Compensation by Populations of Rodent Pests to Applied Control Measures in Intensive Cropping Systems

Thesis Title

COMPENSATION BY POPULATIONS OF RODENT PESTS TO APPLIED CONTROL MEASURES IN INTENSIVE CROPPING SYSTEMS

Peter R. Brown

A thesis submitted for the degree of Doctor of Philosophy.

School of Biological, Earth and Environmental Sciences, The University of New South Wales.

Submitted September 2005. ii

Certificate of Originality

I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, nor material which to a substantial extent has been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where any due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis.

I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project’s design and conception or in style, presentation and linguistic expression is acknowledged.

(Signed) …………………………………………………

iii

Acknowledgements

I wish to express my sincere thanks to many people who have assisted me in this endeavour and for their support. I have received invaluable support from Penny and Kieran, and I am eternally grateful. I thank all my colleagues at CSIRO for their assistance with various aspects of this work: Tony Arthur, Lisa Chambers, Stephen Davis, Stephen Day, Micah Davies, David Grice, Lyn Hinds, Jens Jacob, Dean Jones, Alice Kenney, Charles Krebs, Katrina Leslie, Chris Margules, Steve Morton, Roger Pech, Bill Price, Tony Robinson, Grant Singleton and Monica van Wensveen. I am also grateful to Phil Dunbar, Neil Huth and Greg Mutze. I sincerely thank my friends and colleagues in Vietnam (Cảm ơn ông và bà, quá tử đối vối tôi), particularly Mr Tuan, Prof Tuat, Dr Tan, Mrs Hue, Mrs Hoa, Mrs Ha, Mr Kien, Ms Lien, Ms Yen, Mr Hoa (“HP”) and especially Mr Bia Hoi (“mốt, hai, ba, yo!”). I especially thank the hospitality of my farmer friends and families, particularly Mr Coi (and Mrs Coi and Ms Lien), Mr Hung (“it’s raining”), Mr Dung, Mr Thanh, Mr Que, Mr Hung and Mr Manh, and no thanks would be too much for Mr Grabbit. I also thank the following Australian farming families for allowing me to trap mice on their farms: Blacks, Hannigs, Crothers, Hastings, Lesters, Meads, Poles, Saals, Thompsons, Walters, Wilsomores and Ziesemers. Finally, my sincere thanks go to Grant Singleton and Peter Banks. They have challenged me, and I have benefited greatly and enjoyed their intellectual stimulation and discussions. I acknowledge the support provided to me from CSIRO Sustainable Ecosystems, the Australian Centre for International Agricultural Research (ACIAR), and the University of New South Wales. This Thesis was conducted in compliance with the regulations of the CSIRO Sustainable Ecosystems Animal Ethics Committee (01/02 – 15).

Summary

Rodents cause significant damage to crops throughout the world including damage to wheat by feral house mice (Mus domesticus) in southeastern Australia and damage to rice crops by ricefield rats (Rattus argentiventer) in Southeast Asia. Much time and effort goes into controlling these pests, but control is generally applied after damage has already occurred and there is little understanding of how rodent control affects crop yields. This research was designed to understand how rodent populations recover after control, how crops compensate for damage, and to determine the relationship between rodent abundance and yield loss. These findings will lead to better management of rodents through appropriate timing of control, to minimise damage or yield loss of crops and result in better benefit:cost ratios for farmers. Replicated, manipulative, field studies were conducted in Vietnam and Australia. For ricefield rats in Vietnam, after population control had been applied, population abundance on treated sites was lower during summer, but was higher during winter compared to untreated sites. Furthermore, there was an increase in the proportion of juveniles and a decrease in the body weight of animals. The ricefield rats had variable but large home range sizes, allowing for rapid population recovery after control. For house mice in Australia, densities were lower on treated sites after control, there was a bias towards females and a decrease in survival compared to untreated sites. Poor survival was the only demographic characteristic that was exhibited both for rodenticide-induced population crashes and natural population crashes that occur at the end of mouse plagues. Wheat crops compensated when <50% of tillers were damaged early in the growth of the crop, however, when damage occurred later yield loss was proportional to the level of damage imposed. Compensation was through an increased number of tillers produced and through better survival of tillers. A sigmoidal relationship was found between the density of mice and damage to wheat. Most damage occurred at emergence of the crop when densities were >100 mice/ha. These results from Australia and Vietnam highlight the need to conduct rodent management before sowing or before breeding commences, over large areas, and to focus on key refuge and breeding habitats. This will reduce reinvasion and maximise the effect of any control applied. v

COMPENSATION BY POPULATIONS OF RODENT PESTS TO APPLIED CONTROL MEASURES IN INTENSIVE CROPPING SYSTEMS

Peter R. Brown

CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia

This thesis is based on the following publications and manuscripts that are referred to in the text by their Roman numerals:

I. Brown, P. R., Tuan, N. P., and Banks, P. B. (In Press). Movements, habitat use and response of ricefield rats to removal in an intensive cropping system in Vietnam. Belgian Journal of Zoology.

II. Brown, P. R. and Tuan, N. P. (In Press). Compensation of rodent pests after removal: control of two rat species in an irrigated farming system in the Red River Delta, Vietnam. Acta Oecologica.

III. Brown, P. R. (Manuscript). Short- and long-term demographic changes in house mouse populations after control in dryland farming systems in Australia. Manuscript.

IV. Brown, P. R. (2005). The effect of simulated house mouse damage to wheat in Australia. Crop Protection 24, 101-109.

V. Brown, P. R., Huth, N. I., Banks, P. B., and Singleton, G. R. (Manuscript). Relationship between abundance of pest rodents and damage to agricultural crops. Manuscript.

Table of Contents

Thesis Title...... i

Certificate of Originality ...... ii

Acknowledgements...... iii

Summary...... iv

Table of Contents ...... vi

List of Tables ...... x

List of Figures...... xiii

Chapter 1. Introduction...... 1

1.1 Worldwide impact of rodent pests...... 1 1.2 Rodents in SE Asia...... 4 1.3 House mice in Australia ...... 8 1.4 Impact on crops ...... 11 1.5 Response of rodent populations to control or removal...... 13 Key Question 1: How do rodent populations recover from control?...... 14 1.6 Compensation by crops to damage by rodents...... 16 Key Question 2: How do wheat crops compensate for rodent damage? ...... 18 1.7 The relationship between rodent abundance and crop damage...... 19 Key Question 3: What is the relationship between rodent abundance and crop damage or yield loss?...... 20

Chapter 2. Study areas and species ...... 22

2.1 The ricefield rat in Southeast Asia ...... 22 2.2 House mice in Australia ...... 29

Chapter 3. Results and Discussion...... 35

3.1 How do rodent populations recover from control?...... 35 3.1.1 Movements...... 35 3.1.2 Demographic responses ...... 36 3.2 How do wheat crops compensate for rodent damage? ...... 39 vii

3.3 What is the relationship between rodent abundance and crop damage or yield loss? ...... 40 3.4 General Discussion...... 41 3.4.1 Management Implications...... 42 3.4.2 Future directions ...... 44

Chapter 4. Conclusions...... 46

References ...... 48

Study I. Movements, habitat use and response of ricefield rats to removal in an intensive cropping system in Vietnam...... 66

I.1 Introduction ...... 68 I.2 Materials and Methods ...... 70 I.2.1 Study site...... 70 I.2.2 Trapping and radio-tracking...... 71 I.2.3 Implementation of treatments ...... 72 I.3 Results ...... 73 I.4 Discussion ...... 80 I.5 Acknowledgements ...... 83 I.6 References ...... 83

Study II. Compensation of rodent pests after removal: control of two rat species in an irrigated farming system in the Red River Delta, Vietnam...... 87

II.1 Introduction ...... 89 II.2 Methods...... 92 II.2.1 Study site...... 92 II.2.2 Imposition of treatments ...... 93 II.2.3 Population sampling...... 94 II.2.4 Statistical analyses ...... 95 II.3 Results ...... 96 II.3.1 Abundance ...... 96 II.3.2 Breeding...... 100 II.3.3 Proportion of juveniles in the population...... 102 II.3.4 Body mass ...... 105 II.4 Discussion ...... 108 viii

II.4.1 Breeding...... 108 II.4.2 Rates of increase ...... 110 II.4.3 Juveniles...... 110 II.4.4 Body mass ...... 111 II.4.5 Concluding comments...... 111 II.5 Acknowledgements ...... 113 II.6 References ...... 114

Study III. Short- and long-term demographic changes in house mouse populations after control in dryland farming systems in Australia ...... 120

III.1 Introduction ...... 122 III.2 Methods...... 125 III.2.1 Application of rodenticides...... 127 III.2.2 Mouse trapping and demographic assessment ...... 128 III.2.3 Statistical analyses ...... 129 III.3 Results ...... 130 III.3.1 Density ...... 130 III.3.2 Body mass ...... 135 III.3.3 Juveniles...... 138 III.3.4 Sex ratio ...... 141 III.3.5 Recapture rate...... 141 III.3.6 Condition...... 145 III.4 Discussion ...... 148 III.4.1 Efficacy of rodenticides ...... 148 III.4.2 Short-term demographic effects of baiting on mouse populations ...... 149 III.4.3 Long-term demographic effects of baiting and natural decline on mouse populations...... 152 III.4.4 Conclusions...... 154 III.5 Acknowledgements ...... 155 III.6 References ...... 156

Study IV. The effect of simulated house mouse damage to wheat in Australia ...163

IV.1 Introduction ...... 165 IV.2 Materials and methods...... 167 ix

IV.2.1 Mouse trapping ...... 169 IV.2.2 Statistical analyses ...... 169 IV.3 Results ...... 169 IV.3.1 Imposition of treatments ...... 169 IV.3.2 Yield...... 169 IV.3.3 Number of tillers ...... 171 IV.3.4 Yield per tiller ...... 174 IV.4 Discussion ...... 175 IV.5 Acknowledgements ...... 178 IV.6 References ...... 178

Study V. Relationship between abundance of pest rodents and damage to agricultural crops...... 185

V.1 Introduction ...... 187 V.2 Methods...... 189 V.2.1 Validation of clipping experiment ...... 190 V.2.2 Mouse grazing model...... 191 V.2.3 Analysis of damage/density relationship ...... 193 V.3 Results ...... 193 V.3.1 Validation of clipping experiment ...... 193 V.3.2 Mouse grazing model...... 195 V.4 Discussion ...... 202 V.4.1 The damage-density relationship ...... 202 V.4.2 Economic injury level for mice in wheat ...... 203 V.4.3 Further application of the APSIM model...... 206 V.4.4 Summary ...... 208 V.5 Acknowledgements ...... 208 V.6 References ...... 208

List of Tables

Table 1.1. Damage to rice caused by rodents in selected countries of Asia and possible reasons for increase in rodent problems...... 6 Table 1.2. Impact of rodent damage to wheat and rice plants at different stages of growth. Information based on Wood (1971), Buckle et al. (1979), Fulk and Akhtar (1981), Poché et al. (1981), Haque et al. (1986) and Buckle (1994)...... 17

Table I.1. Summary of radio-collared rats in March (non-breeding season) and June (breeding season) at Vinh Phuc, Vietnam. Shown are Control (C1 and C2) and Treatment (T1 and T2) sites, radio-collar frequency, sex, the number of days each animal was tracked, the number of fixes obtained, the fate of the animal, the home range sizes (ha, calculated using 95% and 100% minimum convex polygon, MCP) and home range span (m)...... 75 Table I.2. Habitat selectivity of ricefield rats during March (tillering stage of rice crop; non-breeding season) and June (ripening stage of rice crop; breeding season) for day and night fixes for each habitat, Vinh Phuc province, Vietnam. The selectivity index is calculated by dividing the proportion of observations of rats in each habitat type by the proportion of habitat available. A selectivity value of >1 implies preference while a value of <1 implies avoidance...... 79

Table II.1. Possible changes in four population characteristics after removal of rats from treated sites using TBS. If the response is neutral, then compensation does not occur...... 92 Table II.2. Details of set up of TBS on treated sites from January 2000 until November 2002 (post-treatment period) for each season. TBS were not set up during winter. The number of rats captured in the TBS is the mean ± standard error from two treated sites (see Singleton et al. (2004) for details) with 8-10 TBS set per site. The trapping sessions used for each season are shown (number of months shown in brackets)...... 94 xi

Table II.3. Number of animals of each species captured using live-capture traps on untreated and treated sites from April 1999 to November 2002. There were 44 trapping sessions conducted on U1, 27 on U2, 41 on T1 and 24 on T2...... 97 Table II.4. Summary of generalized linear model table for adult females breeding for R. argentiventer and R. losea showing degrees of freedom, deviance residual, residual degrees of freedom, residual deviance, F value and the P value...... 101 Table II.5. Summary of generalized linear model table for proportion of juveniles trapped in the population for R. argentiventer and R. losea showing degrees of freedom, deviance residual, residual degrees of freedom, residual deviance, F value and the P value...... 103 Table II.6. Summary of ANOVA table for ln-weight (g) for R. argentiventer and R. losea showing degrees of freedom, sums of squares, mean squares, F value and the P value...... 106

Table III.1. Summary of design of studies showing dates when studies were conducted, number of trapping sessions, date of bait application, number of sites used and the main cropping systems involved...... 126 Table III.2. Summary of analysis of variance using linear mixed effects models for ln- density of mice as estimated using the Petersen Method showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets...... 133 Table III.3. Summary of analysis of variance using linear mixed effects models for ln- transformed body mass (g) showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets...... 136 Table III.4. Summary of analysis of variance using linear mixed effects models for juveniles trapped in the population showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets...... 139 xii

Table III.5. Summary of analysis of variance using linear mixed effects models for the sex ratio of mice trapped in the population showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets. Sex ratio was calculated as the proportion of males out of all captures...... 142 Table III.6. Summary of analysis of variance using linear mixed effects models for the relative condition of mice trapped in the population showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets. Relative condition was calculated using regression in Krebs and Singleton (1993) to compare individual mice against a standardised mouse. Calculations for Walpeup and Wudinna were based on the Victorian mallee regression and for Brookstead on the Darling Downs regression of Krebs and Singleton (1993)...... 146 Table III.7. Comparison of changes in population demographic characteristics in response to three different levels of effectiveness of rodenticide baiting for house mice (Mus domesticus) (short-term demographic responses), a natural crash in densities at the end of a mouse plague (long-term demographic changes) and published demographic characteristics for cyclic vole and lemming populations during the crash period (decline phase) from the peak phase to the low phase.....151

Table IV.1. Relative mean percentage yield loss (± SE) for each stage of damage and intensity of damage compared to experimental control plots. Negative numbers indicate a yield gain...... 171 Table IV.2. Relative mean percentage loss in number of tillers (± SE) for each stage of damage and intensity compared to experimental control plots from two replicates only. Negative numbers indicate a gain in tillers...... 173 Table IV.3. Percentage of immature tillers compared to mature tillers counted prior to harvest (160 DAS) from two replicates only (mean ± SE)...... 173

xiii

List of Figures

Figure 1.1. Hypothetical relationships between the density of rodents and damage (% yield loss) to crops. Four potential relationships are presented. Type I shows sensitivity to rodent damage, where relatively high levels of damage occur at low rodent densities; Type II shows a linear relationship between rodent density and damage up to 100% yield loss; Type III shows a sigmoidal relationship where there is tolerance to damage or compensation of the crop at low densities of rodents, but as rodent densities increase, the level of damage incurred increases faster until a plateau is reached at 100% yield loss; and Type IV shows an exponential growth response where the crop can compensate for up to moderately high densities of rodents, but then the level of damage increases rapidly at high densities...... 20

Figure 2.1. Distribution of Rattus argentiventer through Southeast Asia. The study site in Vinh Phuc Province, Red River Delta, Vietnam is shown (x). This coincides with the northern limit of distribution of R. argentiventer (red shading). Figure adapted from Aplin et al. (2003)...... 23 Figure 2.2. Location of study sites in relation to the city of Hanoi and the Provinces of northern Vietnam. Shown are the location of the study sites, main roads (red), lakes and rivers (blue) Provincial boundaries (dashed lines), and the location of the Noi Bai international airport. Inset: Map of Vietnam showing location of Hanoi. 24 Figure 2.3. Schematic layout of the approximate location of the four study sites at Me Linh District, Vinh Phuc Province, Vietnam. There were two treatment sites (where farmers conducted ecologically-based rodent management including TBS) and two control sites (or untreated sites; where farmers continued their typical rodent management practices). Shown are main roads (red line), rivers, streams and major irrigation supply channel (blue line), and villages (light grey)...... 25 Figure 2.4. Rattus argentiventer photographed in Indonesia. R. argentiventer is a medium sized rat with orange-brown dorsal fur with black flecks. The belly fur ranges from dull white to silvery white. Adults typically weigh up to 220 g. Photographs courtesy of Grant Singleton (© CSIRO Australia)...... 26 xiv

Figure 2.5. Typical scene of farming system at Vinh Phuc Province, Vietnam, showing (top) small fields of tillering rice near to field containing vegetables and (bottom) maturing rice next to an irrigation drain and another vegetable field...... 27 Figure 2.6. Long-term average monthly rainfall (mm) and minimum and maximum temperatures (º C) at Vinh Phuc Province, Vietnam. Spring rice crops are planted in February and harvested in early June, summer rice crops are planted early July and harvested mid-late September and winter crops are grown from September through to February...... 28 Figure 2.7. Mus domesticus photographed in a field setting and in a piggery during a mouse plague. The body weight of an adult mouse is 15-20 g. Photographs taken by Peter Brown (© CSIRO Australia)...... 30 Figure 2.8. House mice, Mus domesticus, have an Australia-wide distribution, but mouse plagues generally occur only in the grain growing regions of southern and eastern Australia (bounded by the red line). The grain growing belts of western Australia, southern and eastern Australia are indicated showing different levels of average wheat yields. The different levels of shading represent average yields for different regions (data sourced from the Australian Bureau of Statistics 2003). Widespread mouse plagues do not occur in Western Australia, but small isolated outbreaks have occurred. The four locations mentioned in the Thesis are shown. Figure adapted from Singleton et al. (2005a)...... 31 Figure 2.9. Typical views of dryland winter cereal farms at Walpeup (top), Wudinna (centre) and Brookstead (bottom). Photographs taken by Peter Brown (© CSIRO Australia)...... 33 Figure 2.10. Long-term average monthly rainfall (mm) and minimum and maximum temperatures (º C) at (a) Walpeup, Victoria, (b) Brookstead, Queensland, and (c) Wudinna, South Australia...... 34

Figure I.1. Abundance of rats (number of rats captured per 100 trap nights) on four sites used for a separate village-level study, Vinh Phuc, Vietnam from January to July 2002. Shown is approximate timing of the spring rice crop and rat breeding season (based on pregnant and lactating adult females) (P. R. Brown and N. P. Tuan unpublished data)...... 74 xv

Figure I.2. Box plot of 95% minimum convex polygon home range sizes (ha) for males and females, in treatment and control sites for March (non-breeding season) and June (breeding season). The box encloses the 25th and 75th percentiles, the solid line shows the median and the dotted line the mean home-range size. Vertical lines span the 10-90th percentiles. Sample sizes are shown at the top...... 76 Figure I.3. Habitat use of ricefield rats (sexes and sites combined) showing the percentage of habitats available to rats and the percentage of radio-telemetry location fixes within each habitat type for day and night fixes, Vinh Phuc Province, Vietnam 2002. (a) March, the non-breeding period for rats during the tillering stage of rice, and (b) June, the breeding period for rats during the ripening stage of rice. Error bars represent standard error of means from habitat use of individual rats...... 78 Figure I.4. Box plot of distances (m) moved by rats from first 2 days to last 2 days of tracking on Treatment (T) and Control (C) sites in March and June 2002. The box encloses the 25th and 75th percentiles, the solid line shows the median and the dotted line the mean distance travelled, and the vertical lines span the 10-90th percentiles. Sample sizes are shown at the top...... 79

Figure II.1. Abundance of rats for (a) Rattus argentiventer and (b) R. losea in untreated and treated sites (mean ± SE) from April 1999 to November 2002, Red River delta, Vietnam. The pre-treatment period was from April 1999 to January 2000 after which the rodent management treatments were imposed on treated sites. Vertical cross-hatched bars represent when TBSs were set up on treated sites. TBSs were established 3-4 weeks prior to the sowing of the surrounding rice crops and were dismantled when the crop inside the TBS was harvested, about 3-4 weeks prior to harvesting of the surrounding rice crops. The shaded bars at the bottom of the graph represent the approximate timing of the spring and summer rice crops and winter vegetable crops...... 98 Figure II.2. Abundance of rats captured per month during spring and summer (when TBS were set up), and winter (when no TBS were set up) on treated and untreated sites for R. argentiventer and R. losea (back-transformed means ± SE). Data are from the post-treatment period from February 2000 to November 2002, Red River xvi

delta, Vietnam, excluding periods during land preparation when no TBS were in place (see text for details)...... 99 Figure II.3. Average monthly exponential rates of increase for treated and untreated sites post-treatment only. The shaded bars at the bottom of the graph represent the approximate timing of the spring and summer rice crops and winter vegetable crops. Data are from the post-treatment period from February 2000 to November 2002, Red River delta, Vietnam...... 100 Figure II.4. Predicted proportion of adult females in breeding condition for R. argentiventer and R. losea from pre-treatment samples (April to February 2000) and from post-treatment treated and untreated sites (March 2000 to November 2002), Red River Delta, Vietnam. The shaded bars at the bottom of the graph represent the approximate timing of the spring and summer rice crops and the winter vegetable crops...... 101 Figure II.5. Predicted proportion of juveniles trapped for R. argentiventer and R. losea from pre-treatment samples (April to February 2000) and from post-treatment treated and untreated sites (March 2000 to November 2002), Red River Delta, Vietnam. The shaded bars at the bottom of the graph represent the approximate timing of the spring and summer rice crops and the winter vegetable crops...... 104 Figure II.6. Predicted mean body mass (g) of male and female R. argentiventer and R. losea from pre- treatment samples (April 1999 to February 2000) and from post- treatment treated and untreated sites (March 2000 to November 2002), Red River Delta, Vietnam. The shaded bars at the bottom, of the graph represent the approximate timing of the spring and summer rice crops and the winter vegetable crops...... 107

Figure III.1. Location of the three study sites used. Brookstead on the Darling Downs (Queensland), Walpeup on the Central Mallee (Victoria), and Wudinna on the Eyre Peninsula (South Australia)...... 125 Figure III.2. Change in mean ln-density of mice per hectare (± SE) as estimated using the Petersen Method from pre-treatment to post-treatment from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995), and Wudinna (South Australia, 2002). ..134 xvii

Figure III.3. Change in mean body mass of mice (g ± SE) for baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002)...... 137 Figure III.4. Percentage of juveniles in the mouse populations (mean ± SE; transformed using the arcsine square-root transformation) on baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002)...... 140 Figure III.5. Sex ratio of mice (mean ± SE, males/total captures) on baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002)...... 143 Figure III.6. Recapture rate of mice (mean ± SE) between trap sessions on baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002)...... 144 Figure III.7. Relative condition (mean ± SE) of mice on baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002). Relative condition was calculated using regressions in Krebs and Singleton (1993; see text for details)...... 147

Figure IV.1. Effect of simulated mouse damage on grain yield (g/plot), Ginninderra Research Station, Canberra. Shown are means ± SE for each intensity of damage and crop stage when simulated damage occurred. Columns with different letters are significantly different as determined by Tukey’s multiple comparison tests (P > 0.05)...... 170 Figure IV.2. Effect of simulated mouse damage on number of tillers per plot, Ginninderra Research Station, Canberra. Shown are means ± SE for each intensity of damage and crop stage when simulated damage occurred. Data from two replicates only...... 172 Figure IV.3. Ratio of number of tillers counted prior to harvest and the number of tillers remaining after treatment for damage imposed at the tillering, booting and ripening stages. Data from emergence were not included because they consisted of counts of plants only. A value of 100 means that 100% of tillers that were present after treatment were counted at harvest. Shown are means ± SE for each intensity of xviii

damage and crop stage when simulated damage occurred. Data from two replicates only...... 174 Figure IV.4. Yield of grain per tiller (g), calculated by dividing the yield per plot by the number of tillers counted prior to harvest. Shown are means ± SE for each intensity of damage and crop stage when simulated damage occurred. Data from two replicates only...... 175

Figure V.1. Yields (kg/ha; mean ± SE) observed from field experiment conducted at the Ginninderra Research Station (Brown 2005) and predicted yield from the APSIM model. Simulated mouse damage was applied at four crop stages (emergence, maximum tillering stage, booting and pre-harvest) and at four levels of intensity (5, 10, 25 and 50%)...... 194 Figure V.2. Extrapolation of the APSIM model from the validation of the clipping experiment conducted at Ginninderra covering all ranges on intensity of damage (0- 100%) and all growth stages of the wheat crop (sowing to harvest). One-off damage was applied in the APSIM model at different levels of intensity at each crop stage. Yield is expressed relative to a maximum yield of 4.2 t/ha...... 195 Figure V.3. Comparisons of yields (kg/ha) observed in the field at the Mallee Research Station and predicted yields from the APSIM model to validate the experimental simulation for 1984-1997. Outbreaks of mice occurred in 1984, 1987/88, 1993/94 and 1997 (indicated by an *), so only yield data for non-outbreak years were analysed...... 196 Figure V.4. (a) Changes in mouse density (mice/ha), (b) amount eaten by the mouse population day-1 (kg), (c) fraction of crop eaten by the mouse population, and (d) yields (kg/ha) modelled using the APSIM mouse grazing model with and without mice. The mean density of mice from sowing to harvest also is shown...... 197 Figure V.5. Relationship between mean mouse density (mice/ha from sowing to harvest) and damage to wheat crops (% yield loss) as estimated using the APSIM mouse grazing model from Walpeup, Victoria, 1983-2003. Years where significant damage occurred or where moderate to high densities of mice were present are indicated. The solid line represents the linear regression and the dashed line represents the exponential relationship using a 3-parameter exponential growth curve (see text for details)...... 198 xix

Figure V.6. Comparison of densities of mice/ha at four crop stages (emergence ~ 10 DAS, panicle initiation ~ 81 DAS, booting ~ 135 DAS, and ripening ~ 168 DAS) related to mouse damage (% yield loss) and mean density of mice/ha as estimated using the APSIM model from Walpeup, Victoria, 1983-2003. The dashed line represents 100 mice/ha and 5% yield loss...... 200 Figure V.7. Relationship between mean mouse density (mice/ha from sowing to harvest) and damage to wheat crops (% yield loss) as estimated using the APSIM mouse grazing model from Walpeup, Victoria, 1983-2003, for non outbreak years (black circle: 1983, 1985, 1986, 1989-1992, 1995, 1996, 1998, 1999, 2002 and 2003), the 2nd year of 2-year outbreaks or the 1-year outbreaks (grey square: 1984, 1988, 1994, 1997 and 2001), or 1st year of 2-year outbreaks (white triangle: 1987, 1993 and 2000). The solid line represents a 3-parameter sigmoidal curve excluding data for 1st year of a 2-year outbreak (see text for details)...... 200 Figure V.8. Modelled relationship between density of mice and damage (% yield loss) to wheat crops. A range of mouse densities were held constant from sowing to harvest and were run through the APSIM mouse grazing model (dark line) and a 4- parameter Chapman regression was fitted to these data (light line; see text for details)...... 201 Figure V.9. Modelled relationships between (a) pest damage (fraction on crop eaten) and relative yield and (b) mouse density (mice/ha) and pest damage (fraction on crop eaten) using the Walpeup-APSIM model...... 207

Chapter 1. Introduction

1.1 Worldwide impact of rodent pests

Rodents remain one of the curses of mankind. For thousands of years they have been causing damage to crops, stored grain and infrastructure, and been reservoirs for devastating human diseases such as plague and typhus. Rodents continue to cause serious damage despite advances in methods of control and management techniques. The family Rodentia is one of the most widely distributed families of animals throughout the world. There are more than 2700 rodent species described, and the Rodentia family accounts for 42% of all mammal species (Aplin et al. 2003; Macdonald 2001). Few of these species have been well studied outside the laboratory. The success of the pest rodent species can be attributed to their short life-span, high rates of fecundity (high litter sizes, short gestation period and short duration to sexual maturity), high dispersal distances, complex social hierarchy, and their physiology and body structure that allow them to live in a wide range of environments (Prakash 1988; Buckle & Smith 1994). In some cultures, rodents are held in high regard. In China and Japan, the clever and quick-witted rat is considered a symbol of good luck and wealth. In Vietnam, rodent meat is eaten to bestow good fortune and fertility on the newly-weds. In Myanmar, the rat is one of eight calendar figures – people born on a Thursday carry the rat as their figure and Buddhists visit pagodas to worship their birth figure. In India, there are a few temples where the local rat species is not allowed to be harmed, and people freely feed and provide milk for them. In general, however, rodents are held in poor esteem and considered major pests because of the devastating impacts caused by approximately 10% of the species belonging to the Rodentia family. Rodents are among the world’s most important pests (Prakash 1988; Buckle & Smith 1994; Singleton et al. 1999a) by causing significant damage to agricultural production throughout the world (Elias & Fall 1988; Hoque et al. 1988; Lund 1988; Marsh 1988; Fiedler 1994; Singleton 2003). It has been estimated that just 5-10% of rodent species are major pest species of agricultural and urban areas (Stenseth et al. 2003), however, it is in the poorer developing countries that the impact of rodents is 2

generally greatest. In Indonesia, rodents are considered the most important pre-harvest pest of rice, causing around 15% losses annually (Geddes 1992) and in Tanzania, rodents cause as estimated 5-15% loss of maize (Leirs 2003). Rodents also cause problems through disease transfer to humans and to their livestock. Rodents are the carriers of human diseases (rodent-borne zoonoses) such as the plague, rat-borne typhus, hantaviruses, and arenaviruses (Gratz 1994; Mills 1999). Rodents carried parasites that were responsible for the Black Death (Yersinia pestis), which killed 20 million people from 1347-1350 (between 30-50% of Europe’s population) (Cantor 2001). Rodents also transmit leptospirosis, a disease with a global distribution that has had major impacts in Indonesia, Thailand, Vietnam, Australia and the Pacific Islands in recent years (e.g. Tangkanakul et al. 2005). In developing countries mortality rates range from 2-10% of those infected, but it is easily treated with antibiotics if diagnosed early. The disease affects many rural communities, especially in developing countries and is often misdiagnosed as malaria or dengue fever (because of the influenza-like symptoms). Most farmers come in contact with leptospirosis through infection as they work in their fields. In 2000, there were 14,285 cases of leptospirosis with 362 deaths reported in Thailand (Tangkanakul et al. 2005). There is much known about some aspects of the biology and behaviour of rodents, especially through studies of the laboratory rat (Rattus norvegicus) and the laboratory mouse (Mus domesticus)1. Laboratory mice have advanced our understanding of genetics and reaction systems for mammals and provide an important model for disease and for drug design and monitoring (Berry & Scriven 2005). Despite this good understanding, there is comparatively little knowledge of how wild rodents interact with their environment. This lack of knowledge is particularly evident for rodents that cause damage to rice crops in Asia. Plagues of house mice (Mus domesticus) cause tremendous damage in the grain- belt of southeastern Australia, and the ricefield rat (Rattus argentiventer) is one of the top three agricultural pests in Southeast Asia. Farmers and governments expend much effort on reducing the impact of rodents, yet little is known about what happens to rodents that survive these control efforts, the processes that influence the level of

1 The house mouse, Mus musculus domesticus, will be referred to as Mus domesticus throughout this Thesis, after Schwartz & Schwartz (1943; see Singleton & Redhead 1990 and Berry & Scriven 2005 for details) 3

invasion by animals from nearby untreated areas, or how often control should be applied? The focus of research for this Thesis is directed towards gaining a better understanding of how rodent populations recover from control campaigns and of the interaction between rodents and their impacts on crops. Such knowledge will lead to better management of rodents through appropriate timing of control, which in turn will minimise damage or yield loss of crops. This will further develop the concept of Ecologically-Based Rodent Management (EBRM) that underlies the research conducted by the Rodent Research Group, CSIRO Sustainable Ecosystems (Singleton 1997; Singleton & Brown 1999; Singleton et al. 1999a). This approach is now being considered for other rodent pests species in a range of countries and situations (Leirs 2003; Meerburg et al. 2004), and so the findings of this Thesis should have broad applicability. All to often there is a focus on reducing rodent numbers without understanding what effect this has on the damage to the crop, or indeed how the crop might compensate for rodent damage. These interactions need to be studied in concert: there are few examples where this approach has been used. This Thesis pieces together some important aspects of the understanding of the response of rodent species to control, the response of crops to damage attack by rodents and the relationship between the density of rodents and their subsequent damage to crops. The research was undertaken in Australia and in Southeast Asia (Vietnam), based on projects funded by the Bureau of Rural Sciences, the Australian Centre for International Agricultural Research and CSIRO Sustainable Ecosystems. This Thesis is laid out in two main parts. Part 1 (Chapters 1-4), provides: • Chapter 1: background to the problem of rodents and develops some general key questions and hypotheses, • Chapter 2: description of the Study Sites and Species of rodents studied, • Chapter 3: description of the Results and Discussion to summarise and integrate the results generated from the data chapters/manuscripts in Part 2 of the Thesis, and, • Chapter 4: a Conclusion to bring the different aspects of the Thesis together. Part 2 of the Thesis is presented as five data chapters/manuscripts (Studies I-V). These have been written as stand-alone manuscripts that are ready for submission to a journal (Study III & V), in press (Studies I & II) or already published (Study IV). A 4

range of approaches were used in this Thesis from (1) direct field observation and intervention (Studies I, III and IV), (2) analysis of existing data sets that had not previously been examined to test specific hypotheses (Studies II and III), and (3) modelling of data (Study V).

1.2 Rodents in SE Asia

In Asia, rodents are considered the number one pest of rice in many countries and also the pest over which farmers have least control (Singleton & Petch 1994; Brown et al. 1999; Leung et al. 1999; Parshad 1999; Schiller et al. 1999; Singleton 2003). Rice is grown on more than 150 million ha and is the staple food for half the world’s population. Ninety-seven percent of rice is grown in Asia (IRRI 1997). In Asia, even if rats damaged as little as 5% of rice, it would be enough to feed 181 million people for one year (based on annual rice production of 540 million tonnes per year, assuming each person ate 150 kg of rice per year and the daily calorie intake was 32%) (IRRI 1997). The ricefield rat (R. argentiventer) has been ranked as the most important pest in Indonesia since 1983 (Singleton & Petch 1994; Leung et al. 1999). A survey showed that rats cause pre-harvest losses of around 17% per year to rice crops (Geddes 1992). In Vietnam, rodents are considered one of the three most important problems faced by the agricultural sector (Huynh 1987). In the uplands of Laos, smallholders regard weeds, soil fertility and rodents as their most important constraints on production (Schiller et al. 1999). In China, rodent control is listed in the top three priorities for plant protection (Zhang et al. 1999). The rodent problem is increasing in rice-based farming systems of some countries in Asia (Table 1.1). The most likely reasons for this are increases in the area and intensity of rice production and asynchronous planting of crops (Singleton & Petch 1994; Singleton 2003). New varieties of rice can be grown for a shorter duration and have higher yields, which enables farmers to plant more crops per year if conditions are suitable. Some of the major pests species of rat regulate their breeding to the stage of rice crop. In Malaysia, Lam (1980, 1983) showed for R. argentiventer that if there was one rice crop per year, there was one breeding season, and if there were two rice crops per 5

year, there were two breeding seasons. In Vietnam, the ricefield rat, R. argentiventer, and the lesser ricefield rat, R. losea, both synchronise their breeding to the reproductive stages of the main spring rice crop, but breeding remains high through the summer rice crop (Brown et al. In Press). Therefore, if there are three rice crops per year and if the rice crops are not planted in synchrony, then the breeding seasons would be extended considerably. Furthermore, if there is a short period between breeding seasons, the survivorship of rats will be higher. Climatic factors also have been cited as a reason for rodent problems in some years. This includes dry years and when conditions are favourable for the flowering of certain types of bamboo. Bamboo flowering has been implicated as a trigger for eruptions of rodents in some Asian countries (Chauhan & Saxena 1985; Nag 1999; Douangboupha et al. 2003). However, there are few good studies of historical data on the impact of rodent populations to determine whether climatic conditions have led to more rodent problems in recent decades. Jaksic and Lima (2003) examined the impact of climatic factors and/or bamboo flowering and seeding on South American rodent populations and found some rodent eruptions were association with bamboo blooming and many were associated with rainfall peaks. There is a paucity of good ecological data available to enable this relationship to be examined in detail for SE Asia (Singleton & Petch 1994).

Table 1.1. Damage to rice caused by rodents in selected countries of Asia and possible reasons for increase in rodent problems. Area of Rice Reason for change in Country rice production Rodent species Level of damage Reference damage (000 ha) A (000 t) A Bangladesh 10,470 29,857 Bandicota indica, B. Varies, but >50% Not stated Karim et al. (1987), bengalensis in some districts Catling et al. (1988), (68% in 1987, Islam et al. (1993), 32% in 1988) Singleton (2003) Cambodia 1,961 3,800 Rattus argentiventer, B. Varies (>20%) New varieties, more Jahn et al. (1999), indica, R. exulans dry season rice Singleton (2003) China 31,720 200,499 R. nitidus, R. norvegicus, R. 0.6 – 5.3% (10% Climate change, Zhang et al. (1999), rattoides, R. tanezumi, R. in 1995) increase in rice Singleton (2003) losea, Microtus fortis, B. production indica, Mus musculus, Apodemus agrarius India 44,800 131,200 B. bengalensis, Mus booguda, Varies (generally Increased area of Parshad (1999), R. rattus brunneusculus, R. 5-15%; periodic irrigation, Singleton (2003) nitidus, Millardia meltada outbreaks 30- asynchronous planting 100%) (flowering of bamboo?) Indonesia 11,624 49,534 R. argentiventer 17% (15-25%) Increase in rice Geddes (1992), production, asynchrony Leung et al. (1999), of planting Singleton et al. (2003) Laos 718 2,103 Lowland: R. exulans, R. rattus Lowland: Unusually dry years, Singleton & Petch B. indica, Significant flowering of particular (1994), Schiller et al. Upland: R. exulans, R. losea, (unquantified) 15% bamboo species (1999), R. rattus, Mus spp, B. indica (?) Khamphoukeo et al. Upland: periodic (2003), Aplin et al. outbreaks 30-100% (2003) 6 7

Table 1.1. Continued Area of Rice Reason for change in Country rice production Rodent species Level of damage Reference damage (000 ha) A (000 t) A Malaysia 674 1,934 R. argentiventer, R. exulans 5% (1-11%) Damage often patchy; Hoque et al. (1988), high losses near Lam et al. (1991), abandoned fields Singleton & Petch (1994) Myanmar 5,300 16,400 B. bengalensis, B. savilei, B. 5-40% Increased area of Aplin et al. (2003), indica, R. rattus, Mus spp. double-cropping Singleton (2003) Philippines 3,978 11,388 R. rattus (tanezumi?), R. 1.65-4.53% Level of damage Hoque et al. (1988), argentiventer, R. exulans, R. Reports in 1999 believed to be low Singleton (2003) norvegicus and 2000 of regions with >50% loss Thailand 10,000 23,272 B. indica, B. savilei, R. 6-7%, up to 18% in Reduction due to more Singleton & Petch argentiventer, R. losea, R. 1977, coordinated control (1994), Boonsong et rattus, M. caroli, M. 1.5% in 1993 methods al. (1999) cervicolor Vietnam 7,648 31,394 R. argentiventer, R. losea 10-30%, but up to Increased area and Singleton & Petch 100% intensity of rice (1994), Brown et al. production (1999) Notes: A Source of data IRRI (1997)

7 8

1.3 House mice in Australia

It is generally believed that house mice were introduced to Australia with European settlers (Singleton & Redhead 1990; Redhead et al. 1991). It is likely that there have been numerous introductions to different localities around Australia and to some the offshore islands. There have been well documented cases of recent introductions such as house mouse on to Thevenard Island, off the Western Australian coast, in the late 1980s (Moro 2001). The house mice present in Australia would most likely have come from England, the Netherlands or France and would therefore probably be Mus musculus domesticus, rather than Mus musculus musculus (which occurs further west in Europe) (Sage et al. 1993; Payseur & Nachman 2005). Domestic and feral populations of house mice have done extraordinarily well in Australia by inhabiting almost all available ecological niches. They have done particularly well in highly modified agricultural landscapes, where native rodents have faired poorly (Redhead et al. 1991). Some native rodent species, however, can form plagues, but these occur after favourable climatic conditions in the arid interior of Australia (Newsome & Corbett 1975; Masters 1993; Predevec & Dickman 1994; Dickman et al. 1999). Feral mice have exploited the highly modified agricultural environments to occasionally reach high densities and to cause significant crop damage and losses. In Australia, feral mice have had the advantage of not having specific predators (although there are a range of predatory birds, mammals and reptiles that feed on mice), no small mammal competitors in the crop ecosystems, and they do not have the full suite of diseases that their forebears have back in Europe (Singleton & Redhead 1990; Tattersall et al. 1994; Singleton et al. 2005a). A mouse plague occurs somewhere in Australia once every four years, but could be one year in seven for any particular region (Singleton 1989; Redhead & Singleton 1988; Mutze 1991; Singleton et al. 2005a). These plagues of house mice are a significant problem to agricultural areas of Australia. It has been conservatively estimated from a survey of grain-growers in Victoria and South Australia that the 1993 mouse plague cost AUD$64.5 million (Caughley et al. 1994). Within the wheat belt of southern and eastern Australia there are a number of regions defined by different soil types, cropping systems and climate. Yet each is subject to mouse plagues. On the Darling Downs in southern Queensland for example, winter and summer crops are 9

grown on a continuous basis on self mulching dark clay soils, whereas on the light sandy loam soils of the Victorian Mallee winter cereals are grown in the same paddock only once every 2-3 years. The mechanisms of plague formation in these regions differ markedly (see Singleton 1989; Cantrill 1992; Pech et al. 1999). Curiously widespread mouse plagues do not occur in Western Australia, although localised outbreaks often occur (Plomley 1972; Chapman 1981). In 2003, high densities of mice were recorded for the first time in Tasmania, where they caused some damage to winter cereal crops and farmers used rodenticides to limit damage (M. Statham personal communication). Mouse plagues occur through a combination of factors. One key factors is the timing of rainfall (Saunders & Giles 1977; Mutze 1989b; Brown & Singleton 1999; Pech et al. 1999; Krebs et al. 2004), which appears to drive the quality of the food supply. Other factors include the lack of predators to allow the population to increase (Sinclair et al. 1990), few diseases (Singleton & Brown 1999; Singleton et al. 2005a), good survival of mice over winter and spring, and an extension to the length of the breeding season (Redhead et al. 1985; Singleton 1989; Boonstra & Redhead 1994; Krebs et al. 1995; Singleton et al. 2001, 2005a). Singleton (1989) found that high survival in addition to extended breeding and high reproductive output was necessary for a mouse plague in the cereal-growing region of north-western Victoria in 1984. Survival of mice is generally much higher in refuge habitats compared to cropping areas (Newsome 1969; Twigg & Kay 1995). High quality food remains one of the critical factors for the generation of mouse plagues (Bomford 1987; Bomford & Redhead 1987; Singleton et al. 2001). Therefore the prevailing conditions, food quality and stability of refuge habitats are important for breeding and survival of mice (Twigg & Kay 1994). Mouse populations undergo important changes in reproductive characteristics through the different phases of mouse plagues. Singleton et al. (2001) analysed 18 years of mouse population data and found that litter sizes changed throughout the breeding season with higher litter sizes in spring and lowest in autumn. Litter sizes were lowest in years when densities were high, but remained constant in low years and in years where when rapid increases in density occurred. They also found that breeding began early and was extended by about 5 weeks in years when mouse densities increased rapidly. Finally, Singleton et al. (2001) argued that it was likely that food quality was a key mechanism that caused changes in reproductive characteristics of house mice. 10

Redhead (1982) suggested that mouse plagues were caused by an extension of breeding into winter as a result of high quality food as a consequence of good autumn rainfall. Twigg and Kay (1995) found that breeding of house mice in an irrigated farming system occurred throughout most of the year in key refuge habitats (roadside verges and fencelines), with peaks from October to March. These changes in reproductive characteristics are therefore a key component that drives mouse plagues in Australia and led to much research effort on limiting the high reproductive potential through research on developing fertility control using immunocontraception (Tyndale- Biscoe 1994; Chambers et al. 1997, 1999a, 1999b; Farroway et al. 2002; Singleton et al. 2002; Hinds et al. 2003). A shift in age structure may be necessary for the generation of mouse plagues (Singleton et al. 2001; Sutherland et al. 2004). Sutherland et al. (2004) argued that high survival of young mice might be more important for the formation of plagues than recruitment. During the breeding season, mice were generally site-attached and home ranges for both males and females overlapped and were less than 0.04 ha (Krebs et al. 1995; Chambers et al. 2000), but after breeding ceased, home ranges increased substantially and most mice became nomadic. During a Brodifacoum baiting experiment in wheat fields in Victoria, Australia, Brown and Singleton (1998) found that some mice were active in the wheat crops during the day, which had not previously been recorded. It is therefore necessary to have a good understanding of the population dynamics of mouse populations when developing management strategies to reduce damage to crops. Densities of mice in non-plague years are normally <50 mice/ha, sometimes as low as <5 mice/ha, but at peak densities during mouse plagues, densities can reach in excess of 1,000 mice/ha (Singleton et al. 2001, 2005a); a 200-fold change in density (see Korpimäki et al. 2004 for discussion). The maximum density that has been estimated in crops was 2,716 mice/ha (Saunders & Robards 1983). Densities of mice also can be very high around intensive animal husbandry facilities such as piggeries during mouse plagues. In southern Australia, the interval from a plague “trigger” to peak population densities is 12-18 months (Singleton 1989). However, at a macro-geographic scale, there is high variation in the synchrony of outbreaks or plagues of mice in Australia. In some years, plagues occur from South Australia, through the grain belt of southern and 11

eastern Australia up to the Darling Downs in Queensland (a range of 1,500 km), whereas, in other years, they occur in smaller, localised areas (<50 km).

