One mechanism that could underlie the development of such systems is the dual role that erosion can play in influencing soil fertility.Erosion removes soil and associated nutrients from surface soils, often limiting the productive capacity of agricultural ecosystems.However, erosion also exposes rock and little weathered soil near the surface, making the effective age of the soil much younger than that of the geological substrate from which it was formed.This effect has the potential to enhance the fertility of both erosional and depositional areas.Studies that have used strontium isotopes as tracers of nutrient sources within native forest ecosystems in Hawai’i show that although the supply of nutrients derived from the weathering of soil minerals is depleted on stable geomorphic surfaces of the older islands , the weathering source is rejuvenated and soil fertility is enhanced on lower slopes and in alluvial areas of those same landscapes.Palmer and others evaluated soil fertility, erosion, and their potential contribution to precontact agriculture on constructional geomorphic surfaces, slopes and valley bottoms on the wet windward side of Kohala volcano of the Island of Hawai’i–the oldest portion of the youngest island in the archipelago.Leeward Kohala supports a large rainfed agricultural system that is bounded by well-defined thresholds in climate and soil fertility ; the soil fertility threshold occurs where cumulative weathering and leaching have depleted soil minerals to the point that they no longer supply substantial quantities of nutrients or buffer atmospheric acidity.This threshold shifts to progressively lower rainfall levels in progressively older substrates across the Hawaiian archipelago.Based on soil properties at the leeward soil fertility threshold, Palmer and others concluded that: soils on constructional surfaces of windward Kohala are too infertile to support intensive rainfed agriculture, at least as it was practiced in leeward Kohala; erosion has a positive but small effect on soil fertility on slopes and alluvial areas in small stream valleys; and rock-derived nutrients dissolved in the streamwater used to irrigate pond fields sufficed to meet the nutrient demands of intensive pond field agriculture in these smaller valleys.
Palmer and others also analyzed soils in Pololu¯ Valley, a large valley in windward Kohala; there,ebb and flow bench soil fertility on lower slopes and alluvium was enriched substantially relative to constructional surfaces or to smaller valleys.They suggested that most of the material transported by erosion in the small valleys was derived from low-fertility soils in the surrounding uplands, whereas most of the material in the large valley came from its steep and little-weathered walls.Despite the fertility of lower-slope soils in the large Pololu¯ Valley, its slopes are extremely steep all the way to the alluvium on the valley floor, and there is no evidence of intensive rainfed agriculture having been practiced on these slopes.Similar slope profiles occur in Waipi’o and Waimanu Valleys in windward Kohala; these are the largest valleys on the Island of Hawai’i, and their alluvial valley floors supported the most intensively irrigated areas on the island.In contrast, most large valleys on older islands in the archipelago have very different slope profiles, with an accumulation of gradually sloping colluvial material between their steep walls and relatively flat alluvial floors.Why do large valleys on the older versus the younger islands in the Hawaiian Archipelago differ in structure? Are soils of the colluvial lower slopes on older islands fertile enough to support intensive rainfed agriculture? Could differences in valley structure have shaped pathways of agricultural development and intensification in pre-contact Hawai’i? In this article, we compare topography, soil fertility, and associated agricultural potential in two large Hawaiian valleys, Pololu¯ on the island of Hawai’i and Halawa on the much older island of Moloka’i.We test the hypothesis that the geological processes of erosion and subsidence influenced pathways of agricultural development and intensification in these landscapes.We focused on Pololu¯ and Halawa Valleys because they are comparable in size, relatively accessible, and because the archaeological remnants of Hawaiian agriculture have been surveyed in both.Pololu¯ Valley is the westernmost of seven large valleys on the windward northeastern flank of Kohala Volcano, the oldest subaerial portion of the Island of Hawai’i.Most of the lava within Pololu¯ Valley is from the Pololu¯ volcanic formation, older tholeiitic basalts that erupted from 400 to 600 ky before present; younger flows of the later alkalic Hawı formation cover much of the surrounding uplands and spill into the Valley.Wave-cut sea cliffs flank Pololu¯ Valley on both sides, and the valley itself is from 200- to 400-m deep.
