Long Substrate Age Gradient

Island chain rule

This substrate age gradient makes use of the increasing age of Hawaiian volcanoes with increasing distance from the hot spot, from southeast to northwest across the islands. Originally, much of the gradient was developed by Ralph Riley for a study of the regulation of nitrogen trace gas emissions (Riley and Vitousek 1995, Riley 1997); it was later refined by Tim Crews, Darrell Herbert, Kanehiro Kitayama, and Peter Vitousek to consist of six sites that range in age from ~300 to ~4.1 million years. These include the 0.3 ky (ky = 1000 years) Thurston site on Kilauea Volcano, the 2.1 ky Ola'a site on Mauna  Loa, the 20 ky Laupahoehoe site on Mauna Kea, the 150 ky Kohala site on Kohala Volcano, the 1400 ky Kolekole site on East Molokai volcano, and the 4100 ky Koke'e site on Kaua'i (Crews et al. 1995). Not every site has been included in each study along the gradient, and additional sites have been included with the gradient for particular studies.

All of the core sites are between 1120 and 1210 m elevation, and so experience a mean annual temperature near 15.5˚C, and all have mean annual precipitation close to 2500 mm (Table). This combination of elevation and precipitation was selected for study because lower and drier areas have been altered far more extensively by human activity, in many cases initially by the Polynesian discoverers of Hawai'i. In contrast, none of the core sites has been cleared or (as far as we can tell) systematically altered by direct human action. Kanehiro Kitayama established a parallel substrate age gradient across the Islands, in wetter sites (> 4000 mm/yr mean annual precipitation) at the same elevation (Kitayama and Mueller-Dombois 1995, Kitayama et al. 1998); some of the older sites on this wet gradient are closer to bogs than forests.  At times results will be compared from this wet substrate age gradient with those from the main gradient; whenever ‘age gradient’, ‘Hawaiian age gradient’, or ‘substrate age gradient’ are used without modification, this refers to the sites receiving 2500 mm/yr of precipitation.

All of the age gradient sites are located on the constructional surface of shield volcanoes, where the influence of erosion has been minimal. Erosion itself is minimal on the younger mountains; on older mountains, we selected sites on constructional surfaces that remain in inter-stream areas. Metrosideros polymorpha is the dominant tree in every site, and the distribution of many other species spans the gradient. Finally, the underlying rock in all of the sites is basalt, as it is for almost all of the Hawaiian Islands.

Overall, the climate, organism, parent material, and relief state factors (sensu Jenny 1980) are as alike as possible across these sites, while time (substrate age) ranges over four orders of magnitude, from 0.3 to 4100 ky.  It is possible that these six sites represent the best-defined long age sequence of sites on Earth, in terms of ability to control sources of variation other than substrate age, and it will be used extensively here. Nevertheless, it is far from perfect; there are assumptions, uncertainties, and simplifications that go into any study based on environmental gradients.

Age Control

The substrate age gradient is intended to represent a gradient in time — and so knowing the time at which surface biological and geochemical processes began to influence each site is crucial. For the stable constructional surface of a shield volcano, that time should be the age of the substrate (lava flow or deep ash deposit) that underlies the site. That sounds straightforward, but it isn’t always. For example, we know the history of the Thurston site quite well; it is on the flank of a satellite volcanic shield of Kilauea Volcano that erupted actively between about 1450 and 1500 AD (Clague et al. 1999). The lava is tens of meters thick; no vegetation or soil organic matter now on the site predates that eruption. In the first 250–300 years following the eruption, a thin soil developed on the site, made up of organic matter and volcanic tephra from the nearby vents of Kilauea. Finally, an explosive eruption in 1790 killed the forest (along with a Hawaiian army that was passing through the Kilauea area), deposited more tephra, and re-initiated succession — this time with a legacy of some buried soil organic matter. The site is young, only hundreds of years old; we call it 300 years old (0.3 ky) because it must be between 200 and 500 years old, and we have to call it something.