1.4 Impact on crops

Rodents cause damage to crops by consuming grain and plant material. They damage crops by digging out newly planted seeds or germinating seeds (Mutze 1998; Brown et al. 2003d; Mulungu et al. 2003; Sudarmaji et al. 2003). Rodents also attack developing tillers, and seem to focus their effort on the nodes of the developing tillers. Once the plant reaches the booting stage, they can attack the developing head inside the tiller. They do this by either climbing up the tiller and chewing through the tiller (if their body weight is light enough) or cutting the tiller near the base of the plant and feeding on the tiller. Once grain starts to develop, rodents can chew through the tiller to access the grain or climb up the tiller. Rodents usually damage more than they actually consume. A rat might cut a mature tiller, but then only eat a small portion of the grain on the tiller, but effectively the entire tiller is lost. Bandicota species are known to cache tillers of wheat plants with mature grain (Fulk 1977; Poché et al. 1982) with farmers sometimes digging up the rat burrows to retrieve the grain (personal observation). There are two commonly used terms to define damage in a pest management context. Economic Injury Level (EIL; Pedigo et al. 1986) occurs when a pest species causes enough damage to provide an unacceptable economic impact to the producer, and the Economic Threshold (ET; Ramirez & Saunders 1999), is when control should be implemented so that damage is maintained below the EIL. Much research has been conducted on these concepts for insect and weed pests of crops (eg: Nabirye et al. 2003; Torres-Vila et al. 2003), but they have not been adequately explored for rodent pests. One of the biggest challenges in establishing the economic impact or damage to crops by rodents is accurately assessing the damage and yield loss that rodents cause (Brown & Singleton 2002). Methods for assessing damage in rice have been reviewed by Buckle (1994). One of the simplest ways of measuring damage is to count the number of tillers that have been damaged out of the total tillers of a hill. This method is then applied to crops through conducting counts along transects through the rice crop (Brown & Singleton 12

2002). Also, it is important to distinguish rodent damage from other types of damage. Rodent damage is obvious when there is evidence of gnawing on tillers, pods or plants. Often, rodents, especially large ones (>60 g), gnaw at the base of tillers to fell them and then feed on the developing grain. There is generally a characteristic cut in the tiller at a 45º angle and some evidence of teeth marks on the exposed plant tissue. Smaller rodents, such as mice (10-15 g) can climb tillers of rice to feed directly on the grain. Rodent damage can be underestimated especially when seeds or very young plants are removed. Moreover, rodent damage is typically patchy. So no simple categorisation of damage is applicable, especially in areas where many different rodent species coexist. Rice is a crop that is subject to rodent attack at all stages of growth. However, data on economic losses caused by rodents often lack rigour. The vegetative recovery of rice plants that have tillers cut by rats, and the changes in the capacity of plants to regenerate as they mature, further complicates estimates of losses (Buckle et al. 1985; Wood 1994). For example, early damage to tillers may result in minimal yield loss because of compensatory growth of the plant. Some compensation can occur during the middle stages of plant growth, resulting in only partial yield loss in relation to the amount of damage inflicted (Buckle et al. 1979; Islam et al. 1993). Damage during the reproductive stage of rice results in yield loss of equivalent proportions to the level of damage. It is therefore not surprising that estimates by government officials of losses caused by rats may vary considerably within a country and sometimes within a local region. Furthermore, the amount of tiller damage at different crop stages can lead to different levels of yield loss. Cumulative damage to dry season rice in West Java, Indonesia, was estimated to be 54% at the panicle initiation stage, 32% at the booting stage and 16% at the ripening stage (Singleton et al. 2005b). It was proposed that rats needed to eat relatively more tillers during the primordial and booting stages compared to the ripening stage when the food was more nutritious. Moreover, damage assessed at the ripening stage would need to be multiplied by a factor of 4.2 to obtain an estimate of yield loss (Singleton et al. 2005b). Although this Thesis focuses on pre-harvest rodent damage and losses, it is generally acknowledged that the level of post-harvest losses can be just as high as pre- harvest losses (Meyer 1994; Singleton 2003). Post-harvest damage by rodents is further exacerbated by contamination by faeces and urine, leading to human health risks (Meyer 1994; Ahmed et al. 1995; Drummond 2001). The losses, contamination and 13

damage to infrastructure can be high, particularly in developing countries (Singleton 2003), because farmers have already invested in growing the crop, harvesting it and storing it. New methods to quantify losses to poor rural farmers in developing countries are only just being established, but developing methods to control rodents and reduce post-harvest losses are beginning to show promise (Belmain et al. 2003).

1.5 Response of rodent populations to control or removal

There are only a few studies that have monitored the ecological response of small mammal populations to the removal of animals. By having a good understanding of the mechanisms that influence how small mammal populations respond to removal, appropriate management recommendations can be established to maximise the benefits. There are two general approaches to study this: (1) through trapping to remove animals and monitor the response, and (2) by applying rodenticides and monitoring the response. In general, rodents respond to removal trapping through shifts in their home ranges towards removal areas (Boutin et al. 1985; Schieck & Millar 1987; Nakata & Satoh 1994; Efford et al. 2000). These shifts primarily occur when animals were previously in high density areas. This has been described as the “vacuum effect” by Efford et al. (2000), but could also be described as density-dependent dispersal, as reported by Gundersen et al. (2001), where female root voles (Microtus oeconomus) immigrated to habitat patches with a lower density from where they originated. Few studies have considered the responses of small mammals to removal of conspecifics following the application of poison baits. Recolonisation of Mastomys natalensis occurred very rapidly on maize fields after an application zinc phosphide, even at low densities (Leirs et al. 1997). Consequently we know little about the rate of population recovery following the use of rodenticides or about the demographic characteristics that might influence the speed and ability of the population to recover, especially for house mice or for ricefield rats. These are important gaps in our knowledge. A variation to the removal trapping technique is to remove animals using a trap- barrier system (TBS). This is a control technique that is being developed for rodent pests in SE Asia to reduce damage to lowland irrigated rice crops. The TBS is a plastic fence that surrounds a small field (dimensions of between 20x20 m or 50x50 m; 0.04- 0.25 ha; Singleton et al. 2003) that is planted with rice 3-4 weeks earlier than the 14

surrounding fields with multiple capture rat traps placed at the base on each side of the fence (Singleton et al. 1998, 1999b). Rats are attracted to the TBS from about 200 m away (Brown et al. 2003b), and rodent damage is found to be lower and yields are higher in the surrounding 10-15 ha of ricefields (Singleton et al. 1998). This work builds on previous work conducted by Lam (1988, 1993), and is currently being trialled at a large scale involving farmers in Vietnam (Brown et al. 2003e, Submitted) and Indonesia (Jacob et al. 2003, Submitted; Singleton et al. 2005b). Concurrent sociological research was conducted to examine the factors that influenced how the TBS was adopted by farmers (Morin et al. 2003; Palis et al. 2003, 2004). It is also being trialled in other SE Asian countries (Jahn et al. 1999) and there is interest for the TBS to be used in Madagascar where lowland irrigated rice is grown (G. R. Singleton and D. A. Jones personal communication). Again, the response of rodent populations to this form of removal is not well known. At a field scale, the movements of rats in rice fields are not well understood. A radio-tracking study of R. argentiventer in Indonesia showed that the home range size was approximately 3 ha for males and 2 ha for females and that rats moved less as the rice crop matured and the canopy thickened (Brown et al. 2001). During the tillering stage of the crop, rats were located in burrows during the day, but as the crop matured, rats utilised the cover provided by the rice crop and spent large amounts of time foraging in the crop during the day (Brown et al. 2001). Tristiani et al. (2003) radio- tracked R. argentiventer in rice fields in West Java, Indonesia, and found that home ranges were significantly larger in the non-breeding season (3.5 ha; no difference between males and females) than in the breeding season (home range size for males = 3.2 ha; females = 2.4 ha). Of interest is how these rat species use their environment and how they might reinvade areas where farmers have been applying rodent control.

Key Question 1: How do rodent populations recover from control?

H0: Rodent populations do not recover from control.

H1: Rodent populations recover rapidly from control through immigration of young males.

Alternatives:

HA: Rodent populations recover slowly from control through increased breeding of surviving animals

Comments: Rodent control may not lead to a reduction in density. This might occur if the control method was not effective in causing mortality of animals, or because reinvasion occurs rapidly and there is effectively no reduction in density over time. An analysis of the demographics of the populations should be able to determine whether short-term demographic changes occur. For example, a younger age structure would be expected if increased breeding is the main compensatory factor. Studies should be designed to monitor the demographics of populations before and after control has been applied. One problem is that any observed trends in movements or demographic changes may be due to some factor other than the experimental manipulation. Therefore, the condition of surrounding habitats (favourable or unfavourable) needs to be taken into account because it might influence how an animal responds to the removal. Another critical issue is the speed at which the populations return to pre- manipulated levels and whether this recovery is due to immigration/movements or through density dependent breeding or survival. One complicating factor is the rapid decline or crash in mouse populations that is observed at the end of mouse plagues. An important question therefore, is whether the population crash caused by an application of rodenticide is effectively the same in terms of timing, impact on demography and recovery as that which occurs through a natural decline or crash associated with the end of a mouse plagues? Studies would need to take into account natural movements that would have occurred (without manipulation), so experimental controls are required to account for this. Furthermore, directional movements towards areas of lower density (vacuum effect) should be examined. In order to sort out the issue of which are random movements and which are directional movements, it is necessary to have a large effect on the population and large sample sizes. Rodenticides if applied with an appropriate carrier at a rate of 1 kg/ha are generally >80% effective, with some achieving around 16

95% efficacy in the short-term (Mutze 1998; Mutze & Sinclair 2004). Therefore, an 80- 95% reduction should be aimed for in a manipulative field experiment. There could be a combination of factors driving the longer-term population response, and these also may be influenced by the timing or season when control was applied (breeding versus non-breeding seasons). Studies should be designed to compare the responses across different seasons (breeding versus non-breeding) and at different population densities. Because of the different nature of the rodent control practices applied by farmers in Vietnam for rats and in Australia for mice, I am not able to directly compare studies, but I hope to make some generalisations. I will be reporting on the responses of rodent populations to ecologically-based rodent management (EBRM) principally in Vietnam where farmers are using the trap barrier system (TBS) at a community level, and to rodenticide baiting on mice in Australia. Study I examines at how ricefield rats in Vietnam respond to removal of conspecifics through their capture in a TBS and whether their movements are directed towards the TBS itself or are random movements. Studies II and III consider how rat populations in Vietnam and house mouse populations in Australia respond to control programs through examining their demographic responses.

1.6 Compensation by crops to damage by rodents

Many types of pests can affect the yield of crops, including insects feeding on plant tissue or developing grain (Bardner & Flethcer 1974; Rubia et al. 1996; Sadras et al. 1999), slug damage to plant tissue (Hammond 2000) and cattle and other large herbivores stomping on crops as they graze (Goryńska 1981; Putman & Moore 1998). In many situations, the crop may compensate for this type of attack or injury, provided the damage is small or early enough for the crop to divert resources to recover from the attack (Bardner & Flethcer 1974; Buckle et al. 1979; Fulk & Akhtar 1981; Poché et al. 1981; Haque et al. 1986; Oyediran & Heinrichs 2002). In Southeast Asia, the crop that suffers most damage by the ricefield rat is rice (Oryza sativa) (Hoque et al. 1988; Parshad 1999; Singleton 2003), and in Australia, the crop that suffers the most damage by mice is wheat (Triticum aestivum) (Redhead 1988; Brown & Singleton 2002). 17

The potential compensatory mechanisms and response of crops to rodent damage is outlined in Table 1.2. This assumes that levels of damage are relatively low. When damage levels are high, the impact on the crops and the resulting yield loss is much greater.

Table 1.2. Impact of rodent damage to wheat and rice plants at different stages of growth. Information based on Wood (1971), Buckle et al. (1979), Fulk and Akhtar (1981), Poché et al. (1981), Haque et al. (1986) and Buckle (1994). Crop Plant part Impact on stage attacked plant Compensatory mechanism Sowing Seed Plant is Surrounding plants utilise soil nutrients, destroyed moisture and light resources, near full compensation if damage is not high Emergence Emerging Eaten parts Surviving parts will compensate fully, if all stems and die, or whole parts consumed, plant dies, but surrounding leaves plant dies plants may compensate Tillering Tiller or leaf Tiller or leaf Resources moved to other tillers and leaves, eaten dies near full compensation Maximum Tiller or leaf Tiller or leaf Resources moved to other tillers and leaves, tillering eaten dies near full compensation if harvest is delayed Booting Tiller or leaf Tiller or leaf Resources moved to other tillers and grain eaten dies weight is higher, near full compensation if damage is not high. Flowering Tiller or leaf Some loss of Plant moves resources to remaining eaten grain on tiller developing tillers, near full compensation if damage is not high Heading Tiller or leaf Loss of grain Plant cannot compensate eaten on tiller Ripe Tiller cut or Loss of grain Plant cannot compensate seeds eaten on tiller from head

Because of the nature of mouse plagues, the problems associated with high densities of mice and damage to crops tends to be acute. In outbreak years, the wheat crops may have little opportunity for compensation because feeding activity on the crop would remain high. In this situation, farmers desperately try to minimise losses by using readily available acute rodenticides such as strychnine or zinc phosphide. In other years when densities are low to moderate, mice cause no perceivable damage. To move from a reactive to a preventative management strategy, it is necessary to have a good understanding of the mechanisms that govern the development of mouse plagues and then to develop models that can forecasts outbreaks. Few studies adequately summarise 18

the impacts of mice on crops. Brown and Singleton (2002) summarised information on the damage caused by mice to different crop types and the density of mice if known. The available published data covered the full spectrum of damage (0.5-100%) and a wide range of mouse population densities (50-2,716 mice/ha), yet it was not possible to develop damage/abundance relationships using these published data. There was a striking paucity of rigorous data on damage or yield losses or adequate data on the population densities of mice (Ryan & Jones 1972; Mutze 1993, 1998; Singleton et al. 1991; Twigg et al. 1991; Kay et al. 1994; Caughley et al. 1994; Brown et al. 1997). This makes it difficult to understand the actual economic losses caused by mice or the relationship between the density of mice and damage to crops. This is an essential element for developing management guidelines.

Key Question 2: How do wheat crops compensate for rodent damage?

H0: Wheat crops cannot compensate for rodent damage.

H1: Wheat crops can compensate for rodent damage, but the compensation depends on the timing of damage.

Alternatives:

HA: Wheat crops can compensate for damage, but it depends on the intensity of damage as well as the timing of damage.

Comments: Rodent damage to crops might be so high that compensation is not possible. Studies need to be designed to look at the response of crops to different intensities (levels) of damage at a range of crop stages. All plants and tillers in experimental plots will need to be counted and simulated damage (clipping) done by hand. All rodents in the vicinity of the experiment should be removed so that rodent damage is limited. The number of tillers damaged and yield of crop will be used to determine the impact of the simulated rodent damage. 19

Study IV was designed to determine the impact of mouse damage on wheat crops. It was based on other published studies of rodent damage to rice crops in Asia, but this was the first time that such a study has been conducted on wheat crops for mice.

1.7 The relationship between rodent abundance and crop damage

Prakash and Mathur (1988) identified that little was known about losses due to field rodents. Have we progressed any further in 17 years? To gain a better understanding we need to determine the relationship between the number of rodents present and the level of damage caused because the shape of the relationship will determine the nature of management goals (Figure 1.1). Unfortunately, this has been done for too few rodent species. A good understanding of the relationship between damage, yield loss and abundance of rodents requires a large data set over a wide range of rodent densities and accurate assessments of rodent damage. It would be reasonable to expect that different species of rodents and different types of crops would have different responses. Indeed, this is the case. A linear relationship was found for ricefield rat damage to rice in Indonesia (Singleton et al. 2005b), and between the abundance (percent trap success) of the oriental vole, Microtis fortis, and the loss rate of rice crops in China (Zhang et al. 1999). A sigmoidal relationship was reported for Mastomys damage within two weeks of sowing of corn in Tanzania (Mulungu et al. 2003). The available published data for house mice in Australia is insufficient to determine the shape of the damage/abundance relationship (Brown & Singleton 2002). There is an express need for this relationship to be developed for SE Asian countries and specific agroecosystems, particularly for house mice in Australia. This would lead to a better understanding of the level of reduction required to achieve tolerable crop losses, and when control is both feasible and economic. Furthermore, the shape of the relationship could be used in calculations of EILs and ETs. There still is limited information of this type available for any rodent pest.

High ) ss lo d el i y % ( e g IIII IIIV ma a D

Low Low High

Mouse density Figure 1.1. Hypothetical relationships between the density of rodents and damage (% yield loss) to crops. Four potential relationships are presented. Type I shows sensitivity to rodent damage, where relatively high levels of damage occur at low rodent densities; Type II shows a linear relationship between rodent density and damage up to 100% yield loss; Type III shows a sigmoidal relationship where there is tolerance to damage or compensation of the crop at low densities of rodents, but as rodent densities increase, the level of damage incurred increases faster until a plateau is reached at 100% yield loss; and Type IV shows an exponential growth response where the crop can compensate for up to moderately high densities of rodents, but then the level of damage increases rapidly at high densities.

Key Question 3: What is the relationship between rodent abundance and crop damage or yield loss?

H0: There is no relationship between the abundance of rodents and crop damage or yield loss.

H1: The relationship between rodent abundance and crop damage or yield loss is linear.

Alternatives:

HA: The relationship between rodent abundance and crop damage or yield loss is positive, but non-linear.

Comments: There is a complex relationship between rodent abundance and crop damage or yield loss. The growing crop can compensate for low levels of damage, or in the early stages of growth there may be little or no yield loss because of compensatory growth, but when rodent densities are high, very high levels of yield loss can occur. The relationship would therefore be expected to be sigmoidal. One method to look at this is to model the response of a crop to simulated damage using crop modelling software to develop theoretical damage/abundance relationships. Study V uses the APSIM (Agricultural Production Systems Simulator; McCown et al. 1996; Keating et al. 2003) crop simulation framework to examine the response of mouse densities over time using actual densities of mice, and field data on diet composition and feeding rates to generate a grazing effect to kill the crop on a daily basis from sowing through to harvesting. This modelling approach was used to examine the effect on yield loss to generate damage/abundance curves. After establishing the damage/abundance relationship, another closely related key question could be asked: At what levels do rodent control lead to less damage or improved yields? This is touched on in Study V where I modelled the effect of controlling mice and examined the resulting response of the wheat crop.

Chapter 2. Study areas and species

2.1 The ricefield rat in Southeast Asia

Studies I and II were both conducted on rats in Vinh Phuc Province, in northern Vietnam (Figures 2.1, 2.2 & 2.3). This was the location of a 4-year project looking at the impact of village-level rodent management on rats, funded by the Australian Centre for International Agricultural Research. Study I was designed to monitor changes in the movements and habitat use of the ricefield rat, Rattus argentiventer (Figure 2.4), in response to village-level rodent management, particularly to the population effects of the trap-barrier system (TBS; a small field is planted 3 weeks early and enclosed by a plastic fence set with multiple-capture rat traps – rats are attracted from up to 200 m from the early-planted crop inside TBS and are caught in the traps; for details see Singleton et al. 1998, 2003). In Study II, I used the population data collected over the 4-year study to examine the response of the rodents to the control applied by farmers. The impact of the village-level rodent control using ecologically-based rodent management has been reported in Brown et al. (Submitted), and general information on the breeding biology and demographics of the main pest species has been reported in Brown et al. (In Press). The analysis in Study II was solely conducted for this Thesis. These two studies were conducted in typical lowland irrigated rice ecosystems for Southeast Asia (Figure 2.5). Rice is the principal crop grown, but because of the close proximity to Hanoi (Figure 2.2; 30 km by road), many vegetable and flower crops are grown also. This area of northern Vietnam is classified as sub-tropical, because of the generally hot and humid weather and the seasonal rainfall pattern (annual rainfall of 1600 mm) (Figure 2.6). There are three main crop seasons: spring rice crop, planted in February and harvested in June; summer rice crop, planted in July and harvested in September; and the winter vegetable crop, planted in October and harvested in January. Summers are hot and wet, and winters are cool and dry. This area of Vietnam is strongly influenced by cool winds coming off the Tibetan Plateau to the northwest of Vietnam during the winter season. There is generally enough rainfall from May to September for crop production, but at other times farmers rely upon irrigation water supplied by a network of channels and pumping stations. Farmers have to physically 23

pump water from the irrigation channels into their fields for crop production. Rice is transplanted in fields after seedlings are grown in nursery plots and the fields are flooded. Vegetable and flower crops are grown on beds of soil and the irrigation water is run down the furrows between the beds. In addition, some watering is done by hand using large watering cans. Many of the physical activities involved in preparing soil is done either by hand or by cattle pulling ploughs. Some farmers own or have access to small tractors for ploughing fields, but it is not a widespread practice.

x Vietnam

Hanoi

BAC GIANG

Noi Bai Airport

VINH PHUC BAC NINH Study site

Re d Riv HANOI er

Hanoi

HA TAY

N HUNG YEN km R ed 011 5 0 R iv er

Figure 2.2. Location of study sites in relation to the city of Hanoi and the Provinces of northern Vietnam. Shown are the location of the study sites, main roads (red), lakes and rivers (blue) Provincial boundaries (dashed lines), and the location of the Noi Bai international airport. Inset: Map of Vietnam showing location of Hanoi. 25

Treatment 2

Ap Mot Village

N m

0 100 500 1000

l ne an ch ply up n s tio iga irr jor Ma To Hanoi

r e v Ri

Control 2

Tien Phong Village

Control 1

Treatment 1

Figure 2.3. Schematic layout of the approximate location of the four study sites at Me Linh District, Vinh Phuc Province, Vietnam. There were two treatment sites (where farmers conducted ecologically-based rodent management including TBS) and two control sites (or untreated sites; where farmers continued their typical rodent management practices). Shown are main roads (red line), rivers, streams and major irrigation supply channel (blue line), and villages (light grey). 26

Figure 2.4. Rattus argentiventer photographed in Indonesia. R. argentiventer is a medium sized rat with orange-brown dorsal fur with black flecks. The belly fur ranges from dull white to silvery white. Adults typically weigh up to 220 g. Photographs courtesy of Grant Singleton (© CSIRO Australia). 27

400 40 Rainfall (mm) 350 35 Maximum temperature )

300 Minimum temperature 30 C ) º ( m

250 25 e r m u ll (

200 20 at a f er in

150 15 p a m R

100 10 e T 50 5 0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month

Figure 2.6. Long-term average monthly rainfall (mm) and minimum and maximum temperatures (º C) at Vinh Phuc Province, Vietnam. Spring rice crops are planted in February and harvested in early June, summer rice crops are planted early July and harvested mid-late September and winter crops are grown from September through to February.

The average size of a field is 0.36 ha (Figure 2.5), or as the Vietnamese call them, a “Sao”. The total average land each family would own is 1-2 ha and is made up of Sao in different locations around the village. At the study site in Vinh Phuc Province, the main pest rodent species in the ricefield rat, Rattus argentiventer, which generally accounted for 52% of captures, but R. losea (29%) and R. rattus (11%) also are likely to cause significant damage to crops (Brown et al. 2003e, Submitted). The other species captured were Bandicota indica and B. savilei (3%), Mus species (2%), R. norvegicus (1%), and Suncus murinus (<1%). Less than 2% of captures could not be identified. These rodents generally live in burrows around the edges of ricefields in the banks of irrigation channels or banks of small roads or paths. For many species, such as R. argentiventer, the small bunds between rice paddies are too small for them to construct burrows, particularly if they are <300 x 300 mm in size. It is unclear what interactions there were between the three main species of Rattus. These species have not been well studied, and there is likely to be interspecific competition for food resources and burrows or nesting sites. R. argentiventer and R. losea are poor climbers, but R. rattus is an excellent climber, so there is likely to be 29

some resource partitioning occurring through this difference. Further research is required to look at this complex association. Study I was designed to look at the response of individual R. argentiventer to removal through trap barrier systems (TBS) using radio-telemetry techniques on two sites were TBS were set up and on two sites where TBS were not set up. Study II was designed to look at the demographic responses of populations on treatment sites (where farmers undertook EBRM activities, especially the TBS) and untreated sites (where farmers continued their typical rodent management practices).

2.2 House mice in Australia

Studies III, IV and V were conducted on house mice (Figure 2.7) at four locations (Figure 2.8), three of which where mouse plagues periodically occur in Australia (Wudinna, South Australia; Walpeup, Victoria; Brookstead, Queensland). Data for Study III were collected from three locations. During a strychnine baiting program in the Central Mallee region of Victoria (on sites around Walpeup) trapping was conducted from May to August 1994 to monitor mouse populations prior to baiting and trapping one- and two-months post baiting on four baited and four unbaited sites. Some of these data were reported in Brown et al. (1997), but here I conducted a more thorough analysis and compared the results to the other studies that were conducted and to look at the short- and long-term demographic changes associated with the baiting and also a natural crash in population densities. During a strychnine baiting program on the Darling Downs, Queensland, in 1995, I conducted live-trapping on two baited sites and two unbaited sites from August to December 1995 (trapping was conducted once before baiting and four times after baiting). Finally, I took advantage of high mouse densities at Wudinna, South Australia, in June and July 2002 to examine the response of mouse populations to zinc phosphide baiting that was instigated by a farmer. Trapping was conducted prior to baiting, immediately after baiting and 1 month after baiting. There were 3 baited and 2 unbaited sites. These data from Brookstead and Wudinna have not been analysed or published before.

Figure 2.7. Mus domesticus photographed in a field setting and in a piggery during a mouse plague. The body weight of an adult mouse is 15-20 g. Photographs taken by Peter Brown (© CSIRO Australia).

Extent of NT mouse-plague prone region Qld WA

SA Brookstead NSW

Wudinna Ginninderra Average yield of wheat Vic 1 to 10 Tonnes per sq km Walpeup 10 to 40 Tonnes per sq km Tas > 40 Tonnes per sq km

Figure 2.8. House mice, Mus domesticus, have an Australia-wide distribution, but mouse plagues generally occur only in the grain growing regions of southern and eastern Australia (bounded by the red line). The grain growing belts of western Australia, southern and eastern Australia are indicated showing different levels of average wheat yields. The different levels of shading represent average yields for different regions (data sourced from the Australian Bureau of Statistics 2003). Widespread mouse plagues do not occur in Western Australia, but small isolated outbreaks have occurred. The four locations mentioned in the Thesis are shown. Figure adapted from Singleton et al. (2005a).

Study IV was conducted at the CSIRO Plant Industry’s Ginninderra Research Station in Canberra, Australian Capital Territory, to simulated mouse damage to wheat. A wheat crop was specifically grown for the study and mouse damage was simulated by hand-cutting tillers at different stages of the crop and at different intensities to examine the response of the crop to the damage. This study was published in February 2005 in Crop Protection. Data for Study V came from a 21-year population study of mice from Walpeup and incorporated the use of a crop modelling computer program (Agricultural Production Systems Simulator, APSIM) to determine the relationship between damage and density of mice on wheat crops. The results from the study on simulated mouse damage to wheat (Study IV) were used to calibrate the modelling in APSIM. Feral house mice are virtually the only small mammal trapped on these farms. From greater than 250,000 trap nights at Walpeup in the 21-years of live-trapping mice, 32

there has been only one capture of another species of small mammal (Smithopsis crassicaudata). At Brookstead, Rattus tunneyi were infrequently captured, but these captures were along grassy verges around the perimeter of the crop. On the Darling Downs in Queensland, the Queensland Department of Natural Resources and Mines conduct regular monitoring for house mice using snap traps baited with bacon and catch a significant number of fat-tailed dunnarts (Smithopsis crassicaudata). If live-capture traps are used baited with wheat, the risk of catching a native small mammal as by-catch is negligible. Mice generally construct burrows in the undisturbed habitats adjacent to crops, such as along fencelines. When conditions are favourable, mice move into crops and build burrows once cover is sufficient (Singleton & Redhead 1990; Krebs et al. 1995, Chambers et al. 1996, 2000; Ylönen et al. 2002). The sites at Walpeup and Wudinna are typical broadacre dryland mixed cereal cropping and sheep enterprises (Figure 2.9). Farmers normally implement a 2-3 year crop rotation with a wheat crop followed by sheep grazing, the field is then returned to cropping or left fallow for another year. In any year, around 50% of the fields are sown to a winter cereal. Winter cereals are generally planted in autumn, around April-July and harvested in November-December, depending on rainfall. The rainfall is winter dominated and the mean annual rainfall is around 340 mm at Walpeup and 320 mm at Wudinna (Figure 2.10a,c). At Brookstead many farmers have removed boundary fences so little livestock grazing is practiced. These farmers grow summer (sorghum) and winter (wheat) crops on a rotational basis. At any time, around 90% of the fields are sown to a crop, and fallow periods are short. The rainfall is summer dominated and the mean annual rainfall is around 700 mm (Figure 2.10b). The average farm size for Walpeup, Brookstead and Wudinna is approximately 1000-2000 ha.

Figure 2.9. Typical views of dryland winter cereal farms at Walpeup (top), Wudinna (centre) and Brookstead (bottom). Photographs taken by Peter Brown (© CSIRO Australia). 34

160 40 Rainfall (mm) 140 (a) Walpeup 35 Maximum temperature )

120 Minimum temperature 30 C ) º m

100 25 e ( r (m u l l 80 20 at a f er in

60 15 p a m R

40 10 e T 20 5

0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month

160 40 (b) Brookstead Rainfall (mm) 140 35

Maximum temperature ) C

) 120 30

Minimum temperature º ( m

100 25 e (m ur

ll 80 20 at a r f e n

i 60 15 p a m R 40 10 Te 20 5 0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month

160 40 Rainfall (mm) 140 (c) Wudinna 35 Maximum temperature )

120 Minimum temperature 30 C ) º m

100 25 e ( r m ( u ll

80 20 at a f er in

60 15 p a m R

40 10 e T 20 5 0 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month

Figure 2.10. Long-term average monthly rainfall (mm) and minimum and maximum temperatures (º C) at (a) Walpeup, Victoria, (b) Brookstead, Queensland, and (c) Wudinna, South Australia.

Chapter 3. Results and Discussion

This section is framed around the Key Questions posed in Sections 1.5 to 1.7, and draws on the findings of the five research chapters (Studies I-V).

3.1 How do rodent populations recover from control?

3.1.1 Movements

In Study I, R. argentiventer were radio-collared and tracked in areas neighbouring trap-barrier systems (TBS; treatment sites) and on areas where no TBS were established (control sites), during a non-breeding period and a breeding period to examine whether rats were attracted to the TBS in a lowland rice ecosystem in Vietnam. The study was conducted at a time when the population abundance of R. argentiventer was low (<2% trap success). Although there were some shifts in home ranges over the periods of tracking, there was no observable shift towards areas where rats were being removed by the TBS. Movements appeared to be random in direction, and there was no difference in the home range size of rats between treatment and control sites. However, the mean distances moved from the first 2 days to the last 2 days of tracking, to determine shifts in centres of activity, were twice as large on treatment sites than on control sites, but this was not statistically significant. The conclusion from the study was that because some individual movements of rats were over large areas (mean = 2.4 ha during the non-breeding period and 9.8 ha during the breeding period; maximum = 33 ha), it is likely that rats would be opportunistic in moving about the environment and therefore recolonisation is essentially a chance event. Furthermore, even at low densities, rodent control would need to be applied over large areas because of the large home range sizes and distances moved by rats. In lowland irrigated rice systems in Indonesia female ricefield rats have home range size of around 1 ha (Tristiani et al. 2000, 2003; Jacob et al. 2004), and up to 2 ha (Brown et al. 2001), but males have slightly larger home ranges (around 3 ha; Brown et al. 2001). The distance between successive locations or fixes were generally much 36

larger for males than females (Brown et al. 2001). The home range sizes for ricefield rats in Vietnam were generally larger than that for rats in Indonesia. The other significant difference was that home range sizes were higher during the breeding season to that of the non-breeding season, which was opposite to that found in other studies of rodent species inhabiting cropping systems (Krebs et al. 1995; Chambers et al. 2000). I have found only one other study in cropping systems where the home range sizes were larger during the breeding period, and this was for Calomys venustus in Argentine agroecosystems (Priotto et al. 2002). Given that many species of small mammal shift their home ranges towards areas of lower density, it would be likely that the ricefield rat would also shift home ranges towards areas of lower density. This did not appear to be the case in Study I, mainly as a result of the low densities at the time of the study. In response to the hypotheses raised as part of Key Question 1 and in view of the results from Study I, the recovery of rat populations at low densities would occur relatively slowly through natural dispersal of animals moving randomly about the environment. Furthermore, some adult females had very large home range sizes, and these could potentially recolonise areas where control had been conducted. Juveniles were not radio-collared, so the likelihood that juveniles are important dispersers could not be assessed. Based on my data, I reject the null hypothesis (H0), and accept a combination of H1 and HA, which I now restate as:

Revised hypotheses: Rodent populations recover from control through immigration of rats. The speed of the recovery depends on the level of random movements through the environment and the population density of rodents.

3.1.2 Demographic responses

Study II analysed population data collected over a four-year period as part of a study to examine the effectiveness of applying ecologically-based rodent management practices on two treatment sites (where farmers were encouraged to adopt and implement a range of targeted rodent control practices including the use of the trap- barrier system; TBS) and two control sites (where farmers maintained their 37

conventional rodent control methods) in lowland rice ecosystems in Vietnam. Analyses were conducted on R. argentiventer and R. losea. Although there was no overall difference in the abundance of rats (all species combined) between treatment and control sites over time (Brown et al. Submitted), when the abundance of rats at different crop stages were taken into account, it was evident that compensation was occurring. The abundance of rats on treated sites was lower during the summer rice crop (when TBS were established) but higher during the winter period (when no TBS were established) compared to control sites. The response for the spring rice crop was mixed. Compensation appeared to be occurring through better survival of juveniles on treated sites and from high immigration from surrounding areas. Furthermore, it seemed that older and larger rats were being captured in the TBS and therefore removed from the rodent population. The responses shown were similar for both R. argentiventer and R. losea, therefore the factors that govern the compensatory mechanisms appeared to be the same for both species. There were no significant changes in the proportion of adult females in breeding condition. There are only two other studies that examine the demographic responses of rodents to the TBS. In Indonesia, Jacob and Wegner (In Press) found that females captured in the TBS had a lower body weight, fewer were in breeding condition and the litter size was smaller. The findings from Study II were opposite to that of Jacob and Wegner (In Press), because I found that older, heavier animals were removed by the TBS because both male and female rats captured in monthly population sampling were lighter on the sites with TBS than on the control sites. I was unable to detect any changes in breeding activity of adult females, but my samples were based on assessment of external breeding characteristics of live-captures rather than autopsies to examine pregnancy rates and litter sizes. On sites where TBS were established at another location in Indonesia, Jacob et al. (Submitted) found that the abundance of rats was reduced, there was a decrease in body size of rats, and fewer reproducing females in treated villages versus control villages. These results are similar to those found in Study II. In Study III, the responses of mouse populations at moderate to high densities (range = 172 – 1299 mice/ha) to broadacre application of rodenticides in Australia were assessed. The demographic changes induced by the baiting were compared across three baiting studies. In each case the baiting occurred just prior to a natural decline or crash 38

in densities, and so the demographic changes in the populations induced by the natural decline were compared to that induced by baiting. The efficacy was 0-39% for strychnine baiting at Walpeup, north-western Victoria in 1994, 92% for strychnine baiting at Brookstead, Darling Downs, Queensland in 1995 and 97% for zinc phosphide baiting at Wudinna, Eyre Peninsula, South Australia in 2002. In general, when the baiting was efficacious, there was a mixed response in body weigh (increase and decrease), mixed response for the proportion of juveniles in the population (increase and decrease), a bias towards females, a decrease in survival of mice and a mixed response in the relative body condition of mice (increase or no change). When baiting was not efficacious, the demographic changes were minor or there was no detectable change. It is generally considered that juveniles or sub-adult males are the type of animals most likely to disperse (Wolff 1993), and particularly for house mice, where males generally have larger home range sizes (Krebs et al. 1995; Chambers et al. 2000). Since there were no strong changes in body weights after baiting, I could not determine the type of animal dispersing into baited areas. This was further compromised when all populations went through a crash in numbers shortly after baiting, so any compensatory response of the mouse population was overwhelmed. Taking the findings from Studies II and III together in relation to Key Question 1, populations compensated through a range of strategies. For R. argentiventer and R. losea populations in Vietnam, there were strong responses through a relative increase in juveniles with an associated reduction in body weight. Because older and heavier rats were captured in TBS and were removed, the populations were compensating through better survival of juveniles and through immigration into treated areas from nearby areas. It appeared that there was no recovery through changes in the percentage of adults in breeding condition, so it was considered that the survival of juveniles was important. For Mus domesticus populations in Australia after baiting, there were some increases in juveniles after baiting, but on some sites there was a decrease. However, it was not possible to examine the reinvasion dynamics of the mouse population after baiting in Study III, because each of the three data sets included a widespread crash in numbers that signalled the end of a mouse plague. It is therefore difficult compare the results from Studies II and III.

Based on my data from Studies II and III, I reject the null hypothesis (H0), and accept a combination of H1 and HA, which I now restate as: 39

Revised hypothesis: Rodent populations recover from control through immigration of young animals (juveniles or sub-adults?) and through better survival and recruitment of juveniles. The speed of recovery is reasonably rapid and continuous.

3.2 How do wheat crops compensate for rodent damage?

Study IV examined the response of a wheat crop to simulated mouse damage. The crop could compensate much of the damage imposed early in the crop (up to 50% intensity of damage), whereas damage that occurred later in the crop, particularly just prior to harvest, generally was in proportion to the level of damage imposed. These results are similar to other published studies of damage to rice crops (Buckle et al. 1979; Poché et al. 1981; Haque et al. 1986). Study IV also looked at the numbers of tillers per plot, the ratio of tillers counted at different crop stages when treatments were applied with the final count of tillers, and the yield per tiller. These data provide important insights into the mechanisms by which the wheat plants were compensating for the simulated mouse damage. In response to the simulated damage, wheat plants increased grain production and the number of tillers or the survival rate of remaining tillers. Re-tillering also occurred, as was found in other studies (Buckle et al. 1979; Fulk & Akhtar 1981; Poché et al. 1981; Haque et al. 1986). The yield response surface provided in Study V (Figure V.2), where all levels of damage levels across all stages of crop growth were modelled, provided a good understanding of how wheat compensates for single-point (one-off) damage. This approach could be used and applied for a range of other vertebrate (eg rabbit, kangaroo and duck) and invertebrate pests of wheat (and other crop types for which primary data are available in the APSIM modelling framework) to assist with management. A good understanding of the relationship between pest abundance and yield loss is required in order to relate pest abundance to pest damage and yield loss. The key results from Study IV are that the wheat crop compensated to a range of intensities of damage and at different levels depending on the timing of damage. 40

Therefore with regard to Key Question 2, I reject the null hypothesis (H0) and accept a combination of H1 and HA, which I now restate as:

Revised hypothesis: Wheat crops can compensate for rodent damage, but if the intensity of damage is high or if high damage occurs late in the growth of the crop, then the crop cannot compensate readily and injury (damage) will equal yield loss.

3.3 What is the relationship between rodent abundance and crop damage or yield loss?

In Study V, a relationship between the density of mice and damage (yield loss) to wheat was established through analysing long-term mouse trapping data from Walpeup (1983 to 2003), in the crop modelling framework, APSIM (Agricultural Production Systems Simulator). A grazing effect of mice on the wheat crop was established based on feeding rates of mice and diet composition and was then modelled in APSIM as crop death. The mean yield loss caused by mice was 12.4%, with a maximum yield loss of 96% in 1984. There were seven out of 21 years where yield loss was >5% (one in three years on average). Three patterns of density of mice from sowing to harvest were categorised: 1. one-year mouse outbreaks (n = 3) and the second-year of two-year outbreaks (n = 2): in these situations densities of mice were high during the early crop stages (emergence; >100 mice/ha), but then crashed to low densities, and the high densities at sowing and emergence alone appeared to be sufficient for relatively high mouse damage (>5% yield loss; mean = 41%), 2. the first-year of two-year outbreaks (n = 3): in this situation densities of mice were relatively high throughout the duration of crop growth (emergence, panicle initiation, booting and to a lesser extent at ripening; 60-350 mice/ha) resulting in relatively high mouse damage (>5% yield loss; mean = 16%), and 41

3. during non-outbreak years (n = 13): mouse densities were low (<25 mice/ha) and the subsequent mouse damage was low (<5% yield loss; mean = 1%). The relationship between these different trajectories of mouse population density through the growth of the crop and crop damage were best explained by a sigmoidal curve that explained 97% of the variation. Most of the yield loss occurred early in the crop growth (around sowing and emergence) when mouse densities were >100 mice/ha. Furthermore, significant damage occurred when densities remained high during the growth of the crop, as was found in the first year of two-year outbreaks. Support for a sigmoidal relationship was subsequently provided when densities of mice were manipulated in the model. This was the first time that a strong sigmoidal relationship has been developed for the relationship between pest abundance and yield loss. Mulungu et al. (2003) found there was a sigmoidal relationship between densities of Mastomys natalensis and damage to corn, but this was only at the sowing stage and was not related to final yield loss. Other studies have found or assumed a linear relationship between pest density and damage (Poché et al. 1982; Buckle et al. 1984; Lefebvre et al. 1989). In view of the results from Study V under Key Question 3, I reject the null hypothesis (H0) and H1, and accept HA, but with some modifications:

Revised hypothesis: The relationship between rodent abundance and crop damage or yield loss is positive and sigmoidal.

3.4 General Discussion

The findings from this Thesis indicate that rodents have a strong capacity to recover from control operations through a combination of strategies including good survival and recruitment of juveniles, immigration of animals, and from large-scale random movements of animals across the landscape (Studies I, II & III). There were few changes in the breeding dynamics of adult females, however, the study was not comprehensive. I examined changes in the percentage of adult females in breeding condition in Study II, but this was through assessing the reproductive condition of adults by external examination (evidence of lactation detected by raised nipples without 42

fur at the base and pregnancy detected by palpation). These methods do not identify pregnancies in the first 8-10 days (first trimester) because the embryos are generally too small to be detected by palpation. A better and more rigorous method would be to use breeding data obtained by autopsied animals to obtain data on pregnancy rates and litter sizes. Unfortunately, this was not an option for my studies as I was using capture-mark- release methods, and not destructive sampling. This would be an area of future research. Wheat crops showed that they could compensate for damage by mice provided the damage was not too high or late in the growth of the crop (Study IV). The modelling conducted to establish the density/damage relationship also strongly supported the ability of wheat crops to compensate for damage by mice (Study V). The modelling suggested one important difference: when densities of mice were high at sowing and at emergence of the crop, the crop suffered significant yield losses (>5%). This was not borne out in Study IV (simulation of mouse damage by hand-cutting wheat plants with a maximum of 50% intensity of damage) because the intensity of damage may not have been high enough at emergence. Higher levels of damage were modelled using APSIM, and needs to be tested in field experiments.

3.4.1 Management Implications

The findings highlight the need to conduct rodent control over large areas to limit reinvasion, and in the case of house mice in Australia before sowing, if densities are >100 mice/ha. Most damage to crops in southern Australia is to newly planted cereal crops when population densities of mice normally peak each year. The current control strategies available to farmers in Australia are cheap and effective at reducing population densities of mice (Mutze & Sinclair 2004). However, the farmers are reactive in their actions and generally their control actions are conducted after there have been high crop losses. In Study V, the effect of reducing the population density of mice on yield loss was modelled using published data for a zinc phosphide baiting experiment in 1997 (Brown et al. 2002). The modelling suggested that only a 20% reduction in density would be required to minimise yield loss, but mouse population densities were relatively low, at 88 mice/ha at sowing. It would be valuable to examine the impact of mouse control at different crop stages when densities are high 43

and to monitor both the response of the crop and the mouse population to these control actions. This also should be modelled in APSIM to compare field data with the model to verify the outcomes of the model. Much research has been conducted on examining the efficacy of various types of rodenticides to reduce population abundance of mice in Australia, including strychnine (Brown et al. 1997; Mutze 1989a, 1993, 1998; Mutze & Sinclair 2004; Saunders 1983), zinc phosphide (Brown et al. 2002; Caughley et al. 1998; Mutze & Sinclair 2004), bromadiolone (Kay et al. 1994; Saunders 1983; Twigg et al. 1991), brodifacoum (Brown & Singleton 1998) and others (Mutze & Hubbard 2000; Twigg & Kay 1992). There is a range of alternative methods to reduce mouse damage, particularly at sowing. These include sowing seeds deeper and/or cross-harrowing after sowing, both actions would make it more difficult for mice to dig and locate sown seed (Brown et al. 2003d), and spraying or grazing vegetation along key refuge and breeding habitats such as fencelines in early spring to reduce the breeding success of mice and thence the level of their subsequent invasion of crops (Brown et al. 1998; Brown et al. 2004). These practices are considered to be cost effective (Brown et al. 2004), but they have not been fully tested when mouse population densities are high. In order for farmers to know when it is best to apply rodenticides, they need to be able to estimate mouse densities, particularly before sowing. Some research has been directed at developing simple methods that farmers could use to monitor mouse numbers in their fields, but a reliable method that can be applied across a range of densities has yet to be found. Pieces of card or paper soaked in vegetable oil have been trialled (Brown et al. 2003a; Mutze 1998; Mutze & Sinclair 2004), but the results are variable, and a rigorous comparison with live-trapping mouse populations is required. Another method could be to use the number of active burrows in an area, but the effectiveness of this method would depend on soil conditions and ground cover. Again, it needs to be evaluated. There is a range of activities that can be conducted by farmers in lowland irrigated rice agroecosystems in Southeast Asia to reduce the damage caused by rodents. Since rats have large home range sizes, management needs to be conducted over large areas, farmers need to work together as a community to control rats, and a combination of control methods is necessary including the following (after Singleton et al. 2004): 44

1. Focus control actions at key times according to the biology of the rodent species (eg before the onset of breeding). 2. Hunt rats and destroy burrows before the maximum tillering stage of the rice crop (kill rats before they start breeding). 3. Keep fields clean of weeds (reduce food supply and cover). 4. Keep bund size small between fields (<300 x 300 mm in size to limit establishment of burrows). 5. Synchronise planting and harvesting of rice crops (to restrict the length of the breeding season of rats). The TBS remains one of the key control methods in countries such as Vietnam and Indonesia because the benefits far outweigh the costs associated with its establishment and maintenance, even during periods when rodent densities are low (Singleton et al. 2005b; Brown et al. Submitted; Jacob et al. Submitted).

3.4.2 Future directions

The findings from this Thesis highlight a few areas that require additional research. These include: 1. Develop density/abundance relationships for rats in rice crops. There are good data on how rice crops can compensate for rodent damage, but what is the relationship between rodent abundance and yield loss? Is it linear or is it sigmoidal? The nature of the timing of damage is different for rats in rice in SE Asia (Singleton et al. 2005b) compared to mice in wheat in Australia (Brown & Singleton 2002), so a good understanding of the nature and timing of damage would be required. 2. The findings from the APSIM modelling need to be verified with additional field data. This would be useful to make the modelling framework more robust, but also help with predictions and in further

developments of a decision support system, such as MOUSER (Brown et al. 2003c). 3. Examine the response of breeding in females to different levels of population control and examine at how compensation occurs. Although I found no evidence for compensation by females, my data were based on 45

live-captures only, and not from autopsies of adult females, which would provide better estimates of their responses.

Chapter 4. Conclusions

This Thesis draws together a range of activities to refine the management of rodents, based on an understanding of how rodent populations respond to control, how crops respond to rodent damage and the interaction between rodent damage and yield loss. Furthermore, it is a cross-disciplinary study utilising population ecology of small mammals, crop agronomy and modelling to achieve the aims. Rodents can sometimes cause serious damage to crops. Ricefield rats cause chronic damage to rice crops in Southeast Asia, but sometimes outbreaks of rats occur and damage can be much higher. During outbreaks of house mice in Australia, damage can be very high, but in most years damage goes unnoticed by farmers. There is a range of effective management practices that farmers can implement for ricefield rats in SE Asia and for house mice in Australia, but they are generally applied only after damage has occurred. For ricefield rats in SE Asia, management needs to be conducted early during the crop, before the onset of breeding by rats and control should be focussed on areas where rats have established burrows. Once the crop matures and the canopy thickens, rats seem to spend more time in the crop itself, so targeting the refuge habitats is not very effective. A combination of methods integrating habitat modification, crop synchrony, focussed campaigns to kill rats at certain times and the use of the TBS is recommended. One of the critical features is getting the community of farmers to work together so that management can be applied over large areas. For house mice in Australia, management also needs to be conducted early, but it needs to be linked to some measure of the density of mice. Traditionally, most damage occurs prior to sowing of crops in southern Australia, when mouse densities are generally highest. If densities are above 100 mice/ha, then action is required to reduce the level of damage at sowing. Densities can also be high during other times of the year, and damage can also occur at other crop stages resulting in yield losses, so having a robust method to estimate densities of mice would be useful. It would also be good to continue research on developing predictive models to estimate when densities of mice are likely to reach levels when significant damage would occur. Zinc phosphide is a readily available and cheap form of rodenticide that can achieve good efficacy. Positive 47

benefit : cost ratios can be achieved when using zinc phosphide, mainly because it is so cheap to apply. The Economic Injury Level that was determine during the APSIM modelling framework was very low, again because of the high efficacy of zinc phosphide and because it is so cheap. Finally, management for mice needs to be applied over large areas so that the chance of rapid reinvasion is reduced and the effectiveness of management is maximised.

