planetwater

• RT @highlyanne: We can't really know this. RT @MichaelKodas: NOAA confirms this is a "1,000-year event" in Boulder http://t.co/rjjSo9fMWP … 2013-09-13
• "the longer a cow has been lying down, the more likely that cow will soon stand up," #ignobel 2013-09-13
• and "once a cow stands up, you cannot easily predict how soon that cow will lie down again " #ignobel http://t.co/wTHV63Yg7o 2013-09-13
• RT @mmoyr: Nice! RT @FQXi: Feynman Lectures on Physics now freely available online: http://t.co/VCgfPFQYky 2013-09-13
• RT @EricHolthaus: Nearly 100 hours on, the Big Thompson River is *still* within inches of its all-time record level. #ColoradoFlood http://… 2013-09-17
• RT @hmason: whoa, http://t.co/8WX8WNHSp6 is very useful 2013-09-17

There was a really bad hail storm a couple of weeks ago in Tübingen. Lots of cars are still damaged.

Here is a video taken by the owner of a local punting boat, who got his customers back to (more or less) safe shore just in time.

via Oliver Gassner

• RT @PetersonGIS: Who needs Tissot when you've got http://t.co/2l64QYspkb 2013-09-12

Some years ago, I helped a fellow student doing fieldwork for a day at the Experimental Lake Area (ELA) in northwestern Ontario.

It was a beautiful winter day, we used snowshoes to get to some weirs to get our samples. I remember that I was amazed that those weirs were heated so they would not freeze. I was also amazed about the big sampling network in a relevant real world setting.

In short, the ELA is a set of lakes that are heavily instrumented in various kinds of the water cycle. Hence, the area is quite large. The video below explains that there were some funding issues in the recent past, which seem to be resolved now. The video gives an overview, what the ELA is.

There is an interesting quote in the video that relates about the relevance of real world field work and resulting data: “the results from the whole ecosystem experiment typically were very different from what you would have predicted from small scale experiments”.

• Spuren eines Hochwassers aus dem Jahr 1342 ("Magdalenenhochwasser") noch heute in der Landschaft sichtbar: http://t.co/LvfzgIIgj8 2013-08-30
• RT @AdviceToWriters: Writing…keeps me from believing everything I read. GLORIA STEINEM #amwriting #writing 2013-09-05
• RT @helmholtz_de: … Hätte der Architekt vor dem Bau mal einen Astronomen gefragt! Hintergrund: http://t.co/odFsHguYRc (hk) 2013-09-05

• Is anybody experiencing issues to start #IPython notebook with google chrome on #mac? http://t.co/HZySQV3PLh 2013-08-29
• RT @sc_k: Jim Geelen (University of Waterloo) and his colleagues have proved Rota’s Conjecture https://t.co/QqMmtIfKAy (http://t.co/9Qgyqb1… 2013-08-29

I was playing with the Gamma Distribution Function in python. As typically, I used a combination of bumpy numpy arrays and the scipy.special functions for Gamma and incomplete gamma functions. At one point I realized that the Gamma distribution function (bottom panel on the figure below) was not reaching unity for large x values and for some parameters (k=0.63, theta=0.05). In the given case, the deviation was about 0.3 in probability space, which is significant.

Note: The Gamma Distribution Function (its cdf or pdf) is not the same thing as the Gamma function!

The solution I came up with is to use the module mpmath. When I do, everything looks fine.

I googled only briefly, but did not find any description of this error in scipy.special. Also I am not sure, if using mpmath is the best/fastest/most elegant solution. However, at least it seems to work.

Catherine Devlin demonstrates a fun way to show a flipped coin programmatically in an IPython workbook adapting the html representation of an object to show the correct picture (“heads” or “tails”) of a coin:

class Coin(object):

  def __init__(self, state=None):

      if state in ('heads', 'tails'):

          self.state = state

      else:

          self.flip()

  def flip(self):

      self.state = random.choice(('heads', 'tails'))

  def __repr__(self):

      return "Coin('%s')" % self.state

  coin_faces = {'heads': 'CENT_OBV.jpg', 'tails': 'CENT_REV.jpg'}

  def _repr_html_(self):

      return '<img src="files/images/%s" />' % self.coin_faces[self.state]

Which results in this random output in IPython:

Just a quick heads up: The Water Science Alliance has a blog: the “water science blog“! The goal of the Water Science Alliance is to “Water Science Alliance is to create synergies between the different water research institutions in Germany and the development and elaboration of interdisciplinary thematic clusters of German water research on the national and international level.”

I followed the weather prediction for the city of Stuttgart between June 3rd and June 11th a little closer. Amongst other sites, I checked the GFS ensembles at wetterzentrale for the city of Stuttgart at seven snapshots in time.

The charts show 20 GFS runs in different colours for temperature at 850hPa (top portion) and 6h-precipitation (bottom portion). In the images I flagged June 15th with a vertical orange line.

At the beginning there seems to be quite good agreement in both temperature and precipitation for the next three to four days.

The prediction from June 4th predicts a temperature decrease on June 10/11, together with significant precipitation. This temperature decrease is not anymore predicted on the 5th, but is predicted again on the 7th.

A significant drop in temperature happened on the 14th, together with a precipitation event, which was not predicted until the 10th.

This is only one example, analyzed by a non-meteorologist. But despite ensembles there is a lead-time of about three days. This also did not change when looking at other models.