Improving Sustainable Hawaiian Sandalwood Silviculture and Endemic Species Conservation with Mixed Stand Management

Objective 1: Optimize species composition and ratio of ʻiliahi and host plants

The first experiment will estimate the optimum ratio of ʻiliahi and host plants to balance competition, facilitation, and parasitism related to long-term survival and growth. We will monitor the growth of ʻiliahi in a host suitability experiment in which seedlings of ʻiliahi were planted in 2019 alone or with seedlings of either koa or ʻaʻaliʻi at one of four spacings (0-2 m)  to evaluate the suitability of host species and spacing (9 total treatments). Experimental units were planted randomly in 40 rows with 9 experimental units per row. Each row was 3 m apart, and ʻiliahi were planted 6 m apart within a row. In total, there were total 360 seedlings of ʻiliahi, 160 of koa, and 160 of ‘a‘ali‘i planted in this experiment.

Results of this initial experiment were published in Thyroff et al. (2023). Three years after planting, ʻiliahi survival and growth were significantly greater when paired with koa and at a closer distance. Given the size of surviving plants at the end this experiment, we expect that the root networks of ʻiliahi and host plants extend beyond the distance between adjacent rows and experimental units within a row. There has also been mortality of some of the ʻiliahi and host plants and natural regeneration of koa and other woody plants within the site. This provides an opportunity to investigate interactions among plants at the community level, which we assume is becoming more important over time. 

We will combine the experimental units within and across rows into adjacent 20 x 20-m plots containing four of the original experimental units within a row and seven rows. There will be one experimental unit or one row between adjacent plots, resulting in 10 total plots. Plant composition in each plot varies depending on previous treatments, plant mortality, and plant natural regeneration. The layout of the original experiment will be mapped over these plots to indicate the position of plants that died in the original experiment and to differentiate between the original experimental plants and natural regeneration (Figure 2). The map of all current plants will be used to identify the nearest neighborhood plant and to calculate plant competition.

Ground-line diameter (D) and plant height (H) will be measured for all plants in every plot to evaluate the growth of individual plants and the plots as a whole. For koa saplings with a stem diameter > 1 cm at breast height (~1.35 m), diameter at breast height (DBH) will also be measured. Additional parameters will also be measured for ʻiliahi, such as crown radius (CR), leaf chlorophyll concentration (CHL), leaf nitrogen (N) concentration (%N), and leaf water potential (LWP). Five fully expanded leaves will be used to estimate foliar chlorophyll, %N, and pre-dawn leaf water potential. Chlorophyll will be estimated using a atLEAF chlorophyll meter (FT Green, Wilmington, DE). Foliar %N will be analyzed by the University of Hawaiʻi at Hilo (UH-Hilo) Analytical Laboratory using a high temperature elemental analyzer. Leaf water potential will be estimated using a leaf pressure chamber. Soil samples will also be collected from every plot at a depth of 0−20 cm to identify the environmental characteristics of the plot. Soils will be analyzed for pH, total nitrogen, and soil organic matter, also by the UH-Hilo Analytical Laboratory.

Competition, facilitation and parasitism will be estimated using mixed-species growth models as in Forrester et al. (2013). We will use spatially explicit competition indices for the individual level due to the irregular spacing in this site (Maleki et al. 2015). Competition will be calculated using the basal area and distances of the nearest neighboring plants surrounding each ʻiliahi. Stand-level competition will be calculated from the total basal area of all plants in the plot. Greater basal area is assumed to equal greater competition, which would reduce growth of all plants. Facilitation and parasitism will be calculated using the ratio of the number of host to ʻiliahi plants and the relative growth rate of ʻiliahi compared to host plants. The importance of facilitation is indicated by greater growth of ʻiliahi when the ratio of host to ʻiliahi plants is greater at the same competition level. The importance of parasitism is indicated by greater relative growth of ʻiliahi to host plants at lower host:ʻiliahi plant ratios, since the parasitic resource drain on the hosts is expected to be greater 

Hypothesized outcomes of the growth of ʻiliahi and host plants under these different scenarios are illustrated in Figure 3. Of note is the hypothesized lower growth of the host under a 1:1 as opposed to a 2;1 ratio of host:ʻiliahi. Without parasitism, host plants would be expected to grow faster at a lower ratio, since this reduces intraspecific competition (Forrester et al. 2013); however, parasitic resource transfer is likely to be much greater at this ratio. With two host species in this experiment (a shrub and an N-fixing tree), optimizing the species mixture will need to take into account the growth rate of each species, the differing architecture of the plants, and the effect of increased N transfer to ʻilahi from koa as compared to ʻaʻaliʻi.