References

Ahmed, E., Hussain, I., and Brooks, J. E. (1995). Losses of stored foods due to rats at grain markets in Pakistan. International Biodeterioration & Biodegradation 36, 125-133. Aplin, K. P., Brown, P. R., Jacob, J., Krebs, C. J., and Singleton, G. R. (2003). Field methods for rodent studies in Asia and the Indo-Pacific. ACIAR Monograph 100. 223 pp. (Australian Centre for International Agricultural Research: Canberra.) Australian Bureau of Statistics (2003). Year Book Australia. (Australian Bureau of Statistics: Canberra.) Bardner, R. and Fletcher, K. E. (1974). Insect defoliations and their effects on the growth and yield of field crops: a review. Bulletin of Entomological Research 64, 141-160. Belmain, S. R., Meyer, A. N., Timbrine, R., and Penicela, L. (2003). Managing rodent pests in households and food stores through intensive trapping. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 440-445. (Australian Centre for International Agricultural Research: Canberra.) Berry, R. J. and Scriven, P. N. (2005). The house mouse: a model and motor for evolutionary understanding. Biological Journal of the Linnaen Society 84, 335- 347. Boonsong, P., Hongnark, S., Suasa-ard, K., Khoprasert, Y., Promkerd, P., Hamarit, G., Nookarn, P., and Jäkel, T. (1999). Rodent management in Thailand. In Ecologically-based Management of Rodent Pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 338-358. (Australian Centre for International Agricultural Research: Canberra.) Boonstra, R. and Redhead, T. D. (1994). Population dynamics of an outbreak population of house mice (Mus domesticus) in the irrigated rice-growing area of Australia. Wildlife Research 21, 583-598. 49

Bomford, M. (1987). Food and reproduction of wild house mice II. A field experiment to examine the effect of food availability and food quality on breeding in spring. Australian Wildlife Research 14, 197-206. Bomford, M. and Redhead, T. D. (1987). A field experiment to examine the effects of food quality and population density on reproduction of wild house mice. Oikos 14, 304-311. Boutin, S., Gilbert, B. S., Krebs, C. J., Sinclair, A. R. E., and Smith, J. N. M. (1985). The role of dispersal on the population dynamics of snowshoe hares. Canadian Journal of Zoology 63, 106-115. Brown, P. R., Chambers, L. K., and Singleton, G. R. (2002). Pre-sowing control of house mice (Mus domesticus) using zinc phosphide: efficacy and potential non- target effects. Wildlife Research 29, 27-37. Brown, P. R., Davies, M. J., Croft, J. D., and Singleton, G. R. (2003a). Impact of farm management practices on house mouse populations and crops in an irrigated farming system. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 338-342. (Australian Centre for International Agricultural Research: Canberra.) Brown, P. R., Davies, M. J., Croft, J. D., and Singleton, G. R. (2004). Can farm management practices reduce the impact of house mouse populations on crops in an irrigated farming system? Wildlife Research 31, 597-604. Brown, P. R., Hung, N. Q., Hung, N. M., and van Wensveen, M. (1999). Population ecology and management of rodent pests in the Mekong River Delta, Vietnam. In Ecologically-based management of rodent pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 319-337. (Australian Centre for International Agricultural Research: Canberra.) Brown, P. R., Leung, L. K-P., Sudarmaji, and Singleton, G. R. (2003b). Movements of the ricefield rat, Rattus argentiventer, near a trap-barrier system in rice crops in West Java, Indonesia. International Journal of Pest Management 49, 123-129. Brown, P. R. and Singleton, G. R. (1998). Efficacy of brodifacoum to control house mice in wheat crops in southern Australia. Crop Protection 17, 345-352. 50

Brown, P. R. and Singleton, G. R. (1999). Rate of increase as a function of rainfall for house mouse Mus domesticus populations in a cereal-growing region in southern Australia. Journal of Applied Ecology 36, 484-493. Brown, P. R. and Singleton, G. R. (2002). Impacts of house mice on crops in Australia - costs and damage. In Human Conflicts with Wildlife: Economic Considerations. (Eds. L. Clark, J. Hone, J. A. Shivik, R. A. Watkins, K. C. VerCauteren, and J. K. Yoder.) pp. 48-58. (National Wildlife Research Center: Fort Collins, Colorado.) Brown, P. R., Singleton, G. R., Dunn, S. C., and Jones, D. A. (1998). The management of house mice in agricultural landscapes using farm management practices: An Australian perspective. Proceedings of the 18th Vertebrate Pest Conference. (R. O. Baker and A. C. Crabb Eds.). Costa Mesa, California. Davis. University of California. pp. 156-159. Brown, P. R., Singleton, G. R., Kearns, B., and Griffiths, J. (1997). Evaluation and cost- effectiveness of strychnine for control of populations of wild house mice (Mus domesticus) in Victoria. Wildlife Research 24, 159-172. Brown, P. R., Singleton, G. R., Norton, G. A., and Thompson, D. (2003c). Mouser (Version 1.0): an interactive CD-ROM for managing the impact of house mice in Australia. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 264-266. (Australian Centre for International Agricultural Research: Canberra.) Brown, P. R., Singleton, G. R., and Sudarmaji (2001). Habitat use and movements of the rice-field rat, Rattus argentiventer, in West Java, Indonesia. Mammalia 65, 151-166. Brown, P. R., Singleton, G. R., Tann, C. R., and Mock, I. (2003d). Increasing sowing depth to reduce mouse damage to winter crops. Crop Protection 22, 653-660. Brown, P. R., Tuan, N. P., Singleton, G. R., Ha, P. T. T., Hoa, P. T., Hue, D. T., Tan, T. Q., Tuat, N. V., Jacob, J., and Muller, W. J. (Submitted). Ecologically-based management of rodents in the real world: application to a mixed agro-ecosystem in Vietnam. Ecological Applications. Brown, P. R., Tuan, N. P., Singleton, G. R., Hue, D. T., Hoa, P. T., Ha, P. T. T., Tan, T. Q., and Tuat, N. V. (In Press). Population dynamics of Rattus argentiventer, R. 51

losea and R. rattus inhabiting a mixed farming system in the Red River Delta, Vietnam. Population Ecology. Brown, P. R., Tuan, N. P., Singleton, G. R., Tuat, N. V., Tan, T. Q., and Hoa, L. T. (2003e). Impact of village-level rodent control practices on rodent populations and rice crops in Vietnam. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 197-202. (Australian Centre for International Agricultural Research: Canberra.) Buckle, A. P. (1994). Damage assessment and damage surveys. In Rodent Pests and Their Control. (Eds. A. P. Buckle and R. H. Smith.) pp. 219-248. (CAB International: Wallingford.) Buckle, A. P., Huson, L. W., and Rowe, F. P. (1984). The frequency distribution of rat damage in rice fields and its implications for the cost-benefit of control programmes. Proceedings of a Conference on The Organisation and Practice of Vertebrate Pest Control. Elvetham Hall, Hampshire, England. Surrey, England. ICI Plant Protection Division. pp. 289-296. Buckle, A. P. and Smith, R. H. (1994). Rodent Pests and Their Control. (CAB International: Wallingford.) Buckle, A. P., Yong, Y. C., and Abdul Raman, H. (1985). Damage by rats to rice in South east Asia with special reference to an integrated management scheme proposed for Peninsular Malaysia. Acta Zoologica Fennica 173, 139-144. Buckle, A. P., Yong, Y. C., and Rowe, F. P. (1979). Yield response of the rice variety Improved Mashuri to simulated rat damage. Malaysian Agriculture Journal 52, 135-144. Cantor, N. F. (2001). In the Wake of the Plague: The Black Death and the World it Made. 245 pp. (Simon & Schuster UK Ltd: London.) Cantrill, S. (1992). The population dynamics of the house mouse (Mus domesticus) in a dual crop agricultural ecosystem. Unpublished PhD Thesis, School of Life Science, Queensland University of Technology. Catling, H. D., Islam, Z., and Rahman, L. (1988). Status of the pests and diseases in Asian deepwater rice. Proceedings of the International Deepwater Rice Workshop. Bangkok, Thailand. Los Baños, Philippines. International Rice Research Institute. pp. 525-537. 52

Caughley, J., Monamy, V., and Heiden, K. (1994). Impact of the 1993 mouse plague. GRDC Occasional Paper Series No.7. (Grains Research and Development Corporation: Canberra.) Caughley, J., Strong, K., and Hinchliffe, P. (1998). Report on the zinc phosphide baiting program to control mice in Central Queensland in 1997. Queensland Department of Natural Resources, Brisbane. pp. 57. Chapman, A. (1981). Habitat preference and reproduction of feral house mice, Mus musculus, during plague and non-plague situations in Western Australia. Australian Wildlife Research 8, 567-579. Chambers, L. K., Lawson, M. A., and Hinds, L. A. (1999a). Biological control of rodents - the case for fertility control using immunocontraception. In Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 215-242. (Australian Centre for International Agricultural Research: Canberra.) Chambers, L. K., Singleton, G. R., and Hinds, L. A. (1999b). Fertility of wild mouse populations: the effects of hormonal competence and an imposed level of sterility. Wildlife Research 26, 579-591. Chambers, L. K., Singleton, G. R., and Hood, G. M. (1997). Immunocontraception as a potential control method of wild rodent populations. Belgian Journal of Zoology 127, 145-156. Chambers, L. K., Singleton, G. R., and Krebs, C. J. (2000). Movements and social organization of wild house mice (Mus domesticus) in the wheatlands of Northwestern Victoria, Australia. Journal of Mammalogy 81, 59-69. Chambers, L. K., Singleton, G. R., and van Wensveen, M. (1996). Spatial heterogeneity in wild populations of house mice (Mus domesticus) on the Darling Downs, South-eastern Queensland. Wildlife Research 23, 23-38. Chauhan, N. S. and Saxena, R. N. (1985). The phenomenon of bamboo flowering and associated increase in rodent population in Mizoram. Journal of the Bombay Natural History Society 82, 644-647. Dickman, CR., Mahon, P. S., Masters, P., and Gibson, D. F. (1999). Long-term dynamics of rodent populations in arid Australia: the influence of rainfall. Wildlife Research 26, 389-403. 53

Douangboupha, B., Aplin, K. P., and Singleton, G. R. (2003). Rodent outbreaks in the uplands of Laos: analysis of historical patterns and the identity of nuu khii. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 103- 111. (Australian Centre for International Agricultural Research: Canberra.) Drummond, D. C. (2001). Rodents and biodeterioration. International Biodeterioration & Biodegradation 48, 105-111. Efford, M., Warburton, B., and Spencer, N. (2000). Home-range changes by brushtail possums in response to control. Wildlife Research 27, 117-127. Elias, D. J. and Fall, M. W. (1988). The rodent problem in Latin America. In Rodent Pest Management. (Ed. I. Prakash.) pp. 13-28. (CRC Press: Boca Raton, Florida.) Farroway, L. N., Singleton, G. R., Lawson, M. A., and Jones, D. A. (2002). The impact of murine cytomegalovirus (MCMV) on enclosure populations of house mice (Mus domesticus). Wildlife Research 29, 11-17. Fiedler, L. A. (1994). Rodent Pest Management in Eastern Africa. Plant Production and Protection Paper No. 123. FAO, Rome. pp. 83. Fulk, G. (1977). Food hoarding of Bandicota bengalensis in a rice field. Mammalia 41, 59-61. Fulk, G. W. and Akhtar, M. T. (1981). An investigation of rodent damage and yield reduction in rice. Tropical Pest Management 27, 116-120. Geddes, A. W. M. (1992). The Relative Importance of Pre-harvest Crop Pests in Indonesia. (Natural Resources Institute: Chatham, UK.) Goryńska, W. (1981). Method of determining relations between the extent of damage in farm crops, big game numbers, and environmental conditions. Acta Theriologica 26, 469-481. Gratz, N. G. (1994). Rodents as carriers of disease. In Rodent Pests and Their Control. (Eds. A. P. Buckle and R. H. Smith.) pp. 85-108. (CAB International: Wallingford, UK.) Gundersen, G., Johannesen, E., Andreassen, H. P., and Ims, R. A. (2001). Source-sink dynamics: how sinks affect demography of sources. Ecology Letters 4, 14-21. Hammond, R. B (2000). Simulation of moderate levels of slug injury to soybean. Crop Protection 19, 113-120. 54

Haque, M. E., Ahmed, M. S., and Brooks, J. E. (1986). Effect of simulated rat damage on deep water rice in Bangladesh. Bangladesh Journal of Agriculture 11, 63-69. Hinds, L. A., Hardy, C. M., Lawson, M. A., and Singleton, G. R. (2003). Developments in fertility control for pest animal management. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 31-36. (Australian Centre for International Agricultural Research: Canberra.) Hoque, M. M., Sanchez, F. F., and Benigno, E. A. (1988). Rodent problems in selected countries in Southeast Asia and islands in the Pacific. In Rodent Pest Management. (Ed. I. Prakash.) pp. 85-99. (CRC Press: Boca Raton, Florida.) Huynh, N. V. (1987). Common vertebrate pests of deepwater rice in the Mekong Delta of Vietnam. In Proceedings of the 1987 International Deepwater Rice Workshop. pp. 599-603. (International Rice Research Institute: Manila.) IRRI (1997). IRRI Rice Facts. Information brochure, International Rice Research Institute, Los Baños, Philippines. Islam, Z., Morton, R. G., and Jupp, B. P. (1993). The effects of rat damage on deepwater rice yields in Bangladesh. International Journal of Pest Management 39, 250-254. Jacob, J., Matulessy, J., and Sudarmaji (2004). Effects of imposed sterility on movement patterns of female ricefield rats. Journal of Wildlife Management 68, 1138-1144. Jacob, J., Sudarmaji, and Singleton. G. R. (2003). Ecologically-based management of rice-field rats on a village scale in West Java: experimental approach and assessment of habitat use. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 191-196. (Australian Centre for International Agricultural Research: Canberra.) Jacob, J., Sudarmaji, Singleton G. R., Rahmini, Herawati, N. A., Brown, P. R., and Muller, W. J. (Submitted). Ecologically-based management of rodents in the real world: application to a rice-based agro-ecosystem in Indonesia. Ecological Applications. Jacob, J. and Wegner, R. E. (In Press). Does continuous removal of individuals separate high and low quality ricefield rats? Journal of Wildlife Management. 55

Jahn, G. C., Mak, S., Cox, P. G., and Nel, C. (1999). Farmer participatory research on rat management in Cambodia. In Ecologically-based Management of Rodent Pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 358-371. (Australian Centre for International Agricultural Research: Canberra.) Jaksic, F. M. and Lima, M. (2003). Myths and facts on ratadas: Bamboo blooms, rainfall peaks and rodent outbreaks in South America. Austral Ecology 28, 237- 251. Karim, A. N. M. R., Ahmed, M. S., Mion, M. Y., and Hoque, M. E. (1988). Rodent ecology and control in deepwater rice. Proceedings of the International Deepwater Rice Workshop. Bangkok, Thailand. Los Baños, Philippines. International Rice Research Institute. pp. 605-617. Kay, B. J, Twigg, L. E., and Nicol, H. I. (1994). The strategic use of rodenticides against house mice (Mus domesticus) prior to crop invasion. Wildlife Research 21, 11-19. Keating, B. A., Carberry, P. S., Hammer, G. L., Probert, M. E., Robertson, M. J., Holzworth, D., Huth, N. I., Hargreaves, J. N. G., Meinke, H., Hochman, Z., McLean, G., Verburg, K., Snow, V., Dimes, J. P., Silburn, M., Wang, E., Brown, S., Bristow, K. L., Asseng, S., Champan, S., McCown, R. L., Freebairn, D. M., and Smith, C. J. (2003). An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy 18, 267-288. Khamphoukeo, K., Douangboupha, B., Aplin, K. P., and Singleton, G. R. (2003). Pest and non-pest rodents in the upland agricultural landscape of Laos: a progress report. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 284-289. (Australian Centre for International Agricultural Research: Canberra.) Korpimäki, E., Brown, P. R., Jacob, J., and Pech, R. P. (2004). The puzzles of population cycles and outbreaks of small mammals solved? BioScience 54, 1071-1079. Krebs, C. J., Kenney, A. J., and Singleton, G. R. (1995). Movements of feral house mice in agricultural landscapes. Australian Journal of Zoology 43, 293-302. 56

Krebs, C. J., Kenney, A. J., Singleton, G. R., Mutze, G. J., Pech, R. P., Brown, P. R., and Davis, S. A. (2004). Can outbreaks of house mice in southeastern Australia be predicted by weather models? Wildlife Research 31, 465-474. Lam, Y. M. (1980). Reproductive behaviour of the rice field rat, Rattus argentiventer and implications for its control. Proceedings of the National Rice Conference. Kuala Lumpur, Malaysia. pp. 243-257. Lam, Y. M. (1983). Reproduction in the ricefield rat Rattus argentiventer. Malayan Nature Journal 36, 249-282. Lam, Y. M. (1988). Rice as a trap crop for the rice field rat in Malaysia. Proceedings of the 13th Vertebrate Pest Conference. (Eds. A. C. Crabb and R. E. Marsh.) pp. 123-128. Lam, Y. M. (1993). An environmentally friendly system for rodent control. Teknolgi Padi 9, 25-28. Lam, Y. M. (1991). Cultural control of rice field rats. In Rodents and Rice. Report and Proceedings of an Expert Panel Meeting on Rice Rodent Control. (Ed. G. R. Quick.) pp. 65-72. (International Rice Research Institute: Manila, Philippines.) Lefebvre, L. W., Engeman, R. M., Decker, D. G., and Holler, N. R. (1989). Relationship of roof rat population indices with damage to sugarcane. Wildlife Society Bulletin 17, 41-45. Leirs, H. (2003). Management of rodents in crops: the Pied Piper and his orchestra. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. Singleton. G. R., L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 183- 190. (Australian Centre for International Agricultural Research: Canberra.) Leirs, H., Verhagen, R., Sabuni, C. A., Mwanjabe, P., and Verheyen, W. N. (1997). Spatial dynamics of Mastomys natalensis in a field-fallow mosaic in Tanzania. Belgian Journal of Zoology 127, 29-38. Leung, L. K. P., Singleton, G. R., Sudarmaji, and Rahmini (1999). Ecologically-based population management of the rice-field rat in Indonesia. In Ecologically-based Management of Rodent Pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 305-318. (Australian Centre for International Agricultural Research: Canberra.) Lund, M. (1988). Rodent problems in Europe. In Rodent Pest Management. (Ed. I. Prakash.) pp. 29-33. (CRC Press: Boca Raton, Florida.) 57

Macdonald, D. W. (2001). The New Encyclopaedia of Mammals. 2nd Edition. 961 pp. (Andromeda: Oxford, UK.) Masters, P. (1993). The effects of fire-driven succession and rainfall on small mammals in spinifex grassland at Uluru National Park, Northern Territory. Wildlife Research 20, 803-813. Marsh, R. E. (1988). Rodent problems on the North American continent. In Rodent Pest Management. (Ed. I. Prakash.) pp. 1-11. (CRC Press: Boca Raton, Florida.) Meerburg, B. G., Bonde, M., Brom, F. W. A., Endepols, S., Jensen, A. N., Leirs, H., Lodal, J., Singleton, G. R., Pelz, H.-J., Rodenburg, T. B., and Kijlstra, A. (2004). Towards sustainable management of rodents in organis animal husbandry. Netherlands Journal of Agricultural Sciences 52, 195-205. Meyer, A. N. (1994). Rodent control in practice: food stores. In Rodent Pests and Their Control. (Eds. A. P. Buckle and R. H. Smith.) pp. 273-290. (CAB International: Wallingford.) McCown, R. L., Hammer, G. L., Hargreaves, J. N. G., Holzworth, D., and Freebairn, D. M. (1996). APSIM: a novel software system for model development, model testing and simulation in agricultural research. Agricultural Systems 50, 255- 271. Mills, J. N. (1999). The role of rodents in emerging human disease: examples from the hantaviruses and arenaviruses. In Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 134-160. (Australian Centre for International Agricultural Research: Canberra.) Morin, S. R., Palis, F. G., Chien, H. V., Chi, T. N., Magsumbol, M., and Papag, A. (2003). A sociological perspective on the community-based trap-barrier system. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 380-388. (Australian Centre for International Agricultural Research: Canberra.) Moro, D. (2001). Evaluation and cost-benefits of controlling house mice (Mus domesticus) on islands: an example from Thevenard Island, Western Australia. Biological Conservation 99, 355-364. 58

Mulungu, L. S., Makundi, R. H., Leirs, H., Massawe, A. W., Vibe-Petersen, S., and Stenseth, N. C. (2003). The rodent density-damage function in maize fields at an early growth stage. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. Singleton. G.R., L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 301-303. (Australian Centre for International Agricultural Research: Canberra.) Mutze, G. J. (1989a). Effectiveness of strychnine bait trails for poisoning mice in cereal crops. Australian Wildlife Research 16, 459-465. Mutze, G. J. (1989b). Mouse plagues in South Australian cereal-growing areas. I. Occurrence and distribution of plagues. Australian Wildlife Research 16, 677- 683. Mutze, G. J. (1991). Mouse plagues in South Australian cereal-growing areas. III. Changes in mouse abundance during plague and non-plague years, and the role of refugia. Wildlife Research 18, 593-604. Mutze, G. J. (1993). Cost-effectiveness of poison bait trails for control of house mice in mallee cereal crops. Wildlife Research 20, 445-456. Mutze, G. J. (1998). The 1993 strychnine baiting program for mouse control in South Australian grain crops. I. Efficacy. Wildlife Research 25, 533-546. Mutze, G. J. and Hubbard, D. J. (2000). Effects of molluscicide baits on field populations of house mice. Agriculture, Ecosystems and Environment 80, 205- 211. Mutze, G. J. and Sinclair, R. (2004). Efficacy of zinc phosphide, stychnine and chlorpyrifos as rodenticides for the control of house mice in South Australian cereal crops. Wildlife Research 31, 249-257. Nabirye, J., Nampala, P., Kyamanywa, S., Ogenga-Latigo, M. W., Wilson, H., and Adopala, E. (2003). Determination of damage-yield loss relationships and economic injury levels of flower thrips on cowpea in eastern Uganda. Crop Protection 22, 911-915. Nag, S. (1999). Bamboo, rats and famines: famine relief and perceptions of British paternalism in the Mizo hills (India). Environment and History 5, 245-252. Nakata, K. and Satoh, K. (1994). Distance of movements of voles after removal in Clethrionomys rufocanus bedfordiae (Rodentia : Cricetidae). Applied Entomology and Zoology 29, 97-103. 59

Newsome, A. E. (1969). A population study of house-mice permanently inhabiting a reed-bed in South Australia. Journal of Animal Ecology 38, 361-377. Newsome, A. E. and Corbett, L. K. (1975). Outbreaks of rodents in semi-arid and arid Australia: causes, preventions, and evolutionary considerations. In Rodents in Desert Environments. (Eds. I. Prakash and P. K. Ghosh.) pp. 117-153. (Junk: The Hague.) Oyediran, I. O. and Heinrichs, E. A. (2002). Response of lowland rice plants to simulated insect defoliation in West Africa. International Journal of Pest Management 48, 219-224. Palis, F. G., Morin, S. R., Chien, H. V., and Chi, T. N. (2003). Socio-cultural and economic assessment of community trap-barrier system adoption in southern Vietnam. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 422-425. (Australian Centre for International Agricultural Research: Canberra.) Palis, F. G., Sumalde, Z. M., and Hossain, M. (2004). Assessment of Rodent Control Projects in Vietnam: Adoption and Impact. Impact Assessment Series Report No. 24. Australian Centre for International Agricultural Research, Canberra. pp. 62. Parshad, V. R. (1999). Rodent control in India. Integrated Pest Management Reviews 4, 97-126. Payseur, B. A. and Nachman, M. W. (2005). The genomics of speciation: investigating the molecular correlated of X chromosome introgression across the hybrid zone between Mus domesticus and Mus musculus. Biological Journal of the Linnaen Society 84, 523-534. Pech, R. P., Hood, G., Singleton, G. R., Salmon, E., Forrester, R., and Brown, P. R. (1999). Models for predicting plagues of house mice (Mus domesticus) in Australia. In Ecologically-based Management of Rodent Pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 81-112. (Australian Centre for International Agricultural Research: Canberra.) Pedigo, L. P., Hutchins, S. H., and Higley, L. G. (1986). Economic injury levels in theory and practice. Annual Review of Entomology 31, 341-368. 60

Plomley, N. J. B. (1972). Some notes on plagues of small mammals in Australia. Journal of Natural History 6, 363-384. Poché, R. M., Haque, M. E., Main, M. Y., Sultana, P., and Karim, M. A. (1981). Rice yield reduction by simulated rat damage in Bangladesh. Tropical Pest Management 27, 242-246. Poché, R. M., Main, M Y., Haque, M. E., and Sultana, P. (1982). Rodent damage and burrowing characteristics in Bangladesh wheat fields. Journal of Wildlife Management 46, 139-147. Priotto, J., Steinmann, A., and Polop, J. (2002). Factors affecting home range size and overlap in Calomys venustus (Muridae: Sigmodontinae) in Argentine agroecosystems. Mammalian Biology 67, 97-104. Prakash, I. (1988). Rodent Pest Management. (CRC Press: Boca Raton, Florida.) Prakash, I. and Mathur, R. P. (1988). Rodent problems in Asia. In Rodent Pest Management. (Ed. I. Prakash.) pp. 67-84. (CRC Press: Boca Raton, Florida.) Predavec, M. and Dickman, C. R. (1994). Population dynamics and habitat use of the long-haired rat (Rattus villosissimus) in south-western Queensland. Wildlife Research 21, 1-10. Putman, R. J. and Moore, N. P. (1998). Impact of deer in lowland Britain on agriculture, forestry and conservation habitats. Mammal Review 28, 141-164. Ramirez, O. A. and Saunders, J. L. (1999). Estimating economic thresholds for pest control: an alternative procedure. Journal of Economic Entomology 92, 391-401. Redhead, T. D. (1982). Reproduction, Growth and Population Dynamics of House Mice in Irrigated and Non-Irrigated Cereal Farms in New South Wales. Unpublished PhD Thesis, Department of Zoology, Australian National University. Redhead, T. D. (1988). Prevention of plagues of house mice in rural Australia. In Rodent Pest Management. (Ed. I. Prakash.) pp. 191-206. (CRC Press: Boca Raton, Florida.) Redhead, T. D., Enright, N., and Newsome, A. E. (1985). Causes and predictions of outbreaks of Mus musculus in irrigated and non-irrigated cereal farms. Acta Zoologica Fennica 173, 123-127. Redhead, T. D. and Singleton, G. R. (1988). An examination of the "PICA" strategy for the prevention of losses caused by plagues of the house mouse (Mus domesticus) in rural Australia. In Vertebrate Pest Management in Australia: Decision 61

Analysis/systems Analysis Approach. (Eds. G. A. Norton and R. P. Pech.) pp. 18-37. (CSIRO: Melbourne.) Redhead, T. D., Singleton, G. R., Myers, K., and Coman, B. J. (1991). Mammals introduced into southern Australia. In Biogeography of Mediterranean Invasions. (Eds. R. H. Groves and F. Di Castri.) pp. 289-304. (Cambridge University Press: Cambridge.) Rubia, E. G., Heong, K. L., Zaluck, M., Gonzales, B., and Norton, G. A. (1996). Mechanisms of compensation of rice plants to yellow stem borer Scirpophaga incertulas (Walker) injury. Crop Protection 15, 335-340. Ryan, G. E. and Jones, E. L. (1972). A Report on the Mouse Plague in the Murrumbidgee and Coleambally Irrigation Areas, 1970. NSW Department of Agriculture, pp. 60. Sadras, V. O., Fereres, A., and Ratcliffe, R. H. (1999). Wheat growth, yield, and quality as affected by insect herbivores. In Wheat: Ecology and Physiology of Yield Determination. (Eds. E. H. Satorre and G. A. Slafer.) pp. 183-227. (Food Products Press: New York.) Sage, R. D., Atchley, W. R., and Capanna, E. (1993). House mice as models in systematic biology. Systematic Biology 42, 523-561. Saunders, G. R. (1983). Evaluation of mouse plague control techniques in irrigated sunflower crops. Crop Protection 2, 437-445. Saunders, G. R. and Giles, J. R. (1977). A relationship between plagues of the house mouse, Mus musculus (Rodentia: Muridae) and prolonged periods of dry weather in south-eastern Australia. Australian Wildlife Research 4, 241-147. Saunders, G. R. and Robards, G. E. (1983). Economic considerations of mouse-plague control in irrigated sunflower crops. Crop Protection 2, 153-158. Schieck, J. O. and Millar, J. S. (1987). Can removal areas be used to assess dispersal of red-backed voles? Canadian Journal of Zoology 65, 2575-2578. Schiller, J. M., Boupha, B. D., and Bounnaphol, O. (1999). Rodents in agriculture in the Lao PDR - a problem with an unknown future. In Ecologically-based Management of Rodent Pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 372-387. (Australian Centre for International Agricultural Research: Canberra.) 62

Sinclair, A. R. E., Olsen, P. D., and Redhead, T. D. (1990). Can predators regulate small mammal populations? Evidence from house mouse outbreaks in Australia. Oikos 59, 382-392. Singleton, G. R. (1989). Population dynamics of an outbreak of house mice (Mus domesticus) in the mallee wheatlands of Australia - hypothesis of plague formation. Journal of Zoology, London 219, 495-515. Singleton, G. R. (1997). Integrated management of rodents: A Southeast Asian and Australian perspective. Belgian Journal of Zoology 127, 157-169. Singleton, G. R. (2003). Impacts of Rodents on Rice Production in Asia. IRRI Discussion Paper Series No. 43. International Rice Research Institute, Los Baños, Philippines. pp. 30. Singleton, G. R. and Brown, P. R. (1999). Management of mouse plagues in Australia: integration of population ecology, bio-control and best farm practice. In Advances in Vertebrate Pest Management. (Eds. D. P. Cowan and C. J. Feare.) pp. 189-203. (Filander-Verlag: Berlin.) Singleton, G. R., Brown, P. R., and Jacob, J. (2004). Ecologically-based rodent management; its effectiveness in cropping systems in South-East Asia. Netherlands Journal of Agricultural Sciences 52, 163-171. Singleton, G. R., Brown, P. R., Pech, R. P., Jacob, J., Mutze, G. J., and Krebs, C. J. (2005a). One hundred years of eruptions of house mice in Australia - a natural biological curio. Biological Journal of the Linnaen Society 84, 617-627. Singleton, G. R., Farroway, L. N., Chambers, L. K., Lawson, M. A., Smith, A. L., and Hinds, L. A. (2002). Ecological basis for fertility control in the house mouse (Mus domesticus) using immunocontraceptive vaccines. Reproduction Supplement 60, 31-39. Singleton, G. R., Krebs, C. J., Davis, S. A., Chambers, L. K., and Brown, P. R. (2001). Reproductive changes in fluctuating house mouse populations in southeastern Australia. Proceedings of the Royal Society of London Series B 268, 1741-1748. Singleton, G. R., Leirs, H., Hinds, L. A., and Zhang, Z. (1999a). Ecologically-based management of rodent pests - re-evaluating our approach to an old problem. In Ecologically-based Management of Rodent Pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 17-29. (Australian Centre for International Agricultural Research: Canberra.) 63

Singleton, G. R. and Petch, D. A. (1994). A review of the biology and management of rodent pests in Southeast Asia. ACIAR Technical Reports No. 30. Australian Centre for International Agricultural Research, Canberra. pp. 65. Singleton, G. R. and Redhead, T. D. (1990). Structure and biology of house mouse populations that plague irregularly; an evolutionary perspective. Biological Journal of the Linnaen Society 41, 285-300. Singleton, G. R., Sudarmaji, and Brown, P. R. (2003). Comparison of different sizes of physical barriers for controlling the impact of the rice field rat, Rattus argentiventer, in rice crops in Indonesia. Crop Protection 22, 7-13. Singleton, G. R., Sudarmaji, Jacob, J., and Krebs, C. J. (2005b). Integrated management to reduce rodent damage to lowland rice crops in Indonesia. Agriculture, Ecosystems and Environment 107, 75-82. Singleton, G. R., Sudarmaji, Jumanta, Tan, T. Q., and Hung, N. Q. (1999b). Physical control of rats in developing countries. In Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 178-198. (Australian Centre for International Agricultural Research: Canberra.) Singleton, G. R., Sudarmaji, and Suriapermana, S. (1998). An experimental field study to evaluate a trap-barrier system and fumigation for controlling the rice field rat, Rattus argentiventer, in rice crops in West Java. Crop Protection 17, 55-64. Singleton, G. R., Twigg, L. E., Weaver, K. E., and Kay, B. J. (1991). Evaluation of bromadiolone against house mice (Mus domesticus) populations in irrigated soybean crops. II. Economics. Wildlife Research 18, 275-284. Stenseth, N. C., Leirs, H., Skonhoft, A., Davis, S. A., Pech, R. P., Andreassen, H. P., Singleton. G. R., Lima, M., Machangu, R. M., Makundi, R. H., Zhang, Z., Brown, P. R., Shi, D., and Wan, X. (2003). Mice, rats and people: the bioeconomics of agricultural rodent pests. Frontiers in Ecology and the Environment 1, 367-375. Sudarmaji, Singleton, G. R., Herawati, N. A., Djatiharti, A., and Rahmini (2003). Farmers’ perceptions and practices in rat management in West Java, Indonesia. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. 64

Spratt.) pp. 389-394. (Australian Centre for International Agricultural Research: Canberra.) Sutherland, D. R., Banks, P. B., Jacob, J., and Singleton, G. R. (2004). Shifting age structure of house mice during a population outbreak. Wildlife Research 31, 613- 618. Tangkanakul, W., Smits, H. L., Jatanasen, S., and Ashford, D. A. (2005). Leptospirosis: an emerging health problem in Thailand. Southeast Asian Journal of Tropical Medicine and Public Health 36, 281-288. Tattersall, F. H., Nowell, F., and Smith, R. H. (1994). A review of the endoparasites of wild House Mice Mus domesticus. Mammal Review 24, 61-71. Torres-Vila, L. M., Rodríguez-Molina, M. C., and Lacasa-Plasencia, A. (2003). Impact of Helicoverpa armigera larval density and crop phenology on yield and quality losses in processing tomato: developing fruit count-based damage thresholds for IPM decision-making. Crop Protection 22, 521-532. Tristiani, H., Murakami, O., and Kuno, E. (2000). Rice plant damage distribution and home range distribution of the ricefield rat Rattus argentiventer (Rodentia: Muridae). Belgian Journal of Zoology 130, 83-91. Tristiani, H., Murakami, O., and Watanabe, H. (2003). Ranging and nesting behaviour of the ricefield rat Rattus argentiventer (Rodentia: Muridae) in West Java, Indonesia. Journal of Mammalogy 84, 1228-1236. Twigg, L. E. and Kay, B. J. (1992). Evaluation of Quintox® for control of feral house mice. Journal of Wildlife Management 56, 174-185. Twigg, L. E. and Kay, B. J. (1994). The effects of microhabitat and weather on house mouse (Mus domesticus) numbers and the implications for management. Journal of Applied Ecology 31, 651-663. Twigg, L. E. and Kay, B. J. (1995). The ecology of house mice (Mus domesticus) in and around irrigated summer crops in western New South Wales. Wildlife Research 22, 717-731. Twigg, L. E., Singleton, G. R., and Kay, B. J (1991). Evaluation of bromadiolone against house mice (Mus domesticus) populations in irrigated soybean crops. I. Efficacy of control. Wildlife Research 18, 265-274. Tyndale-Biscoe, C. H. (1994). Virus-vectored immunocontraception of feral mammals. Reproduction, Fertility and Development 6, 281-287. 65

Wolff, J. O. (1993). What is the role of adults in mammalian juvenile dispersal? Oikos 68, 173-176. Wood, B. J. (1971). Investigation of rats in rice fields demonstrating an effective control method giving substantial yield increase. PANS 17, 180-193. Wood, B. J. (1994). Rodents in agriculture and forestry. In Rodent Pests and Their Control. (Eds. A. P. Buckle and R. H. Smith.) pp. 45-83. (CAB International: Wallingford.) Ylönen, H., Jacob, J., Davies, M. J., and Singleton, G. R. (2002). Predation risk and habitat selection of Australian house mice Mus domesticus during an incipient plague: desperate behaviour due to food depletion. Oikos 99, 284-289. Zhang, Z., Chen, A., Ning, Z., and Huang, X. (1999). Rodent pest management in agricultural systems in China. In Ecologically-based Management of Rodent Pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 261-284. (Australian Centre for International Agricultural Research: Canberra.)

Study I. Movements, habitat use and response of ricefield rats to removal in an intensive cropping system in Vietnam.

Accepted for publication in the Belgian Journal of Zoology.

Brown, P. R.A,B, Tuan, N. P.C, and Banks, P. B.B

A CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia B School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia C National Institute of Plant Protection, Chem, Tu Liem, Hanoi, Vietnam

Movements, habitat use and response of ricefield rats to removal in an intensive cropping system in Vietnam

Peter R. BrownA,B, Nguyen Phu TuanC and Peter B. BanksB

Abstract: Rapid post-control reinvasion typically hampers attempts to manage rodent pests, yet little is known about the demography or behaviours of re-invaders. Here we study the habitat use and movement of Rattus argentiventer using radio- telemetry during a non-breeding season (tillering growth stage of rice) and a breeding season (ripening stage of rice) in lowland irrigated rice in Vietnam. On two treatment sites, farmers removed rats by hunting, digging up burrows and by using trap barrier systems (early planted field of rice surrounded by a plastic fence set with multiple capture rat traps), and on two control sites, farmers conducted their normal control practices. The 95% minimum convex polygon home range size of rats during the non- breeding period was 2.4 ha (n = 12) and significantly smaller than during the breeding period (9.8 ha; n = 10). There was no difference in home range size between treatment (removal sites) and control sites. During the non-breeding period, rats preferred to use the bank/channel habitat during the day, and preferred vegetable habitats at night. During the breeding period, rats preferred using rice habitats both during the day and at night. This preference during the breeding period was strongly influenced by the availability of abundant cover and food offered by the mature rice crops. Rats were moving about the rice fields in random directions and were not influenced by the removal of rats at nearby locations. We conclude that even at low population densities, rodent control would need to be conducted over large areas to prevent recolonisation 68

through random dispersal events and that rodent burrows should be destroyed during the non-breeding season when little cover is provided by crops.

Key words: habitat use, movements, ricefield rat, removal, trap barrier system, bounty system

I.1 Introduction

Rodents are a significant problem for agriculture in Vietnam. They are considered the number one pre-harvest pests of lowland irrigated rice crops, especially in the Mekong and Red River Deltas (Brown et al. 1999, 2003b). In particular, the ricefield rat, Rattus argentiventer (Robinson & Kloss 1916), is the most common rodent found in rice crops in Vietnam, and it is an important pest of rice crops in other parts of Southeast Asia including Malaysia and Indonesia. In Indonesia, it causes annual pre- harvest losses of around 17% (Geddes 1992; Leung et al. 1999). Other rodent species inhabiting rice fields in Vietnam include R. losea, R. rattus and Bandicota indica (Brown et al. 1999, 2003b). Little is known about how these rodent pest species interact with each other within the rice growing areas or how management should be implemented in a palliative manner to reduce damage to rice crops. Currently, most farmers are reactive in their control actions, only implementing management once the rat problem is moderate to severe. Methods for controlling damage caused by rodents in rice agroecosystems include application of rodenticides (Buckle 1999), hunting, fumigation, physical barriers such as the trap barrier system (TBS, Singleton et al. 1998, 1999), and cultural practices such as synchronised cropping, sanitation of fields and encouraging predators (such as barn owls) (Leung et al. 1999). There are few data on how rat populations respond to such control actions or how quickly reinvasion occurs. Also of interest is how rats respond to control at different stages of crop development when the availability of food and cover changes, and whether there are differences in response in breeding and non- breeding seasons of the rats. A key strategy for animals successfully reinvading areas is to have high rates of dispersal. One method for measuring rates of dispersal of small mammals into vacant areas is to experimentally remove animals from grids. Schieck and Millar (1987) 69

studied the response of red-backed voles (Clethrionomys gapperi) to removal trapping in a montane fir forest in Alberta Canada and found that about 80% of the voles caught in the removal area originated from a distance of less than two home ranges away. Nakata and Satoh (1994) studied the response of individual gray-backed voles (Clethrionomys rufocanus bedfordiae) to removal trapping to determine the source of animals moving into the removal grid and the distance that these animals moved from source areas. After 2 weeks, over 90% of the voles initially located within 30 m of the edge of the removal grid were making single-direction movements towards the removal grid. Conversely, Boutin et al. (1985) found that only 28% of snowshoe hares (Lepus americanus) dispersed into removal areas and that most animals died on their home range rather than dispersing, while Sullivan and Sullivan (1986) found that the colonization rate was 25 – 58% per four-week period.

There are few examples where researchers have monitored changes in movements of pest small mammals in natural field conditions using contemporary control methods. Efford et al. (2000) looked at home range changes in feral brushtailed possums (Trichosurus vulpecula) in New Zealand after applying an 80% control in one half of their experimental plot. They found that possums on the edge of the control area moved their home ranges towards the removal area and that the “vacuum effect” in the possums was largely confined to home range adjustments by individuals that had ranges overlapping the area of reduced density. Leirs et al. (1997) found that recolonisation of maize fields by the multimammate rat (Mastomys natalensis) occurred very rapidly after a rodent control operation. Despite the economic and social costs caused by R. argentiventer, its habitat use and movements in the rice agro-ecosystems of Vietnam is not well understood. In West Java, Indonesia, there are two lowland irrigated rice crops produced each year corresponding with the wet and dry seasons. Rattus argentiventer accounted for >95% of rodent species captured (Leung et al. 1999) and were found to have home ranges of 1-3 ha, with little differences between males and females, with smaller home range in the breeding season compared to the non-breeding season (Brown et al. 2001). They mostly utilised banks (burrows) during the tillering stage of the rice crop (non-breeding period), but switch to daytime use of rice paddies throughout the ripening stage of the rice crop (breeding period) (Brown et al. 2001). Research is therefore required in Vietnam because the cropping system and composition of rodent species are different. 70

As part of a large project examining the population response to a range of rodent control methods at a village-level (>100 ha), we examined how individual rats used their environment and how they might respond to removal of other rats through the application of control techniques conducted by farmers. Specifically, we considered whether rats moved in a random pattern (classical diffusion) or directed their movement towards areas of lower population density (“vacuum effect” Efford et al. 2000). This was done by radio-tracking individual rats occupying rice fields on sites where farmers conducted a range of recommended rodent management practices (treated sites) and sites where farmers were not influenced in their rodent management techniques (control sites).

I.2 Materials and Methods

I.2.1 Study site

The study was conducted in Vinh Phuc Province, in northern Vietnam, 40 km north of Hanoi (21°08’ N; 105°45’ E). Four study sites were selected to comprise part of a main village or sub-villages. Each site was 0.5 to 1 km apart and about 100 - 150 ha in size. The sites were set up in March 1999 to monitor the population dynamics of rats before implementation of ecologically based rodent management (Brown et al. 2003b). Within each site families manage small plots of land, each 0.04 ha, and each family generally owns a total of 0.5 – 0.7 ha of land. The principal crop grown in the area is rice. There are two main rice-growing seasons each year, the spring rice season (transplanted late February and harvested mid June), and the summer rice season (transplanted mid July and harvested late September). Rice is not grown in winter because it is too cold. Other crops are vegetables (broccoli, cabbage, kohlrabi, onion, pumpkin, tomato) and flowers (chrysanthemum, rose). Summers are hot and wet, and winters are cool and dry. The annual average rainfall is approximately 1600 mm, most falling during May to September. Farmers irrigate their crops from channels using water supplied from large storage dams in nearby hills. The soil type is heavy red clays. The radio-tracking study was conducted during the spring rice season in 2002. Rice was sown in February 2002 and then harvested in late June and early July 2002. Two sessions of radio-tracking were conducted to coincide with the non-breeding 71

season of rats (during the tillering stage of the rice crop; March) and during the breeding season of rats (during the ripening stage of the rice crop; June).

I.2.2 Trapping and radio-tracking

At each site, rats were caught using single-capture wire cage traps. Traps were baited with fresh vegetables and set strategically at sites where there was obvious rodent activity to catch as many rats as possible over an area of 250 x 250 m. At each site, fifty traps were set per night for eight consecutive nights in March and six consecutive nights in June. All adult female R. argentiventer rats were collared on treatment and control sites, and all adult male R. argentiventer were collared on control sites only. Resources and labour were limited so we chose not to monitor males on treatment sites. Traps were checked hourly during the first few hours after sunset and early each morning. At capture, each rat was weighed (± 2 g), sexed, and breeding condition determined and to confirm species identification and condition. Females with raised teats and perforated vagina were classified as adults, and males with descended testes were classified as adult. Prior to release, at point of capture, each rat was fitted with a single-stage radio transmitter (Sirtrack, New Zealand) attached to a nylon cable tie which functioned also as a collar around the animal’s neck. A 250 x 250 m grid of bamboo poles set 25 m apart was used to provide reference points for locating radio-collared rats. Radio tracking at all sites was conducted for up to 14 days in both March and June. Four locations or “fixes” were sought each day: one during daylight hours (0800-1400 hrs) for location of rat nests; and three after dusk (1900-2400 hrs) when rats were most active. Night fixes were 1 to 1.5 hours apart. It was not always possible to obtain three fixes for each rat after dusk. Collared rats were tracked with a hand-held 3-element Yagi antenna connected to a radio receiver. More than 80% of location fixes were tracked to within 1 m of their actual location, based on sightings of collared rats. For others, it was not possible to obtain more accurate fixes, because rats were moving around in rice paddies and would swim away before we could obtain an accurate fix. The habitat type (large roadside bank or channel bank, rice paddy, vegetable crop, fallow, flower crop) and activity (eg sighted in field or known to be in a burrow) were recorded for each fix. 72

Home ranges were calculated from 95% and 100% minimum convex polygons (MCP) using RANGES V (Kenward & Hodder 1996). We calculated 95% and 100% MCP because the 100% MCP may include forays from their core areas to explore new areas and thus relevant to our hypotheses. Analyses were performed on rats that had >15 fixes, the minimum number of fixes required to estimate 80% of the home range size as found by Brown et al. (2001) and confirmed with these data. Home ranges were ln-transformed to reduce the skewed distribution for statistical analysis. The range span was also calculated using RANGES V, and is defined as the largest distance across the MCP. The habitat use for each rat was determined within each individual animal’s home range by examining the proportion of fixes within each habitat type (Otis & White 1999). Log ratios of usage/availability were calculated for each habitat for each rat as the basis for compositional analysis of proportional habitat use (Aebischer et al. 1993). Habitat availability and use was compared between months (March and June) and time of day (Day or Night). During each tracking session, the crop types grown in each field within the 250 x 250 m grid (6.25 ha) area were recorded by walking through each site. The area of channels, banks and paths was estimated by measuring their widths and lengths. The area of each habitat type was then calculated and converted to a proportion of habitat available.

I.2.3 Implementation of treatments

On Treatment sites (T1 and T2), two areas were set up: 1) where rats were captured and collared for radio-tracking (non-removal area, as described above), and 2) where rats were removed. Each area was 6.25 ha in size. The removal areas were 225- 250 m from the non-removal areas based on average home range sizes and distances that rats would travel and get caught in a TBS (Brown et al. 2001, 2003a). Rats were removed by the use of a tactical bounty system (Singleton et al. 1999), where farmers were paid 200 dong (USD$0.02 per rat) to hunt and dig rats from the removal area at a stage when rat populations abundance was low. The bounty system operated during both March and June on both Treatment sites. In addition, on T2 in June, two trap- barrier systems (TBS; Singleton et al. 1998, 1999) were present with sticky rice as the lure crop (variety Khang Dan, 140-150 days duration, established in late March and 73

harvested after we concluded field work in June). On Control sites (C1 and C2), farmers conducted their normal rodent control practices. To measure the distance and direction of movements of rats from non-removal areas the average location from the first two days of tracking (calculated by averaging X- and Y- coordinates of the first 5-8 fixes) and the average location from the last two days of tracking (last 5-8 fixes) were calculated for each rat. Each period of tracking contained at least two daytime locations. The distances (m) and directions (bearings) moved from the first two days to the last two days were then determined. On Control sites, distances and directions towards the principal compass points (± 45° of each of N, E, S, and W) were calculated, and on Treatment sites, distances and directions towards the removal area (± 45° of N for Treatment 1, ± 45° of E for Treatment 2) and away from removal areas, were calculated.

