USA: Hawaii: Kaupulehu Preserve Tropical Dry Forest Restoration

Invasive Fountain Grass (Pennisetum setaceum); Credit: Hawaii Invasive Species Council
Native Hopbush (Dodonaea viscosa); Credit: Native Plants Hawaii
Native Hawaiin goosefoot (Chenopodium oahuense); Credit: Native Plants Hawaii


Although tropical dry forests were once the most common of all tropical forest types, today they are among the most endangered and degraded of all ecosystems in the world. Grass invasions of fountain grass have catalyzed positive feedback loops in which grasses increase the frequency and intensity of the fire regime, which leads to further reductions of woody vegetation and corresponding increases in the spatial distribution and biomass of fire-adapted grasses. The project sought to investigate under what conditions native dry forest species might effectively compete with fountain grass. Because alien grass invasions have been particularly detrimental tropical dry forest ecosystems, the project sought to uncover effective methods to control these grasses and re-establish native populations. The project nvestigated how light availability (full sun and 50% shade), alien grass control (bulldoze, herbicide, plastic mulch, and trim treatments), and native species additions (outplanting and direct-seeding) affected the establishment of native plants and the suppression of a dominant invasive bunchgrass (fountain grass, Pennisetum setaceum).

Quick Facts

Project Location:
19.6826772, -155.9229469

Geographic Region:
Pacific Ocean

Country or Territory:
United States of America

Tropical Forest, Desert/Arid Land

Tropical Forest - Coniferous, Grasslands & Savannas - Tropical, Other/Mixed

Area being restored:
~3 hectares

Organization Type:

Project Partners:
Kona Dryland Forest Restoration Working Group


Project Stage:

Start Date:

End Date:

Primary Causes of Degradation

Agriculture & Livestock, Invasive Species (native or non-native pests, pathogens or plants)

Degradation Description

Introduction of fountain grass has proven catastrophic to these native arid ecosystems, because the grass invasions often catalyze a positive feedback loop, increasing the frequency and intensity of the fire regime. The dense root systems of many invasive grasses also inhibit nutrient and water acquisition by native species, which their extensive above ground grass biomass may thwart their ability to germinate and establish. The end result is the conversion of formerly diverse native ecosystem to near monoculture of fountain grass. The site of the project was, at the outset, a continuous, highly degraded, treeless fountain grass population with virtually no other native or alien species, except occasional clumps of the invasive shrub Lantana camara. Continued grazing and trampling by non-native ungulates also often destroyed remnant stands of native vegetation and increases vulnerability of the system for further alien species invasions.

Reference Ecosystem Description

Prior to the invasion of fountain grass and the promotion of non-native ungulate grazing, the dry forests of Hawaii were thought to have more total and native tree species that any other region in the state. From native hibiscus trees to rare shrub species such as Gouania almost one-fourth of native Hawaiian species are found in the dry forests. Many of the vines, shrubs, trees, and birds of the dry forests are threatened. Hawaiian owls hunt in the daytime, while the endangered palila, a beautiful finch-like bird, feeds on the seeds and flowers of mamane trees high up on the slopes of Mauna Kea. Most lowland forests native to Hawaiian are either seasonal or sclerophyllous to some degree and mesic transition forests occur where conditions are favorable. Dry forests can be closed or open canopied, exceeding 20 meters in height in montane habitats.

Project Goals

The project’s underlying intent was to establish standards for the further restoration of the dry forest ecosystems found in this part of Hawaii. After first removing non-native ungulate herbivory, the project leaders sought to establish what further interventions would be necessary to restore the native vegetation. This particular project sought out under what conditions native dry forest species could effectively compete with fountain grass.


The project does not have a monitoring plan.


These ecosystems contain a highly diverse and unique species assemblage that formerly covered most of the region, ~128,000 hectares. Although this area maintains some of the largest and highest quality native dry forest remnants in the Hawaiian archipelago, this project was considered essential to addressing the multiple problems of degradation and invasion simultaneously. Most of the remaining dry forest along the North Kona coast consists of fragmented, degraded, and senescent patches, threatening to increase the risk of devastating fires. The need for developing some effective, competitive strategies for dealing with fountain grass was seen as a necessity for the future establishment and sustainability of native plant populations.