The next-youngest site, Ola'a, is covered with a thicker blanket of the same 1790 tephra that covers Thurston. However, this 1790 tephra overlaps deep layers of older tephra, the oldest and thickest of which is ~2100 yrs old. Trees at Ola'a are much larger than at Thurston; we believe that many survived the 1790 tephra, and in any case their roots reach into the deep tephra layers. We define this site to be 2.1 ky old, although certainly many of our soil measurements are based primarily on the influence of the 1790 tephra layer.

The older sites no doubt had similarly complicated histories; we know that is true of the 20 ky Laupahoehoe site (web site address). However, the older the site, the less that uncertainties of hundreds or even thousands of years are relevant.

Climate History

Another important — and interesting — source of variation among sites (other than substrate age) is differences in climate history. Not differences in current climate; that is defined reasonably well. However, the chronosequence approach not only assumes that sites have the same current climate; it further assumes that older sites had the same climate (and other conditions) as the younger ones, when they were young themselves. For example, it assumes that when our 4100 ky site was 0.3 ky old, it had the same climate that the 0.3 ky one does now. That is certainly not true here; or likely anywhere on Earth.

Hotchkiss et al. (2000) evaluated past variations in the climate of the LSAG sites, concentrating on three causes of change over time:

(1) Differences due to the local influence of global climate change. While the Hawaiian Islands’ maritime tropical location buffers them against the most extreme variations in Earth’s glacial-interglacial cycle, palyonological and geomorphological analyses show that montane forest areas of the islands were substantially cooler and drier during glacial times, and the trade wind inversion was at a lower altitude (Porter 1975, Gavenda 1992, Hotchkiss 1998, Science tropical glacier ref.). While the youngest sites have experienced only relatively warm, wet interglacial conditions like the present, the older sites have been through multiple glacial-interglacial cycles.

(2) Differences due to island subsidence. The mass of a growing volcano depresses the underlying oceanic crust, leading to subsidence of the volcano. Consequently, the final elevation of a site — once subsidence has ceased, a million years or more after a volcano is built — can be 1200–1500 meters lower than its initial elevation (Ludwig et al. 1991, Moore and Clague 1992, Price 2002). Therefore, much of the early history of the older sites was spent at a higher, cooler, drier elevation than they occupy at present. Moreover, if the youngest sites persist without being covered by additional lava for another million years, they will no longer be montane sites; they may even be submerged.

(3) Differences due to landslides and erosion upwind, which can expose sites that were in leeward rain shadow positions to much greater precipitation. For example, the 4100 ky Koke'e site began as a drier leeward site, but the removal of much of the island’s windward side by sliding and fluvial erosion has exposed it to the trade winds and caused increased precipitation there (Hotchkiss et al. 2000).

Hotchkiss et al. (2000) also integrated the climatic histories of sites along the age gradient, using cumulative degree-days that have accumulated in stand development (cf., Johnson et al. 1999) for temperature and the cumulative volume of precipitation inputs, and of water leaching through the soil, for precipitation and leaching (and for the 4100 ky site, guessing that Pliocene climates were warmer and wetter than present climates). In this analysis, the cumulative volume of water leached through soils differs substantially from an assumption of constant climate — but the sites remain a monotonic sequence , and the differences among them are far from subtle. Accordingly, this set of sites will be treated as a simple gradient in substrate age, and used as such in analyses of the regulation of nutrient availability, cycling, and limitation. However, the difference between a constant climate and one that has varied substantially and directionally through the history of the site is important to our understanding of soil and ecosystem development. If the analysis by Hotchkiss et al. (2000) is correct, it would require only 240 ky under present climate conditions to leach as much water through the soil profile of the 1400 ky site as is calculated to have been leached during its entire history.

Basic Features of the Gradient


The age-gradient soils were described and characterized by Robert Gavenda (Natural Resource Conservation Service) and Oliver Chadwick (University of California, Santa Barbara). Five pits were dug in each site, and profiles were described in the field and sampled by horizon for a wide variety of structural, mineralogical, elemental, and isotopic analyses. One profile per site was sent to the National Soil Classification Laboratory in Lincoln, Nebraska for characterization.