I.3 Results

In March, we trapped 51 rats from 2800 trap nights (trap success = 1.8%) in total from all sites, and in June we trapped 21 rats from 2400 trap nights (trap success 0.9%). Twenty-one adult R. argentiventer rats were collared for radio-tracking in March, and ten adult R. argentiventer rats were collared in June (Table I.1). The regional abundance of rats at this time (spring) of year was generally low (mean trap success of 0.5% in March and 1.9% in June from our regular trapping locations as part of the village-level study being conducted, Figure I.1), and we believed we captured the majority of R. argentiventer present in the area. In March, 12 and three rats were removed by farmers by hunting and digging burrows from removal areas on Treatment 1 and Treatment 2 respectively, and in June, 63 and 50 rats were removed by farmers from removal areas on Treatment 1 and Treatment 2 respectively, most of which were juvenile animals dug from nests (evidence of active breeding on the sites).

6 Breeding season Control 1 5 Control 2 Treatment 1 4 Treatment 2 (%) 3

Trap success 1

0 Spring rice crop

Jan 02 Feb 02 Mar 02 Apr 02 May 02 Jun 02 Jul 02

Month Figure I.1. Abundance of rats (number of rats captured per 100 trap nights) on four sites used for a separate village-level study, Vinh Phuc, Vietnam from January to July 2002. Shown is approximate timing of the spring rice crop and rat breeding season (based on pregnant and lactating adult females) (P. R. Brown and N. P. Tuan unpublished data).

Table I.1. Summary of radio-collared rats in March (non-breeding season) and June (breeding season) at Vinh Phuc, Vietnam. Shown are Control (C1 and C2) and Treatment (T1 and T2) sites, radio-collar frequency, sex, the number of days each animal was tracked, the number of fixes obtained, the fate of the animal, the home range sizes (ha, calculated using 95% and 100% minimum convex polygon, MCP) and home range span (m). No. Tracking Rat No. 95% 100% Site Sex days Fate Span period No. fixes MCP MCP tracked March C1 37 M 8 28 Alive 1.01 1.15 187 64 M 8 28 Alive 4.99 5.07 496 47 M 12 45 Alive 3.13 3.96 360 241 F 7 21 Poisoned 2.32 2.33 329 15 F 9 30 Alive 1.28 1.85 309 C2 3 M 2 5 Poisoned - - - 6 M 2 6 Poisoned - - - 55 M 11 42 Alive 0.93 0.94 186 60 M 6 24 Poisoned 2.10 2.27 265 24 M 1 2 Poisoned - - - 54 M 1 2 Fatally - - - injured 41 M 2 5 Poisoned - - - T1 75 F 9 28 Poisoned 4.18 4.59 342 68 F 2 5 Poisoned - - - 60 F 2 3 Predation - - - T2 25 F 12 43 Poisoned 0.14 1.33 199 39 F 9 39 Alive 3.79 4.43 381 43 F 2 6 Fatally - - - injured 44 F 14 52 Alive 4.19 4.83 716 45 F 13 50 Alive 0.79 0.91 217 49 F 1 3 Missing - - -

June C1 66 F 10 34 Alive 16.75 18.28 827 26 F 9 30 Alive 3.83 3.85 370 79 F 10 30 Alive 1.04 1.07 193 C2 53 F 13 46 Alive 4.89 9.90 487 T1 33 F 14 48 Alive 1.44 1.49 256 73 F 13 47 Alive 2.56 2.74 287 20 F 12 43 Alive 15.26 16.03 908 T2 23 F 14 45 Alive 2.08 4.16 355 71 F 13 45 Alive 17.07 19.28 755 57 F 9 32 Alive 32.97 43.91 1433

In March, 12 rats had >15 fixes (57% of rats captured), whereas ten rats in June had >15 fixes (100%). Nine rats collared and released in March died from suspected poisoning (small movements, lethargic behaviour observed, or were found lying dead on the ground), one died from predation (found radio collar lying with remains of internal organs) and two were thought to be hit by farmers (fatally wounded by blow to body), whereas in June, there was no mortality of radio-collared rats during the tracking period (Table I.1). If we combine deaths due to rodenticide and injury, we find that in March, farmers caused a mortality rate in collared rats of 20%, 85%, 67%, and 33% for C1, C2, T1, and T2 respectively (52% overall). The average home range size of rats (estimated using the 95% minimum convex polygon method) in March was 2.40 ha (± 0.47 SE) and in June was 9.79 ha (± 3.31 SE) (Figure I.2). The ln-transformed home range size for female rats was significantly larger in June than in March, (F1,13 = 4.781; P = 0.048), but there was no difference between Treated and Control sites (F1,13 = 0.005; P = 0.947). We could not test for differences between males and females, because no males were captured in June.

March - non-breeding season June - breeding season 35 54265 0 30 a)

(h 25 ze i

s 20

nge 15 ra 10 me

Ho 5

0 Male Female Female Male Female Female Control Treatment Control Treatment Figure I.2. Box plot of 95% minimum convex polygon home range sizes (ha) for males and females, in treatment and control sites for March (non-breeding season) and June (breeding season). The box encloses the 25th and 75th percentiles, the solid line shows the median and the dotted line the mean home-range size. Vertical lines span the 10-90th percentiles. Sample sizes are shown at the top.

The home range size of rats calculated using the 100% MCP in March was 2.81 ha (± 0.48 SE) and in June was 12.07 ha (± 4.18 SE). The 100% MCP home range size in March was 0.40 ha larger on average (17.8% increase) and in June was 2.28 ha larger on average (16.3% increase) than 95% MCP. There was no significant difference in size of home ranges between the 95% and 100% MCP (paired t-test; t19 = -0.508, P = 0.618), therefore the 100% MCP did not provide additional information for home range analysis. The home range span of rats was used as an estimate of possible linear movements (Table I.1). There was no significant difference between home range span between months (F1,13 = 1.409; P = 0.256) or treatment (F1,13 = 0.253; P = 0.624). The average home range span in March was 332.3 m (± 44.0 SE, n = 12) and in June was 587.1 m (± 123.4 SE, n = 10). This confirms that the distance between non-removal and the removal areas was set at the right distance (225-250 m). In March, rats spent most time during the day in the bank/channel habitat (82.8% compared to 10% available) and at night most fixes occurred in the vegetable habitat (37.7% compared to 51.3% available) (Figure I.3a). Rats were not located in rice fields during the day at any stage during March. Rats used the flower habitat roughly in proportion with availability (day fixes = 10.3%; night fixes = 15.91; available 16.7%). Some rats consistently had day and night fixes in flower fields suggesting they had constructed a burrow there and were feeding within the field. In June rats were spending more time in rice habitats with 73.2% of day fixes and 73.9% of night fixes in rice paddies compared to the 40.0% available (Figure I.3b). Rats had reduced their use of bank/channel habitats to 25.0% and 18.8% for day and night fixes respectively (compared to 10% available). Very few fixes occurred in vegetable, flower or fallow fields. The availability of crops changed between March and June because of changes in the types of crop grown. The ratios of usage/availability confirm that in March, rats preferred to use bank habitats, and in June, rats preferred to use rice habitats and banks to a lesser extent (Table I.2).

100 a) March Available 80 Day Night

40 Percent (%)

0 b) June Available 80 Day Night

40 Percent (%)

0 Vegetable Bank/channel Flower Rice Fallow

Habitat type Figure I.3. Habitat use of ricefield rats (sexes and sites combined) showing the percentage of habitats available to rats and the percentage of radio-telemetry location fixes within each habitat type for day and night fixes, Vinh Phuc Province, Vietnam 2002. (a) March, the non-breeding period for rats during the tillering stage of rice, and (b) June, the breeding period for rats during the ripening stage of rice. Error bars represent standard error of means from habitat use of individual rats.

The distances moved by rats from the average of the first 2 days to the last 2 days were generally twice as large on treated sites as they were on control sites (Control March = 88.0 m ± 30.6 SE, n = 7; Treatment March = 190.9 m ± 94.0 SE, n = 5; Control June = 184.9 m ± 65.8 SE, n = 4; Treatment June = 411.8 m ± 166.1 SE, n = 6) (Figure

I.4), but the distances moved were not significant (Time F1,18 = 2.86; P = 0.108;

Treatment F1,18 = 0.278; P = 0.604; interaction F1,18 = 0.06; P = 0.811).

Table I.2. Habitat selectivity of ricefield rats during March (tillering stage of rice crop; non-breeding season) and June (ripening stage of rice crop; breeding season) for day and night fixes for each habitat, Vinh Phuc province, Vietnam. The selectivity index is calculated by dividing the proportion of observations of rats in each habitat type by the proportion of habitat available. A selectivity value of >1 implies preference while a value of <1 implies avoidance. Month Habitat type Time Vegetable Bank Flower Rice Fallow March Day 0.12 ± 0.04 8.42 ± 0.81 0.57 ± 0.45 0.00 ± 0.00 0.01 ± 0.00 Night 0.76 ± 0.16 3.07 ± 0.65 0.91 ± 0.43 0.71 ± 0.16 0.53 ± 0.52 June Day 0.02 ± 0.02 2.37 ± 0.48 0.05 ± 0.05 3.65 ± 0.21 0.01 ± 0.00 Night 0.09 ± 0.05 1.89 ± 0.48 0.04 ± 0.02 3.65 ± 0.26 0.86 ± 0.45

1200 7546

1000 )

m 800 ( ed v o 600 stance m

i 400 D

200

0 March C March T June C June T

Time and Treatment Figure I.4. Box plot of distances (m) moved by rats from first 2 days to last 2 days of tracking on Treatment (T) and Control (C) sites in March and June 2002. The box encloses the 25th and 75th percentiles, the solid line shows the median and the dotted line the mean distance travelled, and the vertical lines span the 10-90th percentiles. Sample sizes are shown at the top.

The directions moved by rats on control and treatment sites were proportional with the directions available. On control sites, there were three rats that moved towards North, three to East, five to South and zero to West, with no preference for direction moved (the association between the observed directions used and directions expected 2 was not significant: χ 3 = 4.636; P = 0.2004). On treatment sites, there was one rat that moved towards the removal area and 10 rats that did not move towards the removal area 2 (1/5 rats in March and 0/6 rats in June), with no preference for direction moved (χ 1 = 1.778; P = 0.1824). Therefore, the direction of movements were essentially random on both control and treatment sites. To confirm that rats were indeed breeding in June, the animals that could be retrieved were assessed for breeding condition (presence of embryos or litter of pups in the burrow). Of the three female rats recaptured, one was pregnant and two had young pups in their burrow, confirming that they were indeed breeding (100%). The breeding condition of the other females could not be ascertained because it was not possible to recapture the animals.

I.4 Discussion

The removal of ricefield rats during a low-density phase in an intensive cropping system in Vietnam did not induce movements of neighbouring rats towards the removal area. Rats on non-removal areas were moving randomly with regard to directions, and we believe that for R. argentiventer, recolonisation events during low-density phases occur through random dispersal events (classical diffusion). We could not support the “vacuum effect” proposed by Efford et al. (2000) for brushtail possums in New Zealand, however the density of rats in this study was low. Krebs et al. (1976) found that recolonisation rates were higher when densities of Microtus townsendii were higher because of competition for space, so this study should be repeated at higher densities (>10% trap success) to test this hypothesis. It is likely that populations of ricefield rats are made up of predominantly transient animals with high rates of dispersal, as found for multimammate rats (Mastomys natalensis) in Tanzania (Leirs et al. 1996). We found rats moved around a great deal, and in some cases rats had very large home ranges (>5 ha) and did not consistently use a particular burrow or nest site. In studies conducted in both Vietnam 81

and Indonesia, ricefield rats have recapture rates of less than 1% (Brown et al. 1999, 2003b; Leung et al. 1999; Jacob et al. 2003b), and part of this reason may be because of the high proportion of transient animals. We therefore predict that rats inhabiting these highly modified and intensive rice production systems would have higher rates of dispersal than rats living in stable environments. The home range size of rats during the non-breeding season (tillering stage, March; 2.7 ha) was of the same order as that found for ricefield rats in Indonesia (2-3 ha, Brown et al. 2001). However, the home range size was much larger during the breeding season (ripening stage, June; 10 ha) than in Indonesia. We were surprised to find that home range sizes were larger during the breeding season. In house mice in Australia, for example, home ranges were significantly smaller during the breeding season (Chambers et al. 2000). The home range size of Rattus rattus in macadamia nut orchards in Hawaii did not vary between males and females and did not vary through different stages of nut development (Tobin et al. 1996). Christensen (1996) found no seasonal variation in home range sizes of Mastomys natalensis in Tanzania as determined by capture-mark-release data. However, both male and female Calomys venustus in Argentine agroecosystems had larger home ranges during the breeding season compared to the non-breeding season (Priotto et al. 2002). It is not clear why R. argentiventer might have a larger home range during the breeding season (June), but it could be related to the farming activities or farmers preparing for harvest. We expected that adult female rats, if they are actively breeding, would have stable, small home ranges particularly if they are caring for young in the nest. The recaptures of rats in June confirmed that the rats were indeed breeding (pregnant or suckling new born pups). We could not prevent farmers from undertaking extraneous rodent control on our study sites. On our control sites in March, farmers poisoned nine radio-collared rats with rodenticide and two other rats died through farmers causing fatal injury. This reflects the rat control efforts employed by farmers during the tillering stage of the rice crop. Farmers are generally busy with preparations for harvest in June, so they have little time for undertaking rodent control. No deaths of radio-collared rats occurred on any site in June. The impact of these activities on this study is difficult to determine. Rats in this intensive rice growing agroecosystem are subject to a wide array of disturbances including ploughing of fields, harvesting of crops, irrigation of crops and 82

application of chemicals for weed or insect control. Rat populations have developed strategies for survival under these conditions through high reproductive output (Lam 1983; Tristiani et al. 1998) and through their ability to recolonise areas. Rats were using a range of habitats that were available to them, and their choice of habitat was related to cover and availability of food. When cover from tillering rice was low (March), rats were spending time in burrows in the bank/channel habitat, and when rice was ripening (June), rats were spending their time in the rice fields. We could not measure availability of food for rats, but observations made at the time showed that abundant food was always available through ripening vegetable crops such as kohlrabi, tomatoes, cabbage and broccoli, and particularly in June, abundance of maturing rice. Food was therefore probably not a limiting resource. In March, rats preferred to use the bank/channel habitat during the day, but preferred vegetable habitats at night. In June, rats preferred using rice habitats during the day and at night. This preference in June was strongly influenced by the availability of abundance cover and food offered by the maturing rice crops. These findings are similar to that found for R. argentiventer in Indonesia (Brown et al. 2001). These results suggest that there would be little point in destroying rat burrows along channels and bank habitats during the later stages of crop growth (after maximum tillering stage of rice) because rats were predominantly utilising rice crops (Brown et al. 2001). It would be interesting to monitor the changes in habitat use and movements of rats after harvest of the rice crop to see whether they revert back to using the channel/bank habitat or disperse to other habitats offering sufficient food and cover. Jacob et al. (2003a) found that the home range size of R. argentiventer in Indonesia decreased by 67% after harvest. The findings from the current research will help in refining appropriate management practices that farmers can use on a large scale (eg village level) (Singleton 1997; Brown et al. 2003b; Leirs 2003; Jacob et al. 2003b). Further research is required to examine recolonisation when population densities are higher and to look at other compensatory mechanisms such as breeding performance and recruitment.

I.5 Acknowledgements

We sincerely thank the farmers who assisted us with the implementation of the treatments and for their help with the radio tracking. We thank M. Davies (CSIRO), D.T Hue, P.T.T. Ha, P.T. Hoa, L.T. Hoa, P.T. Lien, P.V. Kien, and Ms Yen (NIPP) for assistance with this research. We thank D. Grice and G. Singleton for comments on early drafts. The research was supported by the Australian Centre for International Agricultural Research (AS1/98/36) and CSIRO. The research was conducted in accordance with the Australian code of practice for the care and use of animals for scientific purposes (SEAEC No. 01/02 - 15).

I.6 References

Aebischer, N. J., Robertson, P. A., and Kenward, R. E. (1993). Compositional analysis of habitat use from animal radio-tracking data. Journal of Wildlife Management 74, 1313-1325. Boutin, S., Gilbert, B. S., Krebs, C. J., Sinclair, A. R. E., and Smith, J. N. M. (1985). The role of dispersal on the population dynamics of snowshoe hares. Canadian Journal of Zoology 63, 106-115. Brown, P. R., Hung, N. Q., Hung, N. M., and van Wensveen, M. (1999). Population ecology and management of rodent pests in the Mekong River Delta, Vietnam. In ‘Ecologically-based management of rodent pests’. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 319-337. (Australian Centre for International Agricultural Research: Canberra.) Brown, P. R., Leung, L. K-P., Sudarmaji, and Singleton, G. R. (2003a). Movements of the ricefield rat, Rattus argentiventer, near a trap-barrier system in rice crops in West Java, Indonesia. International Journal of Pest Management 49, 123-129. Brown, P. R., Singleton, G. R., and Sudarmaji (2001). Habitat use and movements of the rice-field rat, Rattus argentiventer, in West Java, Indonesia. Mammalia 65, 151-166. Brown, P. R., Tuan, N. P., Singleton, G. R., Tuat, N. V., Tan, T. Q., and Hoa, L. T. (2003b). Impact of village-level rodent control practices on rodent populations and rice crops in Vietnam. In ‘Rats, Mice and People: Rodent Biology and 84

Management’. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 197-202. (ACIAR: Canberra.) Buckle, A. P. (1999). Rodenticides - their role in rodent pest management in tropical agriculture. In ‘Ecologically-based Management of Rodent Pests’. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 163-177. (Australian Centre for International Agricultural Research: Canberra.) Chambers, L. K., Singleton, G. R., and Krebs, C. J. (2000). Movements and social organization of wild house mice (Mus domesticus) in the wheatlands of Northwestern Victoria, Australia. Journal of Mammalogy 81, 59-69. Christensen, J. T. (1996). Home range and abundance of Mastomys natalensis (Smith, 1834) in habitats affected by cultivation. African Journal of Ecology 34, 298- 311. Efford, M., Warburton, B., and Spencer, N. (2000). Home-range changes by brushtail possums in response to control. Wildlife Research 27, 117-127. Geddes, A. W. M. (1992). ‘The relative importance of pre-harvest crop pests in Indonesia’. (Natural Resources Institute: Chatham, UK.) Jacob, J., Nolte, D., Hartono, R., Subagja, J., and Sudarmaji (2003a). Pre- and post- harvest movements of female rice-field rats in West Javanese rice fields. In ‘Rats, mice and people: rodent biology and management’. ACIAR Monograph 96. (Eds. Singleton. G. R., L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 277- 280. (ACIAR: Canberra.) Jacob, J., Sudarmaji, and Singleton. G. R. (2003b). Ecologically-based management of rice-field rats on a village scale in West Java: experimental approach and assessment of habitat use. In ‘Rats, Mice and People: Rodent Biology and Management’. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 191-196. (ACIAR: Canberra.) Kenward, R. E. and Hodder, K. H. (1996). ‘RANGES V – An analysis system for biological location data.’ (Institute of Terrestrial Ecology: Dorset.) Krebs, C. J., Wingate, I., LeDuc, J., Redfield, J. A., Taitt, M., and Hilborn, R. (1976). Microtus population biology: dispersal in fluctuating populations of M. townsendii. Canadian Journal of Zoology 54, 79-95. Lam, Y. M. (1983). Reproduction in the ricefield rat Rattus argentiventer. Malayan Nature Journal 36, 249-282. 85

Leirs, H. (2003). Management of rodents in crops: the Pied Piper and his orchestra. In ‘Rats, Mice and People: Rodent Biology and Management’. ACIAR Monograph 96. (Eds. Singleton. G. R., L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 183- 190. (ACIAR: Canberra.) Leirs, H., Verheyen, W., and Verhagen, R. (1996). Spatial patterns in Mastomys natalensis in Tanzania (Rodentia, Muridae). Mammalia 60, 545-555. Leirs, H., Verhagen, R., Sabuni, C. A., Mwanjabe, P., and Verheyen, W. N. (1997). Spatial dynamics of Mastomys natalensis in a field-fallow mosaic in Tanzania. Belgian Journal of Zoology 127, 29-38. Leung, L. K. P., Singleton, G. R., Sudarmaji, and Rahmini (1999). Ecologically-based population management of the rice-field rat in Indonesia. In ‘Ecologically-based management of rodent pests’. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 305-318. (Australian Centre for International Agricultural Research: Canberra.) Nakata, K. and Satoh, K. (1994). Distance of movements of voles after removal in Clethrionomys rufocanus bedfordiae (Rodentia : Cricetidae). Applied Entomology and Zoology 29, 97-103. Otis, D. L. and White, G. C. (1999). Autocorrelation of location estimates and the analysis of radiotracking data. Journal of Wildlife Management 63, 1039-1044. Priotto, J., Steinmann, A., and Polop, J. (2002). Factors affecting home range size and overlap in Calomys venustus (Muridae: Sigmodontinae) in Argentine agroecosystems. Mammalian Biology 67, 97-104. Schieck, J. O. and Millar, J. S. (1987). Can removal areas be used to assess dispersal of red-backed voles? Canadian Journal of Zoology 65, 2575-2578. Singleton, G. R. (1997). Integrated management of rodents: A Southeast Asian and Australian perspective. Belgian Journal of Zoology 127, 157-169. Singleton, G. R., Sudarmaji, and Suriapermana, S. (1998). An experimental field study to evaluate a trap-barrier system and fumigation for controlling the rice field rat, Rattus argentiventer, in rice crops in West Java. Crop Protection 17, 55-64. Singleton, G. R., Sudarmaji, Jumanta, Tan, T. Q., and Hung, N. Q. (1999). Physical control of rats in developing countries. In ‘Ecologically-based Management of Rodent Pests’. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 178-198. (Australian Centre for International Agricultural Research: Canberra.) 86

Sullivan, T. P. and Sullivan, D. S. (1986). Resiliency of snowshoe hares to population reduction. Journal of Applied Ecology 23, 795-806. Tobin, M. E., Sugihara, R. T., Koehler, A. E., and Ueunten, G. R. (1996). Seasonal activity and movements of Rattus rattus (Rodentia, Muridae) in a Hawaiian macadamia orchard. Mammalia 60, 3-13. Tristiani, H., Priyono, J., and Murakami, O. (1998). Seasonal changes in the population density and reproduction of the ricefield rat, Rattus argentiventer (Rodentia : Muridae), in West Java. Mammalia 62, 227-239. 87

Study II. Compensation of rodent pests after removal: control of two rat species in an irrigated farming system in the Red River Delta, Vietnam.

Accepted for publication in Acta Oecologica

Peter R. Brown1,2 and Nguyen Phu Tuan3,

1 CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia 2 School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia 3 National Institute of Plant Protection, Chem, Tu Liem, Hanoi, Vietnam

Compensation of rodent pests after removal: control of two rat species in an irrigated farming system in the Red River Delta, Vietnam.

Peter R. Brown1,2 and Nguyen Phu Tuan3

Abstract: Rodent pests have a strong capacity to recover rapidly from imposed reductions in abundance, but it is unclear how populations compensate to removal when farmers apply rodent control. The response of two rat species to rodent control was monitored using regular live-trapping in an irrigated lowland mixed rice agroecosystem. Ricefield rats, Rattus argentiventer (52% of rodent captures), and lesser ricefield rats, R. losea (29%), were removed using trap-barrier systems (TBS) at two sites while rice crops were present. The TBS was a plastic fence that enclosed a small field planted 3 weeks prior to the surrounding fields and set with multiple-capture cage-traps to capture rats that were attracted to the early-planted crop. Demographic responses of rats were compared to two untreated sites. There was a reduction in abundance of rodents on treated sites relative to untreated sites during the summer rice crop (by 45% and 28% for R. argentiventer and R. losea, respectively) an increase in abundance during the winter season (31% and 69%), and a mixed response during the spring rice crop (39% decrease and 41% increase). There was an increase in the proportion of juveniles captured on treated sites relative to untreated sites post-treatment (148% and 158%) and the body mass was lower on treated sites post-treatment (males: 13% and 41%; females: 22% and 22%). Older, larger animals were removed by the TBS and the rodent populations were compensating through high recruitment of young and high immigration into treated sites. No clear patterns were observed in the occurrence of 89

adult breeding females for either R. argentiventer or R. losea. Rodent management should therefore occur over large areas (>100 ha) to reduce the chance of reinvasion.

Keywords: demography, management, Rattus argentiventer, Rattus losea, rice crop

II.1 Introduction

The response of a population to a severe shock or catastrophe depends on the mechanisms by which the population can recover. If densities are high then compensatory mechanism should return the population to pre-shock densities relatively quickly, but this depends on the severity of the shock (Caughley 1977; Hansson 1992). If densities of a population are low, it may not be possible for a population to compensate and therefore the population may become extinct (Caughley & Sinclair 1994, Sinclair 1996). Extinction is common for small populations on islands subjected to shocks such as predation by introduced animals (Pimm 1987; Alcover et al. 1998; Dowding & Murphy 2001; Álvarez-Castañeda & Ortega-Rubio 2003). A population responds to a decrease in density through density dependent mortality and fecundity, leading to a reduction in interspecific or intraspecific competition for food resources, reduction in predation pressure or the influence of disease. This has been well studied in the sustainable harvesting literature (Caughley 1977; Caughley & Sinclair 1994; Hilborn et al. 1995; Boyce et al. 1999). The rate at which this occurs can affect the speed of population recovery. These compensatory mechanisms are observed through changes in breeding, survival and recruitment (Sullivan & Sullivan 1986; Choquenot 1991; White & Bartmann 1998; Boyce et al. 1999; Byrom 2002). Density dependent dispersal also can be important in enabling populations to recover (Efford et al. 2000; Andreassen & Ims 2001; Gundersen et al. 2002). Rodent pests in particular appear to have an acute ability to recover rapidly after an induced shock (such as imposed control) has been applied (Drummond 1970). Recovery of rodent populations after control mainly has been identified through immigration (Myers & Krebs 1971; Sullivan 1979; Hansson 1992; Montgomery et al. 1997; Mwanjabe & Leirs 1997), but could also occur through higher breeding capacity of the remaining animals (pregnancy rates, litter sizes) (Davis & Christian 1958; 90

Montgomery 1981), and high rates of survival of adults or juveniles in the population (Gliwicz 1981; Montgomery 1981; Gundersen et al. 2001). One way of studying compensatory effects is to remove animals from an area and to monitor responses of the remaining animals. Krebs et al. (1976) found after removing animals from high density populations of Microtus townsendii that colonisers were smaller in size than resident animals, with about 10-15% more subadult males and females in breeding condition on the removal plots than on the control plots. Also, more males colonised the removal area than females. Montgomery (1981) found that in sympatric populations of Apodemus sylvaticus and A. flavicollis, individuals of both species entered reproductive condition earlier on the removal grid than on control grids. Schieck and Millar (1987) found that captures of red-backed voles on a removal area in a montane fir forest in Canada were biased towards young of the year females and against overwintered females, and it was likely that immigration occurred from less than two home range sizes away. Therefore, recovery on removal areas was through immigration of smaller, younger animals and by commencing reproduction earlier. For agricultural rodent pests in Southeast Asia, the level of compensation by populations to control and other disturbances and the associated mechanisms by which they accomplish this compensation are not well understood. Rodents are considered the number one pre-harvest pests of lowland irrigated rice crops, especially in the Mekong and Red River Deltas of Vietnam (Brown et al. 1999, 2003; Tuan et al. 2003). In Indonesia, rodents cause annual pre-harvest losses of around 17% (Geddes 1992; Singleton & Petch 1994; Leung et al. 1999; Singleton 2003). The main pest in Vietnam is the ricefield rat, Rattus argentiventer, but the lesser ricefield rat, R. losea, is commonly found also (Brown et al. 1999; Lan et al. 2003). Farmers, land managers and governments spend much time and effort controlling rats to reduce damage (Singleton & Petch 1994; Tuan et al. 2003). However, control efforts often do not reduce damage, either because control is conducted after damage has already occurred (Buckle 1994), or because the rodent populations recover rapidly after control and continue to cause damage (Sullivan 1979). In order to examine the demographic changes in a rodent population subjected to control actions applied by farmers under “natural conditions”, we use data from a four- year study where rodents were removed using a trap-barrier system (TBS, Singleton et al. 2003). TBS is a method designed to reduce the damage caused by rodents in 91

lowland irrigated rice fields (Lam 1988, Singleton et al. 1998). A plastic fence was erected around a small plot of rice (approximately 20 x 20 m), and the area inside the fence was planted with rice about 3-4 weeks prior to the surrounding fields. Multiple- capture rat traps were placed at the base of the fence at regular intervals adjacent to small holes in the fence. A water-filled moat approximately 1 m wide surrounded the fence and small mounds of mud led to the entrances of the traps to facilitate capture of the rats. These TBSs were used during the spring and summer rice crop seasons on two treated sites and removed up to 2,000 rats each season and effectively acted as a source of mortality (Brown et al. unpublished data). However, despite intensive efforts to control rodents, there was no overall difference in the number of rats captured on regular trap lines (capture-mark-release study) between treated and untreated sites even though the abundance of rats captured on regular trap lines was relatively low. Furthermore, there was no difference in the damage observed to the rice crops or to yield. The rodent populations on the treated sites were compensating for the removal of rats through the TBSs so this study looks at the mechanisms by which they achieved the compensation. Part of the success of the TBS in reducing populations and subsequent damage to rice crops is that animals removed early in the crop stage, prior to the commencement of the breeding season, do not contribute to breeding (Singleton et al. 1999). For example, removing one adult female prior to the breeding season will prevent 24 young being born, based on 3 litters during a breeding season (Lam 1983; Leung et al. 1999) and an average of 8 young per litter (Brown et al. 2003b). However, this assumes no compensation by the remaining rats. Removing animals might leave more resources for resident animals that do not get caught in the TBS and/or it may encourage the successful establishment of immigrants. We therefore consider four aspects of the demography of the two common rodent species at our study site, Rattus argentiventer and R. losea, to determine how rodent populations might compensate for rodent control. Possible increases and decreases due to various factors in demographic changes are outlined in Table II.1. If the response is neutral, then compensation does not occur for that characteristic. This study provides a unique opportunity to consider these mechanisms because the population monitoring and removal methods were independent, and had the added advantage of quantifying the number of animals removed. 92

Table II.1. Possible changes in four population characteristics after removal of rats from treated sites using TBS. If the response is neutral, then compensation does not occur. Population Response Demographic consequence Characteristic Abundance Higher Immigrants removed by TBS, high immigration rate, high survival of remaining animals Lower Resident animals removed by TBS, little immigration, poor survival Breeding Higher Less competition for resources, breeding animals not removed by TBS, immigrants removed by TBS Lower More competition for resources, breeding animals removed by TBS, non-breeding immigrants not removed by TBS Juveniles Higher High recruitment rates (survival) of young, high immigration of juveniles, poor survival of adults, older rats captured in TBS Lower Poor survival of young, poor immigration of young, high survival of older animals, juveniles captured in TBS Body mass Higher Relatively more immigrants of poor condition captured in TBS, small animals captured in TBS (large residents not captured in TBS) Lower Immigrants of good condition or residents of poor condition captured in TBS, large animals (residents) captured in TBS

II.2 Methods

II.2.1 Study site

The study was located on four sites near Tien Phong village, Me Linh District, Vinh Phuc Province, 40 km north of Hanoi in the Red River Delta of northern Vietnam (21°08’ N; 105°45’ E). This location was chosen because there was an outbreak of rodents in the area in 1997/98 that caused substantial damage to rice crops (Nguyen Phu Tuan, unpublished data). The principal crop grown in the area is rice. There are two main rice-growing seasons each year, the spring rice season (transplanted late February and harvested mid June), and the summer rice season (transplanted mid July and harvested late September). Other crops grown are vegetables (broccoli, cabbage, chilli, kohlrabi, onion, pumpkin, tomato) and flower crops (chrysanthemum, rose), and these 93

are grown throughout the year (Tuan et al. 2003). Summers are hot and wet, and winters are cool and dry. The annual average rainfall is approximately 1600 mm, most falling between May and September. Farmers irrigate their crops using water supplied by channels originating from large storage dams in nearby hills.

II.2.2 Imposition of treatments

Two sites were classified as treated sites (T1 and T2), where farmers followed a series of recommended practices to reduce the impact of rodents on their crops following the principles of ecologically-based rodent management (Singleton 1997; Singleton & Brown 1999). Ecologically-based rodent management utilises a solid understanding of the ecology of the main rodent pest species to develop appropriate control strategies that combine cultural, chemical and biological control to provide cost- effective, socially acceptable management that is environmentally benign. Two sites were classified as untreated sites (U1 and U2), where farmers were left to conduct their own conventional rodent control activities. One of the main activities conducted on the treated sites was the construction of the TBS during the spring and summer rice crops. There were 8-10 TBSs set up on each treated site, which protects a total area surrounding the TBSs of approximately 100 ha (Singleton et al. 2004). Farmers removed rats captured in the traps early each morning. The mean number of rats removed from treated sites in each season is shown in Table II.2. Other rodent control practices conducted included erecting plastic fences around individual fields to exclude rats from fields (without traps), using rodenticides to kill rats, and hunting and digging up burrows. These practices were conducted on both treated and untreated sites (Singleton et al. 2004).

Table II.2. Details of set up of TBS on treated sites from January 2000 until November 2002 (post-treatment period) for each season. TBS were not set up during winter. The number of rats captured in the TBS is the mean ± standard error from two treated sites (see Singleton et al. (2004) for details) with 8-10 TBS set per site. The trapping sessions used for each season are shown (number of months shown in brackets). No. rats Trap Season Dates Days set up captured sessions Spring 2000 16 Jan – 25 May 130 481 ± 247 11-14 (4) Summer 2000 16 Jun – 24 Aug 69 1,883 ± 51 16-17 (2) Winter 2000 18-20 (3) Spring 2001 1 Dec – 10 Apr 131 915 ± 88 21-25 (5) Summer 2001 15 Jun – 27 Aug 73 1,682 ± 56 28-29 (2) Winter 2001 30-34 (5) Spring 2002 5 Feb – 15 Jun 130 206 ± 31 35-38 (4) Summer 2002 10 Jun – 20 Aug 71 941 (T2 only) 40-41 (2) Winter 2002 42-44 (3)

II.2.3 Population sampling

Lines of 20 live-cage traps were set in three main habitats to monitor changes in the abundance of rats: (1) big channel banks, >2 m high and >2 m wide carrying irrigation water; (2) medium banks, about 1 m high and 1 m wide, smaller channels or paths between fields; and (3) small banks, <300 mm high and <300 mm wide separating rice fields. There were two replicates of each trap line, and trapping occurred for four consecutive nights for a total of 480 trap nights per site per month. Trapping occurred each month from April 1999 to November 2002, although, not each site was trapped each month because of a shortage of labour (U1 = 44 trap session, U2 = 27, T1 = 41, T2 = 24). Trapping results are presented as the number of individual animals captured per trapping session. More sophisticated abundance estimates could not be calculated because there were no recaptures of rats. Upon capture, each animal was identified to species, marked with an ear punch, weighed, and the external breeding condition was assessed for females (pregnancy by palpation and lactation).

II.2.4 Statistical analyses

As there was no overall difference observed in the monthly abundance of rats between treatments (Brown et al. unpublished data), the number of rats captured during the periods when the TBSs were set up (spring and summer rice crop periods) was compared to the number of rats captured when the TBSs were not set up (over the winter crop period) to determine whether rat populations were compensating through the removal of rats by the TBSs. The dates when the TBSs were set up for spring and summer and the intervening period during winter defined the trapping sessions used in the analysis of abundance for spring, summer and winter periods (Table II.2). The effects of treatment (treated or untreated), season (spring, summer, winter), and species (R. argentiventer, R. losea) and all two-way interactions with treatment were examined by ANOVA models of ln-abundance for each trapping session. To determine whether the number of rats captured in TBSs reflects the number of rats found in the fields, a simple linear regression was calculated for the number of rats captured in the TBS and the number of rats captured from live-traps each season. The exponential rate of increase per month (r) was calculated for both treated and untreated sites as r = ln (rat abundance for month t / rat abundance for month t – 1). Average values were used for each treatment because not all sites were trapped each month. These data were split into increase phases (breeding seasons) for the spring rice crop (c. April to July) and the summer rice crop (c. September to October) and decrease phases (non-breeding periods) during the winter period (c. October to April) and over summer (c. July to September). Paired t-tests on r were performed for breeding and non-breeding periods separately to compare differences between treatments. The occurrence of adult females in breeding condition (lactating and/or pregnant as determined by palpation) and occurrence of juveniles was compared between treatments using generalized linear models (analysis of deviance) for binomially distributed data (breeding yes or no; juvenile yes or no) using the logit link. Probabilities were determined using the F statistic using degrees of freedom/residual degrees of freedom. The effect of treatment, month and pre- versus post-treatment (time) and all two-way interactions of treatment with the other factors were examined with the addition of the factor sex for juveniles. Separate analyses were conducted for 96

R. argentiventer and R. losea. R. argentiventer <50 g and R. losea <45 g were classified as juveniles (Brown et al. unpublished data). Ln-transformed data for body mass of rats trapped on regular trap lines was compared between treated and untreated sites using analysis of variance. We ran a model including the factors treatment, sex, time (pre- and post-treatment), year and month, all two-way interactions with treatment, and the three-way interaction with treatment, sex and time. This model was used to obtain predictions for means and their standard errors that had a significant effect. The relative change in breeding, juveniles and weight between pre- and post- treatment and for treated and untreated sites was calculated by (using means from each treatment, Saunders 1983): (PostTreat – (PreTreat * (PostUntreat / PreUntreat))) Relative change = * 100 (PreTreat * (PostUntreat / PreUntreat))

II.3 Results

II.3.1 Abundance

There were 996 R. argentiventer captured (52.0% of captures) and 554 R. losea captured 28.9% of captures) from regular monthly live trapping from all sites (a total of 1550 captures or 80.9% of all captures) (Table II.3). The remaining species of rats captured included Rattus rattus, R. norvegicus, Bandicota spp., Mus spp., Suncus murinus and some other captures that could not be identified or escaped prior to identification. There was little difference in the abundance of each species among each site, although there were slightly more R. argentiventer captured on T2 (58.9%) than on other sites (49.1%), but this difference mainly occurred during the pre-treatment period. The fluctuations in abundance of R. argentiventer and R. losea were similar, in that peaks in abundance occurred after harvest of the spring and summer rice crops each year (Figure II.1).

Table II.3. Number of animals of each species captured using live-capture traps on untreated and treated sites from April 1999 to November 2002. There were 44 trapping sessions conducted on U1, 27 on U2, 41 on T1 and 24 on T2. Untreated Sites Treated Sites Species U1 U2 T1 T2 Total % Rattus argentiventer 216 208 230 342 996 52.0% Rattus losea 145 116 157 136 554 28.9% Rattus rattus 27 54 52 70 203 10.6% Rattus norvegicus 0 8 3 7 18 0.9% Bandicota spp. 3 44 10 6 63 3.3% Mus spp. 12 18 8 3 41 2.1% Suncus murinus 0 1 1 0 2 0.1% Unidentified 8 9 4 17 38 2.0% Total 411 458 465 581 1915 100.0%

There were significantly more rats caught by monthly regular trapping during the summer rice crop (7.16 ± 0.14 SE) than during the winter vegetable crop (2.94 ±

0.09 SE) or the spring rice crop (1.47 ± 0.07 SE) (F1,162 = 41.29; P < 0.001). Significantly more R. argentiventer (3.56 ± 0.08 SE) than R. losea were caught (2.30 ±

0.09 SE) (F1,162 = 11.21; P < 0.05). There was no difference between treatments (P =

0.99), however, there was a significant interaction between treatment and season (F2,162 = 3.78; P = 0.03) (Figure II.2). The treatment by species interaction was not significant, accordingly the response shown by R. argentiventer and R. losea was similar (P = 0.69).

TBS TBS TBS TBS TBS TBS 80 (a) R. argentiventer Treated 60 Untreated

40 s rat

20 ce of 0

ndan (b) R. losea

Abu 20

0 Winter crop Winter crop Winter crop

AMJ JASONDJFMAMJ JASONDJFMAMJ JASONDJFMAMJ JASOND 1999 2000 2001 2002 Month Figure II.1. Abundance of rats for (a) Rattus argentiventer and (b) R. losea in untreated and treated sites (mean ± SE) from April 1999 to November 2002, Red River delta, Vietnam. The pre-treatment period was from April 1999 to January 2000 after which the rodent management treatments were imposed on treated sites. Vertical cross- hatched bars represent when TBSs were set up on treated sites. TBSs were established 3-4 weeks prior to the sowing of the surrounding rice crops and were dismantled when the crop inside the TBS was harvested, about 3-4 weeks prior to harvesting of the surrounding rice crops. The shaded bars at the bottom of the graph represent the approximate timing of the spring and summer rice crops and winter vegetable crops.

12 Untreated 10 Treated s 8 rat

4 Abundance of

0 Spring Summer Winter Spring Summer Winter R. argentiventer R. losea Season and Species

Figure II.2. Abundance of rats captured per month during spring and summer (when TBS were set up), and winter (when no TBS were set up) on treated and untreated sites for R. argentiventer and R. losea (back-transformed means ± SE). Data are from the post-treatment period from February 2000 to November 2002, Red River delta, Vietnam, excluding periods during land preparation when no TBS were in place (see text for details).

There was a strong positive relationship between the number of rats captured on treated sites over the period when the TBS were set up (total number of rats captured during the spring and summer rice crop) and the number of rats captured in TBSs (y = - 2 691.3 + 56.1x; r = 0.8603; F1,4 = 24.4; P < 0.01). Rates of increase were less variable and closer to zero on treated than on untreated sites (Figure II.3). During the increase phases (breeding periods), monthly rates of increase were higher on untreated sites (mean = 1.17 ± 0.20 SE) compared to treated sites (mean = 0.80 ± 0.18 SE) (Paired t5 = 2.33; P = 0.067), and during the decreases phases (non-breeding periods), monthly rates of increase were lower on untreated sites (mean = -0.98 ± 0.29 SE) compared to treated sites (mean = -0.65 ± 0.28

SE) (Paired t5 = -4.59; P = 0.006).

100

3 Untreated al

i Treated t 2 ) r e ( ponen 1 ex eas y

r 0 l h nt inc -1 o

e m -2 g rate of -3 era

Av -4 Spring rice Summer rice Winter crop Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Month Figure II.3. Average monthly exponential rates of increase for treated and untreated sites post-treatment only. The shaded bars at the bottom of the graph represent the approximate timing of the spring and summer rice crops and winter vegetable crops. Data are from the post-treatment period from February 2000 to November 2002, Red River delta, Vietnam.

II.3.2 Breeding

There were significant differences in the occurrence of adult female R. argentiventer in breeding condition between treatments (P < 0.05) between pre-and post-treatment (P < 0.05) and between months (P < 0.001) (Table II.4). The treatment by time interaction was not significant (P = 0.218) even though there was a slight decrease in the occurrence of adult females in breeding condition on treated sites (pre = 0.28, post = 0.20) and a large reduction on untreated sites (pre = 0.51, post = 0.32). This gives a relative increase of 17% on treated sites from pre- to post-treatment (Figure II.4). There were significant differences in the occurrence of adult female R. losea in breeding condition between treatments (P < 0.05) and between months (P < 0.001) (Table II.4). There were no differences between pre- and post-treatment (P = 0.130) or between years (P = 0.237). The treatment by time interaction was marginally significant (P = 0.073) with a reduction in the occurrence of adult females in breeding condition on untreated sites (pre = 0.47, post = 0.41) and an increase on treated sites (pre = 0.16, post = 0.38) (Figure II.4). This gives a relative increase of 167% on treated sites from pre- to post-treatment. 101

Table II.4. Summary of generalized linear model table for adult females breeding for R. argentiventer and R. losea showing degrees of freedom, deviance residual, residual degrees of freedom, residual deviance, F value and the P value. Source df Deviance Residual Residual F Value P Value Residual df Deviation R. argentiventer Null 307 380.59 Treatment 1 6.82 306 373.77 6.38 0.012 Time 1 5.15 305 368.62 4.81 0.029 Year 3 1.88 302 366.74 0.59 0.624 Month 11 62.33 291 304.42 5.30 < 0.001 Treatment*Time 1 1.63 290 302.79 1.52 0.218

R. losea Null 250 327.60 Treatment 1 5.16 249 322.44 5.32 0.022 Time 1 2.24 248 320.20 2.31 0.130 Year 2 2.81 246 317.39 1.45 0.237 Month 11 76.68 235 240.71 7.19 < 0.001 Treatment*Time 1 3.15 234 237.56 3.24 0.073

1.0 R. argentiventer 0.8

g 0.6

0.4 edin

bre 0.2 s 0.0 Treated Untreated R. losea male 0.8 e Pre-treatment f t l 0.6

Adu 0.4

0.2

0.0 Spring rice Summer rice Winter crop

JFMAMJJASOND

Month Figure II.4. Predicted proportion of adult females in breeding condition for R. argentiventer and R. losea from pre-treatment samples (April to February 2000) and from post-treatment treated and untreated sites (March 2000 to November 2002), Red River Delta, Vietnam. The shaded bars at the bottom of the graph represent the approximate timing of the spring and summer rice crops and the winter vegetable crops. 102

II.3.3 Proportion of juveniles in the population

There was no difference in the occurrence of juvenile R. argentiventer between treatments (P = 0.324) and no difference between pre-and post-treatments (P = 0.375). There were significant differences among years (P = 0.015), and there was a higher occurrence of juvenile females than juvenile males in the population (P < 0.01) and significant differences between months (P < 0.001) (Table II.5). There were significant interactions between treatment and time with an increase in the occurrence of juveniles on treated sites (pre = 0.09, post = 0.15) and a reduction in the proportion of juveniles on untreated sites (pre = 0.16, post = 0.11) (P < 0.001). This led to a 148% relative increase on treated sites from pre- to post-treatment. There was a significant difference between treatment and year (P < 0.05), between treatment and months (P < 0.01) (Figure II.5), but not for the treatment by sex by time interaction (P = 0.390).

103

Table II.5. Summary of generalized linear model table for proportion of juveniles trapped in the population for R. argentiventer and R. losea showing degrees of freedom, deviance residual, residual degrees of freedom, residual deviance, F value and the P value. Source df Deviance Residual Residual F P Value Residual df Deviation Value R. argentiventer Null 991 735.75 Treatment 1 0.83 990 734.93 0.976 0.324 Sex 1 9.01 989 725.91 10.62 0.002 Time 1 0.67 988 725.25 0.79 0.375 Year 3 8.90 985 716.35 3.50 0.015 Month 11 69.69 974 646.66 7.47 < 0.001 Treatment*Sex 1 0.43 973 646.23 0.50 0.478 Treatment*Time 1 10.12 972 636.11 11.93 < 0.001 Treatment*Year 2 5.87 970 630.24 3.46 0.032 Treatment*Month 11 24.10 959 606.14 2.58 0.003 Treatment*Sex*Time 2 1.60 957 604.53 0.944 0.390

R. losea Null 553 485.03 Treatment 1 0.35 552 484.68 0.31 0.575 Sex 1 3.40 551 481.28 3.05 0.081 Time 1 8.50 550 472.78 7.63 0.006 Year 3 14.49 547 548.29 4.34 0.005 Month 11 32.06 536 426.23 2.62 0.003 Treatment*Sex 1 1.53 535 424.70 1.38 0.241 Treatment*Time 1 4.92 534 419.78 4.42 0.036 Treatment*Year 2 0.70 532 419.08 0.32 0.729 Treatment*Month 11 19.53 521 399.55 1.59 0.097 Treatment*Sex*Time 2 0.08 519 399.47 0.04 0.964

104

1.0 (a) Male R. argentiventer Pre-treament 0.8 Treatment Control 0.6 eniles 0.4 Juv 0.2

0.0 (b) Female R. argentiventer 0.8 s

e 0.6 nil e 0.4 Juv 0.2

0.0 (c) Male R. losea 0.8 s

e 0.6 nil e 0.4 Juv 0.2

0.0 (d) Female R. losea 0.8

s 0.6 e nil

e 0.4

Juv 0.2

0.0 Spring rice Summer rice Winter crop

JFMAMJJASOND

Month Figure II.5. Predicted proportion of juveniles trapped for R. argentiventer and R. losea from pre-treatment samples (April to February 2000) and from post-treatment treated and untreated sites (March 2000 to November 2002), Red River Delta, Vietnam. The shaded bars at the bottom of the graph represent the approximate timing of the spring and summer rice crops and the winter vegetable crops.

