Introduction

ʽŌhiʽa lehua (Metrosideros polymorpha) is the most common and widely distributed native forest tree in the Hawaiian Islands. It provides essential habitat to a diverse assemblage of native invertebrates, plants, and forest birds from sea level to tree line, has enormous cultural significance among native Hawaiians, and is essential for maintaining healthy watersheds (Mueller-Dumbois et al. 2016). Consequently, threats to this keystone species from invasive disease, insect pests, and non-native competitors can lead to devastating impacts on biodiversity and the functional integrity of native ecosystems.

Several recent introductions of fungal pathogens to the Hawaiian Islands threaten ‘ōhi‘a and other island endemics. These pathogens include the pandemic biotype of Austropuccinia psidii (Uchida et al. 2006; Stewart et al. 2018), and Ceratocystis lukuohia and Ceratocystis huliohia (Barnes et al. 2018). Austropuccinia has spread rapidly across the Pacific basin over the past 15 years and poses significant threats to biodiversity of Myrtaceae in Australia and the Indo-Pacific (Fensham and Radford-Smith 2021). Introduced to Hawaiʻi in 2005 (Uchida et al. 2006), Austropuccinia has had only modest statewide impact on ‘ōhi‘a to date (Anderson 2012; Loope and Uchida 2012), but there are recent reports of severe defoliation and mortality associated with A. psidii in bands of montane ‘ōhi‘a forest in the Koʻolau Mountains of Oʻahu (Weaver 2018, 2021). In addition, the rust is causing significant impacts to endangered (Eugenia koolauensis) and indigenous (Eugenia reinwardtiana) myrtle species in the islands (Loope 2010). There is ongoing concern that accidental introduction of one or more new strains of the rust fungus could significantly increase pathogenicity and ecological impacts of the disease (Silva et al. 2014; Stewart et al. 2018).

By contrast, the more recently recognized fungal pathogens of ‘ōhi‘a, C. lukuohia and C. huliohia, have caused high mortality and extensive loss of native forest canopy in Hawaiʻi (Mortenson et al. 2016; Perroy et al. 2021). While their origins are still unknown, their genetic homogeneity and similarity to Latin American (C. lukuohia) and Asian-Australian (C. huliohia) clades of Ceratocystis suggests that they were introduced to Hawaiʻi (Barnes et al. 2018). High mortality of ‘ōhi‘a can lead to increased invasion by non-native species and loss of biodiversity (Mortenson et al. 2016; Camp et al. 2019; Fortini et al. 2019). Since the disease was first discovered in the Puna District of Hawaiʻi Island in 2012 (Cannon et al. 2022), both pathogens can now be found throughout Hawaiʻi Island and Kauaʻi Island (Heller et al. 2019). Only the less virulent of the two fungal pathogens, C. huliohia, has been detected on the islands of Oʻahu and Maui (Brill et al. 2019).

While windblown dispersal of infective urediniospores is the primary mechanism for natural transmission of A. psidii (Carnegie and Cooper 2011; McTaggart et al. 2018), there are significant unanswered questions about natural pathways for dispersal and transmission of both C. lukuohia and C. huliohia. The movement of Ceratocystis-contaminated ambrosia beetle (Coleoptera: Curculionidae) frass is the leading hypothesis for the spread of ROD (Roy et al. 2019, 2020a, 2021), although long distance wind dispersal is not well understood. Early patterns of ROD spread suggested that airborne movement of potentially infective propagules in frass was responsible for spread of the disease on prevailing trade winds from the Puna District of Hawaiʻi Island to the South Kona District of the island (Harrington et al. 2021). This hypothesis has recently gained favor as the primary mechanism for long distance dispersal of ROD fungi (Cannon et al. 2022) with little supporting data other than circumstantial observations. Movement of Ceratocystis-contaminated frass by feral ungulates, human foot traffic, and runoff from heavy rainstorms (Harrington 2013; Fortini et al. 2019; Perroy et al. 2021) are possible alternate pathways. Direct transmission through the dispersal of ambrosia beetles or other insect vectors carrying sticky ascospores (Wingfield et al. 2017) and use of contaminated tools and equipment may also be important. In this study, we used inexpensive equipment designed for passive monitoring of airborne particulates to evaluate effectiveness of airborne monitoring of environmental DNA (eDNA) associated with windblown urediniospores of A. psdii for detecting spatial and temporal distribution of this fungus across large geographic areas. We also evaluated whether windblown dispersal of propagules associated with ambrosia beetle frass or spores may contribute to long distance transport of the fungi that cause ROD. These approaches may have widespread application in other island and mainland ecosystems that are threatened by invasive fungal pathogens.

