Distribution of Phytophthora cinnamomi at different spatial scales: When can a negative result be considered positively?

Introduction

Phytophthora cinnamomi is a destructive and widespread soil-borne pathogen that infects woody plant hosts (Zentmyer 1980). It has been recorded from all Australian States and Territories, and is known to cause loss of native species from plant communities in temperate and subtropical areas. Under appropriate conditions P. cinnamomi can cause large-scale ecological damage. Studies in Victoria (Kennedy & Weste 1986; Weste & Marks 1987), Tasmania (Barker & Wardlaw 1995), Western Australia (Shearer & Dillon 1996), South Australia (Davison 1970; Davison & Brumbieris 1973) and Queensland (Brown 1999) have all demonstrated devastating effects of P. cinnamomi on the ecology and conservation of native flora (Gadek 1999) and fauna (Er 1997; Newell 1997). Phytophthora cinnamomi was recognized by the Australian Federal Government as one of only five key threatening processes originally listed under Schedule 3 of the Endangered Species Protection Act 1992 (Environment Australia 1999).

In October 1999, patches of dead and dying trees were identified in the rainforest (simple notophyll vine forest) from air photographic interpretations throughout the Tully Falls area in north Queensland. These patches were scattered across the region and putatively attributed to P. cinnamomi infection. In contrast to southern ecosystems, the behaviour of patch death (in general) and P. cinnamomi (in particular) in tropical environments is poorly understood. There is only one study of the distribution and effects of P. cinnamomi in tropical rainforest in Australia conducted between 1975 and 1982 (Brown 1982). That study examined 3019 randomly collected soil samples in logged and intact forest between Ingham and Cooktown, Queensland, detecting P. cinnamomi in approximately one-third of the samples taken under rainforest canopies. Brown (1982, 1999) suggested that the pathogen was widely distributed, but not ubiquitous, across the sites he sampled.

Conclusions about ecological pattern and the distribution of organisms in space (or time) are, however, not free from observer bias. Indeed, because the observed variability of any system is dependent to some extent on the scale of the description, the method of sampling can often determine the pattern detected (Havry et al. 1978; Hoekstra et al. 1991; Levin 1992). Descriptions of method and scale have generally been poorly reported in studies of the distribution of P. cinnamomi. For example, few recent studies actually outline the size of the sample or the spatial arrangement of the sampling regime. Lack of evidence for any systematic protocol for soil sample collection suggests that earlier studies have concentrated on determining single positive responses that are interpreted to indicate the presence of the pathogen within entire ‘sites’. These studies have assessed P. cinnamomi distribution at single scales, corresponding to a ‘user-defined’ site size. We are unaware of any study that has examined the distribution of P. cinnamomi without averaging within-site variability in this manner. Thus, we know of no estimate of the distribution of P. cinnamomi at very fine scales. With this in mind, the present study was designed to examine the distribution of P. cinnamomi at both broad and fine spatial scales.

Methods

We detected 152 patches (total area = 2618 ha) of apparent canopy death or thinning from aerial photography, rectified against the 1: 50 000 Tully Falls topographic map sheet. Each patch did not define an area of uniform impact or effect. Rather, the patch boundaries delineated areas within which canopy loss could be distinguished. This did not necessarily correspond to 100% canopy loss (D. Stanton, pers. comm.). Site locations were then transferred to a geographic information system (GIS) database and analysed against vegetation and substrate (Gillieson et al. in press). Patches were found to be more or less confined to a single rainforest type (simple notophyll vine forest) on a single substrate type (granite). We used this information to limit all sampling to sites meeting these criteria.

We assessed the presence/absence of P. cinnamomi following a ‘paired-site’ sampling design. Seven patch death sites were matched to seven control sites. The patch death sites were spread across the Tully Falls 1:50 000 topographic map sheet and were randomly selected from the subset of accessible patches with canopy loss. Control sites matched each patch death site. These were within 1 km from the matched patch death sites, a similar distance from roads and watercourses and had similar slope and aspect.

The centre point of each site, located using a hand held global positioning system (GPS) receiver, was used as the centre of a circular quadrat (20-m radius). The quadrat contained two 40-m perpendicular line transects. Soil samples of approximately 250 g were taken at the central point and at 1-m intervals along each transect, resulting in 81 samples per site.

We took soil samples following Shearer and Dillon (1996) by scraping leaf litter from the soil and digging an 80-mm diameter × 100-mm hole. We did not use stratified sampling of soil depth classes (Kennedy & Weste 1986). Soil samples were placed in individual sealed plastic bags. Trowels were sterilized between each sample by immersion in 5% biodegradable bleach and rinsing in fresh water.