105

The proportion of juvenile R. losea in the population was significantly greater pre-treatment than post-treatment (P < 0.01) (Table II.5). There were significant differences between years (P < 0.01) and between months (P < 0.01), but no difference between treatments or sex (P > 0.05). There was a significant interaction for treatment by time with a small reduction in the proportion of juveniles on treated sites (pre = 0.16, post = 0.14) and a large reduction on untreated sites (pre = 0.31, post = 0.11) (F1,534 = 4.42, P < 0.05) (Figure II.5). This led to a 158% relative increase on treated sites from pre- to post-treatment.

II.3.4 Body mass

There were significant differences in the body mass of R. argentiventer between treatments, between sexes, over time (pre- and post-treatment), over different years and between months and all 2-way interactions with treatment, except for treatment by sex (Table II.6). Males were 4.6 g lighter and females were 5.3 g lighter on treated compared to untreated sites post-treatment, despite males being 11.8 g heavier and females being 20.1 g heavier on treated compared to untreated sites pre-treatment (relative reduction of 13% for males and 22% for females) (Figure II.6). There also were significant differences in body mass of R. losea over time (pre- and post-treatment), between months and between years (Table II.6). There were significant 2-way interactions for treatment by time, treatment by month and the 3-way interaction of treatment by sex by time. Males were 5.2 g lighter and females were 8.1 g lighter on treated compared to untreated sites post-treatment, despite males being 32.9 g heavier and females being 13.8 g heavier on treated compared to untreated sites pre- treatment (relative reduction of 41% for males and 22% for females) (Figure II.6).

106

Table II.6. Summary of ANOVA table for ln-weight (g) for R. argentiventer and R. losea showing degrees of freedom, sums of squares, mean squares, F value and the P value. Source df SS MS F P R. argentiventer Treatment 1 2.63 2.63 10.30 0.002 Sex 1 2.79 2.79 10.89 0.002 Time 1 26.69 26.69 104.38 < 0.001 Year 3 9.95 3.32 12.96 < 0.001 Month 11 41.19 3.74 14.64 < 0.001 Treatment*Sex 1 0.04 0.04 0.14 0.706 Treatment*Time 1 3.23 3.23 12.63 < 0.001 Treatment*Month 11 21.16 1.92 7.52 < 0.001 Treatment*Year 2 1.48 0.74 2.88 0.056 Treatment*Sex*Time 2 1.63 0.81 3.19 0.042 Residuals 957 244.71 0.26

R. losea Treatment 1 0.34 0.34 1.24 0.267 Sex 1 0.86 0.86 3.17 0.076 Time 1 4.00 3.99 14.72 < 0.001 Year 3 6.67 2.22 8.19 < 0.001 Month 11 20.00 1.82 6.70 < 0.001 Treatment*Sex 1 0.04 0.04 0.13 0.719 Treatment*Time 1 2.80 2.80 10.30 0.002 Treatment*Month 11 5.88 0.53 1.97 0.029 Treatment*Year 2 0.14 0.07 0.25 0.778 Treatment*Sex*Time 2 3.25 1.63 5.99 0.003 Residuals 519 140.77 0.27

107

300 (a) Male R. argentiventer Pre-treament 250 Treated )

g Untreated 200 ss ( a 150 m 100

Body 50

0 (b) Female R. argentiventer 250 ) g ( 200 ss a 150 m 100

Body 50

0 (c) Male R. losea 250 ) 200 ss (g a 150 m 100

Body 50

0 (d) Female R. losea 250

) 200

ss (g 150 a

m 100

50 Body

0 Spring rice Summer rice Winter crop

JFMAMJJASOND

Month Figure II.6. Predicted mean body mass (g) of male and female R. argentiventer and R. losea from pre- treatment samples (April 1999 to February 2000) and from post- treatment treated and untreated sites (March 2000 to November 2002), Red River Delta, Vietnam. The shaded bars at the bottom, of the graph represent the approximate timing of the spring and summer rice crops and the winter vegetable crops.

108

II.4 Discussion

The implementation of control activities, particularly the use of the TBS, had significant demographic impacts on the two dominant species of rodents in lowland irrigated rice cropping systems in the Red River delta in Vietnam. Both species of rodents showed changes in their demography after the treatments were applied that suggest there were strong compensatory responses at the population level. After the TBSs were set up during the spring and summer rice crops, there was an associated reduction in abundance from monthly live-trapping sessions during summer (45% and 28% for R. argentiventer and R. losea respectively) and an increase in abundance during the winter season (31% and 69% respectively), and a mixed response during spring (39% decrease and 41% increase respectively). There was also an increase in the proportion of juveniles captured post-treatment (148% and 158% respectively) and a lowering of body mass on treated sites post-treatment (males: 13% and 41% respectively; females: 22% and 22% respectively). Older, larger animals were removed by the TBS and the rodent populations were compensating through high recruitment of young, high immigration into treated sites and high survival of residents during the winter period. This is not simply a behavioural shift but new animals were coming into the treated sites. This accords with the prediction outlined in Table II.1, with lower abundance, higher juveniles and lower body mass on treated sites after the introduction of TBS to control rats post-treatment.

II.4.1 Breeding

There was no obvious pattern in terms of compensation through breeding. Although there was an overall treatment effect, a more appropriate effect is shown by the treatment by time (pre- versus post-treatment) interaction for R. argentiventer and R. losea. In both cases this interaction term was not significant suggesting that there was no difference between the occurrence of adult females in breeding condition over time between treatment, even though there was a 17% increase for R. argentiventer and a 167% increase for R. losea. Therefore, any difference observed in abundance cannot be explained by differences observed in breeding. Montgomery et al. (1997) found that reproductive success of female immigrants was enhanced on grids where females were 109

removed, but did not result in an increase in the overall proportion of adult females that were reproductively active. In the current study, we found there were only low levels of breeding during the winter period therefore breeding could not be the main compensatory mechanism at this time. We also found that immigration was higher and the survival of residents was higher during the winter period, confirming the results observed in the changes in population dynamics. Adult female rodents are generally thought to be site attached and are not likely to disperse (Myers & Krebs 1971; Krebs et al. 1995; Chambers et al. 2000). However, Brown et al. (In Press) found that some adult female R. argentiventer moved large distances during non-breeding and breeding seasons, and so could play an important role in dispersal. We did not consider the breeding condition of males because once males reach sexual maturity they remain sexually active throughout their life, particularly for R. argentiventer (Lam 1983), although Jacob et al. (unpublished data) found some differences in the proportion of adult males in breeding condition over different crop stages. In a study comparing the demographics of rats captured in TBS to rats captured from the surrounding area and an area without TBS (Jacob & Wegner In Press), females captured in the TBS had lower body weight, fewer were in breeding condition and litter size was smaller. Furthermore there was no difference in abundance between the treated areas (with TBS) and the untreated areas (where no TBS were set). Jacob and Wegner (In Press) concluded that the TBS removed lower quality dispersing animals without affecting the density of rats in the surrounding irrigated rice fields. This finding is opposite to that found in this study, where it seems that adult animals were removed by the TBS. However, in a similar study in Indonesia, Brown et al. (2003a) could not find any differences in the demographic parameters of rats captured from fields near the TBS or from animals captured in the TBS and therefore could not determine if transients or residents were being captured in the TBS. One weakness of our study was that we were not able to compare the demographics of rats captured in the TBS and those captured on our live-trap lines. Our findings in this study suggest that at low population abundance, residents are being removed by the TBS. We recommend that in the future, detailed information about species, sex, breeding condition, weight and evidence of whether animals were marked during regular population trapping be assessed. We were not able to do this during this study because it was the responsibility of the farmers who operated the TBSs to check the traps in the TBSs each morning. 110

They could confidently count the number of rats captured, but could not be relied on to take other measurements.

II.4.2 Rates of increase

The removal of rats by TBS on treated sites led to a lower rate of increase during the increase phase (breeding seasons) suggesting that breeding was lower on treated sites. Although breeding rates were lower on treated sites, there was no treatment by time (pre- versus post-treatment) interaction, so there was little evidence for lower breeding for either species. During the decrease phase (non-breeding season) rodent populations on treated sites decreased at a slower rate suggesting that the that survival of rats was better on treated sites and that the recruitment of juveniles was higher, as was found in the subsequent analysis of juveniles. There was a strong positive relationship between the number of animals captured in TBSs and the number of rats captured in the fields from regular live- trapping, which indicates that the measure we used to estimate abundance of rats was a reliable index of actual abundance.

II.4.3 Juveniles

The response shown in the proportion of juveniles caught between treatments and over time (pre- versus post-treatment) by R. argentiventer and R. losea was slightly different. For R. argentiventer, there was an increase in the proportion of juveniles on treated sites from pre- to post-treatment and a reduction on untreated sites. This indicates that either the survival of juveniles was very high on treated sites compared to untreated sites, that older animals were removed by TBS, or that juveniles were dispersing into treated sites during the post-treatment period. For R. losea, there was a reduction on both treated and untreated sites from pre- to post-treatment, but the reduction from pre- to post-treatment was 12.5% on treated sites compared to a reduction of 64.5% on untreated sites, so the reduction was not as great on the treated sites. The proportion of juveniles in the population is a reflection of the intensity and duration of breeding, and the survival of young and recruitment into the population. It 111

was likely that small juveniles captured in the live-traps for the population study would be residents, since immigration generally begins with larger juveniles and sub-adults (Baird & Birney 1982; Boutin et al. 1985). The increase in R. argentiventer juveniles and apparently better survival of R. losea juveniles suggests that the survival of juveniles must a significant mechanism for compensation. Unfortunately, we could not assess survival or these rodent populations because we could not recapture any animals in our capture-mark-release study to examine this.

II.4.4 Body mass

There were significant differences in the body mass of R. argentiventer and R. losea between treatments, sexes and over time (3-way interaction). Males and females of both species were relatively lighter on treated sites post-treatment suggesting that older, larger animals were removed through the use of the TBS, and that lighter animals in poorer body condition were immigrating into the treated sites. Such an effect has been found at one of two sites where pocket gophers were removed from orchards (Sullivan et al. 2001).

II.4.5 Concluding comments

The speed by which the population can recover from control is an important issue for rodent pests in agricultural systems and will influence control method used and the frequency of control. If compensation of the population is through immigration then recovery will be rapid, but if it is through better survival and increased breeding then recovery will be gradual. Dispersal is often regarded as the main compensatory mechanism because it happens very rapidly before reproduction can occur (Sullivan 1979; Clout & Efford 1984; Hansson 1992). The removal of animals through the TBS in our study occurred over a relatively long period (TBS were set up for 70-130 days) and we observed changes in abundance, juveniles and body mass (but not breeding), so it is most likely that immigration and survival of young were important compensatory mechanisms. There is good evidence from another study that R. argentiventer are good recolonisers because some radio-collared rats had very large home ranges and appeared to be nomadic in their movements (Brown et al. In Press). This would be a key 112

mechanism by which animals reinvade an area, but successful dispersal is dependent on a range of conditions such as food supply, cover, predation, density, aggressive behaviour etc (Hansson 1992; Lidicker & Stenseth 1992). Management should therefore occur over large areas. Rats on all sites were subjected to a range of mortality factors including poisoning and hunting but also from general farming practices (ploughing, sowing, harvesting, spraying). Fewer farmers were using rodenticides on the treated sites (Singleton et al. 2004), but the effectiveness of this as a mortality factor at the population level is unknown. Of 31 rats that were radio-tracked in two periods, 29% (nine rats; six from untreated sites, three from treated sites) obviously died from rodenticide poisoning, while 6.5% (two rats; one each from a untreated and treated site) died from wounds, which presumably occurred when farmers were hunting rats in the fields (Brown et al. In Press). The overall survival rate over two weeks of radio- collared rats was 56% (9/16 rats) and 60% (9/15 rats) on untreated and treated sites respectively. Since there were no recaptures of rats during the monthly trapping sessions in our study, direct survival rates could not be calculated. Recapture rates of R. argentiventer from live-trapping are known to be very low (Brown et al. 1999; Leung et al. 1999; Jacob et al. 2003). There were general similarities in the response of the two main rodent species to the removal of rats using the TBS. However, there were at least six other species inhabiting the lowland irrigated rice fields of Vinh Phuc. There is a strong possibility that interspecific competition exists between R. argentiventer and R. losea in this environment (Brown et al. 2003b), and there may be subtle effects of this competition on the survival and reproduction, as has been found for bank and field voles (Eccard et al. 2002). Extensive manipulative studies are required to examine these in detail. In conclusion, rat populations on treated sites where TBS are used to remove animals led to removal of a high proportion of resident animals in spring and summer. The demographic responses to this removal were increases in the proportion of juveniles and a reduction in body mass of animals, and increased immigration and survival rates during winter. When the abundance of rats is low, as was the case during post-treatment in this study, it is most likely that the TBS is harvesting the rat population (= “sustained yield”; Davis 1988; Caughley & Sinclair 1994), because the population can easily compensate 113

for the removal of animals captured in the TBS. For many populations at low density killing adults increases the chance of extinction because it is more likely that breeders are removed. Local extinction did not occur during this study primarily because of high immigration. Since there was a linear relationship between the abundance of rats captured in the field using capture-mark-release trapping and the number of rats captured in the TBS, we would expect that when the abundance of rodents are moderate to high, the compensatory mechanisms of the population would not cope with the removal of high numbers of rodents using the TBS. This would subsequently lead to significant reductions in abundance and a consequent reduction in damage to crops, as was observed during high densities of Rattus argentiventer in lowland irrigated rice cropping systems in Indonesia (Singleton et al. 2004). Despite the low abundance, it was still found to be cost-effective in lowland irrigated ricefields in Vietnam and Indonesia (Singleton et al. 2004, 2005) mainly through the reduction in use of rodenticides and plastic fences, but also because the costs of using the TBS were spread across 10-15 hectares.

II.5 Acknowledgements

We sincerely thank all the village heads and farmers involved in the project for their support and willingness to participate in the project. Staff of NIPP provided excellent support while in the field and getting the project underway. In particular we thank Le Than Hoa, Pham Thi Lien and Pham Van Kien. Thanks to G. R. Singleton A. D. Arthur and P. B. Banks for comments on a draft of the manuscript. This research is part of the ACIAR funded project “Management of rodents in rice-based farming systems of Southeast Asia” (AS1 98/36) and was conducted in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. The approval numbers were WEAEC 98/99 – 09 and SEAEC 01/02 – 09.

114

II.6 References

Alcover, J. A., Sans, A., and Palmer, M. (1998). The extent of extinctions of mammals on islands. Journal of Biogeography 25, 913-918. Álvarez-Castańeda, S. T. and Ortega-Rubio, A. (2003). Current status of rodents on islands in the Gulf of California. Biological Conservation 109, 157-163. Andreassen, H. P. and Ims, R. A. (2001). Dispersal in patchy vole populations: role of patch configuration, density dependence, and demography. Ecology 82, 2991- 2926. Baird, D. D. and Birney, E. C. (1982). Characteristics of dispersing meadow voles Microtus pennsylvanicus. American Midland Naturalist 107, 262-283. Boutin, S., Gilbert, B. S., Krebs, C. J., Sinclair, A. R. E., and Smith, J. N. M. (1985). The role of dispersal on the population dynamics of snowshoe hares. Canadian Journal of Zoology 63, 106-115. Boyce, M. S., Sinclair, A. R. E., and White, G. C. (1999). Seasonal compensation of predation and harvesting. Oikos 87, 419-426. Brown, P. R., Hung, N. Q., Hung, N. M., and van Wensveen, M. (1999). Population ecology and management of rodent pests in the Mekong River Delta, Vietnam. In Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 319-337. (Australian Centre for International Agricultural Research: Canberra.) Brown, P. R., Leung, L. K-P., Sudarmaji, and Singleton, G. R. (2003a). Movements of the ricefield rat, Rattus argentiventer, near a trap-barrier system in rice crops in West Java, Indonesia. International Journal of Pest Management 49, 123-129. Brown, P. R., Tuan, N. P., and Banks, P. B. (In Press). Movements, habitat use and response of ricefield rats to removal in an intensive cropping system in Vietnam. Belgian Journal of Zoology. Brown, P. R., Tuan, N. P., Singleton, G. R., Tuat, N. V., Tan, T. Q., and Hoa, L. T. (2003b). Impact of village-level rodent control practices on rodent populations and rice crops in Vietnam. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 197-202. (ACIAR: Canberra.) Buckle, A. P. (1994). Control in temperate field crops and forestry. In 'Rodent Pests and 115

Their Control'. (Eds. A. P. Buckle and R. H. Smith.) pp. 291-312. (CAB International: Wallingford.) Byrom, A. E. (2002). Dispersal and survival of juvenile feral ferrets Mustela furo in New Zealand. Journal of Applied Ecology 39, 67-78. Caughley, G. (1977). Analysis of Vertebrate Populations. (John Wiley and Sons: London.) Caughley, G. and Sinclair, A. R. E. (1994). Wildlife Ecology and Management. 334 pp. (Blackwell: Boston.) Chambers, L. K., Singleton, G. R., and Krebs, C. J. (2000). Movements and social organization of wild house mice (Mus domesticus) in the wheatlands of Northwestern Victoria, Australia. Journal of Mammalogy 81, 59-69. Choquenot, D. (1991). Density-dependent growth, body condition, and demography in feral donkeys: testing the food hypothesis. Ecology 72, 805-813. Clout, M. N. and Efford, M. G. (1984). Sex differences in the dispersal and settlement of brushtail possums (Trichosurus vulpecula). Journal of Animal Ecology 53, 737-749. Davis, D. E. (1988). Population principles. In Rodent Pest Management. (Ed. I. Prakash.) pp. 171-178. (CRC Press: Boca Raton, Florida.) Davis, D. E. and Christian J. J. (1958). Population consequences of a sustained yield program for Norway rats. Ecology 39, 217-222. Dowding, J. E. and Murphy, E. C. (2001). The impact of predation by introduced mammals on endemic shorebirds in New Zealand: a conservation perspective. Biological Conservation 99, 47-64. Drummond, D. C. (1970). Variation in rodent populations in response to control measures. Symposium of the Zoological Society of London 26, 351-367. Eccard, J. A., Klemme, I., Horne, T. J., and Ylönen, H. (2002). Effects of competition and season on survival and maturation of young bank vole females. Evolutionary Ecology 16, 85-99. Efford, M., Warburton, B., and Spencer, N. (2000). Home-range changes by brushtail possums in response to control. Wildlife Research 27, 117-127. Geddes, A. W. M. (1992). The Relative Importance of Pre-harvest Crop Pests in Indonesia. (Natural Resources Institute: Chatham, UK.) 116

Gliwicz, J. (1981). Competitive interactions within a forest rodent community in central Poland. Oikos 37, 353-362. Gundersen, G., Andreassen, H. P., and Ims, R. A. (2002). Individual and population level determinants of immigration and success on local habitat patches: an experimental approach. Ecology Letters 5, 294-301. Gundersen, G., Johannesen, E., Andreassen, H. P., and Ims, R. A. (2001). Source-sink dynamics: how sinks affect demography of sources. Ecology Letters 4, 14-21. Hilborn, R., Walters, C. J., and Ludwig, D. (1995). Sustainable exploitation of renewable resources. Annual Review of Ecology and Systematics 26, 45-67. Hansson, L. (1992). Small mammal dispersal in pest management and conservation. In Animal Dispersal: Small Mammals as a Model. (Eds. N. C. Stenseth and W. Z. Lidicker, Jr.) pp. 181-198. (Chapman and Hall: Cornwell, UK.) Jacob, J., Sudarmaji, and Singleton. G. R. (2003). Ecologically-based management of rice-field rats on a village scale in West Java: experimental approach and assessment of habitat use. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 191-196. (ACIAR: Canberra.) Jacob, J. and Wegner, R. E. (In Press). Does continuous removal of individuals separate high and low quality ricefield rats? Journal of Wildlife Management. Krebs, C. J., Kenney, A. J., and Singleton, G. R. (1995). Movements of feral house mice in agricultural landscapes. Australian Journal of Zoology 43, 293-302. Krebs, C. J., Wingate, I., LeDuc, J., Redfield, J. A., Taitt, M., and Hilborn, R. (1976). Microtus population biology: dispersal in fluctuating populations of M. townsendii. Canadian Journal of Zoology 54, 79-95. Lam, Y. M. (1983). Reproduction in the ricefield rat Rattus argentiventer. Malayan Nature Journal 36, 249-282. Lam, Y. M. (1988). Rice as a trap crop for the rice field rat in Malaysia. Proceedings of the 13th Vertebrate Pest Conference. (Eds. A. C. Crabb, and R. E. Marsh). pp. 123-128. Lan, L. P., Aplin, K. P., Hung, N. M., Quoc, N. V., Chien, H. V., Sang, N. D., and Singleton, G. R. (2003). Rodent communities and historical trends in rodent damage in the Mekong Delta of Vietnam: establishing an ecological basis for effective pest management. In Rats, Mice and People: Rodent Biology and 117

Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 290-296. (ACIAR: Canberra.) Leung, L. K. P., Singleton, G. R., Sudarmaji, and Rahmini (1999). Ecologically-based population management of the rice-field rat in Indonesia. In Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 305-318. (Australian Centre for International Agricultural Research: Canberra.) Lidicker, W. Z., Jr and Stenseth, N. C. (1992). To disperse on not to disperse: who does it and why? In Animal Dispersal: Small Mammals as a Model. (Eds. N. C. Stenseth and W. Z. Lidicker, Jr.) pp. 21-36. (Chapman and Hall: Cornwell, UK.) Montgomery, W. I. (1981). A removal experiment with sympatric populations of Apodemus sylvaticus (L.) and A. flavicollis (Melchior) (Rodentia: Muridae). Oecologia 51, 123-132. Montgomery, W. I., Wilson, W. L., and Elwood, R. W. (1997). Spatial regulation and population growth in the wood mouse Apodemus sylvaticus: experimental manipulations of males and females in natural populations. Journal of Animal Ecology 66, 755-768. Mwanjabe, P. S. and Leirs, H. (1997). An early warning system for IPM-based rodent control in smallholder farming systems in Tanzania. Belgian Journal of Zoology 127, 49-58. Myers, J. H. and Krebs, C. J. (1971). Genetic, behavioral, and reproductive attributed of dispersing field voles Microtus pennsylvanicus and Microtus ochrogaster. Ecological Monographs 41, 53-78. Pimm, S. L. (1987). The snake that ate Guam. Trends in Ecology and Evolution 2, 293- 294. Saunders, G. R. (1983). Evaluation of mouse plague control techniques in irrigated sunflower crops. Crop Protection 2, 437-445. Schieck, J. O. and Millar, J. S. (1987). Can removal areas be used to assess dispersal of red-backed voles? Canadian Journal of Zoology 65, 2575-2578. Sinclair, A. R. E. (1996). Mammal populations: fluctuation, regulation, life history theory and their implications for conservation. In Frontiers of Population Ecology. (Eds. R. B. Floyd, A. W. Sheppard, and P. J. De Barro.) pp. 127-154. (CSIRO Publishing: Melbourne.) 118

Singleton, G. R. (1997). Integrated management of rodents: A Southeast Asian and Australian perspective. Belgian Journal of Zoology 127, 157-169. Singleton, G. R. (2003). Impacts of Rodents on Rice Production in Asia. IRRI Discussion Paper Series No. 43. International Rice Research Institute, Los Bańos, Philippines. pp. 30. Singleton, G. R. and Brown, P. R. (1999). Management of mouse plagues in Australia: integration of population ecology, bio-control and best farm practice. In Advances in Vertebrate Pest Management. (Eds. D. P. Cowan and C. J. Feare.) pp. 189-203. (Filander-Verlag: Berlin.) Singleton, G. R., Brown, P. R., and Jacob, J. (2004). Ecologically-based rodent management; its effectiveness in cropping systems in South-East Asia. Netherlands Journal of Agricultural Sciences 52, 163-171. Singleton, G. R. and Petch, D. A. (1994). A Review of the Biology and Management of Rodent Pests in Southeast Asia. ACIAR Technical Reports No. 30. ACIAR, Canberra. pp. 65. Singleton, G. R., Sudarmaji, and Brown, P. R. (2003). Comparison of different sizes of physical barriers for controlling the impact of the rice field rat, Rattus argentiventer, in rice crops in Indonesia. Crop Protection 22, 7-13. Singleton, G. R., Sudarmaji, Jacob, J., and Krebs, C. J. (2005). Integrated management to reduce rodent damage to lowland rice crops in Indonesia. Agriculture, Ecosystems and Environment 107, 75-82. Singleton, G. R., Sudarmaji, Jumanta, Tan, T. Q., and Hung, N. Q. (1999). Physical control of rats in developing countries. In Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 178-198. (Australian Centre for International Agricultural Research: Canberra.) Singleton, G. R., Sudarmaji, and Suriapermana, S. (1998). An experimental field study to evaluate a trap-barrier system and fumigation for controlling the rice field rat, Rattus argentiventer, in rice crops in West Java. Crop Protection 17, 55-64. Sullivan, T. P. (1979). Repopulation of clear-cut habitat and conifer seed predation by deer mice. Journal of Wildlife Management 43, 861-871. Sullivan, T. P. and Sullivan, D. S. (1986). Resiliency of snowshoe hares to population reduction. Journal of Applied Ecology 23, 795-806. 119

Sullivan, T. P., Sullivan, D. S., and Hogue, E. J. (2001). Reinvasion dynamics of northern pocket gopher (Thomomys talpoides) populations in removal areas. Crop Protection 20, 189-198. Tuan, N. P., Williams, S. J., Brown, P. R., Singleton, G. R., Tan, T. Q., Hue, D. T., Ha, P. T. T., and Hoa, P. T. (2003). Farmers' perceptions and practices in rat management in Vinh Phuc Province, northern Vietnam. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 399-402. (ACIAR: Canberra.) White, G. C. and Bartmann, R. M. (1998). Effect of density reduction on overwinter survival of free-ranging mule deer fawns. Journal of Wildlife Management 62, 214-225.

120

III

Study III. Short- and long-term demographic changes in house mouse populations after control in dryland farming systems in Australia

Manuscript

Peter R. Brown

CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia

121

Short- and long-term demographic changes in house mouse populations after control in dryland farming systems in Australia

Peter R. Brown

CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia

Abstract: In Australia, plagues of house mice (Mus domesticus) cause significant damage to agricultural crops and to rural communities. Rodenticides are often applied to high density mouse populations to limit damage to crops, but the demographic consequences of applying rodenticides on rodent populations are generally not well understood. Furthermore, it is not understood whether the reduction induced by rodenticides would be similar to that of a natural decline or crash in abundance observed at the end of mouse plagues. In this paper, I compared the response of acute rodenticides on rodent populations in three grain growing regions of Australia. Mouse populations were trapped on treated and untreated sites prior to baiting, immediately post-baiting, and at different lengths of time post-baiting. During the post-baiting period, a natural crash in densities occurred on all sites over a 2-4 week period. Data were collected to compare the demographic responses to baiting and the crash. There was a range of responses to the rodenticide baiting, with 0-39% reduction in population densities for strychnine baiting at Walpeup, Victoria, 92% reduction for strychnine baiting at Brookstead, Queensland, and 98% reduction for zinc phosphide baiting at Wudinna, South Australia. After baiting there were mixed responses for body weight (no change, increases and decreases), proportion of juveniles in the population (increases and decreases) sex ratio (no change or bias towards females), survival (no change or decreases) and relative body condition (no change or increases). These responses are likely to be due to differences in the efficacy of the rodenticide. The 122

natural crash induced a >85% decline in population densities across all sites and led to increases and decreases in body weights, reduction in the proportion of juveniles in the population, bias towards males, poor survival (low recapture rate) and poor relative body condition. Poor survival was the only demographic parameter that was consistent for baiting and the crash. Five of seven demographic responses for mice during a crash were similar to that found in the literature for the decline phase of northern cyclic vole and lemming populations, the other two having a mixed response.

III.1 Introduction

When rodents reach high densities they can cause serious damage to a range of resources including crops (Buckle & Smith 1994; Singleton et al. 1999; Stenseth et al. 2003). One strategy for attempting to reduce this damage is to control rodent numbers through the use of rodenticides. Some rodenticides have the capacity to significantly reduce the abundance of rodents by up to 95%. Of interest is what happens to the survivors of such a drastic reduction in abundance and what are the mechanisms by which these populations respond or recover. Some of these responses occur rapidly, such as immigration of individuals to an area of low density (Sullivan 1979; Montgomery et al. 1997; Mwanjabe & Leirs 1997), while other mechanisms may take longer, such as density dependent breeding and survival of young or adults (Montgomery 1981; Gundersen et al. 2001). This drastic reduction in density may be similar in many respects to a rapid decline, or crash in densities as is found in Northern Hemisphere vole and lemming species that undergo dramatic 2-4 year cycles (20 to 2000-fold changes in abundance; Krebs & Myers 1974; Hanson & Henttonen 1988; Lidicker 1988; Hanski et al. 1991; Boonstra et al. 1998; Stenseth 1999; Klemola et al. 2002; Korpimäki et al. 2004) or at the end of mouse plagues (Singleton 1989; Singleton et al. 2005). This paper will examine differences in the demography of house mouse populations in cropping systems following a rodenticide-induced crash and that of a natural crash in density as occurs at the end of a mouse plague. Outbreaks of feral house mice are a serious problem for grain growers in southern and eastern Australia ever since the first plague in 1904 (Saunders & Giles 1977; Singleton et al. 2005). They occur somewhere in southeastern Australia on 123

average one year in four, but one year in 7 in any particular region (Singleton et al. 2005). These outbreaks are characterised by concurrent widespread increases in mouse densities over an area of 50 to 1500 km2 (Singleton et al. 2005). These outbreaks cause severe damage to growing crops and intensive animal husbandry facilities (eg piggeries), but also to stored grain, shops, schools, hospitals, and to the rural community (Redhead 1988; Caughley et al. 1994). It was conservatively estimated that the mouse plague that affected South Australia and Victoria in 1993 caused AU$65 million in damage (Caughley et al. 1994). In southern Australia, mouse population abundance normally peaks in autumn each year when farmers plant winter cereal crops. When mouse numbers are low (low phase), the typical densities of mice in fields are 5 to 50 mice/ha. Under good conditions, mouse numbers can increase to more than 1000 mice/ha (a 100 to >2000 fold increase, Singleton et al. 2005). This increase can occur over one or two breeding seasons (approximately 9-18 months), but the crash in numbers can occur over 2-6 weeks. Moreover, this crash occurs over a wide area, so the mechanism for inducing the crash is widespread. There has been little understanding of the factors that govern such a dramatic decline in abundance of mice, but it is generally thought that a number of factors act in combination and include disease, social stress, lack of food, shifting age structures and predation (Singleton & Hay 1983; Sinclair et al. 1990; Singleton et al. 2001; Ylönen et al. 2002; Sutherland et al. 2004). Disease is probably the major cause of population decline in combination with a range of density-dependent factors (McCallum & Singleton 1989; Singleton et al. 1993, 2005). Overcrowding and external parasite infestations that develop during a plague contribute to the spread of disease, through high rates of contact between mice (Singleton 1985; Arthur et al. In Press). Also, food becomes scarce during autumn prior to planting of winter crops, and as the temperature falls during autumn and winter, mice become physically stressed making them more susceptible to disease. Furthermore, Sutherland et al. (2004) argued that a high survival of young mice might be more important for the formation of plagues than recruitment. It is therefore important to understand the demography of mouse populations, especially because of the extended low phase exhibited by mice. Although the mean interval between mouse plagues is about 7 years, they are irregular (the range is 2-26 years between plagues; Singleton et al. 2005). Rainfall can explain 41-70% of the variation in plague occurrence (Mutze et al. 1990; Brown & Singleton 1999; Kenney et al. 2003). However, understanding the 124

demographic changes and resulting Allee effects during a crash may provide a key that unlocks the reason for the high variation in interval between plagues and could therefore be of value in predicting outbreaks. The application of rodenticides has been widely used in Australia to reduce mouse numbers and their impact on crops (Saunders 1983; Mutze 1989 1993, 1998; Twigg & Kay 1992; Twigg et al. 1991; Brown et al. 1997, 2002; Mutze & Hubbard 2000; Mutze & Sinclair 2004). Strychnine was available until 1996 when it was banned because of concerns about contamination of grain for export and from evidence that it was taken up by plants under certain soil conditions (Kookana et al. 1997; Oliver et al. 2000). There were also some non-target issues (Brown & Lundie-Jenkins 1999). Research was then conducted to test the efficacy of zinc phosphide, which was subsequently registered for use in crops. Zinc phosphide is readily available to farmers and is easy to apply by aerial or ground application. The efficacy of zinc phosphide ranges from 0 to 100%, (Caughley et al. 1998; Brown et al. 2002), but can be 95% effective (Mutze & Sinclair 2004), and there are apparently few non-target issues (Johnson & Fagerstone 1994; Brown et al. 2002; Mutze & Sinclair 2004). All classes of animals in the mouse population are considered susceptible to these rodenticides, except for dependent young in the nests. Furthermore, it is likely that juvenile or subadult males would be the first animals to re-invade baited areas because of aggressive interactions with dominant males in neighbouring areas (Krebs et al. 1995; Chambers et al. 2000; Pocock et al. 2005). In this paper, the response of populations of mice to the application of rodenticides is examined from three different grain-growing regions of Australia. In each of these regions, a natural severe crash in abundance of mice occurred across all sites within a few weeks of the baiting program as often occurs. The aim is to determine and describe the short-term demographic changes of mouse populations after baiting and the long-term demographic changes of mouse populations after the natural crash occurred. I tested the specific predictions that in the short-term (<2 weeks): (1) densities would be lower on baited sites, (2) there would be a bias towards young male mice (as these are most likely to immigrate from surrounding unbaited areas), and (3) the recapture rate would be lower in comparison to unbaited sites. In the long-term (1-4 months and after the crash): (4) densities would be similar on rodenticide-induced crash (baited) and natural crash (unbaited) sites, and (5) there would be few demographic 125

differences observed between rodenticide-induced crash (baited) or natural crash (unbaited) sites. Two of the three studies were conducted in late autumn and winter in southern Australia when the mice were not breeding. Therefore this study did not consider changes in breeding dynamics.

III.2 Methods

Baiting programs were conducted at three locations (Table III.1, Figure III.1). (1) Near Walpeup (35º06’S, 142º00’E), in the cereal-growing region of the Central Mallee in northwestern Victoria in 1994, during a strychnine baiting experiment (see Brown et al. 1997). The population naturally crashed shortly after baiting occurred. (2) Near Brookstead (27º46’S, 151º27’E), on the Darling Downs, Queensland, 1995, during a strychnine-baiting program conducted during a mouse plague. (3) Near Wudinna (33º03’S, 135º28’E), in the Eyre Peninsula in South Australia, 2002, during a zinc phosphide-baiting program conducted during a local outbreak of mice.

NT Qld WA SA Brookstead NSW Wudinna Walpeup Vic

Tas

Figure III.1. Location of the three study sites used. Brookstead on the Darling Downs (Queensland), Walpeup on the Central Mallee (Victoria), and Wudinna on the Eyre Peninsula (South Australia). 126

Table III.1. Summary of design of studies showing dates when studies were conducted, number of trapping sessions, date of bait application, number of sites used and the main cropping systems involved. Number of trap Date of bait Number of Main cropping system and percentage of Site name Dates sessions application sites area baited

Walpeup, Vic May-Aug 1994 4: 1 x pre-bait, 3 x From 27 May to 3 4 baited Strychnine applied to pasture paddocks post-bait (monthly) June 1994 4 unbaited (40% of area within each treated paddock) and fencelines around fallow paddocks (100%) Brookstead, Qld Aug-Dec 1995 5: 1 x pre-bait, 4 x 25 August 1995 2 baited, Strychnine applied to sorghum stubble and post-bait (monthly) 2 unbaited wheat (early tillering stage) (100% of area within each treated paddock) Wudinna, SA Jun-Jul 2002 3: 1 x pre-bait, 2 x 22-23 June 2002 3 baited, Zinc phosphide applied to newly sown post-bait (3-days 2 unbaited wheat (100% of area within each treated after baiting & 1- paddock) month after baiting) `

126 127

Eight study sites near Walpeup (4 baited, 4 unbaited, Table III.1) were located on sandy loams (Land Conservation Council 1974). The topography is flat to mildly undulating with 5-m-high sand dunes. The region has a Mediterranean climate, with hot summers and predominantly winter rainfall. Mean annual rainfall in the district is approximately 340 mm (64 year average). Rainfall was below average for 1994 (48.5% of the long-term mean, LTM). The crops grown are mainly winter cereals (wheat, barley, oats and rye) and some peas. Growers in the Mallee generally implement a three-year crop rotation, which consist of a winter cereal or legume crop, grazing and bare fallow stages. Sites were approximately 2 km apart. Four study sites near Brookstead (2 baited, 2 unbaited, Table III.1) were located on part of an extensive open plain of clay alluvia derived mainly from basalt. The soil is composed mainly of deep, self-mulching black cracking clays (Plumb 1963). Two grain crops are grown each year: summer crops primarily consisting of sorghum but also maize, sunflower and cotton planted from September to February; and winter crops primarily wheat and barley planted in June and July. The climate is temperate and sub- humid, with warm-hot summers and cool winters. The mean annual rainfall is 700 mm (117 year average) and is highly variable with most falling in summer (October to March). Rainfall in 1995 was 158% of the LTM. Sites were approximately 5 km apart. Five study sites near Wudinna (3 baited, 2 unbaited, Table III.1) were similar in soil type and topography to that of Walpeup. The mean annual rainfall is 320 mm (73 year average) with most rain falling May through to October. Rainfall was below average during 2002 (235 mm; 74% of LTM). The crops grown are mainly winter cereals (wheat, barley and oats). Growers generally implement a two-year crop rotation using sheep to graze stubble and pastures. Sites were approximately 1 km apart.

III.2.1 Application of rodenticides

In all three studies, bait was intended to be applied at 3 grains/m2. At Walpeup, growers used their own tractors and fertiliser spreaders to distribute strychnine bait in bands of up to a width of 20 m. Bait was applied to pasture paddocks (stubble from previous crops used as pasture for sheep) in swathes approximately 25 m apart, covering approximately 40% of the paddock. Bait also was applied to all fencelines of fallow paddocks (for details see Brown et al. 1997). At Brookstead, planes were used to 128

spread strychnine bait to treated paddocks (containing sorghum stubble and wheat crops) and covered approximately 100% of these paddocks (for details see Fisher 1996, Cooper & Woods 1996). At Wudinna, zinc phosphide was applied to three fields of newly planted wheat by a fertiliser spreader pulled along by a conventional four-wheel- drive vehicle at 15 km/hour, and was calibrated to deliver the bait at a rate of 1 kg/ha (3 grains/m2). The bait was applied to 100% of treated fields. The bait was commercially available “Mouseoff ZP” (Animal Control Technologies, Melbourne) and was made using clean wheat as the base. The zinc phosphide slurry was then mixed so that it adhered to the surface of the grain. The resulting bait was a dark grey colour. The farmer determined which fields were baited. Prior to baiting, each baited and unbaited field appeared to have similar mouse densities and similar damage to newly planted seeds as determined by the presence of small holes where mice had dug out seeds. The farmer applied similar farming practices across each field.

III.2.2 Mouse trapping and demographic assessment

In each study, Longworth live-capture traps (Abingdon, UK) were used to trap mice. Traps were baited with wheat and spaced at 10-m intervals, and were set for two consecutive nights. Grids (either 6 x 6 or 7 x 7) were set 50 m from the edge of the crop and lines of traps (10 or 20) were located along crop margins (fencelines). At Walpeup, trapping was conducted in pasture paddocks, fallow fields that were subsequently planted to wheat and adjacent fencelines (total of 276 trap nights per site per trap session). At Brookstead, trapping was located in sorghum stubble and newly emerged winter wheat fields (total of 184 trap nights per site per trap session). Trapping at Wudinna was conducted only in newly sown wheat in grids and associated fencelines (total of 112 trap nights per site per trap session). After capture, each animal was either fitted with an ear tag (Hauptner, Germany; at Walpeup and Brookstead) or ear punched (at Wudinna). Weight (± 0.1 g) and head- body length (± 1.0 mm) were measured and the sex recorded. For females, as assessment of reproductive status was based on teat condition and whether the animal was pregnant as detected by palpation. The density of mice was estimated using the Petersen Estimate (Caughley 1977). The density of mice/ha was determined by dividing the Petersen estimate by the 129

effective trapping area of grids and fencelines assuming that only mice occurring within 5 m of the grid or fenceline were likely to enter a trap. The effective reduction in density of mice was calculated using the following formula, which takes into account the relative changes in both the baited and unbaited sites (Saunders 1983):

Upost Bpost – (Bpre x ( U )) 100 Effective reduction = pre x Upost 1 Bpre x ( Upre ) Where Bpre is the mean abundance of mice on baited sites pre-treatment, Bpost is the mean abundance of mice on baited sites post-treatment, Upre is the mean abundance of mice on unbaited sites pre-treatment and Upost is the mean abundance of mice on unbaited sites post-treatment.

III.2.3 Statistical analyses

Ln-density estimates were analysed using an analysis of variance (ANOVA) using liner mixed effects models on each data set (Walpeup, Brookstead and Wudinna) with treatment (baited or unbaited) and crop type (for Walpeup and Brookstead data sets) as fixed factors, and time (trapping session) as a random factor, with site as a group variable. Ln-body weight of mice excluding recaptures within a trapping session were analysed using ANOVA using linear mixed effects models on each data set with treatment (baited or unbaited) and sex (male or female) as fixed factors, and time (trapping session) as a random factor, with site as a group variable. Juvenile mice were defined as animals less than 72 mm (Singleton 1983). The proportion of juveniles trapped in the population excluding recaptures within a trapping session were transformed using the arcsine square root transformation and analysed using an ANOVA using linear mixed effects models on each data set with treatment (baited or unbaited) and sex (male or female) as fixed factors, and time (trapping session) as a random factor, with site as a group variable. Sex ratios were calculated as males/all mice with values >50% represents a male bias in the population. Sex ratios were analysed using an ANOVA using linear mixed effects models on each data set with treatment (baited or unbaited) as a fixed factor, and time (trapping session) as a random factor, with site as a group variable. 130

The recapture rate of marked mice between trapping sessions was calculated as a surrogate for survival. Because the intervals between trap sessions were different, recapture rates are presented as the raw recapture rates and were not adjusted for time interval between trapping sessions, and so it was not possible to analyse the effect of time. Recapture rates were analysed between treatments immediately after baiting and one-month after baiting only using Students’ t-tests. Because of low recapture rates, data for males and females were pooled across all trapping sites for each trapping session. The relative condition of mice was calculated to characterise the survivors or recolonisers following baiting or the natural crash in numbers. Krebs and Singleton (1993) developed a method to calculate the relative condition based on regressions of body length and body weight of mice rather than using a standard scaling factor, which was shown to be erroneous. The relative condition of individual mice was calculated using the regressions of Krebs and Singleton (1993) for Walpeup and Wudinna data (using the Victorian mallee regression) and for Brookstead data (using the Darling Downs regression) to determine the individual condition of mice compared against a standardised average mouse with values >100% indicating relatively good condition and values <100% indicating relatively poor condition. Relative condition was analysed using an ANOVA using linear mixed effects models on each data set with treatment (baited or unbaited) and sex (male or female) as fixed factors, and time (trapping session) as a random factor, with site as a group variable. All statistical analyses were conducted using S-Plus 6.1 for Windows (Insightful Corp., Seattle, USA). Uncertainty is stated as standard error of the mean throughout the paper.

III.3 Results

III.3.1 Density

Walpeup – The average density of mice prior to baiting was 311.6 mice/ha (± 37.8 SE; range 235-451) on fallow paddocks and 487.5 mice/ha (± 175.4 SE; range 172- 1299) on pasture paddocks. The reduction of mice from pre-treatment to immediately post-treatment was 39.1% on pasture paddocks and 0.3% in fallow/crop paddocks. The 131

treatment by time interaction was not significant (P = 0.955; Table III.2), but time and the crop by time interaction were significant (P < 0.05) suggesting that while there was variation over time it was related to changes in density of mice in different paddock types (pasture or fallow subsequently sown to wheat crop) rather than an effect of treatment (baiting). All other factors were not significant. In August (3 months post- baiting), the density of mice on treated sites was 6.8 mice/ha (± 0.7 SE; reduction of 97.6%) and 50.8 (± 41.8 SE; reduction of 92.0%) in fallow/crop and pasture paddocks, respectively, and on untreated sites was 20.9 mice/ha (± 18.9 SE; reduction of 93.0%) and 73.8 (± 42.2 SE; reduction of 85.0%) in fallow/crop and pasture paddocks, respectively (Figure III.2a). Brookstead – The average density of mice prior to baiting was 576.0 mice/ha (± 387.8 SE; range 188-964) in sorghum stubble paddocks and 137.0 mice/ha (± 89.8 SE; range 47-227) in wheat paddocks. The reduction of mice from pre-treatment to immediately post-treatment was 91.5% on sorghum stubble paddocks and 90.7% on wheat paddocks. The treatment by time interaction was marginally significant (P = 0.075; Table III.2), and the crop, time and crop by time interactions were significant (P < 0.05) suggesting that variation over time and between different crop types was greater than the effects of the treatment (baiting). All other factors were not significant. In November (3 months post-baiting), the density of mice on treated sites was 2.8 mice/ha (± 2.8 SE; reduction of 99.8%) and 26.4 mice/ha (± 26.4 SE; reduction of 90.3%) in sorghum stubble and wheat respectively and on untreated sites was 2.8 mice/ha (± 2.8 SE; reduction of 99.4%) and 18.1 mice/ha (± 6.9 SE; reduction of 95.9%) on sorghum stubble and wheat respectively. In December (4-months post-baiting), densities had increased on sorghum crops but decreased further on wheat crops with densities <40 mice/ha on all sites (Figure III.2b). Wudinna – The average density of mice prior to baiting was 635.5 mice/ha (± 120.1 SE; range 396-1075) in newly planted wheat crops. The reduction of mice from pre-treatment to immediately post-treatment was 97.8% on wheat paddocks. The treatment by time interaction was significant (P = 0.012; Table III.2) as were the factors for treatment and time (P < 0.01), suggesting that the large reduction in densities post- baiting were related to the effect of the treatment. In July (1 month post-baiting), the density of mice was 3.7 mice/ha (± 0.9 SE; relative reduction of 99.4%) and 16.0 132

mice/ha (± 16.0 SE; relative reduction of 97.3%) on baited and unbaited sites, respectively (Figure III.2c).

