Invasion and Recovery of Vegetation after a Volcanic Eruption in Hawaii
NPS Scientific Monograph No. 5
NPS Logo


Obviously, the patterns of plant invasion and recovery observed depended on a number of specific factors and combinations of factors of which only some can be explained by the environmental measurements made. However, the breakdown of the total area into the six habitats and the study of associated features provided a basis for isolating the most probable causes or limiting factors from the total matrix of factors involved in the plant establishment on these new volcanic materials.

The recorded plant invasion and recovery patterns may be summarized as follows:

  • A directional progression of invasion was recorded on Kilauea Iki crater floor (habitat 1) and the spatter-with-tree-snags area (habitat 3). A directional recovery was also noted in habitats 3 and 5.

  • Differences in rate of invasion were observed in the same climate on different substrates of the same age.

  • Distinct sequences in arrival of plant life forms were recorded in different habitats.

  • Primary community formation was observed through development of synusiae and aggregation.

  • Survival capacity was seen to be a combination of degree of destruction effects and plant life-form characteristics.

  • Observations were made on the relative contribution of native and exotic species to plant recovery.

Factors Related to Directional Invasion and Recovery

At the onset of the study, it was assumed that the climatic gradient may have some influence on the direction of plant invasion. However, edaphic differences along this gradient exerted a much stronger control. Moreover, directional invasion patterns were not influenced by one factor only.

Substrate heat gradient

The concentric inward advance of plant life on the Kilauea crater floor was closely related to the substrate heat gradient that showed initial cooling at the crater floor margin. As shown by the substrate temperature measurements, the temperature increased toward the center of the crater. The temperature measurements made were too general to assign any threshold values that limited the advance of the different life forms. But the algae were found in the first advancing front and here maxima measured on the lava surface ranged from 54°C to 70°C during the warmest part of the day in summer 1968. Apparently, the same surface maxima were tolerated by the lichens which appeared only 2 years after the algae, but then occurred in the advancing front together with the algae. The probable cause of delay in arrival of the lichens will be discussed later. All other life forms advanced behind the algae. Thus, their establishment was possible only at lower temperature maxima. Moreover, they occurred at their advancing front only in lava cracks where surface temperatures were moderated by shade. Direct temperatures of the vapor steam were as high as 98°C and ferns and mosses would grow next to such hot vapor in the same cracks. But the substrate temperatures at the points of plant attachment were quite moderate, always under 45°C The lava surface temperatures were probably a combination of two heat sources: one coming from the glowing magma below, the other from the sun. W. P. Hasbrouck (U.S. Environmental Science Service Administration) made several subsurface temperature measurements in shallow bore holes of the lava rock near the center of the crater floor in 1967. He recorded the following temperatures (Table 11).

Table 11 shows that the heat increased downward as expected. However, a 50°C (122°F) surface temperature can easily be generated on any black surface during periods of sun exposure in temperate climates. Thus, it can be said with certainty that the pioneer mosses, ferns, and seed plants did not invade the crater floor until the lava surface heat was down to a temperature level that could be found in the cracks of any paved road in that climate. It is probable even that purely sun-generated heat kept these plants from entering the lava surface, while this was evidently not so for the lichens and algae. Therefore, the progress of plant life toward the center of the crater floor was related to a decrease in the volcanic heat from below, and plant life invaded only when the heat level had arrived at a temperature range found on nonvolcanic surfaces.

A subsurface heat-related invasion was also observed on the cinder cone habitat. However, here the invasion pattern was not as clearly directional as on the crater floor. Instead, the invasion delay of 2-3 years (Appendix II) was most probably caused by the slow cooling of the cinder cone surface.

The directional invasion recorded in the spatter-with-tree-snags habitat (3) had nothing to do with substrate temperatures. Here, the direction was related to two factors: the nearness of a seed source, and a snag-density gradient.

TABLE 11. Temperatures in shallow bore holes measured in a 2-ft circle on the Kilauea lava floor (near center) in 1967. (Unpublished data courtesy of W. P. Hasbrouck.)


