Determining what controls Verbascum prevalence
and measuring the effectiveness of spraying it
Measurements and analysis by Warren Bond
Summary and Conclusions
Based on five years of observations, there is strong evidence that the total rainfall in the months of November, December and January determines the prevalence of Verbascum in the subsequent nine months.
Successive negative deviations from this observed relationship likely indicate success in controlling Verbascum, providing a quantitative performance indicator.
One of fotpin's target weeds for many years has been Verbascum, both V. thapsus (great mullein) and V. virgatum (twiggy mullein). Both are captured in this pop-up photograph taken by Rosemary Blemings in 2007. V. thapsus are the paler plants with thicker flower stalks, while V. virgatum are the darker, thinner plants. Both have the ability to invade an area and proliferate, dominating the vegetation, as the photograph shows.
After several years of grubbing, lopping and bagging of seed heads, comprehensive and systematic spraying of rosettes across the Reserve with Brushoff (active ingredient metsulfuron-methyl) was commenced in 2011. After the first two years there was a noticeable decline in Verbascum present, time taken and amount of chemical used. In order to better quantify this decline, compilation of effort (hours spent spraying) and volume of chemical sprayed was begun for a 1.3 ha heavily infested trial patch that included the area where the photo above was taken. The exact area of the trial patch is shown in the maps below.
Location (above) and detail (right)
of Verbascum trial patch monitored
Record of effort and chemical usage for the example Verbascum patch
(last updated 18 Dec 2016)
It can be seen that after intensive treatment in 2011 the amount of chemical required to control Verbascum in this trial patch had decreased considerably (> 90%) by 2014. In 2015, however, there was a seven-fold increase in the amount of chemical required, bringing it back to 60% of the value in 2011.
Volume of chemical required to be sprayed to control Verbascum in the trial patch,
separated into spraying in autumn and spring.
It was concluded that the large increase in 2015 was the result of a large amount of germination of Verbascum (and many other weeds) in response to the above average rainfall in November and December 2014.
Closer examination of rainfall records and comparing them with the amount of chemical required each year suggested that rainfall in the three months November to January (hereafter called summer rainfall) may be the controlling factor in Verbascum germination and prevalence in the subsequent 9 months. This can be seen by comparing the chart of summer rainfall for the six years of records to date
Summer rainfall is defined as the sum of rainfall in November, December and January. For 2011, it is the sum of rainfall in November and December of 2010 and January of 2011.
The relationship between the chemical volume required to control Verbascum in the trial patch and summer rainfall is more clearly displayed in the chart below. For the first five years, the correlation between the two produces a regression coefficient of 0.985 and a residual mean square (rms) of the deviation from the regression line of the measured chemical volume for each point of just 9 L (less than 10% of the average annual volume sprayed).
Regression of annual chemical volume required against summer rainfall for the trial patch. The regression line is fitted for the first five years only. The volume for 2016 falls well below this line (37% less than the value predicted from the regression).
This good regression suggests two things. Firstly that there is a strong relationship between summer rainfall and chemical volume required for the trial patch during this five year period. Secondly, if this relationship is causal, the chemical volume required in the first five years was determined purely by rainfall; the effect of spraying on Verbascum prevalence was minimal at best.
It can be seen that the chemical volume required for 2016 falls well below the regression line; the deviation is 4 times the average deviation of the previous five years. One possible explanation for this is that the unusual weather conditions in 2016, when the above average summer rainfall was followed by an extended abnormally hot period, may have added a second variable that determines Verbascum germination and prevalence. Another is that after five years the spraying treatment is finally influencing the prevalence of Verbascum. The analysis of similar data for the whole Reserve (below) suggests that the latter may be the case, but further monitoring is required to confirm this.
Application the the whole of the Reserve
(last updated 31 Jan 2017)
Since 2012, Verbascum has been systematically and thoroughly sprayed across the whole of the Reserve. Data similar to that for the trial patch above is therefore available, and shown below.
Regression of annual chemical volume required against summer rainfall for the whole Reserve. The regression line is fitted for 2012 to 2015 only. The volume for 2015 falls below this line and that for 2016 deviates from it even further (see discussion below)
In this case, the regression is fitted to the four years of observations from 2012 to 2015, resulting in a regression coefficient of 0.92 and an rms of 100 L (16% of the average annual volume sprayed). Even though the 2015 data is included in the regression, it falls below the fitted line (by 176 L, or 19%). If it is excluded an even better regression is obtained (not shown; regression coefficient of 1.000 and an rms of 10 L) and the 2015 point falls even further below the line.
The 2016 volume sprayed falls even further below the regression line than the 2015 volume (by 390 L, or 48%) a value that far exceeds the average deviation for previous years (the rms, 100 L) and is very unlikely to be a random result.
This suggests that negative deviations from the fitted line are indeed evidence for the effectiveness of spraying on Verbascum control, because there were no abnormal weather conditions in 2015 or 2016 to cause a decrease in Verbascum prevalence and therefore volume required. The observation that the deviation increased from 2015 to 2016 supports the hypothesis that this is evidence for a systematic decline in Verbascum prevalence and that this is likely to be a result of our spraying efforts.
Why should the Reserve as a whole show evidence for Verbascum response to spraying before the trial patch?
The trial patch is heavily infested with Verbascum and has been for many years, whereas the Reserve as a whole has patches like the trial patch but also areas that have been infested for a shorter time. It is suggested that the latter are likely to be controlled sooner than the former. There is some qualitative evidence for this from observations made during spraying each year; Verbascum has been found to be less likely to recur from one year to the next on the less densely infested fringes of patches than in the their heavily infested centre.
Comparing the regression lines for the trial patch and the whole Reserve
The slopes of the regression lines shown in the charts above for the trial patch and the Reserve are very different. This is partly because the volumes sprayed are for very different areas, 1.3 ha for the trial patch and 133 ha for the whole Reserve. Even when expressed on a per hectare basis there is still a ten-fold difference between them: 0.55 L/ha/mm* for the trial patch and 0.045 L/ha/mm* for the whole Reserve [* L/ha/mm is Litres of spray per hectare per mm of summer rainfall]. This supports, to some extent, the argument proposed for the earlier onset of control in the Reserve as a whole.
While the regression slopes are very different for the two areas, the intercept on the horizontal axis is almost identical: 85 mm for the trial patch and 90 mm for the whole Reserve. These values represent the amount of summer rainfall below which no spraying is required, i.e. Verbascum does not germinate and survive. The fact that these two values, obtained from very different areas, are not significantly different provides support to the robustness of the hypothesis that there is a causal relationship between summer rainfall and Verbascum growth.