133

Table III.2. Summary of analysis of variance using linear mixed effects models for ln-density of mice as estimated using the Petersen Method showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets. Walpeup Brookstead Wudinna Variable DF F P DF F P DF F P

(Intercept) 1,36 1475.45 <0.001 1,14 179.11 <0.001 1,6 7901.93 <0.001 Treatment 1,6 0.08 0.788 1,2 3.44 0.205 1,3 94.63 0.002 Crop 1,36 2.15 0.151 1,14 6.10 0.027 - - - Time 3,36 18.92 <0.001 4,14 12.57 0.002 2,6 38.54 0.004 Treatment x Crop 1,36 0.01 0.955 1,14 0.01 0.913 - - - Treatment x Time 3,36 0.19 0.900 4,14 2.68 0.075 2,6 10.24 0.012 Crop x Time 3,36 3.23 0.034 4,14 3.28 0.043 - - - Treatment x Crop x Time 3,36 0.33 0.808 4,14 0.86 0.512 - - - 133 134

(a) Walpeup (b) Brookstead (c) Wudinna

Unbaited Sorghum Baited Baited 7 7 Unbaited Wheat 7 Baited Unbaited Baited Sorghum Baited 6 6 Baited Wheat 6 ) ) )

5 a 5 5 h ha ha / e e/ e/ c

4 c 4

c 4 i i i m m m (

( ( 3 3 3 y y t t ity i i 2 2 2 ns ens ens

1 d 1 1 Ln Ln d Ln de Unbaited Fallow/Crop 0 0 0 Unbaited Pasture -1 Baited Fallow/Crop -1 -1 Baited Pasture -2 -2 -2 May Jun Jul Aug Aug Sep Oct Nov Dec June July

Month Month Month

Figure III.2. Change in mean ln-density of mice per hectare (± SE) as estimated using the Petersen Method from pre-treatment to post- treatment from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995), and Wudinna (South Australia, 2002).

134 135

III.3.2 Body mass

Walpeup – There was no overall effect of treatment on body weights of mice and no treatment by time interaction (P > 0.05; Table III.3; Figure III.3a). The factors for time and sex were significant (P < 0.01) suggesting that variation in body mass of mice was related to changes over time (trap sessions 1-4: 13.3 g, 13.3 g, 12.9 g, 12.6 g) and between sexes (males: 13.2 g ± 0.02 SE; females: 12.9 g ± 0.03 SE) rather than the effects of treatment (baiting). No other interactions were significant (P > 0.05). Brookstead – There was a significant difference between treatments (baited = 13.8 g ± 0.1 SE; unbaited = 12.8 g ± 0.1 SE; P = 0.011), but no treatment by time interaction (P = 0.154; Table III.3; Figure III.3b), suggesting that differences in the body mass of mice between treatments was not related to the effect of baiting over time. There was a significant difference between sexes (males: 13.3 g ± 0.1 SE; females: 13.0 g ± 0.1 SE; P < 0.001). No other interactions were significant (P > 0.05). Wudinna – There was a significant difference in the body weights of mice between treatments (baited = 14.6 g ± 0.1 SE; unbaited = 12.4 g ± 0.01 SE; P = 0.013) but there was no treatment by time interaction (P = 0.103; Table III.3; Figure III.3c), suggesting that baiting did not affect the body weights of mice long term. There also were differences over time (pre-baited =13.9 g ± 0.01 SE; immediately post-treatment = 11.9 g ± 0.1 SE; 1-month post-treatment = 13.2 g ± 0.2 SE; P = 0.045) and between sexes (males = 13.5 g ± 0.1 SE; females = 12.6 g ± 0.1 SE; P < 0.001). No other interactions were significant (P > 0.05). 136

Table III.3. Summary of analysis of variance using linear mixed effects models for ln-transformed body mass (g) showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets. Walpeup Brookstead Wudinna Variable DF F P DF F P DF F P

(Intercept) 1,4724 52520.28 <0.001 1,968 141618.4 <0.001 1,1094 157780.0 <0.001 Treatment 1,6 0.66 0.448 1,3 32.3 0.011 1,3 28.5 0.013 Time 3,4724 4.28 0.005 2,968 1.5 0.223 2,1094 3.1 0.045 Sex 1,4724 16.95 <0.001 1,968 38.0 <0.001 1,1094 49.0 <0.001 Treatment x Time 3,4724 0.14 0.937 2,968 1.9 0.154 2,1094 2.3 0.103 Treatment x Sex 1,4724 3.73 0.054 1,968 0.1 0.724 1,1094 0.0 0.902 Time x Sex 3,4724 1.34 0.260 2,968 0.6 0.566 2,1094 0.3 0.749 Treatment x Time x Sex 3,4724 2.50 0.058 2,968 0.9 0.418 2,1094 1.2 0.309

136 137

(a) Walpeup (b) Brookstead (c) Wudinna

2.9 2.9 2.9 Baited Baited Baited 2.8 2.8 2.8 ) ) ) g g 2.7 g 2.7 2.7 ( ( ( ss ss ss a a a 2.6 2.6 2.6 m m m body body 2.5 body n n n 2.5 2.5 l l l

2.4 2.4 2.4 Unbaited Unbaited Unbaited Baited Baited Baited 2.3 2.3 2.3 May Jun Jul Aug Aug Sep Oct Nov Dec Jun Jul

Month Month Month Figure III.3. Change in mean body mass of mice (g ± SE) for baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002).

137 138

III.3.3 Juveniles

Walpeup – There was a significant difference between the proportions of juveniles between treatments (unbaited = 21.2%; baited = 14.3%; P = 0.014), but no treatment by time interaction (P = 0.292; Table III.4; Figure III.4a) suggesting that baiting did not affect the proportion of juveniles in the population over time. There were significant differences over time (trap sessions 1-4 = 18.6%, 14.5%, 19.4%, 16.2%; P = 0.005) and between sexes (males = 13.0%; females = 23.7%; P < 0.001). The treatment by time by sex interaction was significant (P = 0.017) suggesting that the treatment by time interaction was influenced by changes in the proportion of male and female mice. Brookstead – There was no significant difference between treatments (baited = 25.6%; unbaited = 27.2%; P = 0.080; Table III.4), but there were significant time (trap sessions 2-6 = 36.7%, 25.0%, 7.4%, 2.2%, 0%; P < 0.001) and treatment by time interactions (P = 0.018; Figure III.4b), suggesting that the proportion of juveniles was influenced by the baiting over time. No other factors or interactions were significant (P > 0.05). Wudinna – There was no significant difference in the proportions of juveniles between treatments (baited = 9.0%; unbaited = 25.2%; P = 0.107; Table III.4), but there were a significant time (trap sessions 1-3 = 12.1%, 33.8%, 11.5%; P = 0.034) and treatment by time interactions (P = 0.014; Figure III.4c), suggesting that the proportion of juveniles was influenced by the baiting over time. There were also significant sex (males = 10.5%, females = 31.5%; P = 0.001), time by sex, and treatment by time by sex interactions (Table III.4) suggesting that differences in the proportion of juveniles was influenced by baiting, time and sex (P < 0.001).

139

Table III.4. Summary of analysis of variance using linear mixed effects models for juveniles trapped in the population showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets. Walpeup Brookstead Wudinna Variable DF F P DF F P DF F P

(Intercept) 1,36 763.14 <0.001 1,12 169.10 <0.001 1,11 155.75 <0.001 Treatment 1,6 11.86 0.014 1,2 10.98 0.080 1,3 5.18 0.107 Time 3,36 5.12 0.005 3,12 22.11 <0.001 2,11 4.70 0.034 Sex 1,36 15.22 <0.001 1,12 0.96 0.346 1,11 19.42 0.001 Treatment x Time 3,36 1.29 0.292 3,12 4.98 0.018 2,11 6.47 0.014 Treatment x Sex 1,36 1.30 0.262 1,12 1.60 0.230 1,11 0.92 0.359 Time x Sex 3,36 1.49 0.233 3,12 2.69 0.093 2,11 13.70 0.001 Treatment x Time x Sex 3,36 3.89 0.017 3,12 1.97 0.173 2,11 17.02 <0.001

139 140

(a) Walpeup (b) Brookstead (c) Wudinna

0.8 0.8 0.8 ed) ed) ed) Baited Baited Baited

rm Unbaited rm Unbaited rm Unbaited o o o Baited Baited Baited ansf ansf ansf

r 0.6 r 0.6 r 0.6 t t t

t t t o o o o o o r r r - - - e e e r r r a a a u u 0.4 u 0.4 0.4 q q q s s s sine sine sine c c c 0.2 0.2 0.2 s (ar s (ar s (ar e e e nil nil nil e e e v v v u u 0.0 u 0.0 0.0 J J May Jun Jul Aug J Aug Sep Oct Nov Dec Jun Jul

Month Month Month Figure III.4. Percentage of juveniles in the mouse populations (mean ± SE; transformed using the arcsine square-root transformation) on baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002).

140 141

III.3.4 Sex ratio

Walpeup, Brookstead and Wudinna – There was no difference in the sex ratio over time, between treatments or for the interaction between treatment and time (Table III.5; Figure III.5a-c) suggesting that baiting did not affect the sex ratio of mice.

III.3.5 Recapture rate

Walpeup – The average recapture rate immediately post-baiting was 0.18 (± 0.02

SE) and there was no difference between baited and unbaited sites (t6 = -0.199; P = 0.849). There were fewer recaptures on baited sites 1-month post baiting (unbaited =

0.12 ± 0.02 SE; baited = 0.06 ± 0.02 SE; t6 = 2.097; P = 0.081). Recapture rates were low 2-months after baiting (mean recapture rate = 0.01 ± 0.004 SE), and there was no difference between baited and unbaited sites (t6 = 0.745; P = 0.484) (Figure III. 6a). Brookstead – The recapture rate immediately post-baiting on unbaited sites was 0.08 (± 0.04 SE) and on baited sites was zero, however, there was no statistical difference between treatments (t2 = -2.000; P = 0.184). The recapture rate of mice on unbaited sites (0.02 ± 0.02 SE) 1-month after baiting was not significantly different to baited sites (zero) (t2 = -1.000; P = 0.423) (Figure III.6b). There were no recaptures from any site for trap sessions 5 or 6. Wudinna – The recapture rate of mice immediately post-baiting on unbaited sites was significantly greater (0.19 ± 0.001 SE) than on baited sites (0.002 ± 0.002 SE) (t3 = -71.072; P < 0.001). The recapture rate 1-month post-baiting on unbaited sites (0.004 ±

0.004 SE) was not significantly different to that on baited sites was (zero) (t3 = -1.342; P = 0.272) (Figure III.6c).

142

(Intercept) 1,21 720.11 <0.001 1,8 38.08 <0.001 1,8 7.32 0.027 Treatment 1,21 1.56 0.225 1,8 1.82 0.214 1,8 0.01 0.916 Time 3,21 0.13 0.943 4,8 0.51 0.732 2,8 0.01 0.992 Treatment x Time 3,21 0.70 0.564 4,8 1.87 0.210 2,8 0.08 0.921

142 143

(a) Walpeup (b) Brookstead (c) Wudinna

1.0 Male 1.0 Male 1.0 Male Baited Baited Baited

0.8 0.8 0.8 o o 0.6 o 0.6 0.6 rati rati rati Sex Sex 0.4 Sex 0.4 0.4

0.2 0.2 0.2 Unbaited Unbaited Unbaited Baited Baited Baited 0.0 Female 0.0 Female 0.0 Female May Jun Jul Aug Aug Sep Oct Nov Dec Jun Jul

Month Month Month Figure III.5. Sex ratio of mice (mean ± SE, males/total captures) on baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002).

143 144

(a) Walpeup (b) Brookstead (c) Wudinna

0.25 0.25 0.25 Baited Unbaited Baited Unbaited Baited Unbaited Baited Baited Baited 0.20 0.20 0.20 te te te a 0.15 a 0.15 a 0.15 e r e r e r

0.10 0.10 0.10 Recaptur Recaptur Recaptur

0.05 0.05 0.05

0.00 0.00 0.00 May Jun Jul Aug Aug Sep Oct Nov Dec Jun Jul Month Month Trap Session Figure III.6. Recapture rate of mice (mean ± SE) between trap sessions on baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002).

144 145

III.3.6 Condition

Walpeup – There was no difference in the relative condition of mice between treatments (unbaited = 97.9% ± 0.2 SE, baited = 97.1% ± 0.2 SE; P = 0.310), over time (trap sessions 1-4 = 96.8%, 96.2%, 99.3%, 96.8%; P = 0.066) or for the treatment by time interaction (P = 0.465; Table III.6; Figure III.7a), suggesting that baiting had no affect on the condition of mice over time. There was a significant difference between sexes (males = 98.0% ± 0.2 SE, females = 96.6% ± 0.2 SE; P < 0.001), but no other significant interaction terms. Brookstead – There was no difference in the relative condition of mice between treatments (unbaited = 103.9% ± 0.6 SE, baited = 108.3% ± 0.8 SE; P = 0.087), but there was a significant difference over time (trap sessions 2-6 = 108.9%, 97.4%, 106.5% 100.8%, 98.7%; P = 0.002) and there was a significant treatment by time interaction (P = 0.002; Table III.6, Figure III.7b) suggesting that baiting reduced the relative condition of mice over time. There was no difference in relative condition between sexes (males = 104.3% ± 0.6 SE, females = 108.2% ± 0.8 SE; P = 0.197). Wudinna – There was no difference in the relative condition of mice between treatments (unbaited = 97.4% ± 0.6 SE, baited = 104.6% ± 0.8 SE; P = 0.761), but there was a significant difference over time (trap sessions 1-3 = 103.0%, 91.0%, 98.0%; P < 0.001), but not for the treatment by time interaction (P = 0.274; Table III.6; Figure III.7c), suggesting that while there were changes in the condition of mice over time, it was not related to the effect of baiting. There was no difference in relative condition between sexes (males = 99.7% ± 0.6 SE, females = 100.0% ± 0.7 SE; P = 0.927).

146

Table III.6. Summary of analysis of variance using linear mixed effects models for the relative condition of mice trapped in the population showing main factors (variables), degrees of freedom (DF with numerator and denominator), F probabilities and significance (P) values, for Walpeup (Victoria, 1994), Brookstead (Queensland 1995), and Wudinna (South Australia 2002) data sets. Relative condition was calculated using regression in Krebs and Singleton (1993) to compare individual mice against a standardised mouse. Calculations for Walpeup and Wudinna were based on the Victorian mallee regression and for Brookstead on the Darling Downs regression of Krebs and Singleton (1993). Condition Walpeup Brookstead Wudinna Variable DF F P DF F P DF F P

(Intercept) 1,4695 334956.2 <0.001 1,516 32822.90 <0.001 1,506 13973.77 <0.001 Treatment 1,6 1.2 0.310 1,2 10.08 0.087 1,3 0.11 0.761 Time 3,4695 2.4 0.066 4,516 4.30 0.002 2,506 26.80 <0.001 Sex 1,4695 32.7 <0.001 1,516 1.67 0.197 1,506 0.01 0.927 Treatment x Time 3,4695 0.9 0.465 4,516 4.21 0.002 2,506 1.30 0.274 Treatment x Sex 1,4695 0.0 0.961 ------Time x Sex 3,4695 0.5 0.699 ------Treatment x Time x Sex 3,4695 0.9 0.438 ------

146 147

(a) Walpeup (b) Brookstead (c) Wudinna

120 120 120 Baited Baited Baited Unbaited Unbaited Unbaited 115 Baited 115 Baited 115 Baited ) ) )

% 110 % 110 % 110 ( ( ( on on on i i i 105 105 105 ndit ndit ndit o o o c c 100 c 100 100 e e e v v v i i i lat lat 95 lat 95 95 Re Re Re

90 90 90

85 85 85 May Jun Jul Aug Aug Sep Oct Nov Dec Jun Jul Month Month Month Figure III.7. Relative condition (mean ± SE) of mice on baited and unbaited sites from Walpeup (Victoria, 1994), Brookstead (Queensland, 1995) and Wudinna (South Australia, 2002). Relative condition was calculated using regressions in Krebs and Singleton (1993; see text for details).

147 148

III.4 Discussion

III.4.1 Efficacy of rodenticides

The application of acute rodenticides to high population densities of mice produced a range of responses in the mouse populations. Although there were significant reductions in densities of mice over time as a result of the baiting, the treatment by time interactions were only significant for zinc phosphide baiting at Wudinna, where the relative reduction in population density from baiting was >95%. Where strychnine was used, the relative reductions were <95% (39 and 1% at Walpeup; 91 and 92% at Brookstead) and the treatment by time interactions were not significant (although the reduction observed at Brookstead was almost significant; P = 0.075). One complicating factor for each of these studies was that population densities on unbaited sites declined drastically soon after baits were applied (>85% reductions in densities on baited and unbaited sites 1-4 months after baiting) so the treatment by time interaction was often confounded by this reduction on the unbaited sites. The efficacy of zinc phosphide is generally considered to be high, however where alternative food sources are available, bait acceptance and efficacy can be less than optimal (Caughley et al. 1998, Brown et al. 2002, Mutze & Sinclair 2004). Zinc phosphide was found to be as effective as strychnine in controlling mice in an experiment conducted in cereal crops in South Australia (Mutze & Sinclair 2004). In this study, zinc phosphide was more efficacious than strychnine, however, the application methods were different. In 1994 when the strychnine-baiting program in Walpeup was conducted, it was considered that sufficient bait would be available to high densities of mice if around 40% of paddocks were baited. This was based on the availability of bait on the ground (potentially 12,000 poisoned grains/ha; mice need to eat 1 or 2 grains to receive a lethal dose, Mutze 1989) and the likely movements of mice in such paddocks. A radio- tracking study of mice in the same area showed that the home range size was 0.037 ha during the breeding season (approximately 20 x 20 m) and was 0.119 ha during the non- breeding season (approximately 35 x 35 m) (Chambers et al. 2000). The assumption that mice could travel 12.5 m (half the interval between swathes) during the non- breeding period to locate bait should therefore be valid. For fallow crops, only the 149

fencelines (perimeter of paddocks) were baited. The actual amount of bait found on the ground immediately after application was much lower than was expected (Brown et al. 1997). Nonetheless, there appears to be a strong relationship between the percentage of area where bait was applied and the reduction in population densities achieved.

III.4.2 Short-term demographic effects of baiting on mouse populations

On baited sites at Walpeup immediately post-baiting (2 days after baiting), there was little change in body weight and little change in the percentage of juveniles in the population, but not as great as observed on unbaited sites (so a relative increase). The sex ratio, recapture rate and relative body condition remained unchanged. There appeared to be little impact of the baiting on the demography of the mouse populations, but there was a relative increase in juveniles, suggesting that while some mice might have been removed through baiting, juveniles quickly replaced them. It is not clear if these juveniles came from the gaps between the baited swathes, other nearby unbaited areas, or were always present but became more trappable post-treatment when adults were removed. Juveniles are less site-attached (Krebs et al. 1995, Chambers et al. 2000) and so are better colonisers of vacant areas after poisoning. On baited sites at Brookstead immediately post-baiting (9 days after baiting), the body weights of mice were higher than on unbaited sites, the captures of juveniles were lower, there was a bias towards females, and there were no recaptures of marked mice. The relative condition of mice on baited sites remained high immediately post-baiting compared to unbaited sites where condition dropped which showed the same trend as body weight. A density-dependent response may have occurred through increased availability per capita of food supply for surviving animals. On baited sites at Wudinna immediately post-baiting (4 days after baiting), body weights of mice were lower (but they were also lower on unbaited sites), there was as increase in the percentage of juveniles (but there was also an increase on unbaited sites), there was a bias towards females (but not significant), there was a significant reduction in recaptures and no difference in relative condition. Apart from the significant reduction in density, the only significant demographic response shown by the population was the reduction in recaptures, with a possibility that more males than females were removed. Since the reduction in density was highly significant it is likely 150

that baiting removed all types of mice and so short-term demographic changes were difficult to determine. There was no consistent trend across the three studies for the treatment by time interaction for density, body mass, juveniles, survival or condition, although the recapture rate of male and female mice was lower on baited sites after treatment (Table III.7). The low recapture rates would be expected if rodenticides were highly efficacious. There was a bias towards female mice on baited sites immediately post- bating particularly if the efficacy of baiting was high, but the response was variable. Changes in the sex ratio suggest that sex-specific, spatially density-dependent dispersal (Doncaster et al. 1997; Gundersen et al. 2001) might be occurring. It is not known whether this is because (1) females survived the baiting operation better or (2) females immigrated into baited areas more rapidly than males. It is more likely that (1) would occur because male house mice generally have a larger home range size, particularly during the non-breeding season (Krebs et al. 1995; Chambers et al. 2000) and would be more likely to disperse. In several small mammal species, juveniles or sub-adults are the type of animal most likely to disperse (Wolff 1993) but I found no strong evidence for this. Changes in body weight or condition might help to explain which type of mouse was present immediately post-baiting. For pocket gophers on removal sites in Canada, mean body mass was significantly lower compared to the control sites in one area, but in another area, there was no difference (Sullivan et al. 2001). No consistent increase or decrease in body weight was found in my studies.

151

Table III.7. Comparison of changes in population demographic characteristics in response to three different levels of effectiveness of rodenticide baiting for house mice (Mus domesticus) (short-term demographic responses), a natural crash in densities at the end of a mouse plague (long-term demographic changes) and published demographic characteristics for cyclic vole and lemming populations during the crash period (decline phase) from the peak phase to the low phase. House mice (Mus domesticus) Voles and lemmings Demographic Short-term changes Long-term changes characteristic Crash induced by poisoning Natural crash at end of Crash during decline phase plague Density <50% ~91% >95% >85% reduction ~85% reduction (from 300 to 50 C, D reduction reduction reduction (all sites) A voles/ha) (Walpeup) A (Brookstead) A (Wudinna) A Body weight No change A Increases A Decreases A Mixed response (increases & Decreases E decreases) A Juveniles Relative Decreases A Increases A Decreases (related to time Decreases (?) F, G increase A since breeding ceased?) A Sex ratio No change A Bias towards Bias Male biased or parity A Bias towards males G, bias towards females A towards females H, or parity F females A Survival (recapture No change A Decreases A Decreases A Poor (low recapture rate) A Poor F, G, I, J rate) Condition Relative Increases A No change A Poor condition A Lower in decline years E increase A Reproduction Not assessed Not assessed Not assessed In high-density years breeding In decline years delay in onset of season ceases earlier, larger breeding, decline in litter size, decline median size at maturity, in proportion breeding, although no breeding delayed, litter size difference in productivity observed depressed B (mean litter size multiplied by proportion females breeding) G

Source of information: A This study, B Singleton et al. (2001), C Hanski et al. (1991), D Korpimäki et al. (2004), E Norrdahl and Korpimäki (2002a) F Krebs and Myers (1974), G Norrdahl and Korpimäki (2002b), H Krebs and Boonstra (1978), I Myllymäki (1977), J Wilson et al. (1999). 151 152

Overall, it appeared that demographic variation was associated with differences in the percentage reduction in density after application of the rodenticides. The datasets from the three studies could not be combined because of differences in the timing of pre-and post-bait trapping, farming systems and bait types, and therefore I was unable to use the percentage reduction as a covariate. In terms of the predictions outlined in the Introduction, when baiting was efficacious (1) densities were lower on baited sites, (2) there was a mixed result for juveniles (increases and decreases) with a bias towards a female sex ratio and (3) survival (recapture rate) was low.

III.4.3 Long-term demographic effects of baiting and natural decline on mouse populations

The reductions in density from pre-treatment to 1-4 months post-treatment were >85% on all sites (baited and unbaited). At Walpeup, population densities declined to less than 80 mice/ha on all sites two-months after baiting. The body weights of mice were significantly higher on baited sites, there were fewer juveniles, the sex ratios were similar, the recapture rates were similar and the relative condition of mice was lower on baited sites (but not significantly). Breeding ceased in the fourth week of March 1994 (Singleton et al. 2001), 8-9 weeks prior to rodenticide being applied for this study at Walpeup. It is therefore likely that these juveniles were nestling survivors. Since there were fewer juveniles, we would expect body weights of survivors to be heavier, but this does not explain why the relative condition might be poorer compared to unbaited sites. There was an overall decline in relative condition on all sites after the crash which could reflect a lack of food or adequate cover in late winter. At Brookstead, densities declined to <30 mice/ha on all sites 3-4 months post- baiting. There was an overall decline in body weight across both treatments (baited and unbaited sites), no juveniles were captured, the sex ratio was highly male biased (significantly more males on baited sites) and there were no recaptures. There was a reduction in the relative condition of mice. Overall, these mice were in poor condition, there was no evidence of breeding, no young mice were present and very few females were present. There was no demographic differences in response to the baiting 3-4 months previously. 153

At Wudinna 1-month post-baiting, densities were <20 mice/ha on all sites. There was no difference in body mass between treatments, juveniles were still present, but more were present on baited sites than on unbaited sites (a significant switch from pre-baiting), sex ratios were similar (close to parity), there were few recaptures, and there was little difference in relative condition between treatments. The increase in juveniles on baited sites might be a response to the baiting that occurred 1-month previously. It would have been interesting to monitor the fate of the juveniles on the baited sites to see whether they form the nucleus of the breeding population in the following spring and to see if these populations had higher growth rates than unbaited sites, but it was not possible to continue trapping at Wudinna to determine this. Overall, there was little difference between baited and unbaited sites 1-4 months after baiting. The rapid decline in population densities, or crash, induced the same response in the populations as the baiting accomplished except for some differences in the proportion of juveniles and the sex ratio of mice (Table III.7). Apart from the >85% decline in density, the main response was a reduction in the proportion of juveniles (probably related to the time since breeding ceased) a general reduction in condition of mice and low recapture rates, implying that survival was poor. Presumably, if animals are in poor condition and survival is poor the low phase would be prolonged. These studies were conducted during non-breeding periods. Although the breeding season commenced in Brookstead sometime in October 1995, by then the number of adult females captured was very low. On sites where the density was reduced, we would expect compensatory changes in the breeding performance of adult females through higher growth rates, recruitment of young and breeding performance (higher litter sizes and higher percentage of adult females in breeding condition). These occurred in enclosed populations of mice where 67% of females were sterilised (Chambers et al. 1999), but the density was not manipulated. There are few studies of house mice in Australia where these factors have been considered and so is an area for future research. For cyclic vole and lemming populations in the Northern Hemisphere, the decline phase is generally characterised by reduction in proportion of juveniles, changes in the sex ratio (bias towards males or females depending on species examined), poor survival, poor condition of animals and changes in reproductive characteristics (Table III.7) (Krebs & Myers 1974; Myllymäki 1977; Krebs & Boonstra 1978; Wilson et al. 154

1999; Norrdahl & Korpimäki 2002a,b). There were five of seven demographic categories that were the same for a natural crash in mouse populations and the decline phase for cyclic lemming and vole populations from the Northern Hemisphere, with two categories having a mixed response. For my data on house mice in Australia, the crash phase was characterised by a reduction in the proportion of juveniles (but this could be related to the time since the cessation of the breeding season rather than to a characteristic of a declining mouse population), no change or a male bias in the sex ratio, poor survival (low recapture rate), and poor general condition of mice. From my hypotheses outlined earlier comparing rodenticide-induced crashes and natural crashes, I confirm that (1) densities were lower after the crash and that (2) there was one demographic characteristic having the same response (survival), with the other demographic categories having a mixed response (body weight, juveniles and condition), or were different (sex ratio) (Table III.7).

III.4.4 Conclusions

The effect of applying acute rodenticides over large areas of land can markedly reduce the population density of mice. The apparent benefit of applying bait is to reduce damage to crops, but there are few good studies that actually demonstrate this for house mice in Australia. In most of these cases significant damage had already occurred before bait was applied (Singleton et al. 1991; Mutze 1998; Mutze & Sinclair 2004). One of the confounding factors has been that mouse populations on unbaited sites often decline rapidly in density, or crash, such that the benefit accrued through using rodenticides is lost. This short-term benefit of reduced density may proportionally reduce damage to crops and could outweigh costs associated with the application of rodenticides such as zinc phosphide. The cost of applying zinc phosphide is around AU$10-15 /ha (Brown & Singleton 2002; Mutze & Sinclair 2004) and the break-even point was estimated at 1.9% yield loss (Brown 2005), but this depends on the yield of the crop and the efficacy of control achieved. Since peak mouse densities occur at sowing, it is generally only necessary to protect the crop until just after emergence when the nutritional value of the seed is exhausted (Mutze 1998). There was some evidence of reinvasion of mice into baited areas. This mainly occurred in Walpeup, where the effectiveness of the baiting was relatively low. 155

Therefore, provided the rodenticide is effective and that it is applied over large areas of land there should be little opportunity for rapid reinvasion and subsequent damage. Baiting is cheap and relatively effective, but why bait if mouse populations would crash anyway? We need to have a better understanding of when mouse populations will crash. We have a good understanding of the mechanisms that lead to an increase in mouse populations (Singleton 1989; Brown & Singleton 1999) and can model this and predict outbreaks with reasonable confidence (Pech et al. 1999; Kenney et al. 2003), but our understanding of when populations decline is limited. This knowledge on the demography of mouse populations that undergo crashes in abundance will be useful in further developing these models, and also may assist with the understanding of why there might be variation in the length of the low phase.

III.5 Acknowledgements

I thank Lisa Chambers, Micah Davies, Stephen Day, David Grice, Dean Jones, Alice Kenney, Charles Krebs, Greg Mutze, Bill Price, Grant Singleton, and Monica van Wensveen for assisting with data collection, and all the farmers for their willing participation in the studies. This research was conducted in accordance with the Australian code of practice for the care and use of animals for scientific purposes. Institute AEEC/SEAEC approval numbers were 93/94-33, 93/94 – 2, and 01/02 – 15. I thank CSIRO Sustainable Ecosystems, Bureau of Resource Sciences (Vertebrate Pest Program), and the Queensland Department of Environment and Heritage for their financial assistance for these projects.

156

III.6 References

Arthur, A. D., Pech, R. P., and Singleton, G. R. (In Press). Predicting the effect of virally vectored recMCMV immunocontraception on house mice (Mus musculus domesticus) in mallee wheatlands. Journal of Applied Ecology. Boonstra, R., Krebs, C. J., and Stenseth, N. C. (1998). Population cycles in small mammals: the problem of explaining the low phase. Ecology 79, 1479-1488. Brown, P. R. (2005). The effect of simulated house mouse damage to wheat in Australia. Crop Protection 24, 101-109. Brown, P. R. and Lundie-Jenkins, G. (1999). Non-target mortalities during aerial strychnine baiting of house mice. Wildlife Research 26, 117-128. Brown, P. R., Chambers, L. K., and Singleton, G. R. (2002). Pre-sowing control of house mice (Mus domesticus) using zinc phosphide: efficacy and potential non- target effects. Wildlife Research 29, 27-37. Brown, P. R. and Singleton, G. R. (1999). Rate of increase as a function of rainfall for house mouse Mus domesticus populations in a cereal-growing region in southern Australia. Journal of Applied Ecology 36, 484-493. Brown, P. R. and Singleton, G. R. (2002). Impacts of house mice on crops in Australia - costs and damage. In Human Conflicts with Wildlife: Economic Considerations. (Eds. L. Clark, J. Hone, J. A. Shivik, R. A. Watkins, K. C. VerCauteren, and J. K. Yoder.) pp. 48-58. (National Wildlife Research Center: Fort Collins, Colorado.) Brown, P. R., Singleton, G. R., Kearns, B., and Griffiths, J. (1997). Evaluation and cost- effectiveness of strychnine for control of populations of wild house mice (Mus domesticus) in Victoria. Wildlife Research 24, 159-172. Buckle, A. P. and Smith, R. H. (1994). Rodent Pests and Their Control. (CAB International: Wallingford.) Caughley, G. (1977). Analysis of Vertebrate Populations. (John Wiley and Sons: London.) Caughley, J., Donkin, C., and Strong, K. (1998). Managing mouse plagues in rural Australia. Proceedings of the 18th Vertebrate Pest Conference. (R. O. Baker and A. C. Crabb Eds.). Costa Mesa, California. Davis. University of California. pp. 160-165. 157

Caughley, J., Monamy, V., and Heiden, K. (1994). Impact of the 1993 mouse plague. GRDC Occasional Paper Series No.7. (GRDC: Canberra, Australia.) Chambers, L. K., Singleton, G. R., and Hinds, L. A. (1999). Fertility of wild mouse populations: the effects of hormonal competence and an imposed level of sterility. Wildlife Research 26, 579-591. Chambers, L. K., Singleton, G. R., and Krebs, C. J. (2000). Movements and social organization of wild house mice (Mus domesticus) in the wheatlands of Northwestern Victoria, Australia. Journal of Mammalogy 81, 59-69. Cooper, G. and Woods, N. (1996). Efficiency of bait application. In Report on aerial baiting with strychnine during the 1995 mouse plague in the Dalby-Goondiwindi area, Queensland. (Ed. V. Eldershaw.) pp. 87-94. (Queensland Department of Natural Resources: Brisbane.) Doncaster, C. P., Clobert J., Doligez, B., Gustafsson, L., and Danchin, E. (1997). Balanced dispersal between spatially varying local populations: An alternative to the source-sink model. The American Naturalist 150, 425-445. Fisher, G. (1996). Overall co-ordination. In Report on aerial baiting with strychnine during the 1995 mouse plague in the Dalby-Goondiwindi area, Queensland. (Ed V. Eldershaw.) pp. 14-21. (Department of Natural Resources: Queensland.) Gundersen, G., Johannesen, E., Andreassen, H. P., and Ims, R. A. (2001). Source-sink dynamics: how sinks affect demography of sources. Ecology Letters 4, 14-21. Johnson, G. D. and Fagerstone, K. A. (1994). Primary and secondary hazards of zinc phosphide to nontarget wildlife - a review of the literature. DWRC Research Report No. 11-55-005. United States Department of Agriculture, Animal and Plant Health Inspection Service, Colorado. pp. 26. Hanski, I, Hansson, L., and Henttonen, H. (1991). Specialist predators, generalist predators, and the microtine rodent cycle. Journal of Animal Ecology 60, 353- 367. Hansson, L. and Henttonen, H. (1988). Rodent dynamics as community processes. Trends in Ecology and Evolution 3, 195-200. Kenney, A. J., Krebs, C. J., Davis, S. A., Pech, R. P., Mutze, G. J., and Singleton. G. R. (2003). Predicting house mouse outbreaks in the wheat-growing areas of southeastern Australia. In Rats, mice and people: rodent biology and management. 158

ACIAR Monograph 96. (Eds. Singleton. G. R., L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 325-328. (ACIAR: Canberra.) Klemola, T., Tanhuanpää, M., Korpimäki, E., and Ruihomäki, K. (2002). Specialist and generalist natural enemies as an explanation for geographical gradients in population cycles of northern herbivores. Oikos 99, 83-94. Kookana, R. S., Rogers, S. L., and Oliver, D. P. (1997). Sorption and desorption behaviour of strychnine rodenticide in soils. Australian Journal of Soil Research 35, 491-502. Korpimäki, E., Brown, P. R., Jacob, J., and Pech, R. P. (2004). The puzzles of population cycles and outbreaks of small mammals solved? BioScience 54, 1071-1079. Krebs, C. J. and Boonstra R. (1978). Demography of the spring decline in populations of the vole, Microtus townsendii. Journal of Animal Ecology 47, 1007-1015. Krebs, C. J., Kenney, A. J., and Singleton, G. R. (1995). Movements of feral house mice in agricultural landscapes. Australian Journal of Zoology 43, 293-302. Krebs, C. J. and Myers, J. H. (1974). Population cycles in small mammals. Advances in Ecological Research 8, 267-399. Krebs, C. J. and Singleton, G. R (1993). Indices of condition for small mammals. Australian Journal of Zoology 41, 317-323. Land Conservation Council (1974). Report on the Mallee Study Area. (Land Conservation Council, Victoria: Melbourne.) Lidicker, W. Z., Jr (1988). Solving the enigma of microtine cycles. Journal of Mammalogy 69, 225-235. McCallum, H. I. and Singleton, G. R. (1989). Models to assess the potential of Capillaria hepatica to control population outbreaks of house mice. Parasitology 98, 425-437. Montgomery, W. I. (1981). A removal experiment with sympatric populations of Apodemus sylvaticus (L.) and A. flavicollis (Melchior) (Rodentia: Muridae). Oecologia 51, 123-132. Montgomery, W. I., Wilson, W. L., and Elwood, R. W. (1997). Spatial regulation and population growth in the wood mouse Apodemus sylvaticus: experimental manipulations of males and females in natural populations. Journal of Animal Ecology 66, 755-768. 159

Mutze, G. J. (1989). Effectiveness of strychnine bait trails for poisoning mice in cereal crops. Australian Wildlife Research 16, 459-465. Mutze, G. J. (1993). Cost-effectiveness of poison bait trails for control of house mice in mallee cereal crops. Wildlife Research 20, 445-456. Mutze, G. J. (1998). The 1993 strychnine baiting program for mouse control in South Australian grain crops. I. Efficacy. Wildlife Research 25, 533-546. Mutze, G. J. and Hubbard, D. J. (2000). Effects of molluscicide baits on field populations of house mice. Agriculture, Ecosystems and Environment 80, 205- 211. Mutze, G. J. and Sinclair, R. (2004). Efficacy of zinc phosphide, stychnine and chlorpyrifos as rodenticides for the control of house mice in South Australian cereal crops. Wildlife Research 31, 249-257. Mutze, G. J., Veitch, L. G. and Miller, R. B. (1990). Mouse plagues in South Australian cereal-growing areas. II. An empirical model for prediction of plagues. Australian Wildlife Research 17, 313-324. Mwanjabe, P. S. and Leirs, H. (1997). An early warning system for IPM-based rodent control in smallholder farming systems in Tanzania. Belgian Journal of Zoology 127, 49-58. Myllymäki, A. (1977). Demographic mechanisms in the fluctuating populations of the field vole Microtus agrestis. Oikos 29, 468-493. Norrdahl, K. and Korpimäki, E. (2002a). Changes in individual quality during a 3-year population cycle of voles. Oecologia 130, 239-249. Norrdahl, K. and Korpimäki, E. (2002b). Changes in population structure and reproduction during a 3-yr population cycle of voles. Oikos 96, 331-345. Oliver, D. P., Kookana, R. S., Rogers, S. L., Sinclair, R. G., Roberts, G., and Cook, C. (2000). Strychnine uptake by cereals and legumes grown in the glasshouse and field. Journal of Environmental Quality 29, 1329-1336. Pech, R. P., Hood, G., Singleton, G. R., Salmon, E., Forrester, R., and Brown, P. R. (1999). Models for predicting plagues of house mice (Mus domesticus) in Australia. In Ecologically-based Management of Rodent Pests. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 81-112. (Australian Centre for International Agricultural Research: Canberra.) 160

Plumb, T. W. (Ed.) (1963). ‘Atlas of Australian Resources. Soils’. Second Series. (Department of National Development: Canberra.) Pocock, M. J. O., Hauffe, H. C., and Searle, J. B. (2005). Dispersal in house mice. Biological Journal of the Linnaen Society 84, 565-583. Redhead, T. D. (1988). Prevention of plagues of house mice in rural Australia. In Rodent Pest Management. (Ed. I. Prakash.) pp. 191-206. (CRC Press: Boca Raton, Florida.) Saunders, G. R. (1983). Evaluation of mouse plague control techniques in irrigated sunflower crops. Crop Protection 2, 437-445. Saunders, G. R. and Giles, J. R. (1977). A relationship between plagues of the house mouse, Mus musculus (Rodentia: Muridae) and prolonged periods of dry weather in south-eastern Australia. Australian Wildlife Research 4, 241-147. Sinclair, A. R. E., Olsen, P. D., and Redhead, T. D. (1990). Can predators regulate small mammal populations? Evidence from house mouse outbreaks in Australia. Oikos 59, 382-392. Singleton, G. R. (1983). The social and genetic structure of a natural colony of house mice, Mus musculus, at Healesville Wildlife Sanctuary. Australian Journal of Zoology 31, 155-166. Singleton, G. R. (1985). Population dynamics of Mus musculus and its parasites in Mallee Wheatlands in Victoria during and after a drought. Australian Wildlife Research 12, 437-445. Singleton, G. R. (1989). Population dynamics of an outbreak of house mice (Mus domesticus) in the mallee wheatlands of Australia - hypothesis of plague formation. Journal of Zoology, London 219, 495-515. Singleton, G. R., Brown, P. R., Pech, R. P., Jacob, J., Mutze, G. J., and Krebs, C. J. (2005). One hundred years of eruptions of house mice in Australia - a natural biological curio. Biological Journal of the Linnaen Society 84, 617-627. Singleton, G. R., Hinds, L. A., Leirs, H., and Zhang, Z. (1999). Ecologically-based Management of Rodent Pests. 494 pp. (Australian Centre for International Agricultural Research: Canberra.) Singleton, G. R. and Hay, D. A. (1983). The effect of social organization on reproductive success and gene flow in colonies of wild house mice, Mus musculus. Behavioural Ecology and Sociobiology 12, 49-56. 161

Singleton, G. R., Krebs, C. J., Davis, S. A., Chambers, L. K., and Brown, P. R. (2001). Reproductive changes in fluctuating house mouse populations in southeastern Australia. Proceedings of the Royal Society of London Series B 268, 1741-1748. Singleton, G. R., Smith, A. L., Shellam, G. R., Fitzgerald, N., and Muller, W. J. (1993). Prevalence of viral antibodies and helminths in field populations of house mice (Mus domesticus) in southeastern Australia. Epidemiology and Infection 110, 399-417. Singleton, G. R., Twigg, L. E., Weaver, K. E., and Kay, B. J. (1991). Evaluation of bromadiolone against house mice (Mus domesticus) populations in irrigated soybean crops. II. Economics. Wildlife Research 18, 275-284. Stenseth, N. C. (1999). Population cycles in voles and lemmings: Density dependence and phase dependence in a stochastic world. Oikos 87, 427-461. Stenseth, N. C., Leirs, H., Skonhoft, A., Davis, S. A., Pech, R. P., Andreassen, H. P., Singleton. G. R., Lima, M., Machangu, R. M., Makundi, R. H., Zhang, Z., Brown, P. R., Shi, D., and Wan, X. (2003). Mice, rats and people: the bioeconomics of agricultural rodent pests. Frontiers in Ecology and the Environment 1, 367-375. Sullivan, T. P. (1979). Repopulation of clear-cut habitat and conifer seed predation by deer mice. Journal of Wildlife Management 43, 861-871. Sullivan, T. P., Sullivan, D. S., and Hogue, E. J. (2001). Reinvasion dynamics of northern pocket gopher (Thomomys talpoides) populations in removal areas. Crop Protection 20, 189-198. Sutherland, D. R., Banks, P. B., Jacob, J., and Singleton, G. R. (2004). Shifting age structure of house mice during a population outbreak. Wildlife Research 31, 613- 618. Twigg, L. E. and Kay, B. J. (1992). Evaluation of Quintox® for control of feral house mice. Journal of Wildlife Management 56, 174-185. Twigg, L. E., Singleton, G. R., and Kay, B. J (1991). Evaluation of bromadiolone against house mice (Mus domesticus) populations in irrigated soybean crops. I. Efficacy of control. Wildlife Research 18, 265-274. Wilson, D. J., Krebs, C. J., and Sinclair, A. R. E. (1999). Limitation of collared lemming populations during a population cycle. Oikos 87, 382-398. 162

Wolff, J. O. (1993). What is the role of adults in mammalian juvenile dispersal? Oikos 68, 173-176. Ylönen, H., Jacob, J., Davies, M. J., and Singleton, G. R. (2002). Predation risk and habitat selection of Australian house mice Mus domesticus during an incipient plague: desperate behaviour due to food depletion. Oikos 99, 284-289.

163

Study IV. The effect of simulated house mouse damage to wheat in Australia

Crop Protection

Peter R. Brown

CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia

This chapter was published in Crop Protection in 2005.

Brown, P. R. (2005). The effect of simulated house mouse damage to wheat in Australia. Crop Protection 24, 101-109.

164

The effect of simulated house mouse damage to wheat in Australia

Peter R. Brown

CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia

Abstract: House mice, Mus domesticus, cause significant damage to wheat crops in Australia by digging up and eating newly planted seeds, or by cutting stems and eating developing grain. This study was conducted to determine how wheat compensates for damage by physically cutting tillers to simulate mouse damage. Tillers were cut at five intensities: 0, 5, 10, 25 and 50% at each growth stage of emergence, tillering, booting and ripening. There were five replicates of each treatment plus an experimental control. The later the damage occurred, the greater the yield loss. Significant reductions in yield occurred for 50% intensity of damage that occurred at the booting stage (34.4% reduction in yield), and for 25 and 50% intensity for damage that occurred at the ripening stage (29.3% and 54.7% reduction in yield respectively). There was a significant difference in the average weight of grain per tiller between stage of crop damage, but no difference for intensity of damage or interactions. The yield per tiller was greater for the tillering stage (1.50 g tiller-1 ± 0.04 SE) compared to the experimental control sites (1.18 g tiller-1 ± 0.08 SE). This study demonstrated that wheat crops can compensate for simulated damage by increasing grain production and increasing the number of tillers or survival rate of tillers remaining after damage was imposed. These findings have important consequences for managing potential impacts of mice on wheat crops in Australia. There was little yield loss for up to 50% intensity of damage at emergence. This suggests that mouse control is required only when high levels of damage are likely to occur after the tillering stage. However, even a 3.4% reduction in yield as found for the 50% intensity at emergence would result in a US$18 /ha loss. 165

Keywords: Compensation, Damage, Mouse plagues, Mus domesticus, Yield, Yield loss

IV.1 Introduction

Vertebrate pests cause significant damage to agricultural crops (eg: Goryńska 1981; Gillespie 1985; Bruggers et al. 1998; Marsh 1998; Rao et al. 2002; Singleton 2003; Stenseth et al. 2003), pasture (eg: Crawley & Weiner 1991; Bell et al. 1998; Zhang et al. 1999; Croft et al. 2002), and to forestry resources (eg: Sullivan & Klenner 1993; Notle 1998). Much effort is spent on controlling these pests without knowing whether the damage they cause actually reduces yield (Hone 1994). The relationship between damage caused and yield loss must be understood in order to establish appropriate economic injury levels and economic thresholds (Carlson & Wetzstein 1993; Buckle 1994; Mumford & Knight 1997; Wossink & Rossing 1998; Nabirye et al. 2003). However, the full impact of vertebrate pests on agricultural or forest resources is difficult to assess because of the complexity of the resources (Nolte & Dykzeul 2002) and the variable nature of recovery or compensation by plants. Rodents are one of the most important vertebrate pests to cereal crops globally (Stenseth et al. 2003). Rodents cause significant damage to maize (Key 1990; Fiedler 1994; Leirs et al. 1996; Mwanjabe et al. 2002), wheat (Poché et al. 1982; Brooks et al. 1985), and rice (Fulk & Akhtar 1981; Buckle et al. 1985; Islam et al. 1993; Khokhar et al. 1993; Singleton & Petch 1994; Parshad 1999; Singleton 2003; Wood & Chung 2003). In Australia, the introduced house mouse (Mus domesticus) causes serious damage to cereal grain crops (Brown & Singleton 2002). Under favourable conditions, house mice in Australia can rapidly increase in abundance (densities of >1,000 mice/ha) to form mouse plagues (Redhead 1988; Mutze 1989; Singleton & Redhead 1989; Brown & Singleton 2002), and subsequently cause high agricultural losses. These conditions are though to be driven by weather conditions, particularly good rainfall that promotes the growth of grasses and weeds but also influences good crop production (Brown & Singleton 1999; Pech et al. 1999; Singleton et al. 2005). A mouse plague affected South Australia and Victoria in 1993/94 and caused more than US$50 million in damage to crops, machinery, rural 166

businesses, schools and hospitals (Caughley et al. 1994). These mouse plagues have been a feature of the grain growing areas of southern and eastern Australia since the early 1900s (Saunders & Giles 1977; Mutze 1989; Singleton et al. 2005). In any particular region, a mouse plague generally occurs every 4-7 years (Singleton & Redhead 1989; Singleton, et al. 2005). The crop that suffers the most from mouse plagues in Australia is wheat (Redhead 1988; Brown & Singleton 2002). Wheat is the main winter cereal crop grown in southern and eastern Australia, accounting for 62% of the grains export market and was worth US$7 billion in 2001/02 (Hooper et al. 2003). Mouse populations generally peak in abundance at the time of sowing of winter cereals in southeastern Australia, and during mouse plagues, farmers often have to resow their crops because mice have dug up seed from the ground (Mutze 1998; Brown & Singleton 2002). Significant damage can occur at later stages of crop growth, particularly after mice begin their breeding season in early spring (Singleton et al. 2001) and numbers increase. Management of impacts of mouse plagues in Australia generally has been reactive rather than preventative. During mouse plagues, large amounts of poisons are distributed to control mouse damage (see Singleton 2000 for review). Up to 500,000 ha was baited with zinc phosphide to reduce the damage caused by mice in 1999 (David Croft, NSW Agriculture, personal communication). Mouse control such as this is conducted without a good understanding of whether a reduction in mouse numbers actually results in a reduction in damage. Furthermore, there is little understanding of the association between damage caused by mice to wheat crops and to what extent the crop can compensate for this damage. Recommendations on the appropriate timing of control actions could be refined if this information was available. Some cereal crops posses an innate ability to compensate from insect damage, disease, competition for resources from weeds, and from extremes in climate (Bardner & Flethcer 1974; Chiariello & Gulmon 1991; Rubia et al. 1996; Sadras et al. 1999; Oyediran & Heinrichs 2002). Observations of damage by the multimammate rat (Mastomys natalensis) to maize crops in Tanzania, showed that 89% damage after emergence resulted in only 13% damage just prior to harvest (Mwanjabe & Leirs 1997) showing an ability of maize to recover from early damage. Simulation studies of rodent damage to rice crops in Asia have shown that plants compensated for damage during early growth phases but significant yield loss occurred if damage was at later growth 167

phases (Buckle et al. 1979; Fulk & Akhtar 1981; Poché et al. 1981; Haque et al. 1986). The present study was conducted to determine the extent to which wheat crops can compensate for simulated mouse damage at a range of intensities and at various stages of growth. Compensation is defined here as the ability of a plant or crop to recover from initial damage, and while this damage remains, the plant or crop increases grain production to get near to the potential yield.