6.4 cm50°C59°C
7.5 cm59°C
11.4 cm50°C69°C
14.0 cm54°C62°C60°C

Nearness of seed source

As shown on the map (Fig. 2), a relatively undisturbed rain forest bordered habitat 3 on the east side. The invasion began at this border and progressed across the board walk toward the cinder cone habitat in an easterly direction. The disseminule abundance was probably great right at this border and decreased away from it.

The important invaders here were the exotic woody plants, Rubus rosaefolius, R. penetrans, and Buddleja asiatica. All three were of minor quantitative importance in the undisturbed rain forest next to habitat 3. A major factor for their success was an explosive fruit and seed production of the surviving individuals at the margin of the relatively undisturbed forest. This was undoubtedly caused by the sudden openness and increased light availability right after the ash fallout.

Snag-density gradient

The advance of seed plants onto the new volcanic surface was facilitated by the presence of the Metrosideros tree snags. They caused microhabitats around the stem that showed better moisture relations than existed in the ash between the snags. Many of the new invaders became established at the base of such snags. Such an establishment is shown in Fig. 25 for Buddleja asiatica. This establishment pattern was typical also in the pumice with-tree-snags habitat (4). A number of substrate moisture samples were collected at tree-snag bases and in intervening spaces. The values shown in Table 12 are indicative for the differences recorded.

tree snags
Fig. 25.1. Eastern boundary of spatter-with-tree-snags habitat (3) where it joins the cinder cone habitat (2). Photograph taken in year 4 (1963).

tree snags with shrubs
Fig. 25.2. The same location photographed in year 7 (1966). Several Buddleja asiatica individuals had become established at the bases of tree snags.

tree snags with more shrubs
Fig. 25.3. The same location photographed in year 9 (1968). More Buddleja individuals had become established but several were dying where the snags had fallen.

The greater moisture at the snag bases is easily explained. Rainfall occurred usually in association with strong winds from the northeast. During such driving rains, the tree snags acted as interceptors and more water would run to their bases than was normally distributed over the surface. This precipitation moisture concentration effect on objects elevated above the surface occurred over the entire study area. The effect was borne out by the comparison of moisture caught in paired rain gauges, one of which was equipped with a Grunow fog interceptor. Table 13 shows the amounts of monthly rain water caught in such paired rain gauges in habitat 6. The precipitation water caught in the gauge with fog interceptor was consistently higher. The excess quantity ranged from 1.2 to 1.8. The Grunow fog interceptor is a 20-cm-high wire cylinder of 9.6 cm diameter. A 2-m-tall snag of similar diameter would have probably supplied 12-18 times the amount of normally distributed rainfall. This indicates that the moisture concentration at the bases of snags must have been an important factor in plant establishment. This moisture concentration effect is particularly important when viewed in relation to the low capacity of the ash substrates to store water in available form for plant growth.

The density of snags was higher near the border of the undisturbed forest in habitat 3 (Fig. 9) and decreased to the border of the cinder cone (Fig. 25). This snag-density gradient, therefore, was a supporting factor in the directional invasion from east to west in habitat 3.

TABLE 12. Substrate moisture (% by weight)a recorded at the base of tree snags and in the open ash surface 48 hours following rain in excess of 100 mm.

At snag base>1 m away

No. 31-314.112.4

No. 41-318.416.2


a Each value represents the mean of 10 samples.

TABLE 13. Monthly precipitation water (mm) received in a standard rain gauge and an adjacent one equipped with a Grunow fog interceptor in habitat 6 during 1968.

Guage with
fog interceptor
of standard



Ash-depth gradient

A directional recovery of surviving plants was observed only in two habitats: in the spatter-with-tree-snags habitat (3) and in the pumice-with-surviving-trees habitat (5). The obvious cause was the ash-depth gradient which in both habitats started at their outer margins with less than 10 cm depth and increased in direction of the cinder cone habitat (2) to more than 2 m depth (Fig. 2).