Description of Project Activities:
In December 1998, project participants established four randomized experimental blocks at the study site. Each block consisted of two parallel 34 m long x 10 m wide strips, with 2.3-m spacing between the strips within a block, and 3-m spacing between adjacent strips in different blocks. To investigate the effect of available light, researchers erected 50% shade cloth structures over one randomly selected strip within each block, and allowed the remaining strip to receive ambient full sunlight. This treatment approximated the amount of shading produced by mature forest canopies in the restored ecosystem. Each strip contained four 6 x 6 m plots spaced 2 m apart from each other and the edge of the strip, for a total of 32 plots within this experiment (4 blocks x 2 strips/block x 4 plots/strip). Researchers randomly distributed four fountain grass control treatments (bulldoze, herbicide, plastic mulch, and trim) among the four plots within each strip so that each plot received a different grass treatment. For the bulldoze treatment, researchers removed the top 15 - 25 cm of lava substrate and fountain grass by scraping the blade of a D-8 Caterpillar tractor backwards across the surface of each plot. For the herbicide treatment, researchers first weed-whacked the grass to ground level, then sprayed a 1% Roundup solution onto the cut grass in January and April 1999 as the grass began flushing back. For the plastic mulch treatment, researchers first weed-whacked the grass to ground level, then covered the entire plot with a heavy, 100% light-blocking black plastic mesh in an attempt to smother and "˜"˜cook'' the underlying fountain grass clumps and soil seed bank until the plastic was removed one day before the experiment began (120 days later). For the trim treatment, researchers weed-whacked the grass down to 60 cm in height one week before the experiment began. In the latter three treatments, researchers left the weed-whacked fountain grass stems and associated litter in place, and did not remove them from the plots. Researchers did not use unmanipulated, intact fountain grass stands as a control treatment because previous research and observations have shown that few native species can germinate and/or establish within this environment. However, this trim treatment was a relatively moderate form of grass control that resulted in, on average, more than five times more initial fountain grass cover than any of the other treatments. In May 1999, researchers subdivided each plot into four 3 x 3 m quadrats and randomly selected one quadrat within each plot to receive native outplants. Researchers then designated the quadrat located diagonally across from the outplant quadrat to receive native seeds, while the remaining two quadrats within each plot served as controls for these treatments (no native species added). Researchers used this design so that quadrats with added native species would always be adjacent to and bordered by quadrats without added native species. Since each individual quadrat was equally likely to receive the outplant, seeding, or control treatment in a random and nonbiased manner, this restricted randomization design was unlikely to affect the statistical analysis of this experiment. All outplants and seeds used in this experiment originated from seeds collected from plants growing within or close to the upper Kaupulehu Preserve. Researchers chose species for the outplanting and/or direct-seeding treatments based on their seed abundance and ability to establish and grow in the field. To produce the outplants, seeds were sown in a greenhouse between December 1998 and January 1999 in flats containing Pro-Mix BX potting soil, transplanted into conetainers as the seedlings developed, then transported from the greenhouse to the field site one month prior to outplanting to harden off. For each outplant quadrat, researchers transplanted 44 individuals comprised of seven native species into a 1.8 x 1.8 m planting area centered within the quadrat, with 32-cm spacing between each plant. Prior to transplantation, researchers mixed the outplants together so that the different species were haphazardly interspersed with each other. The outplanting and direct-seeding treatments were performed on 17 May 1999. Outplanting was accomplished by creating or exploiting cracks in the lava with metal digging bars and mixing premoistened Pro-Mix soil into the planting hole as needed to support the transplanted seedlings. For the direct-seeding treatment, researchers sowed a mixture of 12 native species evenly across a 1.8 x 1.8 m area centered within each seeded quadrat. Prior to sowing, researchers soaked the seeds of some species for 24 - 48 hr in cold or hot water as necessary to scarify the seeds and facilitate germination as determined by preliminary trials. Researchers fertilized each outplant quadrat with Vita-Start B-12 and Peters 20:20:20 following the manufacturer's recommendations in May, June, and October 1999. To help the establishing plants survive the extreme heat and drought during the summer of 2000 (see Fig. 2), in July 2000 we also fertilized all outplanted and seeded quadrats using Osmocote 14:14:14. They did not fertilize the control quadrats at this time, because they did not contain any establishing plants. Previous research demonstrated that few native species can establish without supplemental water during the extended drought periods that frequently occur within this study system, in this experiment we watered all quadrats (outplant, seeded, and the two controls within each plot) with an automated gravity-fed irrigation system using the following regime: every day for the first three weeks of the experiment, every other day for the next 3.5 months, every third day for the following three months, once a week for the next five months, once every four days during the heat and drought of July - August 2000, then back to once a week until the end of the experiment in December 2000. Water was administered in the early morning for 5 min using sprinkler heads positioned at the corners of each plot (four heads per plot) that delivered 3.8 L per head per min. The total amount of water delivered to each plot over the 20-mo experiment was ~2900 L.

