Hydroclimatology and Solar Explorations

Climate & Weather Representations

Grounding recent climate assessments into the Colorado River

The various tributaries of the Upper Colorado and its adjacent watersheds often have amazing resonance in their time series signature trends, as the featured image demonstrates.  One gage, the Bear River is near the boundary between Wyoming and Utah and it feeds into the Great Salt Lake Basin, while the other gage, the Animas River is within northwest New Mexico.  Both are taken from the premier (and deservedly so) USGS water information site at [1].

The remaining images compare each of those streams to a recent National Climate Assessment/West Wide Climate Assessment product (WWCA) [2].  I was only trying to reproduce the central image of this figure from [2].

Garfin, Overpeck, and Cayan’s (GOC’s) Bias-Corrected WWCA Climate Figure

Strangely enough the average time series from their source [2b] does not match that central graph of the above image, not even in units.  See the next image below.   Naturally the value of inches in the figure for runoff could be adjusted if we only knew the size of the watershed.  That size is not clearly stated in the GOC doc, but I can assume that for Colorado river basin of the central image column, this size is very large.  In fact, the basin is so large (larger than a few states), that there should be little or no need for any statistical downscaling from global climate models.  Accordingly, there would not be a need for much downscaling and/or bias-correcting [3].  Since the GOC source material doesn’t appear to match  the GOC figure above, that makes the figure irreproducible.  Here are two attempts to match their source lines to their figure.  In this first try, I note that their figure average must be visually estimated from the thick and muddy black line, along with the overall line density of any color towards the latter half of the simulation.  This is confusing because another black line is included apparently to represent some other kind of average beyond the publication date (around 2013).  One is left to assume something about their “average” result beyond that date.  If visual context is all that is permitted then one is left with an impression that on average, the runoff rates will decrease montonically until the end of this century, due to man made climate change.  If we look at their actual model mean values which I overlay in magenta, that monotonic trend cannot be seen.

Also the hazy cloud of other lines surrounding the mean line can be compared to their source model information as well.  They are meant to indicate all of the models the GOC authors ran.  That’s a great deal of data to choke MS Excel with so I’m for now limiting an overlay to the initial period from 1950 to 1975.   I’ve placed that overlay into the inset box at upper left of the image below.

Any can confirm at least two discrepencies.  First note the Bottom of the curves and data.  For their actual model set, the bottom is flat.  For their figure of that set, the bottom wiggles all over the place.  Both cannot be right.  Second, the top peaks shown over this period are of interest.  In their figure, the main peak forms around 1960 and a secondary peak at 1975.  In the actual model set, the two peaks are reversed and the main is therefore at 1975.  Again, the figure and the supporting model records cannot both be right.

I routinely need to compare hydroclimatological products of the western US and so this report is one of many under a performance based sample survey.   I can at least compare the data they did produce to the independent USGS data that they did not.  The following charts provide simple comparisons of the WWCA content of [2b] to an actual USGS upper Colorado River record and to a record from an adjacent watershed [1].

Clearly the Garfin-Overpeck-Cayan resource of [2] fails a trend comparison test for the actual records of the upper Colorado River over our collective lifetimes.  But even if a trend were matched, that wouldn’t necessarily mean anything to me, unless the model came with high accuracy, better resolution, and longer term predictive power.  I’ll revisit over time with more comparisons to this image.  Currently the GOC authors’ work appears to offer poor predictive skill, yet remains a cornerstone of the most recent revision of these reports [4].

This is a good time to also explore the utility of the GOC ensemble of models (the smears of gray, green and orange model runs behind the mean curves in the images before these two curves above) which show such poor fidelity to our streamflow history.  Here’s another excerpt from my excel reviews of their model output data.  Again so my laptop doesn’t choke on this data, I’ve chosen to look at the outputs for the first few years of the simulations.

I could not find the observation based record that their simulations are bias corrected against.  So the demonstration of the skill of their projections has been practically impossible.   In any case, when one averages a group of poor simulations, the resulting mean is also going to be a poor fit.  No party benefits from poor accuracy or transparency, no matter how many times a simulation is varied.

I could be biased myself, and think that the hydroclimatologic model producers of the future could consider coupling the vast computational global climate modeling networks to an alternative solar cycle forced hydro-conceptual model.  They might be surprised at the emergence of accuracy in their models.


[1] USGS surface water historical information at

[2] Cayan, D., M. Tyree, K. E. Kunkel, C. Castro, A. Gershunov, J. Barsugli, A. J. Ray, J. Overpeck, M. Anderson, J. Russell, B. Rajagopalan, I. Rangwala, and P. Duffy. 2013. “Future Climate: Projected Average.” In Assessment of Climate Change in the Southwest United States: A Report Prepared for the National Climate Assessment, edited by G. Garfin, A. Jardine, R. Merideth, M. Black, and S. LeRoy, 101–125. A report by the Southwest Climate Alliance. Washington, DC: Island Press.

[2b] I extracted an average for each year of my profile through the file  streamflow_colo_usbr_mon_00015.cvs

[3] Gangopadhyay, S. and Pruitt, T., 2011.  West-Wide Climate Risk Assessments:  Bias-Corrected and Spatially Downscaled Surface Water Projections.  Technical Memorandum No. 86-68210-2011-01  Water and Environmental Resources Division (86-68200).  U.S. Bureau of Reclamation


Finally I note that for Animas I use trend based on records back to 1931.

Draft informal blog work in progress

1455total visits,2visits today

Leave a Reply