Head in a Cloud

troposphere and stratosphere meet blogosphere

Head in a Cloud random header image

Where’d the rain go?

October 27th, 2007 by yolanda · No Comments

I presented Abrupt Changes in Rainfall During the Twentieth Century by Narisma, G. T. et al. for the ATOC Journal Club. The authors of this paper wanted to develop a method to identify abrupt decreases in rainfall between 1901 and 2000 using global observations, since there hadn’t been a study done on the droughts of the twentieth century. The assemblage of the occurrences of these abrupt changes in rainfall will play a part in understanding how these changes occur and give an indication of how the climate has been changing over the past century. To identify these changes in rainfall between 1901 and 2000, the authors developed a method to classify events as being persistent, abrupt, and significant. They used the Climate Research Unit’s high-resolution precipitation time series data, which has a 0.5 by 0.5 degree resolution. This data are presented in monthly climate grids, i.e. the data are given in millimeters of rainfall per grid point per month. Within each grid cell, in some areas, there were several stations, but if there was not a reasonable value available for the grid cell at any point during the time period, the data were relaxed to the 1961-1990 average for that month.

There were several steps that the authors made to narrow down the data to include only the droughts of this time period. Starting with the raw data, temporal trends and spatial noise were removed by linearly detrending the time series for each grid cell and then applying a three by three Gaussian filter. Then a wavelet analysis was performed on the detrended and smoothed data. Wavelet transforms can be used to represent a time series in terms of the time versus the frequency determined by the wavelet transformation (similar to that of a Fourier transform). If you want to look at this in a very simplified way, this is a way of comparing the values of each time series to itself. Wavelet analysis is helpful because it helps to identify where there are climatic modes in the data. Using the results from the wavelet analysis, the authors were able to see where there was a decrease in rainfall, which was clearly represented by a switch from high frequency to low frequency. The following is an example of the results from this step for the Sahel, which experienced a well-known 30-year drought that started in the late 1960s.


After all of these cases were identified, decreases in rainfall amount due to station number anomalies were accounted for by removing any regions which had a 50% change in the number of stations around the time of the shift.

Using the remaining data, it was then determined which results were abrupt, persistent and significant. Persistence was defined as how long the drought lasted, in years. The transition period was the amount of time, also in years, that it took to descend into a drought period. The transition time should be less than the persistence, and these two values were compared to determine the abruptness of the event. The greater the difference between these two values, the more abrupt the event. Finally to determine the significance of the rainfall decrease, a t-test was run to compare the difference of the average rainfall amount between the high frequency period and the low frequency period. If there was a 99% significant difference between the two, the event was considered to be statistically significant.

So finally, after all that filtering of the data, this is what they came up with for droughts throughout the world between 1901 and 2000.


To verify the accuracy of their method, the authors used historical data such as tree ring, river flow, forest fire, and agricultural reports to ensure that they were detecting actual droughts accurately. The authors also found that there were a couple patterns within their data. Oftentimes droughts would drift to neighboring regions over time, as was the case in Russia into Eastern Europe. They also noticed that some droughts hung out in the same region and amplified over time. This amplification could be attributed to some areas having more than one equilibrium state. Areas where this is the case are able to jump from state to state rather quickly (even possibly abruptly), and some of the cases in this study may just be highlighting some of those areas. They also noticed that areas that are already arid or semi-arid are more prone to experiencing droughts, as can be seen from the following probability density function. figure3.jpg

There were a couple questions raised about this plot at the Journal Club meeting. The black line is the global climatological mean of precipitation over 1901 to 2000. It tells us that a region receiving x amount of rainfall has y probability of experiencing an abrupt decline in rainfall amount. They grey line tells us the same, but it is only includes the areas that actually experienced abrupt decreases in rainfall. This observation has been linked to the positive feedback between vegetation in these regions. It should be noted that the authors did account for seasonal, annual, and inter annual variability by only considering events that had transition periods that were longer than 36 months.

That’s pretty much it! The authors basically just wanted to figure out a way to statistically determine where there were abrupt rainfall decreases all over the world to hopefully jump-start what the occurrence of all these events say about recent climate change. It would be interesting to research the cause of these events, a change in the frequency of these events over time, and a change in the severity of these events over time. Whoever does research on any of those things, at least has a place to start from: a survey of the droughts of the twentieth century.

Tags: ATOC Journal Club · climate

0 responses so far ↓

  • There are no comments yet...Kick things off by filling out the form below.

Leave a Comment