Methods

Environmental samplers and field sites

A network of passive environmental samplers (PES) was deployed at twelve locations on the island of Hawaiʻi between August 2016 and November 2017 to collect airborne propagules of A. psidii, C. lukuohia, and C. huliohia (Table 1, Fig. 1). The samplers were constructed of sheet metal ductwork attached to a sheet metal vane and were placed at the top of three-meter metal poles so that they could rotate freely to face the prevailing wind (Fig. 2). Detailed instructions, diagrams, and parts lists for constructing the samplers are available elsewhere (Atkinson et al. 2019). Airborne particulates were captured on four glass microscope slides positioned at a 45° angle within each sampler. Slides were covered with a piece of Scotch™ tape and smeared with a light coating of silicone grease (Beckman Coulter, Berea, California, USA) to make them sticky. The samplers are relatively inexpensive, can be constructed from materials available at local hardware stores, and are effective at capturing particulates ranging from 5 to 500 µm in diameter (Atkinson et al. 2019).

Table 1 Prevalence of Austropuccinia and Ceratocystis lukuohia detections by site
Full size table
Fig. 1
figure 1

Location of sampling sites on Hawaiʻi Island. Sites were located in the Kohala Mountains on the northern tip of the island (PP, Puʻu Pili; EK, ʻEke; HP, High Pressure; WR, White Road; BY, Base Yard), Kaʻu-South Kona in the southern half of the island (KH, Kona Hema; 1328, Kahuku 1328 m; 1070, Kahuku 1070 m; 928, Kahuku 928 m; 770, Kahuku 770 m) and Hawaiʻi Volcanoes National Park (FP, Footprints Trail; KI, Kīlauea Crater). Base layer courtesy of U.S. Geological Survey

Full size image
Fig. 2
figure 2

Passive Environmental Sampler (PES) deployed at the Kahuku 1070 m site in the southern half of Hawaiʻi Island

Full size image

Passive Environmental Samplers were placed at twelve locations on Hawaiʻi Island. Eleven of these were within or adjacent to native ʽōhiʽa forest near Kīlauea Crater at Hawaiʻi Volcanoes National Park, at the Kahuku Unit of Hawai’i Volcanoes National Park and the Nature Conservancy’s Kona Hema Reserve in the districts of Kaʻu and South Kona, and in the Kohala Mountains at the northern end of the island. The single exception was a PES placed at the State of Hawaiʻi Division of Forestry and Wildlife Base Yard approximately 1500 m SW of the town of Waimea in an area surrounded by agricultural fields and open pastureland (Table 1, Fig. 1). Samplers were operated either singly or as pairs at weekly or biweekly intervals between August 2016 and September 2017 with the most continuous period of deployment at 4 elevations in the Kahuku Unit of Hawai’i Volcanoes National Park (Table 1). PES were placed at distances ranging from 0.25 to 7.2 km from trees that were known to be infected with Ceratocystis (\(\overline{X }\) = 2.5 km, N = 12), although it is possible that undetected infected trees may have been closer to the samplers. No attempt was made to locate trees infected with Austropuccinia because of their widespread distribution on Hawaiʻi Island and low intensities of infection (Anderson 2012).