IV.2 Materials and methods

The experiment was conducted at CSIRO Plant Industry, Ginninderra Research Station, Canberra, Australia, on a red podsolic soil (Sleeman 1979). The wheat crop (Triticum aestivum variety H45) was sown on 11 July 2002 at the rate of 70 kg/ha, and harvested on 23 December 2002 (165 days after sowing, DAS). The crop received pre- emergent herbicide (Logran), fertilizer at sowing (110 kg/ha Starter IS), and herbicide treatment at tillering (2.1 L/ha Bromoxynil, 78 DAS). The region went through a drought during 2002. There was 44 mm of rain in June 2002 (147% of long term mean, LTM), 18 mm in July (44% LTM), 25 mm in August (49% LTM), 62 mm in September (111% LTM), 12 mm in October (19% LTM), 10 mm in November (15% LTM) and 17 mm in December (33% LTM) (52% LTM June-December). Sprinklers were used to irrigate the crop at flowering (120 DAS; 35 mm) and the milky dough stage (134 DAS, 35 mm) to allow the crop to develop and mature. Irrigation water came from a water storage dam located on the research station. Wheat was sown in five 120 m strips, each strip being one replicate. The distance between each strip was 0.5 m, which allowed access to the crop by farm machinery for spraying (width of a wheel track). There were 25 plots per strip, with each plot 2 m in length. The beginning and end of each plot was marked with flagging tape held to the ground using small bamboo pegs. A buffer area was maintained between each plot and was generally 2 m in length. Each plot was marked at emergence (56 DAS). Each plot comprised ten rows of plants, each row was 0.2 m apart. The outer two rows on each side of each plot were designated as buffer rows. All plant and tiller counts and all treatments were applied to the central six rows of plants (plot size of 2 x 1.2 m). 168

The treatments were imposed to simulate mouse damage at five levels: 0, 5, 10, 25 and 50%; and at four crop stages: emergence (61-62 DAS, 10-11 September), maximum tillering (90-95 DAS, 9-14 October), booting (112-113 DAS, 31 October – 1 November), two weeks prior to harvest (153-158 DAS, 11-16 December), and an experimental control (no treatment). Mouse damage is likely to occur during all these phases. There were two types of controls: 0% level of damage, and experimental control (no treatment). This gave 25 treatment plots for each of the five replicates. Treatments within each replicate were assigned randomly. Each plot was marked with a small plastic tag denoting timing and intensity of treatment. During the emergence stage, treatment was applied by removing seedlings using a small metal spike to prise the seedling from the soil while ensuring minimal disturbance to surrounding seedlings. This was done to simulate the digging up of seeds by mice after sowing. For the remaining treatments, tillers were cut using a scalpel blade 50 mm above the soil surface (to replicate cut tillers by mice). At each stage of treatment, all plants (at emergence) and tillers (at all other crop stages) were counted for plots where treatments were applied. Every twentieth tiller was removed for the 5% intensity, every tenth tiller was removed for the 10% intensity, every fourth tiller was removed for the 25% intensity and every second tiller was removed for the 50% intensity. The same process was used at emergence, except entire plants were removed rather than tillers only. To determine if wheat plants were compensating by growing new tillers, counts of tillers were made from plots in two replicates just prior to harvest (159-160 DAS). It was not possible to conduct counts on more replicates because of shortage of time before harvest. Counts were made of the number of mature tillers (bearing ripe grains) and immature tillers (short and green tillers bearing immature grain). On the day before harvest, the buffer areas between each plot were flattened to the ground by foot to ensure no contamination of experimental plots with grain from buffer areas. A small conventional combine harvesting machine was used to harvest the grain. Grain from each plot was collected in labelled cloth bags. Each grain sample was cleaned using a seed and grain cleaner (Clipper Model 400 Office Tester and Cleaner) and weighed using an electronic balance.

169

IV.2.1 Mouse trapping

Trapping was conducted at the site to remove any house mice that may cause damage to the crop. Fifty Longworth single-capture live traps were set around the perimeter of the crop and along an adjacent fenceline for two or three consecutive nights when each treatment was imposed. The abundance of mice at the experimental site was low. The trap success at emergence (48 DAS) was 2.7% (4 mice from 150 trap nights), at maximum tillering (90 DAS) 1.3% (2/150), at the booting stage (112 DAS) 2.0% (3/149), and just prior to harvest (159 DAS) 3.0% (3/100).

IV.2.2 Statistical analyses

Yield data were square root transformed and analysed using a two-way analysis of variance, using the factors of stage of crop damage (experimental control, emergence, tillering, booting, and ripening) and intensity of damage (0, 5, 10, 25 and 50%). For significant interactions, Tukey’s pair wise comparisons were conducted to examine significant differences between factors. Data on the number of tillers were analysed in the same manner as for yield data. Means are presented ± one standard error (SE).

IV.3 Results

IV.3.1 Imposition of treatments

There was an average of 1164.4 (± 44.6 SE) tillers per plot counted at the tillering stage, 879.8 (± 27.0 SE) at the booting stage, and 793.4 (± 17.1 SE) at the ripening stage.

IV.3.2 Yield

The average yield of control plots was 989.2 g (± 43.1 SE, n = 25; plot size = 2 x 1.2 m, so 4.12 tonnes/ha). There were significant differences in yield between time of 170

damage (F4,99 = 2.673; P < 0.05), between intensities of damage (F4,99 = 4.238; P <

0.005), and interactions (F16,99 = 1.877; P < 0.05). In general, the later the damage occurred, the less able the wheat plants could compensate for damage (Figure IV.1). Among the crop stages, only those treated at ripening had significantly lower yields than experimental control sites (805.8 g ± 54.8 SE and 989.2 g ± 43.1 SE respectively; P < 0.05). Also, only those plots receiving 50% damage had significantly lower yields than plots with 5% damage (796.1 g ± 58.6 SE and 1043.1 g ± 51.7 SE respectively; P < 0.05). There were significant reductions in yield for 50% intensity compared to 5% intensity for damage that occurred at the booting stage (649.3 g ± 53.3 SE and 1160.8 g ± 197.1 SE respectively; P < 0.05), and for 50% intensity compared to 0, 5 and 10% intensity for damage that occurred prior to harvest (448.3 g ± 15.0 SE, 981.5 g ± 87.0 SE, 1064.5 g ± 113.3 SE and 835.8 g ± 81.8 SE respectively; each P < 0.05) (Figure IV.1). There was no difference in yields within control sites or among the different levels of intensity of damage at emergence or at tillering stages (P > 0.05).

1400 a a a 1200 a a

) a a a a a a a g a a ( 1000 a a ot a a

r pl 800 a,b

pe b

eld 600 i b

n y 0%

ai 400 5%

Gr 10% 200 25% 50% 0 Control Emergence Tillering Booting Ripening

Crop stage Figure IV.1. Effect of simulated mouse damage on grain yield (g/plot), Ginninderra Research Station, Canberra. Shown are means ± SE for each intensity of damage and crop stage when simulated damage occurred. Columns with different letters are significantly different as determined by Tukey’s multiple comparison tests (P > 0.05).

171

At 25% intensity of damage, there was a significant reduction in yield when damage occurred at ripening compared to emergence (698.2 g ± 58.3 SE and 1148.4 g ± 169.6 SE respectively; P < 0.05) (Figure IV.1). At 50% intensity of damage, there were significant reductions in yield when damage occurred at ripening compared with the experimental control, emergence and tillering stages (448.3 g ± 15.0 SE, 1059.7 g ± 61.3 SE, 955.4 g ± 118.4 SE, 867.8 g ± 147.2 SE respectively; P < 0.05) and for damage that occurred at booting compared to the experimental control (648.3 g ± 53.3 SE and 1059.7 g ± 61.3 SE respectively; P < 0.05). There were no significant differences in yield among crop stages with 0, 5, and 10% damage (each P > 0.05). The wheat crop compensated for damage through increasing yield, and the earlier the damage occurred the greater the compensation (Table IV.1). Damage and yield losses only corresponded when damage was imposed at the ripening stage (yield losses of 0.8%, -7.6%, 15.5%, 29.3% and 54.7% for 0, 5, 10, 25 and 50% intensities respectively). Complete compensation occurred for damage at emergence and during tillering. Partial compensation occurred at the booting stage. The large yield loss for 0% intensity at emergence (17.2% yield loss) cannot be explained.

Table IV.1. Relative mean percentage yield loss (± SE) for each stage of damage and intensity of damage compared to experimental control plots. Negative numbers indicate a yield gain. Intensity Emergence Tillering Booting Ripening (%) (%) (%) (%) (%) 0 17.2 ± 11.5 0.3 ± 8.5 -0.7 ± 8.7 0.8 ± 8.8 5 -0.3 ± 7.9 -4.1 ± 11.3 -17.4 ± 19.9 -7.6 ± 11.5 10 4.3 ± 18.1 -3.3 ± 6.5 1.1 ± 8.2 15.5 ± 8.3 25 -16.1 ± 17.1 13.7 ± 10.3 19.5 ± 10.6 29.3 ± 5.9 50 3.4 ± 12.0 12.3 ± 14.9 34.4 ± 5.4 54.7 ± 1.5

IV.3.3 Number of tillers

There was an average of 1164.4 tillers per plot (± 44.6 SE) counted at the tillering stage, 879.8 tillers (± 27.0 SE) at the booting stage, and 793.4 tillers (± 17.1 SE) at the ripening stage. The average number of tillers on all control sites prior to harvest was 710.5 (± 27.5 SE). 172

A similar trend in reductions in yield was evident for the number of tillers counted prior to harvest (Figure IV.2). There were significant differences in the number of mature tillers between timing of damage (F4,25 = 3.821; P < 0.05), between intensities of damage (F4,25 = 13.136; P < 0.001), and interactions (F16,25 = 3.638; P < 0.005). Since tillers from only two replicates were counted, these data were not analysed further. As for the yield results, the later the damage occurred, the wheat plants were less able to compensate for damage. Furthermore there were increases in the number of tillers for 5 and 10% intensities but not at higher intensities (Table IV.2). The counts of tillers prior to harvest show that even at 50% intensity of damage imposed at the tillering stage, there was no statistical difference in the number of mature tillers compared to control sites.

1000

800 plot

per 600 illers t

of 400 0% 5%

Number 200 10% 25% 50% 0 Control Sowing Tillering Booting Ripening

Crop stage Figure IV.2. Effect of simulated mouse damage on number of tillers per plot, Ginninderra Research Station, Canberra. Shown are means ± SE for each intensity of damage and crop stage when simulated damage occurred. Data from two replicates only.

173

Table IV.2. Relative mean percentage loss in number of tillers (± SE) for each stage of damage and intensity compared to experimental control plots from two replicates only. Negative numbers indicate a gain in tillers. Intensity Emergence Tillering Booting Ripening (%) (%) (%) (%) (%) 0 1.3 ± 6.8 -2.3 ± 7.6 -4.8 ± 16.1 -17.0 ± 6.8 5 -3.4 ± 10.5 -5.4 ± 9.4 1.3 ± 21.5 3.8 ± 2.0 10 4.5 ± 9.9 -0.1 ± 1.3 -2.0 ± 2.1 1.6 ± 0.4 25 0.7 ± 5.7 18.8 ± 8.7 22.4 ± 2.9 18.2 ± 5.4 50 12.2 ± 1.6 12.7 ± 6.0 48.1 ± 0.4 45.6 ± 2.3

The number of immature tillers present was counted prior to harvest on two replicates (short green tillers with immature grain). The percentage of immature tillers was highest for damage that occurred at the booting stage and for 50% intensity of damage (8.0%, Table IV.3).

Table IV.3. Percentage of immature tillers compared to mature tillers counted prior to harvest (160 DAS) from two replicates only (mean ± SE). Intensity Control Emergence Tillering Booting Ripening (%) (%) (%) (%) (%) (%) 0 0.1 ± 0.1 0.1 ± 0.1 0.6 ± 0.6 0.3 ± 0.1 0.0 ± 0.0 5 0.1 ± 0.1 0.1 ± 0.1 0.2 ± 0.2 0.5 ± 0.3 0.0 ± 0.0 10 0.0 ± 0.0 0.6 ± 0.6 0.2 ± 0.1 0.8 ± 0.7 0.0 ± 0.0 25 0.1 ± 0.1 0.3 ± 0.0 1.0 ± 0.2 2.9 ± 0.9 0.1 ± 0.1 50 0.2 ± 0.1 0.2 ± 0.1 0.7 ± 0.1 8.0 ± 0.8 0.0 ± 0.0

The number of tillers counted at harvest was compared to the number of tillers remaining after treatments were imposed, at the tillering, booting and ripening stages to determine the percentage of survival of remaining tillers (Figure IV.3). A value of 100% implies that all tillers survived from treatment to harvest. For damage imposed at the tillering stage, an average of 95% of tillers that were remaining on the 50% level of intensity were counted prior to harvest, whereas only 60% of the tillers that were remaining on the 10% intensity were counted prior to harvest (Figure IV.3). Other intensities were approximately 80%. For damage imposed at the booting stage, an average of 85-90% of tillers that were remaining for all levels of intensity were counted prior to harvest. For damage imposed at the ripening stage, 100% of tillers remaining were counted prior to harvest for all levels of damage intensity, as expected. 174

100 (%) t

t n es e rv 80 m t ea r to ha t r ior e 60 r t g af ted p in n

n 40 i

a 0% m cou 5% s 10%

s re 20 ller r i 25% e T ll i 50% t / 0 Tillering Booting Ripening

Crop stage Figure IV.3. Ratio of number of tillers counted prior to harvest and the number of tillers remaining after treatment for damage imposed at the tillering, booting and ripening stages. Data from emergence were not included because they consisted of counts of plants only. A value of 100 means that 100% of tillers that were present after treatment were counted at harvest. Shown are means ± SE for each intensity of damage and crop stage when simulated damage occurred. Data from two replicates only.

IV.3.4 Yield per tiller

The yield per tiller was calculated to determine if this provided a mechanism for compensation by wheat crops (Figure IV.4). There was a significant difference between stage of crop damage (F4,24 = 3.303; P < 0.05), but no difference for intensity of damage or interactions (P > 0.05). The yield per tiller was greater for the tillering stage (1.50 g tiller-1 ± 0.04 SE) compared to control sites (1.18 g tiller-1 ± 0.08 SE). No further analyses were conducted, but the yield per tiller was greatest on 50% intensity of damage from the tillering stage onwards.

175

2.5

2.0 ) g ( 1.5 tiller per 1.0 eld i Y

0% 0.5 5% 10% 25% 50% 0.0 Control Emergence Tillering Booting Ripening

Crop stage Figure IV.4. Yield of grain per tiller (g), calculated by dividing the yield per plot by the number of tillers counted prior to harvest. Shown are means ± SE for each intensity of damage and crop stage when simulated damage occurred. Data from two replicates only.

IV.4 Discussion

Wheat crops were able to compensate for simulated damage by increasing grain production through increasing the number of tillers or a higher survival rate of the remaining tillers after damage was imposed. Significant reductions in yield occurred only when the damage occurred during the ripening stage at high levels of intensity (50% at booting and ripening and 25% at ripening). These results demonstrate that compensation was more than just increased tillering, as found for simulated rodent damage to rice crops in Southeast Asia (Buckle et al. 1979; Poché et al. 1981; Haque et al. 1986) and for simulated insect defoliation to rice crops in West Africa (Oyediran & Heinrichs 2002). The damage to plants at all other stages and intensities in this experiment did not cause a significant reduction in yield. These findings have important consequences for managing potential impacts of mice on wheat crops in Australia. There was no yield loss for up to 50% intensity of 176

damage at emergence. The impact of mice is generally greatest in late autumn and early winter at a time when winter cereals crops are planted across southern Australia (wheat, barley and oats) (Mutze 1998; Brown & Singleton 2002). However, wheat crops could compensate for a high level of damage from planting through to the tillering stage, and up to moderate levels at booting and ripening stages. This suggests that mouse control is required only when high levels of damage occur after the tillering stage (>25% at booting or >10% at the ripening stage). The simulated damage imposed in this experiment was regular in pattern (eg every fourth tiller for 25% damage), but natural damage caused by rodents and other animals to crops follow highly aggregated or clumped patterns (Buckle 1994; Engeman et al. 1994; Tristiani et al. 2000; Mulungu et al. 2003). Therefore, the results from this research could be considered as a maximum response. There have been only a few studies of mouse damage to wheat crops at various crop stages and these were at a subset of mouse population densities (Brown & Singleton 2002; Brown et al. 2003b). More research effort is required to measure damage to crops over a range of mouse densities to determine the impact on yield and whether the compensation would be similar to that found here. There was a decline in the number of tillers that were counted in successive treatments. This was not an effect of the treatments, but of a natural reduction in tillers as the plant matures and has been observed both in wheat (Miralles & Slafer 1999) and rice (Buckle 1994). Oyediran and Heinrichs (2002) conducted artificial insect defoliation and reported that the earlier in the plant growth the injury occurs, the more rapidly the plants can compensate by translocating assimilates from injured to healthy tillers. This partly explains the observed ability of wheat crops to compensate for up to 50% damage early during the growth of the crop (up to tillering stage) because most of the tillers might have become non-effective, but when damage was high the plant was able to redirect resources to the remaining tillers. The ability of wheat plants to compensate from simulated damage also can be partially explained through higher yields per tiller and a higher survival rate of remaining tillers. Tiller cutting by rodents can also stimulate re-tillering (Buckle et al. 1979; Fulk & Akhtar 1981; Poché et al. 1981; Haque et al. 1986). Re-tillering occurred in this experiment on damaged plants (up to 8% of tillers on 50% intensity at booting stage were immature prior to harvest, Table 4). Damage that occurred earlier also 177

initiated the development of new tillers, but these had sufficient time to mature and bear viable grain and so were not counted as immature prior to harvest. However, the majority of the compensation by the wheat plants was through the higher survival of the remaining tillers, and these tillers having higher grain weight. It was not possible to count the number of grains per tiller at harvest, which would help to clarify this. These data were analysed using statistical comparisons at the 95% level of significance, however, a farmer may have a different level of tolerance for mouse damage to their crop. For example, the 12.3% reduction in yield for the 50% intensity of damage at the tillering stage would result in a loss of US$67 /ha assuming a yield of 4.1 tonnes/ha (yield per hectare from this experiment) and a farm gate price of US$132 (average farm gate value for wheat, 1998-2002, Turner et al. 2001). This would be significant for a farmer. Even a 3.4% reduction in yield as found for the 50% intensity at emergence would result in a US$18 /ha loss. This level in yield loss is greater than the cost of broadscale application of zinc phosphide (US$10 /ha, Brown & Singleton 2002), which is widely used in Australia to manage outbreaks of mice (Caughley et al. 1998; Singleton 2000; Brown et al. 2002, 2003a). Based on these data, the break-even level of yield loss if using zinc phosphide would be 1.9%, however, other cost effective management options are available to farmers (Brown et al. 2004). A farmer would need to balance the costs of conducting mouse control against a likely loss in yield and even a small effective level of damage could be economically significant. The wheat crop compensated from simulated mouse damage through increasing grain production, number of tillers and survival of remaining tillers. There was little yield loss for up to 50% intensity of damage at emergence. Mouse control is required only when high levels of damage are likely to occur after the tillering stage, therefore early action would obviate this build up and prevent damage from occurring. One limitation of this study was that simulated damage occurred at discrete periods through the growth of the crop. This may not represent natural conditions in fields when mouse densities are likely to fluctuate or damage may occur over a sustained period at a range of intensities (up to 100%). Therefore, further research is required on the relationship between abundance of mice and damage to crops over a range of mouse population densities. This can be through observations of a range of densities in the field, through manipulation in the field of small enclosures, or by modelling the response using crop models and incorporating a mouse population model. Knowledge of this relationship 178

would enable the development of appropriate management targets (Brown & Singleton 2002).

IV.5 Acknowledgements

I am grateful to Phil Dunbar and staff of CSIRO Plant Industry, Ginninderra Research Station, for arranging the sowing and harvesting of the crop and to the assistance provided to me by Micah Davies and David Grice to impose the treatments. Thanks to Grant Singleton and Peter Banks for their critical review of the manuscript.

IV.6 References

Bardner, R. and Fletcher, K. E. (1974). Insect defoliations and their effects on the growth and yield of field crops: a review. Bulletin of Entomological Research 64, 141-160. Bell, A. C., Byrne, P. M., and Watson, S. (1998). The effect of rabbit (Oryctolagus cuniculus) grazing damage on the growth and yield of winter cereals. Annals of Applied Biology 133, 431-442. Brooks, J. E., Sultana, P., and Poché, R. M. (1985). Damage by Bandicota bengalensis to growing wheat in Bangladesh and methods of control. Acta Zoologica Fennica 173, 135-137. Brown, P. R., Chambers, L. K., and Singleton, G. R. (2002). Pre-sowing control of house mice (Mus domesticus) using zinc phosphide: efficacy and potential non- target effects. Wildlife Research 29, 27-37. Brown, P. R., Davies, M. J., Croft, J. D., and Singleton, G. R. (2003a). Impact of farm management practices on house mouse populations and crops in an irrigated farming system. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 338-342. (ACIAR: Canberra.) Brown, P. R., Davies, M. J., Croft, J. D., and Singleton, G. R. (2004). Can farm management practices reduce the impact of house mouse populations on crops in an irrigated farming system? Wildlife Research 31, 597-604. 179

Brown, P. R. and Singleton, G. R. (1999). Rate of increase as a function of rainfall for house mouse Mus domesticus populations in a cereal-growing region in southern Australia. Journal of Applied Ecology 36, 484-493. Brown, P. R. and Singleton, G. R. (2002). Impacts of house mice on crops in Australia - costs and damage. In Human Conflicts with Wildlife: Economic Considerations. (Eds. L. Clark, J. Hone, J. A. Shivik, R. A. Watkins, K. C. VerCauteren, and J. K. Yoder.) pp. 48-58. (National Wildlife Research Center: Fort Collins, Colorado.) Brown, P. R., Singleton, G. R., Tann, C. R., and Mock, I. (2003b). Increasing sowing depth to reduce mouse damage to winter crops. Crop Protection 22, 653-660. Bruggers, R. L., Rodriguez, E., and Zaccagnini, M. E. (1998). Planning for bird pest problem resolution: A case study. International Biodeterioration & Biodegradation 42, 173-184. Buckle, A. P. (1994). Damage assessment and damage surveys. In Rodent Pests and Their Control. (Eds. A. P. Buckle and R. H. Smith.) pp. 219-248. (CAB International: Wallingford.) Buckle, A. P., Yong, Y. C., and Abdul Raman, H. (1985). Damage by rats to rice in South east Asia with special reference to an integrated management scheme proposed for Peninsular Malaysia. Acta Zoologica Fennica 173, 139-144. Buckle, A. P., Yong, Y. C., and Rowe, F. P. (1979). Yield response of the rice variety Improved Mashuri to simulated rat damage. Malaysian Agriculture Journal 52, 135-144. Carlson, G. A. and Wetzstein, M. E. (1993). Pesticide and pest management. In Agricultural and Environmental Resource Economics. (Eds. G. A. Carlson, D. Zilberman, and J. A. Miranowski.) pp. 221-267. (Oxford University Press: New York.) Caughley, J., Donkin, C., and Strong, K. (1998). Managing mouse plagues in rural Australia. Proceedings of the 18th Vertebrate Pest Conference. (R. O. Baker and A. C. Crabb Eds.). Costa Mesa, California. Davis. University of California. pp. 160-165. Caughley, J., Monamy, V., and Heiden, K. (1994). Impact of the 1993 Mouse Plague. GRDC Occasional Paper Series No.7. (GRDC: Canberra, Australia.) 180

Chiariello, N. R. and Gulmon, S. L. (1991). Stress effects on plant reproduction. In 'Response of plants to multiple stresses'. (Eds. H. A. Mooney, W. E. Winner, and E. J. Pell.) pp. 161-188. (Academic Press: San Diego.) Crawley, M. J. and Weiner, J. (1991). Plant size variation and vertebrate herbivory: winter wheat grazed by rabbits. Journal of Applied Ecology 28, 154-172. Croft, J. D., Fleming, P. J. S., and van de Ven, R. (2002). The impact of rabbits on a grazing system in eastern New South Wales. 1. Ground cover and pastures. Australian Journal of Experimental Agriculture 42, 909-916. Engeman, R. M., Sugihara, R. T., Pank, L. F., and Dusenberry, W. E. (1994). A comparison of plotless density estimators using monte carlo simulation. Ecology 75, 1769-1779. Fiedler, L. A. (1994). Rodent Pest Management in Eastern Africa. Plant Production and Protection Paper No. 123. FAO, Rome. pp. 83. Fulk, G. W. and Akhtar, M. T. (1981). An investigation of rodent damage and yield reduction in rice. Tropical Pest Management 27, 116-120. Gillespie, G. D. (1985). Feeding behaviour and impact of ducks on ripening barley crops grown in Otago, New Zealand. Journal of Applied Ecology 22, 347-356. Goryńska, W. (1981). Method of determining relations between the extent of damage in farm crops, big game numbers, and environmental conditions. Acta Theriologica 26, 469-481. Haque, M. E., Ahmed, M. S., and Brooks, J. E. (1986). Effect of simulated rat damage on deep water rice in Bangladesh. Bangladesh Journal of Agriculture 11, 63-69. Hone, J. (1994). Analysis of Vertebrate Pest Control. (Cambridge University Press: Cambridge.) Hooper, S., Barrett, D., and Martin, P. (2003). Australian Grains Industry 2003 - Grains Industry - Performance and Outlook. 12 pp. (Australian Bureau of Agricultural and Resource Economics: Canberra.) Islam, Z., Morton, R. G., and Jupp, B. P. (1993). The effects of rat damage on deepwater rice yields in Bangladesh. International Journal of Pest Management 39, 250-254. Key, G. (1990). Pre-harvest crop losses to the African striped ground squirrel, Xerus erythropus, in Kenya. Tropical Pest Management 36, 223-229. 181

Khokhar, A. R., Rizvi, S. W. A., and Shaukat, S. S. (1993). Rodent damage and its effects on yield loss in different rice varieties in Punjab, Pakistan. International Journal of Pest Management 39, 397-399. Marsh, R. E. (1998). Historical review of ground squirrel crop damage in California. International Biodeterioration & Biodegradation 42, 93-99. Miralles, D. J. and Slafer, G. A. (1999). Wheat development. In Wheat: Ecology and Physiology of Yield Determination. (Eds. E. H. Satorre and G. A. Slafer.) pp. 13- 43. (Food Products Press: New York.) Mulungu, L. S., Makundi, R. H., and Leirs, H. (2003). Robustness of techniques for estimating rat damage and yield loss in maize fields. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. G. R. Singleton, L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 224-228. (Australian Centre for International Agricultural Research: Canberra.) Mumford, J. D. and Knight, J. D. (1997). Injury, damage and threshold concepts. In Methods in Ecological and Agricultural Entomology. (Eds. D. R. Dent and M. P. Walton.) pp. 203-220. (CAB International: Wallingford, Oxon.) Mutze, G. J. (1989). Mouse plagues in South Australian cereal-growing areas. I. Occurrence and distribution of plagues. Australian Wildlife Research 16, 677- 683. Mutze, G. J. (1998). The 1993 strychnine baiting program for mouse control in South Australian grain crops. I. Efficacy. Wildlife Research 25, 533-546. Mwanjabe, P. S. and Leirs, H. (1997). An early warning system for IPM-based rodent control in smallholder farming systems in Tanzania. Belgian Journal of Zoology 127, 49-58. Mwanjabe, P. S., Sirima, F. B., and Lusingu, J. (2002). Crop losses due to outbreaks of Mastomys natalensis (Smith, 1834) Muridae, Rodentia, in the Lindi Region of Tanzania. International Biodeterioration & Biodegradation 49, 133-137. Nabirye, J., Nampala, P., Kyamanywa, S., Ogenga-Latigo, M. W., Wilson, H., and Adopala, E. (2003). Determination of damage-yield loss relationships and economic injury levels of flower thrips on cowpea in eastern Uganda. Crop Protection 22, 911-915. Nolte, D. L. (1998). Efficacy of selected repellents to deter deer browsing on conifer seedlings. International Biodeterioration & Biodegradation 42, 101-107. 182

Nolte, D. L. and Dykzeul, M. (2002). Wildlife impacts on forest resources. In Human Conflicts with Wildlife: Economic Considerations. (Eds. L. Clark, J. Hone, J. A. Shivik, R. A. Watkins, K. C. VerCauteren, and J. K. Yoder.) pp. 163-168. (National Wildlife Research Center: Fort Collins, Colorado.) Oyediran, I. O. and Heinrichs, E. A. (2002). Response of lowland rice plants to simulated insect defoliation in West Africa. International Journal of Pest Management 48, 219-224. Parshad, V. R. (1999). Rodent control in India. Integrated Pest Management Reviews 4, 97-126. Pech, R. P., Hood, G., Singleton, G. R., Salmon, E., Forrester, R., and Brown, P. R. (1999). Models for predicting plagues of house mice (Mus domesticus) in Australia. In Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 81-112. (Australian Centre for International Agricultural Research: Canberra.) Poché, R. M., Haque, M. E., Main, M. Y., Sultana, P., and Karim, M. A. (1981). Rice yield reduction by simulated rat damage in Bangladesh. Tropical Pest Management 27, 242-246. Poché, R. M., Main, M Y., Haque, M. E., and Sultana, P. (1982). Rodent damage and burrowing characteristics in Bangladesh wheat fields. Journal of Wildlife Management 46, 139-147. Rao, K. S., Maikhuri, R. K., Nautiyal, S., and Saxena, K. G. (2002). Crop damage and livestock depredation by wildlife: a case study from Nanda Devi Biosphere Reserve, India. Journal of Environmental Management 66, 317-327. Redhead, T. D. (1988). Prevention of plagues of house mice in rural Australia. In Rodent Pest Management. (Ed. I. Prakash.) pp. 191-206. (CRC Press: Boca Raton, Florida.) Rubia, E. G., Heong, K. L., Zaluck, M., Gonzales, B., and Norton, G. A. (1996). Mechanisms of compensation of rice plants to yellow stem borer Scirpophaga incertulas (Walker) injury. Crop Protection 15, 335-340. Sadras, V. O., Fereres, A., and Ratcliffe, R. H. (1999). Wheat growth, yield, and quality as affected by insect herbivores. In Wheat: Ecology and Physiology of Yield 183

Determination. (Eds. E. H. Satorre and G. A. Slafer.) pp. 183-227. (Food Products Press: New York.) Saunders, G. R. and Giles, J. R. (1977). A relationship between plagues of the house mouse, Mus musculus (Rodentia: Muridae) and prolonged periods of dry weather in south-eastern Australia. Australian Wildlife Research 4, 241-147. Singleton, G. R. (2000). Ecologically-based rodent management - Integrating new developments in biotechnology. Proceedings of the 19th Vertebrate Pest Conference. (T. P. Salmon and A. C. Crabb Eds.). Mission Valley Hilton Hotel, San Diego, California. Davis, California. University of California. pp. 221-227. Singleton, G. R. (2003). Impacts of Rodents on Rice Production in Asia. IRRI Discussion Paper Series No. 43. International Rice Research Institute, Los Bańos, Philippines. pp. 30. Singleton, G. R., Brown, P. R., Pech, R. P., Jacob, J., Mutze, G. J., and Krebs, C. J. (2005). One hundred years of eruptions of house mice in Australia - a natural biological curio. Biological Journal of the Linnaen Society 84, 617-627. Singleton, G. R., Krebs, C. J., Davis, S. A., Chambers, L. K., and Brown, P. R. (2001). Reproductive changes in fluctuating house mouse populations in southeastern Australia. Proceedings of the Royal Society of London Series B 268, 1741-1748. Singleton, G. R. and Petch, D. A. (1994). A review of the Biology and Management of Rodent Pests in Southeast Asia. ACIAR Technical Reports No. 30. ACIAR, Canberra. pp. 65. Singleton, G. R. and Redhead, T. D. (1989). House mouse plagues. In Mediterranean Landscapes in Australia: Mallee Ecosystems and their Management. (Eds. J. C. Noble and R. A. Bradstock.) pp. 418-433. (CSIRO: Melbourne.) Sleeman, J. R. (1979). The Soils of the Ginninderra Experiment Station, ACT. CSIRO Division of Soils, Divisional Report No. 41. Stenseth, N. C., Leirs, H., Skonhoft, A., Davis, S. A., Pech, R. P., Andreassen, H. P., Singleton. G. R., Lima, M., Machangu, R. M., Makundi, R. H., Zhang, Z., Brown, P. R., Shi, D., and Wan, X. (2003). Mice, rats and people: the bioeconomics of agricultural rodent pests. Frontiers in Ecology and the Environment 1, 367-375. 184

Sullivan, T. P. and Klenner, W. (1993). Influence of diversionary food on red squirrel populations and damage to crop trees in young lodgepole pine forest. Ecological Applications 3, 708-718. Tristiani, H., Murakami, O., and Kuno, E. (2000). Rice plant damage distribution and home range distribution of the ricefield rat Rattus argentiventer (Rodentia: Muridae). Belgian Journal of Zoology 130, 83-91. Turner, S., Barrett, D., Beasley, A., and Gleeson, T. (2001). Outlook for Australian grains industry, 2001-02. Agriculture Australia 2001 Conference. Melbourne. Canberra. Australian Bureau of Agricultural and Resource Economics. Wood, B. J. and Chung, G. F. (2003). A critical review of the development of rat control in Malaysian agriculture since 1960s. Crop Protection 22, 445-461. Wossink, G. A. A. and Rossing, W. A. H. (1998). On increasing returns and discrete choice: integrating production ecological principles in economic analysis of crop management. Journal of Environmental Management 54, 233-247. Zhang, Z., Chen, A., Ning, Z., and Huang, X. (1999). Rodent pest management in agricultural systems in China. In Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. (Eds. G. R. Singleton, L. A. Hinds, H. Leirs, and Z. Zhang.) pp. 261-284. (Australian Centre for International Agricultural Research: Canberra.)

185

Study V. Relationship between abundance of pest rodents and damage to agricultural crops

Manuscript

Peter R. Brown1,2,3, Neil I. Huth4, Peter B. Banks2, Grant R. Singleton1,3

1 CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia 2 School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia 3 Cooperative Research Centre for Pest Animal Control, GPO Box 284, Canberra, ACT, 2601, Australia 4 CSIRO Sustainable Ecosystems, PO Box 102, Toowoomba, Qld, 4350, Australia

186

Relationship between abundance of pest rodents and damage to agricultural crops

Peter R. Brown1,2,3, Neil I. Huth4, Peter B. Banks2, Grant R. Singleton1,3

1 CSIRO Sustainable Ecosystems, GPO Box 284, Canberra, ACT, 2601, Australia 2 School of Biological, Earth and Environmental Sciences, The University of New South Wales, NSW, 2052, Australia 3 Cooperative Research Centre for Pest Animal Control, GPO Box 284, Canberra, ACT, 2601, Australia 4 CSIRO Sustainable Ecosystems, PO Box 102, Toowoomba, Qld, 4350, Australia

Abstract: Vertebrate pests, particularly rodents, can cause significant damage to a range of agricultural crops. In Australia, introduced house mice Mus domesticus have significant impacts on wheat crops from sowing through to harvest by periodically undergoing widespread outbreaks where densities may exceed 1,000 mice/ha. An understanding of the relationship between pest abundance and damage to crops is essential for the calculation of economic injury levels (EIL) and lead to more informed management decisions. The crop modelling framework, APSIM, was used to simulate the impact of mouse damage on yield loss on wheat from Walpeup, northwestern Victoria, where a long term dataset on the density of mice was available (1983-2003). The model was calibrated using results from field trials where wheat plants were hand clipped to imitate mouse damage. The grazing effect of mice was estimated using the population density, daily intake per mouse and the proportion of wheat grain and plant tissue in the diet. The fraction of crop eaten by mice each day was modelled as crop death in APSIM. Yields were determined with and without the grazing effect of mice to determine yield loss. The mean yield loss caused by mice was 12.4% (± 5.4 SE; range – 0.5 to 96%). There were 7/21 years when yield loss was >5%. A damage/mouse abundance relationship was constructed and a sigmoidal curve explained 97% of variation when accounting for different trajectories of mouse densities from sowing to harvest. It appears that the majority of simulated damage occurred around emergence of 187

the crop when mouse densities were >100 mice/ha. The sigmoidal damage/abundance curve was further supported when the model was subjected to a range of mouse densities. This is the first time that field data on mouse density and a crop simulation model have been combined to estimate yield loss. We show how the model could be used to examine the efficacy of baiting and how to estimate EILs. Because the broadscale application of zinc phosphide is cheap and effective, the EIL is very low (<1% yield loss). The APSIM modelling environment is highly flexible and could be used for other vertebrate pests in a range of crops or pastures to develop density/damage relationships and to assist with management.

Key-words: house mouse, Mus domesticus, wheat, yield loss, APSIM, modelling

V.1 Introduction

Vertebrate pests such as wild boars, ducks, moose, deer, rodents and monkeys can cause significant damage to agriculture through browsing foliage, consuming bark of forest resources, grazing on pasture, or eating crops (Goryńska 1981; Gillespie 1985; Buckle & Smith 1994; Putman & Moore 1998; Nolte & Dykzeul 2002; Rao et al. 2002). Much time and effort is spent on controlling these pest animals, but it is often not known whether this control actually results in less damage or prevents further damage (Hone 1994). It is generally assumed that if animals are killed or removed, then there will be a reduction in damage, but few studies actually demonstrate this (eg: Advani & Mathur 1982; Mutze 1993; Sheikher & Jain 1997). In order to know the benefits of control, a sound understanding of the relationship between the abundance of pest animals and the damage they cause is required. This relationship can then be used to assess economic injury levels (EILs) and to establish appropriate management goals. Worldwide, rodents are a group of vertebrates that cause significant damage to a range of agricultural crops. These include damage by Bandicota spp. to wheat (Brooks et al. 1985; Poché et al. 1982) and to rice (Islam et al. 1993), Mastomys natalensis and Xerus erythropus damage to maize (Key 1990; Mulungu et al. 2003; Mwanjabe et al. 2002), Microtus damage to alfalfa (Sterner et al. 1996), Rattus rattus damage to macadamia nuts (Tobin et al. 1997; White et al. 1998), R. sordidus and R. rattus damage to sugarcane (Lefebvre et al. 1989; Whisson 1996), R. tiomanicus damage to oil 188

palms (Wood & Liau 1984), and R. argentiventer damage to lowland irrigated rice (Buckle et al. 1979, 1985; Khokhar et al. 1993; Tristiani et al. 2000; Singleton et al. 2005b). Rodents also cause damage to cauliflower and cabbage crops (Sheikher & Jain 1997). There are only a few cases where the relationship between damage and abundance has been described. Poché et al. (1982) and Lefebvre et al. (1989) described linear relationships and Mulungu et al. (2003) described a sigmoidal relationship. It is therefore necessary to determine whether the relationship is linear or sigmoidal for each species. In Australia, the feral house mouse, Mus musculus domesticus (Schwarz and Schwarz), causes significant damage to agricultural crops, particularly to winter cereals such as wheat (Singleton & Redhead 1989; Mutze 1993, 1998; Caughley et al. 1994; Brown et al. 1997; Brown & Singleton 2002), but also cause damage to other crops (Saunders & Robards 1983; Twigg et al. 1991; Singleton et al. 1991; Brown & Singleton 2002) and to infrastructure in rural areas (Redhead 1988; Caughley et al. 1994). Mice cause damage at all stages of crop development by digging up newly planted seeds, by cutting tillers to gain access to nutrients contained within the tiller, or by accessing the developing grain as the crop matures. Outbreaks of mice have been a feature of the southern and eastern grain growing regions of Australia since 1904 (Singleton et al. 2005a) and are characterised by rapid and widespread increases in densities of mice to >800 mice/ha. Outbreaks occur somewhere in Australia every 4 years or so, but their frequency in any particular region is normally 1 year in 7 (Singleton 1989; Mutze 1991; Singleton et al. 2005a). Damage to crops can be severe during these outbreaks (Redhead 1988; Caughley et al. 1994), and it is not uncommon for farmers to have to resow their crops when mouse densities are high at sowing (Mutze 1998). Farmers currently manage mouse problems and damage by baiting with zinc phosphide (Brown et al. 2002; Mutze & Sinclair 2004). During non-outbreak years, densities are typically <50 mice/ha and little crop damage or economic loss are reported (Singleton et al. 2001; Brown and Singleton 2002). Given the high fluctuations in densities of mice and the damage they can inflict on wheat crops, the relationship between the density of mice and damage to crops needs to be understood. This would enable the establishment of damage thresholds and management targets to prevent both severe damage and reduce the indiscriminate use of rodenticides to control damage that has already occurred. There are two approaches to 189

develop such relationships: (1) to relate field observations of densities of mice and damage to crops, or (2) to develop a theoretical relationship by simulating the effects of mouse damage using crop models. There has been little success in establishing a density-damage relationship for mice using direct field observations because of (1) high variation between sites and over time and (2) difficulties in obtaining reliable measurements of yield loss (Brown & Singleton 2002). Furthermore, collecting these data can be expensive and time consuming. Simulating the effects of mouse damage on crops using crop simulation modelling has many advantages, including (1) predicting yield loss of crops under different mouse population density projection scenarios, (2) predicting the impacts of the effectiveness of rodent control on yield loss, (3) conducting benefit:costs analyses of various management strategies, and (4) determining economic injury levels (EILs) and economic thresholds (ETs) for management. Here we build on a study that simulated mouse damage to wheat crops through hand-cutting tillers at different crop stages (Brown 2005) and extend this using a well-validated crop modelling framework to develop a theoretical density-damage relationship using densities of mice estimated from mark-recapture field studies in wheat fields. Crop models have previously been used to examine the relationship between the abundance of insects and damage to crops (eg Pinnschmidt et al. 1995) and weed density and crop yield (Doyle 1991). This is the first time such an approach has been used for rodents.

V.2 Methods

The Agricultural Production Systems Simulator (APSIM) is widely used to simulate biophysical processes in farming systems and to examine the effect of changes in farm management practices, irrigation schedules, fertiliser applications and can be used to examine economic benefits of different management practices (Hammer et al. 1996; McCown et al. 1996; Keating and Meinke 1998; Muchow and Keating 1998; Probert et al. 1998; Cheeroo-Nayamuth et al. 2000). APSIM reliably predicts yields for a range of crops (Robertson et al. 2000; Keating et al. 2003; Yunusa et al. 2004). Modules for 25 types of crops and pastures can be used and include canola, chickpea, lupin, lucerne, maize, rice, sorghum, soybean, sugar, sunflower and wheat. APSIM-Wheat has been 190

extensively used and tested (Keating et al. 2003; Wang et al. 2003) and was used in this modelling exercise (APSIM Version 3.6). APSIM-Wheat simulates the growth and development of a wheat crop in daily time-steps. The module simulates responses to weather (radiation and temperature), soil water and soil nitrogen. The model simulates phenological development of the plants, leaf area growth, biomass, nitrogen concentration of leaves, stems, roots and grains on a daily basis and also predicts grain size and grain number. Two model configurations were developed. The first was used to calibrate the APSIM model on the experimental damage by mice by hand-clipping wheat plants at the Ginninderra Research Station, Australian Capital Territory (35º12’S, 149º00’E) as reported in Brown (2005). The second model simulated the response of wheat crops to changes in the abundance of wild house mouse populations in agricultural landscapes at Walpeup, Victoria (35º08’S, 142º01’E) using a long-term data set on mouse population changes.

V.2.1 Validation of clipping experiment

APSIM requires information on the latitude, longitude, annual average ambient temperature (ºC), annual amplitude in mean monthly temperature (ºC), maximum and minimum temperatures (ºC), daily radiation (MJ m-2) and daily rainfall. The climatic variables from the Bureau of Meteorology (BOM) station at Hall in 2002 were used to run the model (BOM station number 070045; located 4 km from the Ginninderra Research Station). These data were sourced from APSRU-SILO (http://www.nrm.qld.gov.au/silo/datadrill.html). Fertiliser application and irrigation schedules followed that reported in Brown (2005). Soil properties including water and nitrogen levels were estimated based on general conditions at the Ginninderra Research Station. Wheat, Triticum aestivum, variety H45 was used. A simulated mouse-grazing effect was imposed by killing a certain fraction of the wheat crop at specific times as defined in Brown (2005) (one-off cutting). Simulated damage was applied at emergence, maximum tillering, booting and ripening stages at 0, 5, 10, 25 and 50% intensities. Simulations were run in APSIM for a theoretical 1 ha field for each combination of crop stage and intensity of mouse damage. Mean actual yields from five replicates reported in Brown (2005) were compared with 191

the simulated yields using linear regression to test for differences between actual and simulated yields.

