Identifying and Predicting Cool/Wet and Warm/Dry Climate Regime Shifts for the Columbia River Basin

Overview

Multi-decade, bimodal patterns of below and above average streamflow have been shown to occur in the Columbia River Basin since about 1900 (figure 1). These climatic effects are associated with observed tendencies in the PDO climate index, however it is not clear if this is a cause-and-effect relationship, or whether the PDO index simply tracks the global climate dynamics that cause these persistent trends. Predictability of these persistent features of the climate system remains an active topic of research.

Despite these uncertainties, knowledge of the dominant wet or dry pattern is extremely important for streamflow forecasting in the Columbia River basin because the observed inter-annual variability of streamflow associated with ENSO events is biased up and down according to the decadal patterns. El Niño events during the wet decadal phase, for example, are rather ordinary in terms of streamflow outcome (figure 2), whereas El Niño events in dry decadal periods are associated with extremely low streamflow and multi-year droughts in the Columbia basin.

While it is possible to observe changes in these decadal patterns after a sufficient amount of time has past, identifying transitions in real time presents considerable challenges. From 1977 until at least the mid 1990s, the climate for the Pacific Northwest (PNW) has been in a persistent warm dry phase, however observed streamflow in the past several years (and especially water year 1997) has been strongly above average (figure 3). Note the reversals in the trends near the observed PDO regime shifts in 1925, 1947, and 1977. The flows are sufficiently high from 1996-2000, for example, that a statistically significant change in the mean is observed in comparison with the period from 1977-1995. Does this mean that a shift to a persistent long-term wet cycle has occurred? It cannot be demonstrated conclusively that this is so, but clearly identifies a short term change in the climate.

To look for clues about the nature of transitions in decadal variability, the historic record of April-September average naturalized streamflow data from 1900-2001 were examined (figure 4).

click image to enlarge

Figure 4 Standardized Anomalies for April-September Natural Flow at The Dalles

For the 20th century, transitions between decadal-scale patterns appear to be strongly associated with extreme events in opposition to the dominant trend. The figure identifies summer streamflow events more than 1.5 standard deviations above or below the mean in opposition to the dominant trend. With the exception of 1973-1974, in which the pattern appears to flip-flop from dry to wet, streamflow events of this magnitude in opposition to the established trend appear to "herald" the coming decadal shift. Another way to look at this is to consider the observed probability of a high flow event more than 1.5 std deviations above the mean during PDO positive epochs, and of a low flow event more than 1.5 std deviations below the mean during PDO negative epochs. From 1900-1996, the observed probability of the former event is 0.0, and of the latter event is 1/52 or 1.9%.

If these patterns are indeed general and repeating events for the Columbia Basin, it suggests that a regime shift from a dry cycle to a wet cycle has probably occurred around 1997 (an extremely high flow water year), and that variability in Columbia Basin streamflow associated with ENSO events could be biased towards above average streamflow for the next several decades.

Since the winter of 1998-1999, cold coastal ocean conditions associated with cool phase PDO have largely been present (PDO Index), and biological indicators such as salmon recruitment from Washington and Oregon waters (Mantua et al., 1997) have been consistent with cool phase PDO. Despite these indications of a return to a persistent cool wet cycle in the Pacific Northwest, the climate conditions in the winter of 2000-2001 were extremely dry, and such events are extremely uncommon in the cool phase PDO epochs in the rest of the 20th century.

Several hypotheses are suggested:

  1. The wet cycle observed from 1996-2000 is simply a short-term excursion from a predominantly warm and dry pattern that will continue for an indefinite period
  2. The wet cycle observed from 1996-2000 heralds a persistent wet cycle like previous cool phase PDO epochs in the 20th century, but variablity is much greater than in the last cool-wet cycle (1947-1976).
  3. The current climate cycle marks a departure from the relatively consistent patterns of the 20th century, and coming decades will produce different and less predictable patterns of variability.

None of these hypotheses regarding the role of decadal scale variability can currently be rejected. One approach that has been suggested by several researchers (see e.g. Dettinger, M.D., D.R. Cayan, G.J. McCabe, and K.T. Redmond, 2000. Winter-spring 2001 United States streamflow probabilities based on anticipated neutral ENSO conditions and recent NPO status. Experimental Long-Lead Forecast Bulletin 9(3): 55-60) is to make use of the interannual persistence of the PDO index. This approach does not provide any better understanding of the physical mechanisms involved, but does provide a straight-forward approach to dealing with short term excursions from the dominant trends.