Sample processing and diagnostic qPCR

After weekly or biweekly collection from PES, two-inch tape strips on each of four glass slides were cut into four to six pieces with a sterile scalpel blade, peeled from slides, and then transferred to 2-ml screw-cap tubes containing 0.3 g of 100 µm silica beads and 0.3 g of 800 µm zirconium beads (OPS Diagnostics, Lebanon, New Jersey, USA). During early phases of the study up to 1/3 of each tape strip was left on the slide and archived for future microscopic examination. During later stages of the study, one of the four slides from each collection was archived in its entirety for future microscopic study. DNA was extracted from tape strips using a Machery-Nagel™ Nucleospin™ Plant II Extraction Kit according to the manufacturer protocol after homogenization at 4.5 m/sec for 40 s on a FastPrep-24™ 5G homogenizer (MP Biomedicals, Santa Ana, California, USA). Because of concerns about low DNA template concentrations and subsequent qPCR test sensitivity, extraction protocols were modified after the first 4 months of the study to use QIAamp™ DNA Investigator Kits according to manufacturer protocols (Qiagen Inc., Germantown, Maryland, USA). This method improves DNA recovery from silica extraction columns by incorporating carrier mRNA into the extraction reagents. To make template concentrations comparable between the two extraction methods, samples that were extracted from Machery-Nagel™ Nucleospin™ Plant II Kits were concentrated by ethanol precipitation and resuspended in volumes of elution buffer that were comparable to those used to recover DNA from silica columns with QIAamp™ DNA Investigator Kits.

Extracts from each slide were tested by qPCR for the presence of C. lukuohia and C. huliohia DNA using the qPCR assay described by Heller and Keith (2018). Assays for both C. lukuohia and C. huliohia were multiplexed with a primer pair and probe for MeNu47, a putative glycoside hydrolase identified as a low-copy nuclear gene within ʽōhiʽa and other species of Myrtaceae as an internal positive control to detect airborne pollen, ambrosia beetle frass, and plant trichomes originating from ʽōhiʽa and other species of Myrtaceae. Samples were run in triplicate for either C. lukuohia or C. huliohia in 96-well plates on a CFX96 Real-Time System (BioRad Laboratories, Inc., Hercules, California, USA) with ultrapure water non-template controls and gBlock® oligonucleotide (Integrated DNA Technologies, Coralville, Iowa, USA) positive controls. The gBlock® positive control was a synthetic 250-base pair (bp) double-stranded oligonucleotide containing the cerato-platanin gene qPCR targets for C. lukuohia and C. huliohia and the Myrtaceae glycoside hydrolase gene. Tape strip extractions were tested individually (i.e., 3 or 4 slides/extractions per sampler), and a positive detection for any of the slides was denoted as a positive detection for the entire sampler.

Samples were tested for A. psidii using a Taqman assay described by Baskarathevan et al. (2016). Primer PpsiITS1R (5'-TGA TTT TAG ACA ATA ATA ATA AGG G) and the FAM-labeled Taqman probe (PpsiITS1R, 5'-FAM-AGA TTA ATA TCT TTG CCA CGT ATA CCA-BHQ1) were identical to those described by Baskarathevan et al. (2016), but primer PpsiITS1F (5'-GTA TCT TTA TTG AAA CAT AGT AA) was modified by a single bp to make it specific for isolates of A. psidii that have been identified in Hawaiʽi (Uchida et al. 2006). Samples from each sampler were pooled (3–4 slides/PES) and then run as triplicates in 96-well plates on a CFX96 Real-Time System (BioRad Laboratories, Inc., Hercules, California, USA) with ultrapure water no-template controls and a positive control consisting of DNA extracted from a naturally infected ʻōhiʻa leaf.