Ecological Outcomes Achieved

Eliminate existing threats to the ecosystem:
Throughout the experiment, rain was generally modest and patchy, with 11 of the 20 months having less than 20 mm of total precipitation and only four months with greater than 50 mm. Total precipitation during the three summer months was only 5.3 and 43.2 mm in 1999 and 2000, respectively, while November was the wettest month in both years with 67.6 and 186.2 mm, respectively. The percent cover of native species increased in all light and grass control treatments throughout the study site. The percent cover of native species in the shaded quadrats was nearly twice that of the full-sun quadrats during most of the study. The relative percent cover of native species among the grass control treatments also remained fairly consistent over time, with the greatest cover in the bulldozed quadrats, intermediate cover in the plastic mulch and herbicide quadrats, and the least cover in the trim quadrats. The percent cover of fountain grass also increased in all treatments throughout the experiment. There was little difference in fountain grass cover between the shaded and full-sun quadrats, and between the bulldoze, plastic mulch, and herbicide quadrats, but the percent cover of fountain grass in the trim quadrats was consistently about twice that of any other treatment. The final percent cover of native species differed significantly among the light, grass control, and species addition quadrats, but there were no significant interactions among these three treatments. For the light treatments, there was significantly more native cover in the shaded vs. full-sun quadrats (40.2 ± 3.8% vs. 27.4 ± 3.7%). Native cover among the grass control treatments was significantly greater in the bulldoze quadrats (45.5 ± 5.3%) than the herbicide (33.8 ± 5.2%) or trim quadrats (20.3 ± 4.7%), and marginally greater than in the plastic mulch quadrats (35.7 ± 5.4%). The percent cover of natives did not differ significantly between the herbicide and plastic mulch quadrats, but both of these were significantly greater than the trim treatment. Among the species addition treatments, native percent cover in the outplanted quadrats (69.1 ± 3.5%) was significantly greater than the seeded (35.3 ± 5.6%) quadrats, and both of these treatments were significantly greater than the control (15.5 ± 1.7%). The final percent cover of fountain grass differed significantly among the grass control and species addition treatments, but not between the two light treatments. Final fountain grass cover in the outplanted quadrats (36.3 ± 5.2%) was significantly less than the control (47.7 ± 3.7%), and marginally less than the seeded quadrats (45.2 ± 5.8%), while the latter two treatments did not differ significantly from each other. Among the grass control treatments, the final percent cover of fountain grass was significantly greater in the trim (80.3 ± 3.1%) than in the herbicide (34.3 ± 4.3%), plastic mulch (33.0 ± 4.3%), or bulldoze quadrats (29.3 ± 3.9%), but there were no significant differences in fountain grass cover among the latter three treatments. The initial percent cover of native species and fountain grass at the beginning of the experiment were significantly, positively correlated with the final percent cover of native species and fountain grass at the end of the experiment, respectively. The initial and final percent cover of fountain grass, as well as the initial and final grass height and percent grass in flower were all also significantly, negatively correlated with the final percent cover of native species. Native cover was dominated by three shrub (Chenopodium oahuense, Dodonaea viscosa, and Sophora chrysophylla) and two vine species (Canavalia hawaiiensis and Ipomoea indica), while the percent cover of the remaining outplanted and seeded species (as well as all alien species other than fountain grass) was generally negligible. Outplant survival in all light and grass control treatments declined fairly consistently over the course of the experiment. Analysis of the temporal pattern of outplant survival between the light treatments showed that these two curves differed significantly, with generally higher survival in the shaded quadrats, although by the end of the experiment, survival was similar in both light environments. Outplant survival in the grass control treatments was highest in the herbicide and plastic mulch quadrats and lowest in