Monday, April 19, 2010

The "Bolivia effect"



Trying out V1.4 of TempLS, I looked into the Bolivia effect.
This dwells on the fact that Bolivia hasn't reported any temperatures to GHCN for about 20 years, which is supposed to be a gap in our knowledge. It has had a run at WUWT, and I noticed that it popped up on Google as a suggested search, so someone thinks it is important.


So is it really true that Bolivia creates a big gap? Well, it's true that interior S America is one of the poorly covered regions of the non-Arctic world, so info from there would be useful. But let's look at the maps of V1.4.

Update 15 July - there is a new post which looks at data from the GSOD data base. This has about 30 stations reporting from 1990 to present. GHCN seems to do quite well when compared with this better coverage.



Bolivia


The coords of La Paz are about (-16, -68). So let's see what is within a 1200 km radius. The V1.4 criterion to insert is:

  "Boliv" = getdist(-16,-68)<1200 & isLand, # coords of La Paz

Here's what the analysis reports:


This shows stations that have reported at any time since 1900, and there are lots of stations in Bolivia. However, there is a big drop-off in the 90's and indeed, although this doesn't show it, there were none from Bolivia.


The temperatures themeselves don't show a great trend. The early 20C years look quite warm, but based on very few stations.


So let's see a map of more recently reporting stations. That can be retrieved with this selection:

  "Boliv2" = getdist(-16,-68)<1200 & isLand & tv$endyr>2008,


That filters out stations that haven't reported in 2009-10:


OK, so no drop-off in numbers now (as exoected). The plot since 1901, affected early by small numbers, is broadly similar, and the plot since 1979, with many more stations, is quite similar. The trend since 1979 goes up from 0.08 to 0.12 C/dec.



But there was an objection that the high Andes in Bolivia shouldn't be approximated with humid lowland and coastal areas. OK, let's put in an altitude filter instead, for stations>400m:

   "Boliv1" = getdist(-16,-68)<1200 & isLand & tv$alt>400,




The restriction to recent reporting has gone, so the numbers drop in late years. The temps, though are very similar to the earlier all-altitude figures. And the post-1979 trend is almost the same, at 0.08 C/dec.


So let's look at recent stations, and this time take advantage of another field in the temperature.inv file, called "loc". It characterises stations within 30 km of the coast as "CO". So filtering those with:


  "Boliv3" = getdist(-16,-68)<1200 & isLand & tv$endyr>2008 & tv$loc!="CO",




So station numbers are now getting a bit sparse, though not the total absence that the "Bolivia effect" suggests. The region still has lots of info. And the temperature plot since 1901 is still fairly similar, although there is now a jump at about 1992 which boosts the post-1979 trend to 0.24 C/dec.

The point of this post is really to show the kinds of investigation that can be done. But as far as the Bolivia effect goes - well, if you look calmly to see what is there, it isn't ideal, but it's far from disastrous. The imperfections relate much more to the lack of numbers before 1940. Climate scientists can't change that - they have to work with what we've got.

3 comments:

  1. sure beats looking at rows of numbers

    ReplyDelete
  2. ha, feed the bolivian stations to the KML generator.

    then google earth.

    Then record a tour. click each station in order and when you are done you have a kmz file. a recorder tour. people download it, and they take the tour.
    You can even do a voice over. esp if you have one of them funny accents

    ReplyDelete
  3. You're just about there. I'd humbly request that you overlay some of these plots, pre-1990. One data series would be from neighboring high-altitude stations only; the other would also use the stations within Bolivia. Before 1990, we have enough information to more or less test the interpolation that GISS is doing now. The strength of the correlation between these two series will give you an idea of how good you'd expect the current interpolation to be.

    If somebody out there really wanted to put some effort into it: Some recent CLIMATs are now available for Bolivia. You can see them on the JMA page, or the ogimet page. On their own, I don't think they're short records appearing after a very long gap. But somebody could compile other data sources (SYNOP) into monthly means, and see if they can fill in the missing data.

    ReplyDelete