Strawberry guava (waiawī, Psidium cattleyanum O. Deg., Myrtaceae) is a small tree invasive on oceanic islands where it may alter forest ecosystem processes and community structure. To better understand the dynamics of its invasion in Hawaiian rainforests in anticipation of the release of a biocontrol agent, we measured growth and abundance of vertical stems >= 0.5 cm DBH for 16 years (2005-2020) in an intact Metrosideros-Cibotium rainforest on windward Hawai'i Island. Specifically, we compared the growth and abundance of both shoots (originating from seed or from the root mat) and sprouts (originating above ground from established stems) in four replicate study sites. Mean stem density increased from 9562 stems/ha in 2005 to 26,595 stems/ha in 2020, the majority of which were stems < 2 cm DBH. Mean annual rates of population growth (lambda) varied between 1.03 and 1.17. Early in the invasion, both density and per capita recruitment of shoots was greater than that of sprouts, but as overall stem density increased over time, sprout abundance and recruitment came to surpass that of shoots. Relative growth rates among small stems < 2 cm DBH declined over time for both shoots and sprouts, but relative growth rates of sprouts were consistently greater than that of shoots after the first 3 years. The capacity of strawberry guava to recruit from both rooted shoots and vegetative sprouts contributes to the facility with which it can invade intact rainforest, persist in the forest understory, and respond to canopy opening. Strawberry guava thus poses a considerable risk of stand replacement for Hawaiian rainforests. Stand management will require perpetual efforts of guava control at high priority sites as extreme weather events associated with climate change bring canopy-opening events due to storms, drought and pathogens.

Sites: We measured guava stem diameters annually between 2005 and 2020 at each of four replicate study plots selected to represent early stages of strawberry guava invasions in intact Metrosideros-Cibotium rainforest on windward Hawai'i Island (Juvik and Juvik 1998). Wet forests in Hawai'i are high priority conservation areas because of the biological diversity they harbor and their importance in the water economy of the islands (Jacobi and Warshauer 1992, Tunison 1992). Our study plots were established in the following conservation areas: Kahauale'a Natural Area Reserve (KAH, 19^o10'N, 155^o10'W), Pu'u Maka'ala Natural Area Reserve (MAK, 19^o34'N, 155^o11'W), Ola'a Forest Reserve (OLA, 19^o27'N, 155^o11'W), and Upper Waiakea Forest Reserve (WAI, 19^o35'N, 155^o12'W).  All sites are at approximately 900 m elevation and distances between sites are 2 to 17 km. Estimated annual rainfall is 3000-4000 mm at OLA and KAH and 4000-5000 mm at WAI and MAK (Giambelluca et al. 1996). Projected mean annual temperature based on adiabatic lapse rates is 17-17.5° C for the elevation range of the four study sites (Giambelluca and Schroeder 1998). All sites are on relatively young tholeiitic basalt lava flows that formed 200-1500 years BP (Wolfe and Morris 1996). The forests resemble native lowland (100-1200m elevation) wet forests with an 'ōhi'a lehua (Metrosideros polymorpha Gaud) overstory and an understory dominated by tree fern hāpu'u (Cibotium spp.) as described by Gagne and Cuddihy (1999) and Juvik and Juvik (1998). All areas are under conservation protection by the State of Hawai'i.

Species: Strawberry guava (waiawī, Psidium cattleyanum O. Deg.) is a small tree, 2-8 m tall. The yellow-fruited form (P. cattleyanum f. lucidum), dominant in the forests studied here, is one of three forms common across Hawai'i (Wagner et al. 1999) occurring in similar habitats. Strawberry guava produces 2-3 cm diameter berries with multiple 5 mm long seeds (Wagner et al. 1999) via both sexual reproduction and apomixis. In the wet forests of Hawai'i, seeds germinate within a year and do not accumulate in a soil seed bank (Uowolo and Denslow 2008). In Hawai'i seeds are dispersed by birds, rodents, and pigs as well as humans.  

Strawberry guava also reproduces vegetatively from both above-ground stems and from the root mat. For the purposes of this study, sprouts are defined as arising above-ground from established leaning or vertical stems. Such sprouts may overtake a leaning mother stem, obscuring the origin of older stems. Alternatively rooted shoots may arise via seed germination or directly from the root mat.  In this study we measured and tracked vertical stems standing more than 45 degrees from horizontal and greater than 0.5 cm at breast height (1.37 m, DBH). The population thus contained both shoots, apparently originating from seed or roots, and sprouts, originating as branches from older shoots or sprouts. We were unable to distinguish root sprouts from seedlings non-destructively and thus identified stems with an obvious above-ground connection to a mother stem as sprouts; shoots arising from the soil with no obvious above-ground connection to an existing stem were assumed to have originated from seeds or roots. Huenneke and Vitousek (1990), working in forests in the same area found that the proportion of rooted stems arising from seeds versus from roots varied widely. Thus, our study population is narrowly defined as vertical stems arising directly from the soil (shoots) or vegetatively from previously established stems (sprouts); leaning stems were excluded.