Classification and regression tree analysis (CART) using a random forest algorithm will be used to assess the influence of plant composition, competition, parasitism, facilitation, and soil quality on the growth and leaf physiology of ʻiliahi and host plants. This is to identify the importance of soil resources on growth and the driving forces for parasitism for both ʻiliahi and host in mixed stands. The relationship between selected factors and the growth of ʻiliahi and host will then be evaluated using regression analysis. Linear, power, exponential, polynomial, and threshold models will be tested for best fit using significance of fit parameters, including the significance of ANOVA F-test, the adjusted coefficient determination (R2adj), the residual standard error (RSE), and the Akaike information criterion (AIC)  (Rozendaal et al. 2020). The Normality of residuals and heteroscedasticity assumptions will also be used to evaluate these models (Meng et al. 2021). The validity of each model will be assessed using the mean absolute bias (MAB), and the root mean square error (RMSE) generated from the Leave-One-Out Cross-Validation (LOOCV) method (Tetemke et al. 2019). Finally, a scatter diagram will be created based on the relationships among host density, competition, facilitation, and parasitism to find an optimum planting mixture to balance the growth of ʻiliahi and host plants. This research will provide new information and a deeper understanding of how ʻiliahi and potentially other Santalum species perform in mixed-species stands. It will also provide a novel test of the application of mixed species growth modeling to include hemiparasitic trees or other plants. This will directly inform recommendations for silviculture and restoration of dry forests with ʻiliahi and provide a foundation for application to other Santalum species.

Objective 2: Overstory Thinning

The second experiment will evaluate the response of ʻiliahi saplings to the thinning of overstory koa trees. Our aim is to increase growing space for ʻiliahi (reduce competition), recognizing this may result in the loss of parasitic resource transfer from nearby hosts.  

A koa plantation was established in 2010 at KMR in a fenced 0.81 ha area. Trees were planted on approx. a 5.5-m spacing. Survival was 77% after one year, which has resulted in heterogeneous canopy cover. In 2020, one-year-old nursery-grown ʻiliahi seedlings were planted between the koa trees on a similar 5.5-m spacing. Survival of ʻiliahi after 3 years is > 95%, and growth has been positively related to canopy openness.

For the proposed experiment, a randomized design will be set up to facilitate a thinning trial with three levels and eight replicates. Thinning levels will include a no-thin control, 2-sided release, and 4-sided release (Table 1). A circular area 5.5 m in diameter will be measured around select ʻiliahi saplings, with the sapling designated as the center point of the circle. Koa tree stems that exist within the circular area will be subject to thinning based on the treatment level (Figure 4). After thinning, the natural regeneration of koa seedlings or sprouts within the thinned areas will be removed.

The effectiveness of thinning will be evaluated every year. Similar measurements as in Experiment 1 will be used, e.g. D, H, CR, CHL, %N, and LWP. Similar measurements will also made for the koa trees remaining in the thinning circle. Data will be assessed for homogeneity of variance using the Bartlett test (Beyene 2016). Treatment levels will be compared and evaluated using analysis of variance (ANOVA) and followed by Tukey’s multiple range test. The normality of residuals is examined using the Shapiro-Wilk test (Ghasemi and Zahediasl 2012). The nonlinear least-squares (NLS) regression analysis is also applied to identify ʻiliahi response to thinning level. The best-fit model will be evaluated based on the significance of the fit parameter, the significance of ANOVA, the coefficient of determination (R²_adj), and the Akaike information criterion (AIC) (Guillemot et al. 2015).

Furthermore, an individual growth model of ʻiliahi will be developed using nonlinear regression analysis to project the trend performance of ʻiliahi in the future. We will focus on projecting ground-line diameter (d)  and tree height (h) since both parameters have been continuously monitored from planting until 2023. All three thinning levels will be modeled. Three alternative models will be evaluated for fit: Richards, Schumacher, and Hossfeld (Carrijo et al. 2020). Leave-One-Out Cross Validation (LOOCV) will be used as a validation test due to the relatively small sample size (Bhandari et al. 2021). Models will be evaluated by the significance of ANOVA, R²_adj, RSE, AIC, MAB, and NRMSE (Rozendaal et al. 2020). The normality of residuals and heteroscedasticity assumptions will also be considered (Meng et al. 2021).

In parallel, individual growth models of koa as the host will be developed using a similar method. We will use results from Baker et al. (2008) that measured koa growth rates and responses to thinning in a similar dry forest as a comparison to modeled koa growth rates. This will allow for an estimate of the long-term effectiveness of thinning for growth of ʻiliahi and potentially identify the timeline for when additional thinning may be required relative to the estimate earliest period when koa and/or ʻiliahi are projected to produce a merchantable yield and commercial thinning can be considered feasible.