In habitat 3 this directional recovery was manifested in basal sprouting of Metrosideros snags that were initially believed killed. Basal sprouting started at the habitat border, where the spatter was less than 10 cm deep, and continued eastward to the middle of habitat 3 (near the board walk) where the spatter was about 50 cm deep (Fig. 9.4). The basal sprouting was unique to the spatter habitat and must have been related to the nature of this pyroclastic deposit. The material immediately became welded or cemented into coarser fragments upon deposition. The fallout stripped off nearly all small branches and bark. However, at the base of the stems, molds were formed through a sudden chilling of the glowing ash. A small space was usually left between the scored stem base and the mold. This space must have provided for precipitation moisture penetration and accumulation around the stem and for gaseous exchange which encouraged the basal sprouting from still living tissue at the stem base. A tree mold in habitat 3 is shown in Fig. 26, where the original Metrosideros tree was burned at the base. It is probable that only the older trees with rough bark and a thick epiphytic moss layer surrounding their tree base survived the spatter deposition.

tree mold with fern
Fig. 26. Tree mold in habitat 3 with Sadleria cyatheoides seedling. Tree molds were preferred microhabitats for invasion of pioneer mosses and ferns.

The Metrosideros trees in habitat 5 recovered more or less simultaneously along the ash-depth gradient, and their resprouting was rarely from the base, but instead from the remaining aerial stem part. A directional recovery was observed only for the surviving undergrowth species, the shrubs and herbs. Here a rough relationship of size and pumice depth was noted. The surviving shrubs reappeared generally up to a pumice depth of 50 cm and the herbs up to depths of 25 cm. However, these life-form boundaries of surviving undergrowth plants were not so clearly defined. A directional invasion pattern similar to that of habitat 3, with Rubus rosaefolius and R. penetrans advancing from the undisturbed stand, was also seen in habitat 5. However, this area was not included in the transect system. The invasion at this side was not nearly as vigorous as that observed in habitat 3. The two Rubus species were much more scattered and very few Buddleja seedlings advanced from the east side of habitat 5. This difference in habitat 5 was probably related to the much rarer occurrence of these species in the neighboring, undisturbed seasonal forest (Fig. 2). Also, the somewhat reduced rainfall along this border of habitat 5 may have been a contributing factor.

Factors Related to Differences in Invasion Rate

Rate of invasion was defined previously in three ways: (1) the quantitative spread of plants over the new volcanic surfaces in terms of increasing frequencies; (2) the increase in number of species; and (3) the increase in plant cover. The results can be briefly summarized as follows:

Habitat 1 (crater floor) showed a moderately fast rate of invasion but no significant increase in cover. Its species diversity was moderate, with 30 species in year 9.

Habitat 2 (cinder cone) showed a delay of invasion by 2-3 years and no significant cover increase. Its species diversity was low, with only 18 species in year 9.

Habitat 3 (spatter with snags) showed a fast rate of invasion that was supported by a high rate of plant cover increase. Species diversity was highest, with 64 species established in year 9.

Habitat 4 (pumice with snags) showed an earlier invasion but otherwise the same rate as occurred in habitat 2. Its species diversity was moderate with 34 species, and the cover was small in year 9.

Habitat 5 (pumice with surviving trees) showed a somewhat faster invasion rate than habitat 4. The cover of the invading species was also greater than in habitat 4, but species diversity was even less, with 28 species in year 9.

Habitat 6 (thin fallout area) showed an invasion rate similar to habitat 5, but there were some significant differences in floristic composition. The cover of the invaders was a little greater than in habitat 4, but less than in 5, and the species diversity was relatively low, with 23 species in year 9.

These differences in rate of invasion can be explained by almost the same factors that were discussed for the directional influence of invasion patterns observed.

Substrate cooling, disseminule supply and life form

The 2-3 year delay in plant invasion in habitat 2 was already explained as being primarily the result of the slow cooling of the cinder cone. The low species diversity in year 9 was also related to the late start of plant invasion. In time, species diversity will probably increase and become similar to that of habitats I and 3. The fast rate of invasion in habitat 3 was related to the absence of such a limiting factor as a volcanic heating from below. But this was also true for habitats 4, 5, and 6. A major factor in habitat 3 was the nearby seed source and the favorable microhabitats provided by the tree snags. The fast cover increase was a function of the rapidly growing, exotic woody seed plants and grasses that covered more surface area than the other pioneer life forms, such as the algae, mosses, ferns, and seedlings of Metrosideros trees and native shrubs. The somewhat faster invasion rates in habitats 5 and 6 as compared to 4 and 2 are also related to invasion of exotics and their growth characteristics.