Verification of qPCR products

For samples that tested positive for Ceratocystis lukuohia, we reamplified and sequenced a 305-bp region of the C. lukuohia cerato-platanin gene with two pairs of nested forward and reverse primers to verify species identity. The first external primer pair (60F, 5'-TGG GCC TCT CAC TAA TAG TCT CC; 393R, 5'-CGT TGT CGA CAC GGC CAG) amplified a conserved 334 bp region of the gene. The second internal primer pair (CP_For_A, 5'-TCC TGA CAT CGC CGG C; CP_Rev_A, 5'-CCA ACC AGT TTA GTG AAG GCA GT) amplified a 305 bp sequence that was specific for C. lukuohia. Both nested PCR reactions were run using OneTaq PCR Mastermix (New England Biolabs, Ipswich, MA) according to manufacturer instructions with 4 mM of each primer. Template from extracted tape strips was amplified for 25 cycles (95 °C for 30 s, 50 °C for 30 s, 72 °C each for 30 s) with a final 10-min extension step at 72 °C with primers 60F and 393R. One µl of product was amplified with primers CP_For_A and CP_Rev_A in a 25 µl reaction for 34 cycles (95 °C for 30 s, 55 °C for 30 s, 72 °C each for 30 s) with a final 10-min extension step at 72 °C. Samples were tested with a positive control consisting of genomic DNA extracted from ʽōhiʽa wood infected with C. lukuohia and an ultrapure water non-template control. When the correct size PCR product was present by agarose gel electrophoresis after amplification with internal nested primers, the product was cleaned with a QIAquick™ PCR cleanup kit (Qiagen Inc., Germantown, Maryland, USA), sequenced in both forward and reverse directions with an ABI 3730xl DNA Analyzer (Sequetech, Mt. View, CA), and trimmed and aligned with a C. lukuohia reference sequence (Genbank Accession No. KU43257.1) using Geneious 8.1.9 software (Biomatters LTD, New Zealand).

A subset of samples that tested positive for A. psidii by qPCR were amplified using nested PCR primers described by Langrell et al. (2008). The forward primer of the outer pair of nested primers, Ppsi1 F, was modified by a single bp to match the ITS sequence from Hawaiian isolates of A. psidii (5`-TTC TAC CTT ATT ACA TGT ATC T), while remaining primers Ppsi6 R (5'-GTC ATA TTG ACA GGT TAG AAG C) and the internal nested primers Ppsi2 F (5'- ATA GTA ATT TGG TAT ACG TGG C) and Ppsi4R (5'-GTC AAT CCA AAT CAA AGT ATG) were identical to those described by Langrell et al. (2008). Presence or absence of the 379 bp product amplified by internal primers Ppsi2 F and Ppsi4 R was determined with agarose gel electrophoresis. When the correct size PCR product was present, representative samples from different geographic locations on Hawaiʽi Island were cleaned with a QIAquick™ PCR cleanup kit, sequenced in both forward and reverse directions with an ABI 3730xl DNA Analyzer (Sequetech, Mt. View CA) and trimmed and aligned with an A. psidii reference sequence (Genbank Accession No. EF599768.1) with Geneious 8.1.9 software (Biomatters LTD, New Zealand). Representative sequences for both Ceratocystis and Austropuccinia fungi are deposited in GenBank (Accession numbers MZ836780-MZ836793, MZ889510).

Environmental data

Wind speed and direction data were collected at 1 min intervals with HOBO microstations supporting a Davis® S-WCF-M003 wind speed and direction sensor (Onset Computer Corporation, Bourne, Massachusetts, USA) at the long-term Kahuku study site. Two microstations and sensors were mounted on top of a 3-m length of ¾-inch electrical conduit at the 928 m elevation and 1378 m sampling locations. Instruments were placed at the 928 m site in August 2016 and at the 1,328 m site in December 2016.

Daily rainfall data was obtained from the U.S. National Oceanic and Atmospheric Administration (NOAA) Global Historical Climatology Network (https://www/ncdc.noaa.gov/cdo-web/) from two weather stations that bracketed the Kahuku study sites—Hawaiian Ocean View 4.7 NNE, elevation 1273 m and Naalehu 5.7 W, elevation 679 m). The two stations ranged from 1.6 to 9.7 km from the four sampler locations. Weekly precipitation totals were calculated by adding daily totals from the first day in the sampling period to one day prior to the last date of the sampling period to avoid duplicate counting. A mean weekly precipitation value was calculated by averaging weekly rainfall totals from the two weather stations and used in subsequent analyses.

Measurements of wind speed, wind gusts and wind direction were highly correlated between the 928 m and 1328 m sites (wind speed R2 = 0.8022, P < 0.0001; wind gust R2 = 0.8471, P < 0.0001; wind direction R2 = 0.6730, P < 0.0001), therefore data from the 928 m site was used for analysis since the data set was more complete. Average wind speed and wind gusts (m/sec) and average wind direction were calculated from measurements collected at 1 min intervals during the sampling week and used in subsequent analyses. Wind direction data were converted to mean vectors (µ) using the program Oriana (Kovach 2011).