the bulldoze and trim quadrats; analysis of these survival curves also found significant differences among these four treatments. The proportion of outplants surviving to the end of the experiment differed significantly among the grass control treatments and among the different species outplanted, but it did not differ significantly between the light treatments and there was not a significant light by grass control interaction. Although for most individual species, outplant survival was lowest in the trim quadrats, the grass control treatment with the greatest outplant survival varied considerably among the different species. The results from this study and previous research within this ecosystem suggest that even highly degraded tropical dry forests may be at least partially restored. This experiment began in a treeless section of a small Hawaiian dry forest preserve dominated by dense African fountain grass stands containing virtually no other alien or native species. Yet, by the end of the 20-month study, some experimental quadrats had 80% cover of native species and 15% cover of fountain grass. Even in some control (nonplanted or seeded) quadrats, the vegetative spread of native vines and establishment of native shrubs from newly produced seeds in adjacent quadrats resulted in 30% native cover and 25% cover of fountain grass. Oceanic island communities in general, and Hawaiian ecosystems in particular, have proven notoriously susceptible to the devastating effects of alien species invasions. The inability of island floras to successfully compete with alien plants is often attributed in part to their evolution in a relatively noncompetitive and unique environment. However, the results of this study and other research within and outside of Hawaii suggest that in some cases, island plant species may successfully compete or at least co-exist with aggressive alien species. For example, in another experiment within this study system that occurred under essentially fountain grass-free conditions, researchers found that subsequent weeding did not significantly improve the performance of native species, even though by the end of that experiment many non-weeded plots were completely covered by two notoriously noxious alien species. This project has also shown that some relatively simple and inexpensive techniques can greatly facilitate the establishment of native species in grass-invaded ecosystems. Previous research in Hawaiian dry forests also showed that few native species germinate or establish in unmanipulated fountain grass populations. In the present study we found that, even relative to trimmed fountain grass stands, all three more aggressive grass control treatments had significantly more native cover throughout the experiment. The most radical technique, bulldozing, produced the greatest native cover, even though its fountain grass cover was not significantly less than the herbicide and plastic mulch treatments. This result may have been caused by decreased fountain grass root biomass and/or increased soil water availability (via crushing the lava substrate into smaller and more uniform pieces with the bulldozer tines) in the bulldozed quadrats, although researchers did not measure either of these variables in this experiment. The results of this experiment clearly demonstrate the importance of native species additions to the restoration of degraded ecosystems such as Hawaiian dry forests. Despite creating numerous seemingly favorable microsites of supplemental water, shade, and aggressive initial grass control, there was virtually no "˜"˜natural'' (i.e., unplanted) recruitment of native species into these or any other experimental quadrats over the 20 months of this study. In contrast, both the outplanting and direct-seeding treatments resulted in the establishment of relatively diverse populations of native vine and shrub species that continued to expand right up to the end of the experiment. We also found a significant overall negative relationship between the percent cover of native species and fountain grass cover, height, and percent in flower. This result may imply that in contrast to mature native canopy trees, these native understory species effectively compete with fountain grass for water and/or nutrients, and thus establishing vigorous populations of native vines and shrubs could be an effective technique for at least partially suppressing alien grass populations.