V.2.2 Mouse grazing model

The climatic records for 1983-2003 for Walpeup were sourced from APSRU- SILO, as for the validation model (BOM station number 076064; located at the Mallee Research Station). Fertiliser schedules, soil properties and nitrogen levels were estimated based on general conditions at the Mallee Research Station (O’Leary and Conner 1996) and were reset in early March of each year to replicate those for a wheat crop in a 2-year wheat-pasture rotation. Wheat variety Yitpi was used because it is a common variety for southern Australia. To determine whether the APSIM model was adequately predicting yields of wheat crops when densities of mice were low or when little mouse damage was noted, the observed and predicted yields from the Mallee Research Station for 1984-1997 were analysed using linear regression and paired t-test. The amount eaten by the mouse population each day (A) was calculated as:

A = B x C x D; where: B = Mouse density: The abundance of mice was obtained from Singleton (1989), Brown and Singleton (1999) and Singleton et al. (2005a). Basic trapping information for mice were as follows: Longworth live-capture traps (Longworth Scientific, Abingdon, UK) baited with wheat were set over 3 or 4 consecutive nights, every 6-8 weeks from February 1983 to December 2003. There was a period from November 1990 to October 1992 where no regular trapping occurred. Mouse densities were known to be low at this time (Singleton personal observation) so we estimated the missing densities based on years of low densities and by fitting missing data using a linear time step. A range of trap layouts were used, but typically grids of 6 x 6 traps in crop and pasture fields and lines of 15-20 traps were set along adjacent fencelines with 10 m between each trap. There were 400 to 1,400 trap nights set per trapping session. These trapping data were in the form of adjusted trap success (number of mice captured per 100 trap nights adjusted using the frequency-density transformation of Caughley 192

(1977)). These data were then converted into density estimates. The conversion was based on datasets where data for adjusted trap success and Petersen density estimates were available (n = 52). The density of mice/ha was determined by dividing the Petersen estimate by the effective trapping area of grids and fencelines assuming mice living within 5 m of the grid or fenceline were likely to enter a trap. Density estimates were then divided by 0.52 since Petersen estimates are known to underestimate densities of mice (Davis et al. 2003). The simple linear regression was: Density = 1.565(adjusted trap success)/0.52, with r2 = 0.836. We calculated the daily density of mice by assuming a linear change between trap sessions. C = Amount eaten by mice day-1: An adult mouse needs to eat 3-4 g of food each day in order to survive (Mutze et al. 1991; Moro and Bradshaw 2002). Mice are wasteful eaters so we conservatively assume that mice waste or spoil as much food as they actually consume (0.006 kg day-1). D = Grazing fraction: The proportion of wheat grain and plant tissue in the diet of mice was used to obtain the final mouse feeding effect. Monthly figures of diet composition for wheat grain and plant tissue were obtained from a diet study conducted in the same region (Tann et al. 1991). The estimated daily amount of grazing by the mouse population (A) was imported into APSIM. The fraction of wheat crop eaten by mice was calculated for the vegetative (sowing to maximum tillering stages) and reproductive stages (panicle initiation to harvesting stages) as follows: Vegetative stage: A/wheat biomass. Reproductive stage: A/wheat grain yield. These calculations were conducted in APSIM to obtain the cumulative effect of mouse grazing. An upper limit of the fraction of plants grazed was set at 0.25. This was invoked to prevent crop failure only when mouse densities were high just after emergence of the crop. Farmers sometimes have to resow because mouse damage at sowing is high (Mutze 1998), but we were interested in how the crop would compensate for such damage without resowing included in the model.

193

V.2.3 Analysis of damage/density relationship

The APSIM model was run initially to obtain yields without the influence of mice and was then run to compare the impacts of mouse grazing. The relationship between mouse density and damage to the wheat crop (yield loss) was analysed by regression. A range of linear and non-linear regressions models were ranked using Akaike’s information criterion adjusting for sample size (AICc; Burnham & Anderson 1998). Only the model with the lowest AIC is reported. A change in AIC of less than 2 suggests that the models were equivalent. Analyses were conducted in S-Plus 6.1 for Windows statistical program (Insightful Corporation 2001).

V.3 Results

V.3.1 Validation of clipping experiment

The observed mean yield in control plots was 4121.4 kg/ha (± 179.8 SE), and the APSIM model predicted the yield of control plots as 4034.3 kg/ha. The regression explained 78% of the variation between the observed and predicted yields, but the slope 2 was less than one (linear regression = 0.6693x + 972.87; r = 0.7816; F1,15 = 53.69; P < 0.001; Figure V.1). There was some variability between plots and across the field and the APSIM model did not adequately simulate some of the higher yields. The APSIM model, however, correctly simulated all other intensities of damage through the different crop stages. Despite this result, we are reasonably confident that APSIM can adequately simulate the effects of damage by mice.

194

6000 Observed 5000 Predicted ) a

h 4000 (kg/ d l 3000 e i y n 2000 Grai

1000

0 0 5 10 25 50 5 10 25 50 5 10 25 50 5 10 25 50 Control Emergence Max Tillering Booting Pre Harvest Crop stage and % damage

If we extend this relationship to include a range of damage intensities (0-100%) from sowing to harvest, we obtain a response surface whereby one-off damage can potentially be compensated for early in the crop, but then as the crop develops, there is a diminished capacity to compensate for this damage (Figure V.2). There was little or no yield loss at the early stages of crop development with a sharp decline in yield as the intensity of damage increased (>75% damage intensity). The shape of the slope changes from a concave-down at sowing through to a linear slope just prior to harvest.

195

1.0

0.9

0.8

d 0.7

i y

0.6

v i

t 0.5

a l

e 0.4 R 0.3 Sowing 0.2 0.1 Emergence

0.0 e Max Tillering g 10 a 20 Booting 30 st 40 p 50 o 60 r Inten 70 C sity o 80 f dam 90 Pre-harvest age 100 (%)

Figure V.2. Extrapolation of the APSIM model from the validation of the clipping experiment conducted at Ginninderra covering all ranges on intensity of damage (0- 100%) and all growth stages of the wheat crop (sowing to harvest). One-off damage was applied in the APSIM model at different levels of intensity at each crop stage. Yield is expressed relative to a maximum yield of 4.2 t/ha.

V.3.2 Mouse grazing model

Given we were using average yields from different fields and simulations using average soil conditions at the start of the season after a pasture phase, we considered the model was appropriate to adequately simulate crop growth in the area and was therefore appropriate to examine the effect of mouse damage to the crop. The model initially was run to adequately predict wheat yields based on biophysical properties of the Walpeup region and climate and was compared to actual yields of wheat in non-outbreak years (Figure V.3). The mean observed yield was 2046.2 kg/ha (± 132.9 SE) and the mean predicted yield was 1954.7 kg/ha (± 186.4 SE) (for the non-outbreak years of 1985, 1986, 1989-1992, 1995, and 1996). The linear regression explained 62% of the variation between the observed and predicted yields and the slope was close to one (y = 196

2 1.106x – 309.1; r = 0.6222; F1,6 = 9.882; P = 0.020). There was no significant difference between observed and predicted yields (paired-t7 = 0.792; P = 0.455).

3000 Observed 2500 Predicted * ) -1 a

h 2000 * 1500 eld (kg i * y 1000 *

Grain 500 * *

0 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 Year

Figure V.3. Comparisons of yields (kg/ha) observed in the field at the Mallee Research Station and predicted yields from the APSIM model to validate the experimental simulation for 1984-1997. Outbreaks of mice occurred in 1984, 1987/88, 1993/94 and 1997 (indicated by an *), so only yield data for non-outbreak years were analysed.

Outbreaks of mice occurred in 1984, 1987/88, 1993/94, 1997 and 2001 (Figure V.4a). The highest density occurred in June 1984 with an estimated 2,200 mice/ha. The density of mice in non-outbreak years was generally low (<50 mice/ha). The amount of crop eaten by the mouse population (Figure V.4b) and the fraction of crop consumed (Figure V.4c) fluctuated according to the population density of mice at the time, with highest losses in 1984, 1987/88, 1993/94, 1997 and 2001 within a few weeks of sowing and emergence of the crops (sometime during May-July each year). Smaller peaks of damage occurred when the reproductive stage of the crop commenced in September and October each year with mice damaging the developing tillers. Harvesting occurred in November and December.

197

1500 (a) 1250 /ha e 1000

mic 750 of

ity 500

250 Dens 0 (b) ten 3

(kg) 2 -1

ount ea 1 day m

A 0 (c) of

ten 0.2 n tio 0.1 op ea r Frac c 0.0 (d) -1 Yield without mice ) ha

-1 2500 Yield with mice

400 e Mean density of mice ha-1 ha 2000 300 mic 1500 of eld (kg i 200 1000 n y i a 500 100 Gr an density e

0 0 M 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 Year Figure V.4. (a) Changes in mouse density (mice/ha), (b) amount eaten by the mouse population day-1 (kg), (c) fraction of crop eaten by the mouse population, and (d) yields (kg/ha) modelled using the APSIM mouse grazing model with and without mice. The mean density of mice from sowing to harvest also is shown.

The mean simulated yield of wheat without mice for 1983-2003 was 1516.0 kg/ha (± 156.5 SE; range = 148.3 - 2679.2 kg/ha) and the mean simulated yield of wheat with mice was 1398.0 kg/ha (± 170.0 SE; range = 33.5 – 2612.8 kg/ha) (Figure V.4d). The difference between the APSIM model without mice and with mice is the yield loss caused by mice. There was a relative yield loss of 7.8% based on the mean yields over the 21 years. Differences between yields with and without mice are solely due to the grazing effect of mice on the crop, and not from differences in climate (since the same climate data were used). The mean density of mice from sowing to harvest also is shown in Figure 4d. The mean yield loss caused by mouse damage each year was 12.4% (± 5.4 SE). The maximum yield loss occurred in 1984 (95.8%) and the minimum yield loss occurred in 1992 (-0.5%, effectively a yield gain). There were seven years (7/21 years, 198

or one year in three) when the yield loss was greater than 5%, being 1984 (95.8%), 1987 (16.3%), 1988 (11.0%), 1993 (26.9%), 1994 (61.7%), 1997 (33.6%) and 2000 (5.2%). Using the mean density of mice from sowing to harvest as a simple measure of the density of mice over the period of crop growth, we plotted yield loss and density to get the damage/abundance relationship (Figure V.5). A linear regression gives loss = 2 (0.194 x density) – 1.140 (F1,19 = 42.58; P < 0.001; r = 0.6915; AIC = 174.26). The mean population density of mice from sowing to harvest at which 5% loss occurred was 31.6 mice/ha and 50% loss occurred at 263 mice/ha. Conversely, if the density of mice was 50 mice/ha, then the yield loss would be 8.6%. A 3-parameter exponential growth (b*density) curve was also fitted and was in the form: loss = y0 + a(e ). The parameter values 2 were estimated as a = 16.03, b = 0.005, and y0 = -13.88 (F2,20 = 21.50; P < 0.0001; r = 0.7049; AIC = 175.32). The density of mice at which 5% loss occurred was 34 mice/ha and 50% loss occurred at 288 mice/ha, and if the density of mice was 50 mice/ha, then the yield loss would be 6.5%.

100

1984

60 eld loss)

i 1994 y

1997 Damage (% 1993 20

1988 1987

0 2000 2001

0 100 200 300 400

-1 Mean density of mice ha Figure V.5. Relationship between mean mouse density (mice/ha from sowing to harvest) and damage to wheat crops (% yield loss) as estimated using the APSIM mouse grazing model from Walpeup, Victoria, 1983-2003. Years where significant damage occurred or where moderate to high densities of mice were present are indicated. The solid line represents the linear regression and the dashed line represents the exponential relationship using a 3-parameter exponential growth curve (see text for details).

The mean density of mice from sowing to harvest provided a very coarse measure of mouse numbers and did not provide any detail about when damage occurs in 199

relation to different profiles in changes of mouse density over time through the growth of the crop (ie increasing, stable or decreasing). To clarify the relative contribution of the damage (% yield loss) that mice cause at different crop stages, the densities of mice at emergence (~ 10 days after sowing, DAS), panicle initiation (~ 81 DAS), booting (~ 135 DAS) and ripening stages (~168 DAS) were plotted against the percent yield loss (Figure V.6). Three patterns of mouse abundance and yield loss emerged: (1) in one- year mouse outbreaks (1984, 1997 and 2001) and the second-year of two-year outbreaks (1988 and 1994) densities of mice were high during the early crop stages (emergence; >100 mice/ha) and this alone appears to be sufficient for relatively high mouse damage (>5% yield loss), (2) in the first-year of two-year outbreaks (1987, 1993 and 2000) densities of mice were relatively high throughout the duration of crop growth (emergence, panicle initiation, booting and to a lesser extent at ripening; 60-350 mice/ha) resulting in relatively high mouse damage (>5% yield loss), and (3) during non-outbreak years, mouse densities were low (<25 mice/ha) and the subsequent mouse damage was low (<5% yield loss). Therefore, the majority of the damage appears to have occurred early in the crop stage, except in years when densities remained relatively high throughout the growth of the crop, as found in the first year of a two-year outbreak. The damage/abundance relationship was re-examined after excluding data from the first years of two-year outbreaks (1987, 1993 and 2000). The best fit for this simulation was a 3-parameter sigmoid curve and was in the form: loss = a/(1+e-((Density- b)/c)) (Figure V.7). The parameter values were estimated as a = 94.80, b = 28.65, and c = 2 123.80 (F2,15 = 242.73; P < 0.0001; r = 0.970; AIC = 112.60). The density of mice at which 5% loss occurred was 42 mice/ha and 50% loss occurred at 127 mice/ha, and if the density of mice was 50 mice/ha, then the yield loss would be 6.7%.

200

2000 100 Yield loss Density of mice ha-1 at emergence Density of mice ha-1 at panicle Initiation -1 80 1500 Density of mice ha at booting Density of mice ha-1 at ripening -1 Mean density of mice ha-1 d loss) 60 l ce ha e i i y m 1000 % y of t 40 age ( Densi 500 Dam 20

100 0 0 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Year

100 Low density years 2nd year of 2-year outbreak or 1-year outbreak 1984 1st year of 2-year outbreak 80 loss) 60 1994 eld i y

ge (% 40 a 1997

Dam 1993 20

1988 1987

2000 0 2001

0 100 200 300 400

-1 Density of mice ha Figure V.7. Relationship between mean mouse density (mice/ha from sowing to harvest) and damage to wheat crops (% yield loss) as estimated using the APSIM mouse grazing model from Walpeup, Victoria, 1983-2003, for non outbreak years (black circle: 1983, 1985, 1986, 1989-1992, 1995, 1996, 1998, 1999, 2002 and 2003), the 2nd year of 2-year outbreaks or the 1-year outbreaks (grey square: 1984, 1988, 1994, 1997 and 2001), or 1st year of 2-year outbreaks (white triangle: 1987, 1993 and 2000). The solid line represents a 3-parameter sigmoidal curve excluding data for 1st year of a 2-year outbreak (see text for details). 201

The APSIM model was then run with mouse densities ranging from 1 mouse/ha to 1,000 mice/ha, but densities were held constant each day from sowing to harvest to establish a theoretical damage response curve. Climatic variables from 1983 were used. These data revealed a sigmoid curve, so we modelled the data with a 4-parameter (-b(density) c Chapman regression in the form: loss = y0 + a(1-e ) (Figure V.8). The parameter values were estimated as a = 101.81, b = 0.0394, c = 5888.69, and y0 = -1.59 2 (F3,32 = 17720.27; P < 0.0001; r = 0.9994). The inflection point for 5% loss occurred at 194.8 mice/ha and 100% yield loss was reached beyond ~ 300 mice/ha. There were yield gains (up to 5.4%) when the density of mice was 100-150 mice/ha.

100

80 loss)

d l 60 e i y % 40 age ( Dam 20

0 100 200 300 400 500 600

-1 Mean density of mice ha

Figure V.8. Modelled relationship between density of mice and damage (% yield loss) to wheat crops. A range of mouse densities were held constant from sowing to harvest and were run through the APSIM mouse grazing model (dark line) and a 4-parameter Chapman regression was fitted to these data (light line; see text for details).

202

V.4 Discussion

V.4.1 The damage-density relationship

The APSIM model successfully described the relationship between damage caused by mice to wheat crops in the central mallee region of Victoria. This response was initially described by a linear relationship using observed mouse population density data, but when densities of mice at different crop stages was considered and years when mouse densities remained high were excluded a sigmoidal curve appeared to better explain the relationship between mouse density and yield loss. This model suggested that significant damage (>5%) occurred to crops when the density at emergence was >100 mice/ha, and that significant damage occurred when the mean density from sowing to harvest was >42 mice/ha. There were two years (1987 and 1993) that did not fit the sigmoidal relationship, which were described as the 1st year of a two-year outbreak. In both these cases, densities were above 100 mice/ha at emergence, but then densities remained high through panicle initiation, booting and ripening stages and subsequent simulated damage was relatively high (16.3% in 1987 and 26.9% in 1993). A farmer would therefore need to know whether mouse densities would be above 100 mice/ha at sowing in order to implement appropriate rodent control strategies. The shape and form of the damage/density relationship modelled here for mice in wheat crops using APSIM is similar to that developed by Mulungu et al. (2003) who also found a sigmoidal relationship for Mastomys natalensis damage to maize in Tanzania. Like Mus domesticus in Australia, Mastomys cause most damage at sowing prior to emergence, when they dig out the seeds from the ground, and also when the cobs are maturing (Fiedler 1988; Mwanjabe and Leirs 1997; Mwanjabe et al. 2002). Total yield losses to maize, sorghum, paddy rice and pulses was approximately 50% during an outbreak of Mastomys natalensis in Tanzania in 1989 (Mwanjabe et al. 2002). The study of Mulungu et al. (2003) was based on losses of seeds at sowing as their measure of damage, not final yield loss. Mulungu et al. (2003) found there was a plateau of damage at approximately 70%, but the minimum level of damage (= intercept, ie when densities of Mastomys were zero) was approximately 15% and explained that this was the background germination failure rate. Using the APSIM 203

modelling approach as used here for house mice on wheat effectively minimised these sources of variation and looks at the total yield loss, not just damage at sowing. There has been considerable research on quantifying relationships between damage and abundance for invertebrate pests and weeds in crops (Pedigo et al. 1986; Doyle 1991; Nabirye et al. 2003; Zou et al. 2004), but there are only a few studies for rodent pests (Poché et al. 1982; Buckle et al. 1984; Lefebvre et al. 1989; Mulungu et al. 2003). There has, however, been a recent focus on modelling the bioeconomics of these rodent pests (Stenseth et al. 2003, Davis et al. 2004). The dearth of research on the relationship between damage and rodent abundance for rodent pests is surprising considering the scale and extent of rodent damage to crops worldwide (Prakash 1988; Singleton & Petch 1994; Singleton et al. 1999; Stenseth et al. 2003). Serious rodent problems generally occur in the poorer developing countries where farmers’ livelihoods and food security are of concern (Singleton 2003). For invertebrate pests, a linear relationship between damage and abundance is generally described (Pedigo et al. 1986; Nabirye et al. 2003; Zou et al. 2004). This would indicate that little compensation is shown by the crop to the damage or injury. Our results indicate a complex relationship between damage and abundance is likely to be sigmoidal where the compensatory response of wheat is an important component of the relationship. Furthermore, the timing of damage is important. The results from the APSIM model suggest that a linear relationship exists during the later stages of wheat crop growth, when there is little opportunity for the crop to compensate for damage.

V.4.2 Economic injury level for mice in wheat

Having determined the relationship between the density of mice and the damage they cause to wheat crops, we can turn our attention to the economic injury level (EIL; Pedigo et al. 1986) and economic threshold for management action (ET; Ramirez & Saunders 1999) for mice. There has been some early work on developing economic injury levels and economic thresholds for rodents (Buckle et al. 1984; Buckle 1994), but the concept has not been widely applied. Buckle (1994) described a technique slightly different to that provided by Pedigo et al. (1986), which is the method adopted by entomologists for calculating EIL. EILs have been developed for a range of insect pests in a range of different crop types (Nabirye et al. 2003; Torres-Vila et al. 2003). If we 204

assume these insect pests feed solely on the crop to grow and reproduce then they would cause injury and damage. House mice are slightly different because they feed on a range of alternative food sources in addition to the wheat plants (Tann et al. 1991) and could be considered as herbivores on wheat crops. Kotanen and Rosenthal (2000) argued that through differences in size-related ecology and behaviour of vertebrate and invertebrate herbivores there are differences in the type of direct and indirect damage caused to plants. The response of the plant to vertebrate or invertebrate herbivory, therefore would be different. According to the criteria that are listed by Kotanen and Rosenthal (2000), it would seem that house mice would be considered as invertebrates because of their small body size relative to plant size, long duration of herbivore attack (except at sowing, prior to emergence), low rate of incremental damage by individuals during attack, moderate scale of patchiness and uncommon indirect effects. We define EIL here as the density of the pest population at which the economic damage equals the cost of control measures (Pedigo et al. 1986). The information required is the cost per hectare of applying a management action (C; $ /ha), the market value of produce (V; $ /kg), injury units per animal (I; proportion damage/animal/ha), damage per unit injury (D; kg reduction/ha/proportion damaged) and a measure of the proportionate reduction in injury (K, effectiveness of control operation), to give a general model of EIL = C/VIDK (Pedigo et al. 1986). If the relationship between damage and density is linear, the slope of the function (b) can be used to replace IxD in the EIL calculation (so EIL = C/VbK; Nabirye et al. 2003). Using these factors for managing house mice in Australia, we can estimate C = $15 /ha (the cost of applying zinc phosphide by aerial application; the main control method used by farmers to reduce mouse numbers; Brown and Singleton 2002), V = $315 /ha (mean yield of 1.5 tonnes/ha as modelled here without mice multiplied by the farm gate value of wheat of $210 tonne-1; Turner et al. 2001), b = 0.194 (slope of the linear regression of the damage/density relationship) and K = 0.95 (effectiveness of zinc phosphide on mice in wheat crops; Mutze & Sinclair 2004). We then calculate the EIL as 0.26% (a very low value). Even if the effectiveness of the zinc phosphide baiting was 51% (K = 0.51), as reported by Brown et al. (2002), then the EIL would be 0.48%. In either case, these values are far below the value that southern Australian wheat farmers appear to tolerate of around 5% damage (Singleton and Brown unpublished data). Nabirye et al. (2003) reported an EIL of 6.7% for flower thrips on cowpea in eastern Uganda. Some of the 205

limitations of this approach are that: (1) it is not possible to use a sigmoidal relationship to calculate a single EIL; (2) an EIL for mice in wheat does not represent an easily measurable impact of mice (low levels of mouse damage are very difficult to assess in damage surveys (Brown and Singleton 2002; Brown et al. 2004), so a better measure would be the density of mice in the fields through trapping or burrow counts); (3) the timing of control is another factor that is not adequately covered in the EIL calculation. Furthermore, the EIL is likely to vary depending on climatic conditions, soil types etc, but at least a rudimentary estimate of EIL can be developed to assist in the management of mice. Perhaps a better measure for managing mice is to calculate the minimum density of mice that cause financially significant damage to wheat. A farmer normally does not notice damage less than 5% and they normally are reactive in that they do not conduct control until they observe significant damage (Brown and Singleton unpublished data). Using the figures described above, then the cost of control is 4.8% of the potential income ($15/$315). If mice cause 5% yield loss, then the final income would be $299.25 (a $15.75 reduction). Therefore, any control effort conducted by the farmer would need to reduce the mouse density to less than 35 mice/ha if a linear relationship is used. The sigmoidal relationship gave 5% yield loss at 53 mice/ha, therefore the critical threshold is likely to be 35-50 mice/ha. These findings emphasise the importance of monitoring the densities of mouse populations, particularly prior to sowing, to determine if management is warranted. Furthermore, the APSIM model suggests that damage is likely to be above 5% one year out of three, so farmers need to be aware of the possible impact of mice on yield reduction and to conduct preventative rather than crisis management. There is a range of preventative mouse control options available to farmers other than to apply poison bait (see Caughley et al. 1998; Brown et al. 2004). There are two approaches to validate this model: (1) collect field data covering a wide range of densities of mice and determine damage, injury and yield loss, and (2) exclude mice from replicated fenced plots at sowing and assess differences in yield to areas where mice have access to the crop and regularly monitor these mouse populations. A similar approach using an exclusion fence was tried for an experiment to examine the effect of sowing depth on damage caused by mice (Brown et al. 2003), but the plastic sheeting was removed about 1-month after emergence and other animals 206

such as European hares caused significant damage to plots, so the wheat could not be harvested.

V.4.3 Further application of the APSIM model

This basic APSIM model could be used to examine the benefits in yield response after rodent control has been applied. An initial run of this model was conducted on the Walpeup population data from 1997 and data from a study by Brown et al. (2002). The initial mean density of mice from sowing to harvest was 88 mice/ha. Zinc phosphide was applied to a field prior to sowing and we compare the results for an effective baiting that reduced the mouse density by 0, 50 and 95% (we assumed no immigration from untreated areas or increases in density through breeding). Predicted yields were 548.3 kg/ha (0% density reduction; 66% of yield without mice), 913.8 kg/ha (50% density reduction; 111% of yield without mice) and 833.1 kg/ha (95% density reduction; 101% of yield without mice). There was an increase in yield with a 50% reduction in density of mice. By manipulating the effectiveness of baiting further, significant damage occurred when the effectiveness of baiting was less than 20% and the mean density of mice was greater than 75 mice/ha (>5% yield loss). Therefore, under the conditions in 1997, any control method that reduced the density of mice by more than 20% (to mean densities less than 75 mice/ha) resulted in low levels of damage (potential yield gain). Further work is required to test this model under field conditions and to test its suitability for managing mice. This approach using the APSIM model could also be used to examine the benefits of using fertility control of mice (Chambers et al. 1999; Farroway et al. 2002; Singleton et al. 2002). This basic model can also be used to examine the relationships between relative yield and pest damage (fraction of crop eaten as used here) and between pest damage and pest abundance following the process described by Hone (2004). Our data gives a linear relationship between relative yield and pest damage (fraction of crop eaten) with a suggestion of a concave-down relationship (linear regression: yield = 102.62 – (19.69 2 x damage); r = 0.754; F1,19 = 58.21; P < 0.001) (Figure V.9a), whereas Hone (2004) predicted this curve has a concave-down relationship representing compensation. Furthermore, our data give a sigmoidal relationship between pest damage (fraction of crop eaten) and pest abundance (best fit was a 4-parameter sigmoid function; damage = 207

(–(density – 47.91)/1.20) 2 0.08 + 2.16/(1 + e ); r = 0.856; F3,17 = 33.57; P < 0.001) (Figure V.9b), whereas Hone (2004) gave this as a linear relationship.

)

100 (a) ce 4 (b) mi y b

80 3 ) % (

d 60 l crop eaten

e f o

yi 2 n o ve i t 40 cti a l a e r f R (

e 1 20 mag da

0 st 0 Pe 01234 0 100 200 300 400 500 Pest damage (fraction of crop eaten by mice) Density of mice ha-1

Figure V.9. Modelled relationships between (a) pest damage (fraction on crop eaten) and relative yield and (b) mouse density (mice/ha) and pest damage (fraction on crop eaten) using the Walpeup-APSIM model.

There appears to be some overcompensation of the wheat crop at low mouse densities. Some evidence for this occurs in the model where densities of mice were manipulated, where yields appeared to increase by up to 5% when densities were 100- 150 mice/ha. However, it would be unlikely that such moderately high densities of mice could be maintained in field conditions, so the apparent overcompensation may be purely an artefact of the model. Overcompensation was suggested by Brown (2005) in a field experiment that manipulated levels of damage to the crop, and could occur through the early reduction of canopy vigour leading to more water available for grain filling. A similar response has been shown for skip-row cropping where yields were similar to conventional cropping (Fukai & Foale 1988; Whaley et al. 2000). Complete or overcompensation for small or moderate levels of damage have been found under certain environmental conditions for head-clipping weevil, Haplorhynchites aeneus, on sunflowers (Pilson & Decker 2002), experimental clipping of sunflower (Moriondo et al. 2003), brown planthopper, Nilaparvata lugens, on rice (Rubia-Sanchez et al. 1999), experimental clipping of grass, Bouteloua gracilis (Kotanen & Bergelson 2000), mule deer and elk feeding on scarlet gilia, Impomopsis aggregata (Paige & Whitham 1987), 208

and for experimental clipping of the field gentian, Gentianella campestris (Lennartsson et al. 1998). It is plausible that low densities of mice could increase wheat yields.

V.4.4 Summary

The APSIM model was used to describe the growth of wheat crops in northwestern Victoria, an area subjected to mouse plagues. By using long-term data on the abundance of mice, we have incorporated the effects of these mice by simulating their damage to the wheat crops. A damage-density relationship was then established, and the data suggested a sigmoidal relationship when different categories of mouse outbreaks were considered. This was further supported with a sigmoidal relationship when densities of mice were manipulated. This modelling approach has proved useful in establishing a theoretical relationship, but field data are required to validate the model. There are potentially many uses of this model such as modelling the response of a wheat crop to the management of mice and to include the benefits of fertility control of mice. The model could also be incorporated into predictive models to examine scenarios based on likely densities of mice and look at the benefits and costs of a range of control strategies.

V.5 Acknowledgements

We sincerely thank Peter Carberry, Jim Hone, Mark Howden and Tony Arthur for discussions and comments on this work.

V.6 References

Advani, R. and Mathur, R. P. (1982). Experimental reduction of rodent damage to vegetable crops in Indian villages. Agro-Ecosystems 8, 39-45. Brooks, J. E., Sultana, P., and Poché, R. M. (1985). Damage by Bandicota bengalensis to growing wheat in Bangladesh and methods of control. Acta Zoologica Fennica 173, 135-137. 209

Brown, P. R. (2005). The effect of simulated house mouse damage to wheat in Australia. Crop Protection 24, 101-109. Brown, P. R., Chambers, L. K., and Singleton, G. R. (2002). Pre-sowing control of house mice (Mus domesticus) using zinc phosphide: efficacy and potential non- target effects. Wildlife Research 29, 27-37. Brown, P. R., Davies, M. J., Croft, J. D., and Singleton, G. R. (2004). Can farm management practices reduce the impact of house mouse populations on crops in an irrigated farming system? Wildlife Research 31, 597-604. Brown, P. R. and Singleton, G. R. (1999). Rate of increase as a function of rainfall for house mouse Mus domesticus populations in a cereal-growing region in southern Australia. Journal of Applied Ecology 36, 484-493. Brown, P. R. and Singleton, G. R. (2002). Impacts of house mice on crops in Australia - costs and damage. In Human Conflicts with Wildlife: Economic Considerations. (Eds. L. Clark, J. Hone, J. A. Shivik, R. A. Watkins, K. C. VerCauteren, and J. K. Yoder.) pp. 48-58. (National Wildlife Research Center: Fort Collins, Colorado.) Brown, P. R., Singleton, G. R., Kearns, B., and Griffiths, J. (1997). Evaluation and cost- effectiveness of strychnine for control of populations of wild house mice (Mus domesticus) in Victoria. Wildlife Research 24, 159-172. Brown, P. R., Singleton, G. R., Tann, C. R., and Mock, I. (2003). Increasing sowing depth to reduce mouse damage to winter crops. Crop Protection 22, 653-660. Buckle, A. P. (1994). Damage assessment and damage surveys. In Rodent Pests and their Control. (Eds. A. P. Buckle and R. H. Smith.) pp. 219-248. (CAB International: Wallingford.) Buckle, A. P., Huson, L. W., and Rowe, F. P. (1984). The frequency distribution of rat damage in rice fields and its implications for the cost-benefit of control programmes. Proceedings of a Conference on The Organisation and Practice of Vertebrate Pest Control. Elvetham Hall, Hampshire, England. Surrey, England. ICI Plant Protection Division. pp. 289-296. Buckle, A. P. and Smith, R. H. (1994). Rodent Pests and Their Control. (CAB International: Wallingford.) 210

Buckle, A. P., Yong, Y. C., and Abdul Raman, H. (1985). Damage by rats to rice in South east Asia with special reference to an integrated management scheme proposed for Peninsular Malaysia. Acta Zoologica Fennica 173, 139-144. Buckle, A. P., Yong, Y. C., and Rowe, F. P. (1979). Yield response of the rice variety Improved Mashuri to simulated rat damage. Malaysian Agriculture Journal 52, 135-144. Burnham, K. P. and Anderson, D. R. (1998). Model Selection and Inference: A Practical Information-Theoretic Approach. 353 pp. (Springer-Verlag: New York.) Caughley, G. (1977). Analysis of Vertebrate Populations. (John Wiley and Sons: London.) Caughley, J., Bomford, M., Parker, B., Sinclair, R., Griffiths, J., and Kelly, D. (1998). Managing Vertebrate Pests: Rodents. 130 pp. (Bureau of Resource Sciences and Grains Research and Development Corporation: Canberra.) Caughley, J., Monamy, V., and Heiden, K. (1994). Impact of the 1993 mouse plague. GRDC Occasional Paper Series No.7. (GRDC: Canberra, Australia.) Chambers, L. K., Singleton, G. R., and Hinds, L. A. (1999). Fertility of wild mouse populations: the effects of hormonal competence and an imposed level of sterility. Wildlife Research 26, 579-591. Cheeroo-Nayamuth, F. C., Robertson, M. J., Wegener, M. K., and Nayamuth, A. R. H. (2000). Using a simulation model to assess potential and attainable sugar yield in Mauritus. Field Crops Research 66, 225-243. Davis, S. A., Akison, L. K., Farroway, L. N., Singleton, G. R., and Leslie, K. E. (2003). Abundance estimators and truth: accounting for individual heterogeneity in wild house mice. Journal of Wildlife Management 67, 634-645. Davis, S. A., Leirs, H., Pech, R. P., Zhang, Z., and Stenseth, N. C. (2004). On the economic benefit of predicting rodent outbreaks in agricultural systems. Crop Protection 23, 305-314. Doyle, C. J. (1991). Mathematical models in weed management. Crop Protection 10, 432-444. Farroway, L. N., Singleton, G. R., Lawson, M. A., and Jones, D. A. (2002). The impact of murine cytomegalovirus (MCMV) on enclosure populations of house mice (Mus domesticus). Wildlife Research 29, 11-17. 211

Fiedler, L. A. (1988). Rodent problems in Africa. In Rodent Pest Management. (Ed. I. Prakash.) pp. 35-65. (CRC Press: Boca Raton, Florida.) Fukai, S. and Foale, M. A. (1988). Effects of row spacing on growth and grain yield of early and late sorghum cultivars. Australian Journal of Experimental Agriculture 28, 771-777. Gillespie, G. D. (1985). Feeding behaviour and impact of ducks on ripening barley crops grown in Otago, New Zealand. Journal of Applied Ecology 22, 347-356. Goryńska, W. (1981). Method of determining relations between the extent of damage in farm crops, big game numbers, and environmental conditions. Acta Theriologica 26, 469-481. Hammer, G. L., Holzworth, D., and Stone, R. (1996). The value of skill in seasonal climate forecasting to wheat crop management in a region with high climate variability. Australian Journal of Agricultural Research 47, 717-737. Hone, J. (1994). Analysis of Vertebrate Pest Control. (Cambridge University Press: Cambridge.) Hone, J. (2004). Yield, compensation and fertility control: a model for vertebrate pests. Wildlife Research 31, 357-368. Insightful Corporation (2001). Splus 6.1 for Windows Unsers' Guide. (Insightful Corporation: Seattle, WA.) Islam, Z., Morton, R. G., and Jupp, B. P. (1993). The effects of rat damage on deepwater rice yields in Bangladesh. International Journal of Pest Management 39, 250-254. Jacob, J., Sudarmaji, Singleton G. R., Rahmini, Herawati, N. A., Brown, P. R., and Muller, W. J. (Submitted). Ecologically-based management of rodents in the real world: application to a rice-based agro-ecosystem in Indonesia. Ecological Applications. Keating, B. A., Carberry, P. S., Hammer, G. L., Probert, M. E., Robertson, M. J., Holzworth, D., Huth, N. I., Hargreaves, J. N. G., Meinke, H., Hochman, Z., McLean, G., Verburg, K., Snow, V., Dimes, J. P., Silburn, M., Wang, E., Brown, S., Bristow, K. L., Asseng, S., Champan, S., McCown, R. L., Freebairn, D. M., and Smith, C. J. (2003). An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy 18, 267-288. 212

Keating, B. A. and Meinke, H. (1998). Assessing exceptional drought with a cropping systems simulator: a case study for grain production in northeast Australia. Agricultural Systems 57, 315-332. Key, G. (1990). Pre-harvest crop losses to the African striped ground squirrel, Xerus erythropus, in Kenya. Tropical Pest Management 36, 223-229. Khokhar, A. R., Rizvi, S. W. A., and Shaukat, S. S. (1993). Rodent damage and its effects on yield loss in different rice varieties in Punjab, Pakistan. International Journal of Pest Management 39, 397-399. Kotanen, P. M. and Bergelson, J. (2000). Effects of simulated grazing on different genotypes of Bouteloua gracilis: how important is morphology? Oecologia 123, 66-74. Kotanen, P. M. and Rosenthal, J. P. (2000). Tolerating herbivory: does the plant care if the herbivore has a backbone? Evolutionary Ecology 14, 537-549. Lefebvre, L. W., Engeman, R. M., Decker, D. G., and Holler, N. R. (1989). Relationship of roof rat population indices with damage to sugarcane. Wildlife Society Bulletin 17, 41-45. Lennartsson, T., Nilsson, P., and Toumi, J. (1998). Induction of overcompensation in the field gentain, Gentianella campestris. Ecology 79, 1061-1072. McCown, R. L., Hammer, G. L., Hargreaves, J. N. G., Holzworth, D., and Freebairn, D. M. (1996). APSIM: a novel software system for model development, model testing and simulation in agricultural research. Agricultural Systems 50, 255- 271. Moriondo, M., Orlandini, S., and Villalobos, F. J. (2003). Modelling compensatory effects of defoliation on leaf area growth and biomass of sunflower (Helianthus annus L.). European Journal of Agronomy 19, 161-171. Moro, D. and Bradshaw, S. D. (2002). Diets and predictions of feeding rate of house mice and Lakeland Downs short-tailed mice inhabiting an arid-zone island in Western Australia. Australian Journal of Zoology 50, 249-265. Muchow, R. C. and Keating, B. A. (1998). Assessing irrigation requirements in the Ord Sugar Industry using a simulation modelling approach. Australian Journal of Experimental Agriculture 38, 345-354. Mulungu, L. S., Makundi, R. H., Leirs, H., Massawe, A. W., Vibe-Petersen, S., and Stenseth, N. C. (2003). The rodent density-damage function in maize fields at an 213

early growth stage. In Rats, Mice and People: Rodent Biology and Management. ACIAR Monograph 96. (Eds. Singleton. G.R., L. A. Hinds, C. J. Krebs, and D. M. Spratt.) pp. 301-303. (ACIAR: Canberra.) Mutze, G. J. (1991). Mouse plagues in South Australian cereal-growing areas. III. Changes in mouse abundance during plague and non-plague years, and the role of refugia. Wildlife Research 18, 593-604. Mutze, G. J. (1993). Cost-effectiveness of poison bait trails for control of house mice in mallee cereal crops. Wildlife Research 20, 445-456. Mutze, G. J. (1998). The 1993 strychnine baiting program for mouse control in South Australian grain crops. I. Efficacy. Wildlife Research 25, 533-546. Mutze, G. J., Green, B., and Newgrain, K. (1991). Water flux and energy use in wild house mice (Mus domesticus) and the impact of seasonal aridity on breeding and population levels. Oecologia 88, 529-538. Mutze, G. J. and Sinclair, R. (2004). Efficacy of zinc phosphide, stychnine and chlorpyrifos as rodenticides for the control of house mice in South Australian cereal crops. Wildlife Research 31, 249-257. Mwanjabe, P. S. and Leirs, H. (1997). An early warning system for IPM-based rodent control in smallholder farming systems in Tanzania. Belgian Journal of Zoology 127, 49-58. Mwanjabe, P. S., Sirima, F. B., and Lusingu, J. (2002). Crop losses due to outbreaks of Mastomys natalensis (Smith, 1834) Muridae, Rodentia, in the Lindi Region of Tanzania. International Biodeterioration & Biodegradation 49, 133-137. Nabirye, J., Nampala, P., Kyamanywa, S., Ogenga-Latigo, M. W., Wilson, H., and Adopala, E. (2003). Determination of damage-yield loss relationships and economic injury levels of flower thrips on cowpea in eastern Uganda. Crop Protection 22, 911-915. Nolte, D. L. and Dykzeul, M. (2002). Wildlife impacts on forest resources. In Human Conflicts with Wildlife: Economic Considerations. (Eds. L. Clark, J. Hone, J. A. Shivik, R. A. Watkins, K. C. VerCauteren, and J. K. Yoder.) pp. 163-168. (National Wildlife Research Center: Fort Collins, Colorado.) O'Leary, G. J. and Conner, D. J. (1996). A simulation model of the wheat crop in response to water and nitrogen supply: II Model validation. Agricultural Systems 52, 31-55. 214

Paige, K. N. and Whitham, T. G. (1987). Overcompensation in response to mammalian herbivory: the advantage of being eaten. The American Naturalist 129, 407-416. Pedigo, L. P., Hutchins, S. H., and Higley, L. G. (1986). Economic injury levels in theory and practice. Annual Review of Entomology 31, 341-368. Pilson, D. and Decker, K. (2002). Compensation for herbivory in wild sunflower: response to simulated damage by the head-clipping weevil. Ecology 83, 3097- 3107. Pinnschmidt, H. O., Batchelor, W. D., and Teng, P. S. (1995). Simulation of multiple species pest damage in rice using CERES-rice. Agricultural Systems 48, 193- 222. Poché, R. M., Main, M Y., Haque, M. E., and Sultana, P. (1982). Rodent damage and burrowing characteristics in Bangladesh wheat fields. Journal of Wildlife Management 46, 139-147. Prakash, I. (1988). Rodent Pest Management. (CRC Press: Boca Raton, Florida.) Probert, M. E., Carberry, P. S., McCown, R. L., and Turpin, J. E. (1998). Simulation of legume-cereal systems using APSIM. Australian Journal of Agricultural Research 49, 317-327. Putman, R. J. and Moore, N. P. (1998). Impact of deer in lowland Britain on agriculture, forestry and conservation habitats. Mammal Review 28, 141-164. Ramirez, O. A. and Saunders, J. L. (1999). Estimating economic thresholds for pest control: an alternative procedure. Journal of Economic Entomology 92, 391-401. Rao, K. S., Maikhuri, R. K., Nautiyal, S., and Saxena, K. G. (2002). Crop damage and livestock depredation by wildlife: a case study from Nanda Devi Biosphere Reserve, India. Journal of Environmental Management 66, 317-327. Redhead, T. D. (1988). Prevention of plagues of house mice in rural Australia. In Rodent Pest Management. (Ed. I. Prakash.) pp. 191-206. (CRC Press: Boca Raton, Florida.) Robertson, M. J., Carberry, P. S., and Lucy, M. (2000). Evaluation of a new cropping option using a participatory approach with on-farm monitoring and simulation: a case study of spring-grown mungbeans. Australian Journal of Agricultural Research 51, 1-12. 215

Rubia-Sanchez, E., Suzuki, Y., Miyamoto, K., and Watanabe, T. (1999). The potential for compensation of the effects of the brown planthopper Nilaparvata lugens Stal (Homoptera: Delphacidae) feeding on rice. Crop Protection 18, 39-45. Saunders, G. R. and Robards, G. E. (1983). Economic considerations of mouse-plague control in irrigated sunflower crops. Crop Protection 2, 153-158. Sheikher, C. and Jain, S. D. (1997). Rodents in cauliflower and cabbage: population, damage and control. International Journal of Pest Management 43, 63-69. Singleton, G. R. (1989). Population dynamics of an outbreak of house mice (Mus domesticus) in the mallee wheatlands of Australia - hypothesis of plague formation. Journal of Zoology, London 219, 495-515. Singleton, G. R. (2003). Impacts of Rodents on Rice Production in Asia. IRRI Discussion Paper Series No. 43. International Rice Research Institute, Los Bańos, Philippines. pp. 30. Singleton, G. R., Brown, P. R., Pech, R. P., Jacob, J., Mutze, G. J., and Krebs, C. J. (2005a). One hundred years of eruptions of house mice in Australia - a natural biological curio. Biological Journal of the Linnaen Society 84, 617-627. Singleton, G. R., Farroway, L. N., Chambers, L. K., Lawson, M. A., Smith, A. L., and Hinds, L. A. (2002). Ecological basis for fertility control in the house mouse (Mus domesticus) using immunocontraceptive vaccines. Reproduction Supplement 60, 31-39. Singleton, G. R., Hinds, L. A., Leirs, H., and Zhang, Z. (1999). Ecologically-based Management of Rodent Pests. ACIAR Monograph No. 59. 494 pp. (Australian Centre for International Agricultural Research: Canberra.) Singleton, G. R., Krebs, C. J., Davis, S. A., Chambers, L. K., and Brown, P. R. (2001). Reproductive changes in fluctuating house mouse populations in southeastern Australia. Proceedings of the Royal Society of London Series B 268, 1741-1748. Singleton, G. R. and Petch, D. A. (1994). A Review of the Biology and Management of Rodent Pests in Southeast Asia. ACIAR Technical Reports No. 30. ACIAR, Canberra. pp. 65. Singleton, G. R. and Redhead, T. D. (1989). House mouse plagues. In Mediterranean Landscapes in Australia: Mallee Ecosystems and their Management. (Eds. J. C. Noble and R. A. Bradstock.) pp. 418-433. (CSIRO: Melbourne.) 216

Singleton, G. R., Sudarmaji, Jacob, J., and Krebs, C. J. (2005b). Integrated management to reduce rodent damage to lowland rice crops in Indonesia. Agriculture, Ecosystems and Environment 107, 75-82. Singleton, G. R., Twigg, L. E., Weaver, K. E., and Kay, B. J. (1991). Evaluation of bromadiolone against house mice (Mus domesticus) populations in irrigated soybean crops. II. Economics. Wildlife Research 18, 275-284. Stenseth, N. C., Leirs, H., Skonhoft, A., Davis, S. A., Pech, R. P., Andreassen, H. P., Singleton. G. R., Lima, M., Machangu, R. M., Makundi, R. H., Zhang, Z., Brown, P. R., Shi, D., and Wan, X. (2003). Mice, rats and people: the bioeconomics of agricultural rodent pests. Frontiers in Ecology and the Environment 1, 367-375. Tann, C. R., Singleton, G. R., and Coman, B. C. (1991). Diet of the house mouse (Mus domesticus) in the mallee wheatlands of north-western Victoria. Wildlife Research 18, 1-12. Tobin, M. E., Koehler, A. E., and Sugihara, R. T. (1997). Effects of simulated rat damage on yields of macadamia trees. Crop Protection 16, 203-208. Torres-Vila, L. M., Rodríguez-Molina, M. C., and Lacasa-Plasencia, A. (2003). Impact of Helicoverpa armigera larval density and crop phenology on yield and quality losses in processing tomato: developing fruit count-based damage thresholds for IPM decision-making. Crop Protection 22, 521-532. Tristiani, H., Murakami, O., and Kuno, E. (2000). Rice plant damage distribution and home range distribution of the ricefield rat Rattus argentiventer (Rodentia: Muridae). Belgian Journal of Zoology 130, 83-91. Turner, S., Barrett, D., Beasley, A., and Gleeson, T. (2001). Outlook for Australian grains industry, 2001-02. Agriculture Australia 2001 Conference. Melbourne. Canberra. Australian Bureau of Agricultural and Resource Economics. Twigg, L. E., Singleton, G. R., and Kay, B. J (1991). Evaluation of bromadiolone against house mice (Mus domesticus) populations in irrigated soybean crops. I. Efficacy of control. Wildlife Research 18, 265-274. Wang, E., van Oosterom, E. J., Meinke, H., Asseng, S., Robertson, M. J., Huth, N. I., Keating, B. A., and Probert, M.E. (2003). The new APSIM-Wheat model – performance and future improvements. Solutions for a better environment: Proceedings of the 11th Australian Agronomy Conference, Geelong, Victoria. 217

Available on the World Wide Web: http://www.regional.org.au/au/asa/2003/p/2/wang.htm Whaley, J. M., Sparkes, D. L., Foulkes, M. J., Spink, J. H., Semere, T., and Scott, R. K. (2000). The physiological response of winter wheat to reductions in plant density. Annals of Applied Biology 137, 165-177. Whisson, D. (1996). The effects of two agricultural techniques on populations of the canefield rat (Rattus sordidus) in sugarcane crops of north Queensland. Wildlife Research 23, 589-604. White, J., Horskins, K., and Wilson, J. (1998). The control of rodent damage in Australian macadamia orchards by manipulation of adjacent non-crop habitats. Crop Protection 17, 353-357. Wood, B. J. and Liau, S. S. (1984). A long-term study of Rattus tiomanicus populations in an oil palm plantation in Johore, Malaysia. II. Recovery from control and economic aspects. Journal of Applied Ecology 21, 465-472. Yunusa, I. A. M., Bellotti, W. D., Moore, A. D., Probert, M. E., Baldock, J. A., and Miyan, S. M. (2004). An exploratory evaluation of APSIM to simulate growth and yield processes for winter cereals in rotation systems in South Australia. Australian Journal of Experimental Agriculture 44, 787-800. Zou, L., Stout, M. J., and Ring, D. R. (2004). Density-yield relationships for rice water weevil on rice for different varieties and under different water management regimes. Crop Protection 23, 534-550.