This brings up the question of disseminule supply. This factor was not measured in this study because of technical difficulties and lack of time. It was somewhat equalized. however, by layout of the transect system (Fig. 2). As mentioned above, in habitat 3, disseminule supply was undoubtedly an important factor. Also, habitat 1 adjoined an intact rain forest that surrounded the crater rim. However, because of the substrate-heat gradient on the crater floor, a disseminule-supply gradient from the margin inward was probably only a minor contributory factor here.

An approximate equalization of the disseminule-supply factor existed among habitats 2, 4, 5, and 6 that were cut by transect AA' (Fig. 2). This equalization is supported by the observation that there was no directional invasion noted in the transect quadrats. Transect AA' is about equidistant from the undisturbed rain and seasonal forests to the east of the study area. An additional disseminule supply developed later in habitats 5 and 6 through sexual reproduction of surviving native trees and shrubs. The lateral transects BB' in habitat 2 and DD' and CC' in habitat 4, which were closer to a surviving seed source, can be considered to have counterbalanced the disseminule-supply factor among the habitats. Therefore, we may assume that this factor did not enter significantly into the invasion differences noted among habitats 2, 4, 5, and 6. However, there is another factor that needs special consideration. This is the availability of microhabitats favorable for invasion.

Availability of microhabitats

The relatively fast rate of invasion of mosses, ferns, and seed plants on the Kilauea crater floor, which seemed to be limited only by the cooling rate of the lava surface, was undoubtedly related to the large number of favorable microhabitats. These were primarily the joint cracks and crevices disrupting the smooth and solid pavement-like rock surfaces on the lava floor.

There are several factors involved in rendering the lava floor fissures as favorable microhabitats. The more moderate surface temperature as compared to the flat surfaces has already been emphasized. Another factor was the added precipitation moisture. As mentioned earlier, after rain showers, vapor steaming was vigorous from most cracks. A certain amount of this vapor was thought to condense directly at the fissure-sides where the mosses and ferns became established. This factor is extremely difficult to measure directly, but a certain idea of the added quantity of both, vapor steaming and rain falling at an angle during high winds, was obtained by the paired rain gauges placed on the crater floor during 1967 and 1968. The values shown in Table 14 are for 1968 only in order to save space and to allow for ready comparison with the Kau Desert (habitat 6) values in Table 13.

By comparing the two tables, it can be seen that the relative amounts of added precipitation in the interceptor gauges were not very different in habitats 1 and 6. However, the actual amounts of precipitation per month were about 50-80 mm more in habitat 1. The interceptor-gauge results do not separate the amounts contributed by driving rain and vapor steaming. However, from general observations, the amount added in habitat 6 was entirely from driving rain. In habitat 1, which was in a deep wind-sheltered depression where rains would descend more vertically, the amount added was probably mostly from vapor steaming. Thus, vapor-steam precipitation may have added some moisture to those joint cracks where the sides of the rock sheets jutted higher from the ground. Duly (1967b) also believed that vapor steam was a major factor in maintaining pioneer plant populations on new lava flows. However, a particular concentration pattern of plants at such cracks was not observed. Another moisture-concentration effect was probably much more important on the lava floor habitat. This was the lateral runoff of precipitation water following its fall on the solid pavement pieces of the lava. These pavement pieces must have acted like inverted funnels, causing a greater water supply in the fissures. Since there was no significant water storage capacity for plant growth in the lava rock itself, the added lateral seepage after rains and its temporary trapping in sealed joints and where chipped-off wind- or water-moved rock flakes had become trapped was probably the most important environmental factor in filling these fissures with plants. Yet, after a relatively fast advance of the ferns and native woody seed plants toward the crater center, the cover of these plants was still rather insignificant in year 9. The reason for this was that they remained confined to the fissures throughout this observational period. The invasion of plants with root systems onto the lava rock surface will probably take a very long time. It requires some accumulation of particulate material and probably a reduction in surface temperature maxima.

TABLE 14. Monthly precipitation received in two sets of paired rain gauges on the Kilauea crater floor in 1968 (transects a and b). One gauge equipped with Grunow fog interceptor (f).

