Clouds shuffle across a sullen sky, staring down at a landscape that is waiting in limbo until the autumn rains. The rainwater tank is getting low, plants not heavily watered have shut down growth as they run out of soil moisture to draw on. Rain will come, but when? And how much?
Previously I’ve had a bit of a stab at charting long-term rainfall trends using Bureau of Meteorology historical data for the town, compared to the recent data from Melbourne Water’s nearby weather stations. With a bit more thought and work I think I have improved the analysis. I’ve also included more data.
The result, in short, is that rainfall has been declining over the 137 years since the records began. This appears to be caused mainly by a large decline in autumn rainfall.
I also looked more carefully at whether the south end of town gets less rainfall than the north (as I had previously assumed). It appears that is true and likely to hold with current weather patterns.
There are five sets of data over the years: 1880-1961, 1960-1990, and 1990 on from the Bureau of Meteorology; and the two Melbourne Water sets from December 2005 onward. The set from the 1990s (at the golf course in Darley) has too many months missing to use, apart from the years 1991-98.
Here’s a map showing the locations of the sites where rainfall has been monitored at different times: yellow pins mark the sites (BOM and Melbourne Water).
I compared the data by average rainfall for the decade (or the portion of the decade for which records were available), and also by calendar season averaged over each decade, in the following graphs:
The trend lines (as calulated by LibreOffice, not by me) indicate a long term decline. When broken down by season, it seems that most of the decline in rainfall has occurred in autumn (orange).Just to highlight that, here’s the autumn data on their own:
That’s potentially quite significant, ecologically, as a lot of native plants rely on autumn rainfall to put on a growth spurt before it gets too cold and dark in winter.
Since Melbourne Water has collected data simultaneously from the Parwan Creek and Darley stations, I decided to compare them. Is there really a north-south rainfall gradient across Bacchus Marsh?
The graph above compares the averages for the two sites, across the 11 years of data. Obviously another few years of data could change these averages significantly if they were wetter or drier than the last 11: that is why I have included the error bars, which show the range of Standard Error of the Mean (SEM).
For those who don’t use statistical analysis, the main significance of the SEM error bars is that if they overlap, there is a reasonable probability that the two averages are not significantly different: it’s a way of predicting what the average might be if the same weather patterns continue over time and give us a larger set of data, to calculate a more long-term average. There are far more sophisticated ways to analyse this kind of data than simple error bars, but for now I’m just going to leave it at that.
Anyway, back to the weather report. The comparison of Darley and Parwan Creek for the last 11 years shows that the error bars barely overlap (by less than 1mm of rainfall). I’ll take this as an indication that there probably is a north-south declining gradient to rainfall here.
That leads me to the disclaimer, when we compare rainfall from different recording sites, it is important to remember that the comparisons are possibly not accurate, because differences might be caused by the different location on the rainfall gradient.
The last 3 decades also only have partial data sets, which might upset their averages. However, reproducing the graphs above only up to 1990 still gives the same overall and seasonal trends,and even the wet years of the 1990s had low autumn rainfall, so I’m fairly comfortable with my analysis. It’s not a perfect data set but it’s the only one I have to work with!
I should also add that I haven’t studied the methods that meteorologists use to analyse rainfall trends, which may be quite different from or more powerful than the simple analysis here.