Statistical analysis

Pearson’s chi-squared test was used to test for significance in Austropuccina detections among sites and Ceratocystis detections among sampler elevations at Kahuku. Generalized linear models (GLM) were used to analyze the effect of season and weather on Austropuccinia detections at Kahuku. We modeled the total weekly number of Austropuccinia qPCR detections across PES at the 928 m site assuming a poisson distribution and zero inflation. Diagnostic plots and statistics confirmed that the assumed distribution was appropriate and the need for the zero-inflation correction. The first GLM included only season (September–November, December–February, March–May, June–August) as an independent variable and we ran several GLMs on weather parameters including the effects of average wind speed, average wind direction, average wind gusts, and average rainfall on Austropuccina detection. For the weather GLMs, the most appropriate model was selected based on Akaike information criterion scores. Statistical analyses and visualizations were performed in R version 4.0.2 (R Core Team 2020) using the tidyverse, glmmTMB, DHARMa, AICcmodavg, and ggplot2 packages (Wickham 2016; Brooks et al. 2017; Wickham et al. 2019; Hartig 2020; Mangiafico 2020; Mazerolle 2020). Raw data and metadata are available as a U.S. Geological Survey Data Release (Atkinson and Roy 2022).

Results

Prevalence by site

Prevalence of Austropuccina in weekly and biweekly collections from PES varied widely across the island of Hawaiʽi, ranging from 3 to 14% of weekly collections at locations near Kīlauea Crater in Hawaiʽi Volcanoes National Park to 16–39% of weekly and biweekly collections at locations in South Kona and Kaʽu districts (Kahuku and Kona Hema) to a high of 100% of biweekly collections from four forested locations in the Kohala Mountains at the northern end of the island (Table 1). Differences among sites were significant (X2 = 83.896, df = 16, P < 0.0001). When prevalence by elevation was compared at Kahuku for weeks 10–51 of the study where sample sizes across all four elevations were equal, no significant differences in prevalence were evident (X2 = 8.016, df = 6, P = 0.2370).

In contrast to numerous detections of A. psidii by qPCR, C. lukuohia was detected in a single collection at the 1328 m site at Kahuku during the week of July 19–26, 2017 (Table 1) and in two collections at Kīlauea Crater during the weeks of December 7–14, 2016 and August 9–15, 2017.

Verification of Austopuccinia and Ceratocystis detections

To eliminate the possibility that qPCR detections of Austropuccinia and Ceratocystis were false positives, representative samples with low Cq values were amplified with nested primers described by Langrell et al. (2008) for Austropuccinia or nested primers designed to amplify the cerato-platanin gene of Ceratocystis. Representative samples collected at Kahuku (GenBank Accession No. MZ836782-MZ836788), Footprints Trail (MZ836780-MZ836781), Kona Hema (MZ836793), and High Pressure (MZ836789-MZ836792) were amplified with nested Austropuccinia primers. The resulting 253–290 bp sequences were identical to a reference sequence of Austropuccinia psidii from Hawaiʽi (GenBank Accession No. EU711421) and differed by a single bp at nucleotide 229 (A to G) from a rose apple (Syzgium jambos) isolate collected in Hawaiʻi (GenBank Accession No. EU711420). Only one of three detections of Ceratocystis (August 9 – 15, 2017) could be verified by additional amplification and sequencing with nested cerato-platanin gene primers. The 168 bp product (GenBank Accession No. MZ889510) that was amplified from a sample collected at Kīlauea Iki was 100% identical to C. lukuohia when aligned with a reference isolate from Hawaiʻi (GenBank Accession No. KU043257).

Effect of season and environmental factors

Austropuccinia DNA detections in PES were most likely to occur during the Fall (24 PES detections) and Spring (31 PES detections) and least likely to occur in the Winter (17 PES detections) and Summer (12 PES detections, X2 = 6.790 df = 3 P = 0.079) (Table 2, Fig. 3A). The best-fitted model for predicting Austropuccina DNA detections in PES included average wind speed, where increased wind speed resulted in significantly fewer Austropuccina DNA detections (Table 3, Fig. 3B).

