Department of Primary Industries, Parks, Water & Environment Poppy Growers Tasmania Inc. Analysis of factors influencing poppy yield June 2013 i Macquarie Franklin Administration Office 112 Wright Street | East Devonport | Tasmania | 7310 Phone: 03 6427 5300 | Fax: 03 6427 0876 | Email: [email protected] Web: www.macquariefranklin.com.au Report author: Thom Goodwin, Lance Davey & Michael Lehman An appropriate citation for this Macquarie Franklin, Date, Analysis of factors influencing poppy report is: yield, Devonport TAS Document status: FINAL This report has been prepared in accordance with the scope of services described in the contract or agreement between Macquarie Franklin and the Client. Any findings, conclusions or recommendations only apply to the aforementioned circumstances and no greater reliance should be assumed or drawn by the Client. Furthermore, the report has been prepared solely for use by the Client and Macquarie Franklin accepts no responsibility for its use by other parties. ii Contents 1 Introduction .................................................................................................................................... 4 1.1 Background & Purpose ........................................................................................................... 4 1.2 Methodology ........................................................................................................................... 4 2 Results ............................................................................................................................................. 5 2.1 Morphine ................................................................................................................................ 7 2.1.1 Summary Statistics .......................................................................................................... 7 2.1.2 Regression Results .......................................................................................................... 8 2.2 Thebaine ............................................................................................................................... 12 2.2.1 Summary statistics ........................................................................................................ 12 2.2.2 Regression results ......................................................................................................... 14 3 Overall Conclusions ....................................................................................................................... 18 4 Limitations..................................................................................................................................... 19 5 Recommendations ........................................................................................................................ 19 7 Statistical Appendix ....................................................................................................................... 20 iii 1 Introduction 1.1 Background & Purpose This report presents the results of a statistical regression analysis that was used to examine the variability of active ingredient yield (kg of alkaloid per hectare) in poppy crops grown in Tasmania. In January 2013 Macquarie Franklin prepared a report for the Department of Primary Industries, Parks, Water & Environment and Poppy Growers Tasmania examining ‘Social and economic factors influencing farmer decisions to grow poppies in Tasmania’. Following the conclusion of that project, DPIPWE and PGT requested that Macquarie Franklin undertake additional analysis to assess factors influencing the active ingredient yield of poppy crops in Tasmania. The goal of this project was to better understand factors influencing the variability of poppy yield so that this information can be used to identify government or industry led research, development and extension activities that may assist poppy growers to increase their profitability. During the planning phase of this project it was envisaged that a wide range of agronomic practices and weather observations would be used as part of the analysis. However, much of the agronomic practice information collected by the two main poppy companies participating in this project (Tasmanian Alkaloids & GSK) was incompatible. This incompatibility means that the focus of the analysis was restricted to estimating the effect of region specific weather variables on active ingredient yield, with a limited number of agronomic practices examined where compatible. 1.2 Methodology Multiple regression analysis was used to statistically examine the influence of weather variability and cultural practices on the active ingredient yield of poppy crops grown in Tasmanian. Given the availability of both time series and cross-sectional data, a technique known as panel regression was used, as this approach directly incorporates both types of data. The aim of this approach was to explain the variability of active ingredient yield both directly and through its individual components. More specifically, the aim was to explain the following variables:  assay (% alkaloid content);  yield (dry tonnes per hectare);  and alkaloid per hectare. Paddock level data on assay, yield, alkaloid per hectare and agronomic practices over the period 2002 to 2011 was obtained from the two major poppy companies (Tasmanian Alkaloids and GSK) and meteorological data (daily rainfall, maximum temperature and minimum temperature for 5 km by 5 km grids across Australia) was purchased from the Australian Bureau of Meteorology. It was envisaged that a wide range of agronomic practices would be incorporated within the regression model. However, the need to aggregate data across both companies to maintain privacy, combined with the limited compatibility of the two datasets, meant that only the following variables were considered capable of analysis: 4  temperature o Average temperature between planting and harvest (degree Celsius) o A measure of temperature variability using the summation of maximum and minimum temperatures differences over the growing period (degree Celsius);  rainfall o Total rainfall received from March 1 to planting (mm) i.e. pre-planting rainfall o Total rainfall received from planting until to December 31 (mm) i.e. growing period rainfall o Total rainfall received from January 1 until harvest (mm) i.e. harvest period rainfall;  length of the growing period (days);  a time trend to allow for varietal improvements and changes in cultural practices;  use of growth regulators; and  an allowance for regional specific factors (soil type, topography, etc.). 