Atmospheric River Sequences as Indicators of Hydrologic Hazard in Historical Reanalysis and GFDL SPEAR Future Climate Projections

Abstract

When multiple atmospheric rivers (ARs) occur in rapid succession, the combined effect on the hydrologic system can lead to more flooding and damage than would be expected from the individual events. This temporally compounding risk is a source of growing concern for water managers in California. We present a novel moving average‐based definition of AR “sequences” that identifies the time periods of elevated hydrologic hazard that occur during and after consecutive AR events. This marks the first quantitative evaluation of when temporal compounding is contributing to AR flood risk. We also assess projected changes in sequence frequency, intensity, and duration in California using the Geophysical Fluid Dynamics Laboratory Seamless System for Prediction and EArth system Research (GFDL SPEAR) global coupled model. Sequence frequency increases over time and is fairly uniform across the state under both intermediate (SSP2–4.5) and very high (SSP5–8.5) emissions scenarios, with the largest changes occurring by the end of the century (+0.72 sequences/year in SSP2–4.5, +1.13 sequences/year in SSP5–8.5). Sequence intensity and duration both see increases in the medians and extreme values of their respective distributions relative to the historical baselines. In particular, “super‐sequence” events longer than 60 days are projected to occur 2–3x more frequently and to emerge in places that have never seen them in the historical record. In a world where California precipitation is becoming more variable, our definition of sequences will help identify when and where hydrologic impacts will be most extreme, which can in turn support better management of the state's highly variable water resources and inform future flood mitigation strategies.

Temporal compounding increases economic impacts of atmospheric rivers in California

“…The first set of analyses examines whether higher-intensity ARs are more likely than lower-intensity ones to occur in close temporal proximity to other AR events. We compute the probability of AR events occurring within a specified time window, using a 5-day interval between events as the metric of proximity (35,36), and define two types of compounding as follows: "Adjacent" ARs are those with another AR event within 5 days before or after, and "sandwiched" ARs are those with AR events within 5 days both before and after (Fig. 1).…”

“…The probability of impact on non-AR days within a sequence is 5.4% higher than that on non-AR days outside of a sequence. The persistent effect of sequences on impact likelihood even during non-AR days may be related to the delay that can occur between AR events and their hydrologic impacts (35). Figure 4 illustrates these coefficient values and their associated 90% CIs.…”

“…AR sequences were identified on the basis of the method outlined by Bowers (35) and are described briefly here. In each MERRA-2 grid cell, we calculate the 5-day moving average IVT and identify the intervals when the moving average exceeds the 30-year median IVT for that cell.…”

“…There is therefore a growing need to both (i) identify when temporal compounding is contributing to AR impacts and (ii) quantify how compounding intensifies those impacts relative to the expected consequences of individual AR events. Bowers (35) addressed the first of these two needs by proposing a definition of AR sequences based on moving average IVT. Sequences define the time windows of clustered AR activity where temporal compounding is contributing to amplified risk of flooding and negative consequences.…”

“…However, most of the work on AR families is concerned with differences in the synoptic environments between families and individual ARs. Sequences, on the other hand, focus directly on impacts and have been shown to capture time windows of high hydrologic hazard, namely, elevated runoff and soil moisture (35).…”