I was recently down in Lewiston, Idaho and saw that a cement contractor and placed the message, “Rain go away” on the signage outside their shop. Indeed, this spring might have been too much of a good thing for certain sectors of society. That is not what this post is about. Instead, let’s delve into data. I’ve been a bit skeptical, like a good scientist, of the precipitation and temperature record at the Moscow UI COOP station. It is one of the United States Historical Climate Network stations, and hence has been scrutinized for issues relating to changes in observational practice and station relocations.
Taking a look at the 110 year record we can see a huge increase in precipitation ~ 25% over 100 years. Not conducive to setting cement. Most of us in Moscow acknowledge that Moscow is a bit of an oddity, but a 25% increase in precipitation? Time to bust out the detective lens…
The station has moved a few times in its record according the NCDC MMS system where you can examine the metadata for each station. A nice resource. The station had been located in downtown up until 1960 when it moved to southeast of town (point south of hwy 8), and again moved in 2000 to its current location on the UI experimental farm north of hwy 8. These may not seem like big changes (3 miles east, 30 feet up in elevation); however can be reflected as being non-negligible in the observational record.
We previously noted that this spring’s precipitation was record shattering for a few locales in the inland northwest, Moscow included. However, the increase in precipitation in the graph above does include the multidecadal drought for parts of the western US in the 1920/30s – Idaho included.
So is this increase in precipitation real?
Several other stations in eastern Washington and the Idaho panhandle also show upward trends in precip over the past 60-100 years, particularly during the spring months.
But the increase at the Moscow station is a bit more. To get a sense of the relative change in precipitation for the Moscow station versus some of its neighboring analog stations in eastern WA and the Idaho panhandle I did a double mass analysis. I have performed this sort of analysis (and associated methods) for a set of 200 COOP stations in California that we published in J. Applied Meteorology and Climatology and now serves as the basis for the operational California Climate Tracker. This analysis is typically performed by hydrologists to compare gauge records between neighboring stations as a means of checking for inconsistencies.While there are good reasons for exceptions, at the regional scale (~150km) and in the absence of any large topographic barriers, we should precipitation records to do a decent job at tracking one another (e.g., a wet spring in Moscow better be a wet spring in Pullman, Spokane, etc.). Exceptions to that rule can occur for a year or two (i.e., a big convective downpour over Moscow, avoids nearby gauges), and may be more prevalent for disorganized convective-type precipitation; however the general relationship between precipitation amounts between nearby stations should be relatively stationary through time (at least we think this is the case).
If we assume a relatively stationary relationship between stations through time (with wiggle room for seasonal variations and occasional rouge event or two) there should be a linear fit in the double mass curve (red line shows a proxy for this). Instead we see changes in the slope of this curve, that suggest some nonlinearities. If we instead look at the difference between the assumed linear fit and the observed double mass curve we get the following plot:
Big changes in the slope of the curve suggest changes in the precipitation relationships between stations. Of course, a change in the double mass curve could indicate a change in (a) station #1, (b) station #2, or (c) station #1 and #2. Where there is consistency between independent sites, you have a much better handle on who is the suspect. Here we can see the main change coincident from all three stations in the mid-to-late 1960s, here shown as Moscow getting wetter (or all other stations getting drier, your call). There is a hint of additional changes in the mid 90s as well, but not as pronounced as the mid-late 1960s change.
So is this increase in precipitation real?
Yes, but I can also observe more precipitation if I took my rain gauge further eastward and uphill into the Northern Rockies as we know there is an increase in precipitation due to gentle orographic uplift along a west-east transect on the windward side of the Northern Rockies. At the same time, most long term records in the region show an small increase in precipitation over the last 60-80 years – part of which might be associated with the protracted dry spell at the beginning of the record. So yes, the increase is real, but its likely inflated.
We can see this issue in the temperature record as well, most evident for minimum temperatures as they are most sensitive to landscape position, particularly during the quiescent period of the year in summer/fall.
This is not an isolated issue with the Moscow record, but something that we always are working with. It is always good to have a healthy amount of skepticism about observations and learn from them, rather than simply tossing “bad” data. Rather it is good to think of observations as being “piecewise continuous truths” – Kelly Redmond’s great term. No changes are evident in the metadata (location, means of recording, etc.) at the Moscow UI COOP station, and this an unfortunate issue with climate records. Possible explanations: undocumented station relocations, instrumentation changed, vegetation contamination, changes in measuring practices of snow, sprinklers… I did a quick comparison of the 1 Oct to 31 May precipitation accumulation for the Moscow UI COOP station and four “good” Cocorahs observers located in Moscow proper (myself included). The average of the four in town observers was 15% less than the amount recorded at the official station. Admittedly cocorahs is a citizen science effort, but I’m not sure how to get an extra 4.5 inches of rain between two locations less than a mile apart and at the same elevation.
Methods have been developed to automatically search for abrupt changes in station observations that are not observed in surrounding stations; however, these are not a perfect way of doing this either due to issues associated with “fixing” climate data. Most major global surface temperature datasets have ways in which data is cleaned up to avoid things like missing observations and station moves. However, these are constantly under scrutiny. There is a big effort underway by Berkeley Earth Surface Temperature trends to reanalyze temperature records transparently to see if such data “fixing” has cooked the records and results. Having played with the good, bad and ugly data, my guess is that this will be a wash and the global land surface temperature records will stand the test and support NASA/NOAA/CRU records.
*Thanks to Kelly Redmond, Western Regional Climate Center for discussions on this topic*