Month Guage labels identifying transects on Fig. 2

March134.2199.11.5 124.8185.11.5
April623.2745.81.2 570.5699.51.2
May139.6258.51.9 130.7245.01.9
June226.3348.31.5 196.6285.91.5
July140.7209.21.5 130.2182.51.4
August97.7165.81.7 92.8150.11.6
September97.0139.41.4 93.8140.01.5
October151.8169.11.1 143.8172.31.2
November111.7122.41.1 101.4108.31.1
December523.9646.51.2 457.1575.11.3

Totals2246.13004.1 2041.72743.8
Means1.4 1.4

The cinder cone habitat was also criss-crossed by surficial fissures. Here, lateral seepage was probably less important. However, many fissure-profiles showed a cemented layer within 50 cm beneath the upper loose rubble of cinder and ash. This cemented sublayer may have acted similarly to the lava rock by channelling rain water laterally to the fissures, because most mosses, ferns, and seed plants became established in the fissures.

Tree molds, such as are shown in Fig. 26, were characteristic microhabitats in the spatter-with-snags habitat (3). These were also preferentially invaded where the snags had fallen right after spatter deposition. Such tree molds did not receive added moisture by either seepage or interception. In most tree molds, ferns and mosses became established on the side of the walls below the general surface of the habitat or even on the bottom of the mold. Shading and wind protection and thus conservation of moisture as well as lower surface temperature maxima were probably the main factors that made these microhabitats favorable for fern and moss invasion. Seed plants were found mostly at the molds with standing snags, or where the snags had fallen long after the seedlings had become established. In several such cases where the snags had fallen, the taller Buddleja individuals died back. This die-back phenomenon is shown in Fig. 25.3. It was probably related to edaphic drought.

Depressional microhabitats, such as cracks and molds, were found only in the first three habitats. Habitats 4, 5, and 6 had generally smooth, even surfaces covered with pumice. In the pumice-with-snags habitat (4), seed plant and fern invasion was decidedly associated with the standing snags. Here, the added precipitation moisture from interception was the main factor. The invasion pattern in habitat 5 (pumice with surviving trees) seemed more haphazard than in any of the other habitats. Some Rubus shrubs became established at tree bases, but others invaded the spaces between the trees. Accumulation of leaf litter was noted under the larger surviving trees in year 7 and 9 (Fig. 12.4), but this did not seem to have any specific influence on plant invasion in year 9. In habitat 6, invasion of new plants was concentrated somewhat around surviving shrubs but the open barren surface was invaded also by a few scattered sedges (Bulbostylis capillaris) and grasses (Andropogon virginicus and Rhynchelytrum repens). Thus, in general, invasion patterns associated with recognizable microhabitats decreased with the severity of the volcanic disturbance.

Climatic gradient

Contrary to an initial hypothesis, the climatic gradient had very little influence on the invasion patterns in the "Devastation Area." If there had not been an edaphic heat gradient on the Kilauea crater floor, the establishment of mosses, ferns, and seed plants in the joint cracks and fissures would probably have been haphazard over the entire floor to begin with. Then also, the invasion rate may have been considerably faster than at the drier end of the climatic gradient. The same cannot be said for the cinder come (habitat 2). Elimination of the invasion delay caused by the prolonged volcanic heating showed that the invasion rate on the cone after cooling was not faster than in habitat 4. The invasion rates in habitat 5 and 6 were also not slower than those in habitat 4. Therefore, the effect of decreasing rainfall and increasing desiccating power along transect AA' (NW to SE) was counterbalanced by the influence of the microhabitats described previously.

However, a certain floristic variation was correlated with the climatic gradient. This was the appearance of a few pioneer seed plants in the upper Kau Desert habitat (6) which did not appear in the moister zone. These were Rumex giganteus, Rhynchelytrum repens, and Bulbostylis capillaris. Moreover, a few rain forest associated pioneers did not enter habitat 6, such as Hedyotis centranthoides, Pipturus albidus, Vaccinium calycinum, and the exotics Eupatorium riparium and Paspalum dilatatum. From an island-wide distribution study of grasses on Oahu, Rhynchelytrum repens is known to be associated with seasonal climates (Karrawinata and Mueller-Dombois 1972).

<<< Previous <<< Contents >>> Next >>>

Last Updated: 1-Apr-2005