Soils in the five younger sites are classified as Andisols, while the 4100ky site is an Oxisol (Table). Andisols are influenced by volcanic parent material and its weathering products, while Oxisols are highly weathered soils of a type that is widespread in the humid tropics (Sanchez 1976). Among the Andisols, the two youngest have poorly developed profiles and tend towards Inceptisols; the 20 ky and 150 ky sites are classic Andisols dominated by gel-like non-crystalline secondary minerals; and the 1400 ky site tends towards an Ultisol.

The mineralogy of each horizon was determined using X-ray diffraction and Fourier transform infrared spectroscopy, and quantified using a sequence of increasingly harsh wet chemical extractions, following the procedures described by Chadwick et al. (1994). The most abundant soil minerals change substantially along the sequence (Vitousek et al. 1997b), with the two youngest sites mostly containing primary minerals (olivine, glass, and plagioclase feldspar) derived from the volcanic parent material. The first two of these minerals weather rapidly, and even feldspar nearly disappears by the 20 ky site. The young sites also contain secondary non-crystalline minerals (primary ferrihydrite, allophane, and imogolite) that form as weathering products of primary minerals (Shoji et al. 1993); these become the dominant minerals in the 20 ky and 150 ky sites. Allophane, imogolite, and ferrihydrite are metastable, X-ray amorphous minerals characterized by high degree of hydration, short-range crystal order, and a very high surface area. Over long time scales, these minerals continue to weather, eventually forming secondary kaolin and crystalline sesquioxide minerals characteristic of highly weathered tropical soils. These crystalline minerals become important in the 1400 ky site, and they dominate the clay fraction of the oldest site on the sequence. The progression from allophane-imogolite-ferrihydrite to kaolin-sesquioxide is important because non-crystalline minerals have large, reactive, variable-change surfaces (cf., Uehara and Gillman 1988) that bind phosphorus and soil organic carbon much more effectively than do more oxidized crystalline minerals (Wada 1989, Schwertmann and Taylor 1989).


The structure and composition of vegetation along the age gradient were determined by Kanehiro Kitayama, in an analysis that included the parallel, wetter sequence of sites (Kitayama and Mueller-Dombois 1995). Results from the six core sites are summarized in Crews et al. (1995). Five 20x20 m plots were established in each site, the height and dbh of the five tallest trees in each plot were determined, and vegetation was stratified into structural layers and inventoried with Braun-Blanquet cover-abundance scales (Mueller-Dombois and Ellenberg 1974). Detrended correspondence analysis (DECORANA-Hill 1979), a multivariate technique, was used to evaluate overall shifts in community composition.

The height of the tallest trees increased from the youngest into intermediate-aged sites, and then declined substantially in the older sites. Metrosideros polymorpha overwhelmingly dominated all the sites, accounting for > 75 percent of tree cover, and several additional species occurred in most or all sites. Native tree ferns in the genus Cibotium (C. glaucum and C. chammissois) dominated the subcanopy in the younger sites, but declined monotonically to very low abundance in the oldest site.

The youngest site on the sequence had the smallest number of species per 0.2 ha (28), while the oldest site had the greatest species richness (66); intermediate sites displayed no consistent pattern (Crews et al. 1995). Detrended correspondence analysis showed that quantitative shifts in community composition are highly correlated with substrate age (Fig. 3-17). The first axis of variation arrayed the sites in chronological order, while subsequent axes explained little of the variation in composition. These vegetation, like the soils (which contribute substantially to the vegetation patterns), are consistent with the assumption that the sites represent a clear and monotonic age sequence.

Characteristics of sites on the substrate age gradient across the Hawaiian Islands. In addition to the information here, all sites have mean annual temperatures of 15.5–16˚ C, and all average ~2500 mm/yr of precipitation.


Substrate age (ky)

Elevation (m)



Soil classification






 Lithic Hapludand





 Mauna Loa

 Thaptic Udivitrand





 Mauna Kea

 Hydric Hapludand






 Hydric Hydrudand





 East Moloka'i

 Hydric Hydrudand






 Plinthic Kandiudox

Reference; Nutrient Cycling and Limitation: Hawai'i as a Model System by Peter Vitousek
Princeton University Press 2004