Table 2 Estimated coefficients from the best-fitted General Linear Mode of season and Austropuccina DNA detections
Full size table
Fig. 3
figure 3

A Monthly Austropuccina psidii DNA detections in Passive Environmental Samplers and B average speed over time at Kahuku, Hawaiʻi from September 2016 to August 2017

Full size image
Table 3 Estimated coefficients from the best-fitted General Linear Model of weather variables and Austropuccina DNA detections
Full size table

Discussion

We demonstrated that a simple unpowered device capable of tracking wind direction can be used in combination with sensitive molecular diagnostics to detect seasonal patterns and geographic differences in numbers of airborne urediniospores of Austropuccinia psidii. When combined with molecular detection methods, PES may be suitable for monitoring incipient invasions of Austropuccinia and other fungal pathogens that disperse in the air column in remote locations or topographically difficult areas where access by foot or motorized vehicle is limited.

In Hawaiʻi, the pandemic strain of Austropuccinia psidii (Stewart et al. 2018) rapidly spread after introduction to susceptible species of Myrtaceae from sea level to 1500 m elevation in areas with mean annual rainfall that exceeds 750 mm (Loope 2010). Early spread of the rust was driven primarily by high rates of infection and mortality in the invasive rose apple (Uchida and Loope 2009). Since the widespread collapse of rose apple populations, Austropuccinia infections are likely maintained at low to moderate levels in ‘ōhi‘a and non-native Myrtaceae species (Anderson 2012). The most noteable impacts of the rust have been on the federally listed species, Eugenia koolauensis, which is facing extinction due to high susceptibility (Phillipson et al. 2019). While mortality of ʽōhiʽa has been less than initially feared, a widespread infection rate of up to 5% of trees with from 5 to 10% of leaves carrying the fungus has been reported on Hawaiʻi Island (Loope 2010; Anderson 2012). More recent outbreaks leading to defoliation and mortality of infected trees on Oʻahu (Weaver 2018, 2021) have led to heightened concern about impacts of the disease.

Regional differences in detections of Austropuccina were evident, with highest numbers in the wet, cool montane northern Kohala mountains and lowest numbers in the Kīlauea Crater area of Hawaiʻi Volcanoes National Park. This difference is curious because both areas are similar in terms of rainfall, temperature, humidity, and elevation. Additional field work may help to determine whether differences are due to lower infection rates in ‘ōhi‘a at Kīlauea Crater, differences in abundance of potential non-native secondary hosts for the rust, or differences in environmental factors that can affect germination of spores.

We found seasonal detections of Austropuccinia at our Kahuku long-term study site, where numbers of positive PES were highest in the Spring (March–May) and Fall (September–November) and lowest in the Summer (June–August) (Table 2). Given the high susceptibility of young leaves and shoots to Austropuccinia (Glenn et al. 2007) and peaks in urediniospore production that match periods of new growth in rose apple (Tessmann et al. 2001), we suspect that this may be related to seasonal patterns of flushing or new growth in ‘ōhi‘a that have been reported in the Spring and Fall seasons in Hawaiʻi (Porter 1973). Seasonal declines in rainfall and lower humidity that can occur during the summer months at the Kahuku Study sites (Giambelluca et al. 2013) may have also led to lower urediniospore production and detection (Tessmann et al. 2001). We found a significant relationship between wind speed and the detection of Austropuccinia DNA. This may be related to increased atmospheric turbulence and vertical diffusion of urediniospores (Zauza et al. 2010), but finer scale studies would be needed to define these relationships more precisely. Future research including localized temperature and relative humidty as well as presence and density of host plants may further explain patterns of urediniospore dispersal in Hawaiʻi. The seasonal pattern of urediniospore detection we described may have value in designing monitoring strategies for the rust in Hawaiʻi.