Surveys: At the start of the study (2005) all four sites had established populations of strawberry guava with a range of stem diameters represented. With one exception (OLA), we established one 0.25 ha plot at each site.  The study plot at OLA, with an initially higher-density guava population, was 0.15 ha. All vertical stems at least 2 cm DBH were tagged in each plot and their diameters measured. In addition, we tagged and measured all vertical stems >= 0.5 cm DBH and < 2 cm DBH in a stratified random set of 5 x 5 m subplots at each site (KAH: 6 subplots; MAK: 5 subplots; OLA: 5 subplots; WAI: 11 subplots). Diameter was re-measured annually, and new recruits tagged. Stems dying and leaning to less than 45 deg from horizontal were noted and not included in the study population going forward. The population of strawberry guava at each site reported here thus comprised only vertical stems.

 Analyses: We calculated basal area and yearly relative growth rates (RGR=log (BA _t+1/BA _t) based on basal area for individual stems. Density (stems/ha) was calculated from sample plots to allow comparisons among sites with different sample areas; estimates of total population density was based on the sum of the density of stems >=2 cm DBH from the entire plot plus the estimate of density of small stems (>=0.5 cm DBH and < 2cm DBH) from the subplots. Thus, total estimated density comprised all vertical stems >=0.5 cm DBH for each site. Total basal area per hectare was calculated similarly. Lambda (N(t+1)/ N(t)) was calculated from total population densities. To better understand the pattern differences in shoots and sprouts, we focused on sources of variation among small stems < 2 cm DBH, which comprised the majority of the population. Per capita annual recruitment and per capita stem death plus initial leaning of stems were calculated for both shoots and sprouts as a function of the total number of stems of all sizes present in the previous year at each site.

To determine whether lambda varied over time, we used linear mixed effects models using the lme() function in the nlme package (Pinheiro et al. 2023) in R (R Core Team, 2022). Year coded as a factor) was the fixed effect and site was the random effect. We accounted for temporal autocorrelation using AR1() auto correlation structure. We used a likelihood ratio test to assess whether the random effect of site was significant. To determine how shoots and sprouts differed over time in their densities, relative abundances, total basal area per ha, relative growth rates of stems < 2cm DBH, per capita recruitment, and per capita dying/leaning, we used linear mixed effects models using the lme() function in the nlme package.. For all models, we included stem type (shoot, sprout), year, and the stem type x year interaction as fixed effects (with year coded as a factor) and site as a random effect. Additionally, for the relative growth rate model we included a random effect of a stem ID nested within site. All models accounted for temporal autocorrelation using AR1() autocorrelation structure. When needed, we also accounted for heteroscedasticity by fitting different variances for each stem type in each year using varIdent(). Type III ANOVAs were run on the models to test significance of fixed effects. Post hoc tests were used to test for the difference between stem types in each year using the multcomp package (Hothorn et al. 2008). Marginal means and standard errors for plotting relative growth rates were calculated using the emmeans package (Lenth 2023). ANOVA results are provided in the figure captions.

In addition, we estimated the ages of shoots in the population based on annual basal area increments. For this estimate we pooled data from the shoots at all four study sites under the assumption that the sites were samples of a forest-wide population of strawberry guava. Only shoot growth was used in this estimate because sprout growth is in part dependent on the mother stem. Shoot growth rates vary as a function of their DBH as well as a function of light availability and other microsite characteristics; thus, we used four estimates of annual growth increment to provide a range of age estimates. Excluding stems with zero or negative relative growth rates, we estimated smallest, largest, mean and median basal area increments of shoots in 1 cm DBH size-classes using data for the year 2019-2020. For the few stems larger than 12 cm DBH we pooled data for all remaining large stems to estimate the increment. For each 1 cm growth ring and estimated annual basal area increment, we calculated the number of years a stem would take to reach the next size class. The total lapsed time to pass through size class 1 (0.5-2 cm DBH) through size class 21 (20-21 cm DBH) was an estimate of the years elapsed between the smallest and largest shoot size.