2 Results This section presents summary statistics and results from the regression analysis for the combined company data on Morphine and Thebaine varieties. 2.1 Interpretation of summary statistics Summary statistics are presented using boxplots, with each boxplot representing the seasonal distribution of the data. Combined company data on Morphine varieties are displayed for seasons 2001-02 to 2010-11 and combined company data on Thebaine varieties are displayed for seasons 2001-02 to 2008-09. Boxplots have been used to present the summary statistics as they are a convenient way to present variation contained in the data. A guide to interpreting boxplots is presented in Figure 1. As this project is concerned with the variability of poppy yield, presenting the actual values for each of the variables is not necessary. Therefore, the vertical axis on all of the boxplots has been suppressed to preserve confidentiality of individual company data. The values reported on the vertical axis are index values with the average value during the 2001-02 season set equal to one. 2.2 Interpretation of regression results Only the direction and magnitude of the regression results are displayed within the body of this report, numerical regression results are left for the Appendix. Schematics diagrams are used to present the estimated effect of weather and agronomic practice variables. These diagrams report the estimated effects as either ‘Positive’ or ‘Negative’, with ‘–‘ indicating that the effect was found to be insignificant. A series of charts are also used to present the estimated effects of weather variables over a range of possible scenarios. Figure 2 is an example of the estimated effect on assay, yield and alkaloid, of rainfall during the harvest period. Estimated changes in output are set equal to zero at 190 mm of rainfall as this was the average harvest period rainfall within the sample analysed. 5 Regression results should be interpreted as the effect of individual variables holding all the other variables constant. This means that the effect of say, rainfall during the harvest period, is the estimated effect after making a statistical allowance for the different varieties grown and the different regions in which they were grown. Figure 1: Illustrative example of a Boxplot Figure 2: Example of the estimated effect of rainfall during the harvest period 25% 20% 15% t 10% u p t 5% u o n 0% i eg -5% 0 50 100 150 200 250 n a h -10% c -15% -20% -25% rainfall (mm) Morphine Assay Yield Alkaloid Chart range includes the 2nd and 3rd quartile. Average harvest period rainfall in the sample was 190 mm. 6 2.3 Morphine This section presents summary statistics and results from the regression analysis for the combined company data on Morphine varieties. A guide to the interpretation of the boxplots and the regression results is presented at the beginning of Section 2. 2.3.1 Summary Statistics  Average Morphine assay has shown a slight increased over the study period (excluding 2011) and there has been significant variability (Figure 3)  Average yield per hectare for Morphine crops showed some increase over the study period before falling away in later years and there has been significant variability (Figure 4)  Average Morphine alkaloid per hectare showed some increase over the study period before falling away in later years and there has been significant variability (Figure 5) Figure 3: Morphine assay by season (Index: Average 2002 assay = 1) 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 7 Figure 4: Morphine yield by season (Index: Average 2002 yield = 1) 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Figure 5: Morphine alkaloid by season (Index: Average 2002 alkaloid = 1) 5 4 3 2 1 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2.3.2 Regression Results The goodness of fit measures for the Morphine regression models reveal that:  34% of the variation in Morphine assay (%) can be explained by the variables examined. o Weather variables explain 19% of the variation in Morphine assay. o When regional variation is not taken into account only 28% of the variation in Morphine assay can be explained by the variables examined.  45% of the variation in yield (t/ha) can be explained by the variables examined. o Weather variables explain 6% of the variation in yield. 8 o When regional variation is not taken into account only 8% of the variation in yield can be explained by the variables examined.  45% of the variation in alkaloid (kg/ha) can be explained by the variables examined. o Weather variables explain 8% of the variation in alkaloid. o When regional variation is not taken into account only 10% of the variation in alkaloid can be explained by the variables examined. The estimated effect of the time trend, season length and weather variables on assay, yield and alkaloid is displayed in Figure 6. Only the direction of the estimated effects are reported here (Positive, Negative or Insignificant) with the estimated values from the regression model reported in the Appendix. Figure 6: Estimated effect on Morphine varieties of variables considered in the regression analysis Figure 7 illustrates the estimated effect of the time trend variable. This variable has been used as a proxy for varietal improvements and improvements in cultural practices over time. Figure 8, Figure 9 and Figure 10 provide examples of the estimated effects of changes in rainfall (relative to the seasonal average) on Morphine assay, yield and alkaloid. Increases in both pre-season and growing period rainfall above average values are estimated to have a positive effect on assay, yield and alkaloid (Figure 8 & Figure 9). Above average rainfall in the harvest period is estimated to have a negative effect on the output of these variables (Figure 10). 9 Figure 7: Effect of the time trend variable on Morphine varieties 2.5 2 ) 1 = 2 1.5 0 0 2 ( x 1 e d n I 0.5 0 2002 2003 2004 2005 2006 2007 2008 2009 Morphine Assay Time Trend Yield Time Trend Alkaloid Time Trend Figure 8: Effect of pre-season rainfall on Morphine varieties 25% 20% 15% t 10% u p tu 5% o n 0% i eg -5% 0 100 200 300 400 500 600 n a h -10% c -15% -20% -25% rainfall (mm) Morphine Assay Yield Alkaloid Chart range includes the 2nd and 3rd quartile. Average pre-season rainfall in the sample was 433 mm. 10