We tested soil samples for the presence of P. cinnamomi using the blue lupin baiting technique of Chee and Newhook (1965) as modified by Pratt and Heather (1972). Lupin roots showing signs of infection were plated onto 4% water agar containing 50 p.p.m. streptomycin sulphate. This technique isolates P. cinnamomi rather than other Phytophthora species (Ribiero 1978). Microscopic examination after plating was used to verify the presence of P. cinnamomi. To assess the likelihood of cross-contamination, 10 control samples were randomly placed among the 81 samples from each site.

We first compared the total number of positive responses in patch death sites and control sites. We used paired t-tests to compare the mean number of positive responses between matched sites. We also investigated the pattern of distribution within each patch death and control site by performing a simple runs test to detect deviation from random in the distribution of positive responses. We tested each transect independently, and compared the number of transects in patch death and control sites that displayed aggregated, random and over-dispersed distributions (Zar 1984).

Results

Of the 1134 field samples plated, 783 were positive for P. cinnamomi, and P. cinnamomi was detected in every patch death and control site. All 140 control samples were negative for P. cinnamomi infection suggesting that cross-contamination rarely occurred in the laboratory.

The mean number of samples per site showing positive responses was not significantly greater in patch death sites (58.71, SE = 7.99) than in control sites (46.14, SE = 2.18; paired t-test: t = 1.4, P = 0.21).

The pattern of distribution of P. cinnamomi at 1-m intervals along transects was most commonly random for both patch death (12 tests) and control sites (11 tests; Table 1). The five significant deviations all suggested aggregation rather than over-dispersion.

Discussion

Phytophthora cinnamomi was present in all seven patch death and all seven control sites sampled across the Tully Falls area. We found no significant difference in the total number of positive responses in samples taken from patch death and unaffected sites, suggesting that P. cinnamomi does not occur in greater concentration in patch death sites. Taken together, these two results suggest a uniform distribution of the pathogen at this spatial scale (1300 m² sites) and on this vegetation type and substrate. An implication is that the pathogen can be present within the landscape without causing visible signs of patch death at any particular time. Wilson et al. (2000) also suggested the presence of P. cinnamomi need not coincide with patch death, because four sites in the Eastern Otway Ranges in Victoria, with positive responses for P. cinnamomi, had vegetation that appeared healthy. This should not, however, be interpreted to suggest that locations where P. cinnamomi is present without apparent effects will remain free from canopy disease. Long-term studies in Victoria have shown episodic resurgence of canopy death over a period of 30 years (Weste & Ashton 1994; Weste 1997), leading Weste (1997) to assert that disease and recovery constitute recurring cycles.

In contrast to the broad-scale uniform pattern, when the resolving power of the sampling unit was 1-m intervals, the distribution of P. cinnamomi was random in both patch death and in control sites. The observed pattern therefore appears dependent upon the scale of sampling and is defined by the scale of the observations (Meentemeyer 1989; O'Neill et al. 1989; Fox 1992; Ludwig et al. 2000).

That results are different depending on scale raises important implications for comparisons of distribution between studies. The correct description of pattern can only be made when there is confidence in the ability of the sampling strategy to assess the distribution. Our estimate for the mean number of positive and negative responses can be used to determine the appropriate number of samples needed to test for the presence of P. cinnamomi in sites of similar size to transects used here. For example, 56.6% of samples taken in control sites returned positive records. A single sample has a probability of producing negative evidence for P. cinnamomi of 43%, even though it may be present in the immediate area. Thus, three or four samples are needed to reduce the chance of a false negative to less than 5%.

An earlier survey of P. cinnamomi distribution in the wet tropics inferred a wide, but not ubiquitous, distribution (Brown 1982). This pattern of distribution was based on the absence of P. cinnamomi in two thirds of the soil samples collected across a large proportion of the wet tropics area. However, Brown (1982, 1999) examined, on average, approximately two samples per location; less than our estimated required sample size of three to four. He thus may have overestimated the number of negative locations.

Strategies for soil sampling to detect P. cinnamomi are poorly detailed in the literature. Our approach in taking a definable quantity of soil in a spatially explicit arrangement was an attempt to standardize this procedure. We recommend that future studies be similarly designed to detect variance in the spatial distribution of this pathogen. This approach would allow comprehensive comparisons to be made between the spatial extent of P. cinnamomi and the spatial extent of patch death between vegetation types.

Acknowledgements

This project was partially funded by the Wet Tropics Management Authority and was performed as partial fulfilment of the requirement of J. Pryce's BSc Honours degree. We thank D. Gillieson and J. Landsberg for discussion on early drafts of the manuscript. We also thank T. Roberts, B. Paulus, R. Jensen and S. McKenna for assistance in the field and North Queensland TAFE for use of a laminar flow cabinet. The editor plus two anonymous reviewers helped to clarify the manuscript greatly.