This result is shaped by differential changes in annual mean total pre-128 cipitation ( Figure S5) and surface air temperature within individual simulations (Fig-129 ure S6). In the Southern Hemisphere, models project (4/6 simulations) increases in mean 130 peak SWE ( Figure S7). This results from increases in annual mean total precipitation 131 ( Figure S8) combined with weaker, though significant, annual mean surface air temper-132 ature increases compared with the Northern Hemisphere ( Figure S9). However, a sig-133 nificant decrease in mean peak SWE in both hemispheres is found by the two models 134 that project out to 2100, save for certain portions of the Canadian Rockies and north-     • C. This asymmetry in low-to-no snow emergence is not only a function of hyp-252 sometric differences along the American Cordillera, but also changes across spatial scales.

253
At planetary scales, a shift in the atmospheric general circulation (e.g., meridional slow-254 down in the Northern Hemisphere) was found and is related to asymmetric hemispheric 255 warming. At regional scales, changes brought about by alterations to landfalling storm 256 characteristics (e.g., heterogeneous changes to annual total precipitation and ubiquitous 257 reductions in snowfall fraction) and more localized elevation-dependent warming were 258 identified. Our findings suggest that the prevention of low-to-no snow emergence in ei-259 ther hemisphere requires global warming to be limited to, at most, +2.5 • C.

260
In the midlatitudes of the American Cordillera, mountain runoff is critically im-  Table S1. The only model to satisfy Tier 1-3 (MRI-AGCM3-

333
With that said, the analysis was limited by the lack of Earth system models that 334 simulated a future climate scenario to 2100 and output daily SWE to appropriately es- The model horizontal resolutions across the three models used in our analysis spanned 357 ∼250 km to ∼20 km (Table S1). Yet, even at the horizontal resolutions offered in High- in ERA5 and CNRM ( Figure S13 and S16), EC-Earth3P ( Figure S14 and S17), and MRI 385 ( Figure S15 and S18) are provided in the Supplemental Material. The variables analyzed 386 coincide with those presented in the analysis of future projections, namely annual mean 387 surface air temperature, annual total precipitation, peak SWE, and annual total runoff.

388
The variables are also masked with the American Cordillera mask at their native res-389 olutions. Summary statistics, namely annual median, spatial-mean median bias, mean, 390 spatial-mean mean bias, and standard deviation, and the spatial-mean range of bias, are 391 also provided (Table S2 and S3).

392
Relative to ERA5 all HighResMIP models have a cold-bias (mean bias between -393 0.4 to -1.4 • C relative to ERA5's 4.6 • C) in the Intermountain West that amplifies at 394 higher-resolution (Table S2). Most models (4/6 simulations) also produce more precip-  To quantitatively isolate the persistence of low-to-no snow years we develop sev-432 eral definitions and apply them to the HighResMIP simulation over the American Cordillera.

433
Low-snow is defined as peak SWE ≤30th percentile and (virtually) no-snow is peak SWE

444
We define a back-to-back low-to-no snow year as "extreme", five years in a row as 445 "episodic", and 10 years in a row as "persistent". These temporal persistent low-to-no snow (10-years in a row) has yet to occur in the historical record.

457
This would likely be virtually impossible to meet historical water demand assuming no 458 changes to water management practices and infrastructure.

459
Using both the low-to-no snow definition and the temporal definitions, we then estimate, at each latitude band (mean conditions) along the American Cordillera when peak SWE percentiles met (1) or didn't meet (0) the ≤30th percentile.
A common statistical analysis when dealing with data such as Z Extr t , Z Epis t , and Z Pers t is that of logistic regression, which seeks to quantify relationships between some -19-independent or explanatory variable of interest (e.g., surface air temperature [tas]; X t ): X t = tas value (K) in year t, t = 1950 − 2099, and a dependent or response variable (e.g., low-to-no snow; Y t ) that is {0, 1}. For sim-460 plicity of notation, we drop the superscript and simply outline an analysis for Z t , which 461 generically refers to extreme, episodic, and persistent low-to-no snow. Eventually, we will 462 fit a separate logistic regression to each of Z Extr t , Z Epis t , and Z Pers t .

463
Unlike ordinary least squares regression, which quantifies linear relationships between an explanatory variable and some continuous response variable, logistic regression instead quantifies a relationship between an arbitrary explanatory variable and the probability of the response variable being 1, denoted p t = Prob(Z t = 1). However, since probabilities are bounded by 0 and 1, the regression is conducted on the so-called "log odds" of Z t = 1, often referred to as the logit function, which is defined as: Importantly, for p t ∈ (0, 1), the logit(p t ) ∈ (−∞, ∞), which ensures that the estimated probabilities lie between zero and one. In this case, we set up the logit statistical model using the tas time series {X t : t = 1950, . . . , 2099} as an explanatory variable as follows: where we model where X t = the tas value in year t. Numerical methods are used to estimate the statistical parameters {β 0 , β 1 }, call these { β 0 , β 1 }, which can then be used to derive estimated probabilities p t in each year using the inverse logit function We can then use these fitted probabilities to identify a date-of-emergence for each low-      1950-2000 is used as the historical reference period to compute percentile bins and annual mean surface air temperature anomalies. Low-to-no snow conditions are defined as annual peak SWE ≤30th percentile.
-36- Figure S2. Same as in Figure S1, however the magnitude of peak SWE change (in mm) is shown relative to the historical reference period . -37-