Factors limiting recovery of the ecosystem:
Plant performance in the outplanting and direct-seeding treatments showed considerable variation both among the different species and among the light and grass control treatments. For example, outplant survival ranged from 23% for the canopy tree Diospyros sandwicensis to 91% for the shrub Chenopodium oahuense, and while the survival of Diospyros outplants was more than three times greater in the shade, Chenopodium survival was virtually identical in the full-sun and shade environments. Similarly, four of the 12 species seeded into the experiment produced no surviving individuals, three species yielded an average of 0.3 individuals/quadrat, and the remaining five species produced between 3.8 - 21.9 individuals. Researchers also did not find a consistent relationship between the performance of individual species in the outplant and seeding treatments, or between the natural abundance of species and their performance in the experiment. For instance, seeds of the most common native shrub in this study system, Nototrichium sandwicense, essentially failed to germinate in any experimental quadrat, but had 70% outplant survival. Conversely, Chenopodium produced nearly half of the total number of sur viving seeded individuals, its outplants had the highest survival, and were on average twice as long as the next largest species, but its natural abundance and dominance is far less than three of the four other shrub species used in this experiment. Researchers also found an inverse relationship between the performance of the two endangered canopy trees seeded into this experiment, Colubrina oppositifolia and Kokia drynarioides, and their natural abundance. Although the final survival of both species in this experiment was poor, at their peak before the heat and drought of the summer of 2000, there were a total of 13 Colubrina and 80 Kokia seedlings. Yet, while there are still bands of Colubrina trees scattered across the study site and 280 individuals left in the state, the Hawaiian-island endemic Kokia has only three wild individuals remaining. The response of these species to the experimental treatments was also distinct: whereas at its peak there were nearly three times more Kokia individuals in the trimmed-grass quadrats relative to the next most abundant grass treatment, and there was little difference in abundance between the two light environments, Colubrina abundance was more than three times greater in the shade relative to full sun conditions, but did not differ dramatically among the grass control treatments. These results highlight the importance of investigating species- and treatment-specific responses before attempting larger scale restoration projects, particularly when using rare species with limited seed availability. For example, this experiment suggests that direct-seeding vigorous sun-tolerant shrubs and vines might be a cost-effective technique for both suppressing fountain grass and creating microenvironments to facilitate the establishment of slower growing, rarer species like Colubrina, while carefully managed fountain grass stands might serve as a better nurse environment for Kokia.

Socio-Economic & Community Outcomes Achieved

Economic vitality and local livelihoods:
Because of the highly fragmented and degraded condition of the native dry forests in Hawaii and throughout the tropics, even successful restoration projects of even a few hectares could create urgently needed new habitat for rare species. Such restorations could serve as a much-needed propagule source and catalyst for even larger projects.

Long-Term Management

With the conclusion of this project, it is possible that there is now sufficient knowledge to attempt tropical dry forest restoration projects at ever larger spatial scales. Based on this research and the work of others, it is clear that dry forest restoration in Hawaii and other regions with similar ecological dynamics should focus on three major objectives: (1) reduction or complete removal of non-native ungulates, (2) control of dominant invasive grasses, and (3) exploitation of existing or creation of new favorable microsites combined with reintroduction of carefully selected native species to these areas. This project was a part of a larger Kona Dryland Forest Restoration Working Group dedicated to collaborative management and restoration of the dry forests in Kona, the creation of this group indicates a broader social commitment to the restoration of the dry forests in Kona.

Sources and Amounts of Funding

This research was supported in part by the U. S. Fish and Wildlife Service and NSF grant DEB 9610413.

Other Resources

Cabin, Robert J. et al. 2002. Effects of light, alien grass, and native species additions on Hawaiian dry forest restoration. Ecological Applications 12(6): 1595-1610.

Primary Contact

Organizational Contact