In contrast to numerous Austropuccinia and Myrtaceae detections in PES, there were only three detections of C. lukuohia and none of C. huliohia. One detection was at the 1328 m sampling site at Kahuku where the closest known infected tree was over 3 km away. Two other detections were at Kīlauea Iki in Hawaiʻi Volcanoes National Park in an area that has been intensively surveyed for infected trees by National Park Service Resources Management staff. The closest known ROD-infected ‘ōhiʻa was approximately 1.7 km away. The two detections at Kīlauea Iki were adjacent to a parking area for field vehicles and private cars used by park staff. While field vehicles were routinely pressure washed after use, private cars that were driven to residential areas within ROD outbreaks were less likely to be decontaminated. We can not rule out the possibility that infective frass may have originated as contamination from these sources, especially given the low numbers of infected trees in the area. By contrast, the single C. lukuohia detection at the Kahuku study site may be an example of rare, yet possible long-distance dispersal.

Our findings provided only marginal support for the hypothesis that long distance airborne dispersal of Ceratocystis is a pathway for movement of the fungus. While airborne detections of infected frass are relatively common near infected trees, with detections of up to 9% of PES in ROD outbreak areas within 50-m of infected trees (Atkinson et al. 2019), detections fall quickly as distance from positive trees increases (Roy et al. 2020b, 2021, Table 4). Other studies have found similar patterns for dispersal of frass associated with C. platani with detections falling exponentially as distance from infected trees increased (Luchi et al. 2013). While it is possible that major storm events and high winds may contribute to spread of the disease, particularly if atmospheric turbulence leads to increased vertical diffusion of frass (Hirst et al. 1967; Zauza et al. 2010), there is little direct evidence to support this. Atkinson et al. (2019) used PES to sample airborne inoculum during two major tropical storms in the Puna District of Hawaii Island in 2016 and found no evidence of increased airborne detections of Ceratocystis. This question may only be settled by additional investigations that use more sensitive diagnostics (Luchi et al. 2013) and that also demonstrate that airborne inocula are viable and capable of causing new infections. In fact, several biological characteristics of the fungus are inconsistent with airborne dispersal. First, production of sticky ascospores in galleries of wood-boring ambrosia beetles and presumable generation of fruity smells by the Ceratocystis fungi suggest that the fungi may be better adapted to dispersal by insect vectors (Wingfield et al. 2017). Second, association of Ceratocystis propagules with ambrosia beetle frass would suggest that size of the particles (approximately 500 µm on average (Atkinson et al. 2019)) and intimate association with the wood would lead to rapid removal from the air column by both gravity and rainfall. Finally, unlike Austropuccinia, Ceratocystis propagules require presence of a recent wound on a tree, either from a broken branch or insect, to initiate a new infection. While additional work on the potential airborne dispersal of Ceratocystis may help to define precise distances frass can move in the air column, investigations of the role that ambrosia beetles may play as direct vectors of the disease and localized spread of contaminated soil containing infective frass by feral ungulates (Fortini et al. 2019, Perroy et al. 2021) may be more likely to lead to practical management solutions to limit the spread of ROD.

Table 4 Airborne detections of Ceratocystis on Hawaiʻi Island as distance from infected trees increases
Full size table

In summary, environmental sampling was an effective method to monitor airborne spread of Austropuccinia psidii when combined with sensitive molecular detection methods. We also demonstrated that long distance dispersal of Ceratocystis under normal wind and weather regimes is rare with only a single detection after widespread sampling. While it is possible that storm and high wind events may increase the likelihood of long-distance dispersal of Ceratocystis, limited available evidence does not support this (Atkinson et al. 2019). By contrast, environmental sampling is effective for detecting airborne urediniospores of Austropuccinia and may be used to monitor for this rust in rugged, remote locations where thick vegetation, steep terrain, and deep stream canyons make access difficult for vegetation surveys. Environmental sampling might be useful for measuring outplanting risk to exposure to disease for threatened and endangered species of Myrtaceae that are highly susceptible to Austropuccinia and might also be useful for monitoring for introductions and movement of new genetic strains of the fungus. In Hawaiʻi, PES may be of some value in selecting outplanting sites for restoration of Eugenia koolauensis, a federally listed species that is particularly vulnerable to Austropuccina (Phillipson et al. 2019).