Pololu¯ has a long history of Hawaiian occupation ; it was colonized by AD 1200 , and evidence of both irrigated pondfields and rainfed agricultural systems can be found on the broad, relatively flat floor of the valley.Halawa Valley is the easternmost of four large valleys on the windward flank of east Moloka’i.The east Moloka’i volcano emerged approximately 1.8 million y before present, and later alkalic eruptions covered most of its surface around 1.4 my ago.Like Pololu¯ , Halawa Valley is flanked by sea cliffs, and the main valley itself is approximately 300-m deep.Halawa has a pre-contact cultural sequence dating from at least AD 1300.Its extensive irrigation systems are intact; a total of 693 pondfield terraces have been recorded in nine separate irrigation complexes.The colluvial slopes above the irrigation systems on the valley floor exhibit dense archaeological landscapes of residential and rainfed agricultural features.We characterized the modern topography of Pololu¯ and Halawa Valleys using a 10-m digital elevation model provided by the National Elevation Dataset.Four 200-m-wide topographic swaths were selected to represent slope profiles in each valley, one near the valley outlet to the ocean, two in the center, and one just below the major waterfalls that bound the upper margin of the valley cores.Together, these transects sample more than 25% of each valley.The minimum, maximum, and mean elevations in the across-swath direction were computed for each point along the long dimension of the topographic swath.We compared mean elevations from each of the swaths by normalizing elevation and distance along the swath, and centering the swaths on the minimum swath elevation.Second, we calculated the distribution of slope angles across the full surface of each valley using a low-pass filter on the NED-DEM with a cut-off frequency of 0.05 m-1 to characterize the overall valley morphology by removing topography related to small individual gullies and ephemeral streams.From this filtered DEM, we extracted all filtered slope values within Pololu¯ and Halawa, summarizing these values as a cumulative distribution that denotes the fraction of the mapped valleys whose slopes are less than a given value.We sampled alluvial and colluvial soils within each of the valleys, and along transects on upland soils on both sides of each valley; locations where soils were collected are shown in Figure 1.We collected integrated soil samples from 0- to 30-cm depth following protocols similar to those of Vitousek and others , using a tiling shovel to expose a 30-cm profile and collecting an integrated sample across this depth.This depth generally encompassed the soil that was churned by cultivator’s digging sticks —and in earlier studies , the chemistry of these integrated 30- cm samples correlated well with deeper profiles collected across Kohala Volcano.Moreover, analyses of the 30-cm samples provided consistent thresholds that defined the distribution of pre-contact rainfed agricultural systems.Samples on the upland transects were collected systematically at approximately 500-m intervals, and sample positions were recorded via GPS.In Halawa Valley, slope and alluvial samples were collected on transects reaching from alluvial soils near the main stream up to sloping colluvial soils, continuing upslope above any remnants of Hawaiian agriculture to the base of the cliffs that surround the valley.The basal slopes of Pololu¯ are much steeper, and transect sampling was not feasible; instead, we collected alluvial and lower-slope soils widely across the valley and at the lowest portion of its steep slopes.
Most of the alluvial and lower-slope samples were collected within long-abandoned Hawaiian agricultural systems.Soils were air-dried, sieved , and then divided into three homogenous sub-samples, and all analyses were carried out as described in the supplemental material to Vitousek and others.Briefly, one sub-sample was analyzed for total C and N using an elemental analyzer; a portion of this sub-sample also was extracted using the method of Kuo and analyzed for resin-extractable phosphorus using an Alpkem RFA/2 Auto Analyzer.A second sub-sample was analyzed for cation exchange capacity and exchangeable Ca, Mg, Na, and K at the University of California, Santa Barbara, using the NH4OAc method at pH 7.0.The third sub-sample was shipped to ALS Chemex and analyzed for total concentrations of Ca, Mg, Na, K, P, Sr, and Nb using lithium borate fusion and X-ray fluorescence spectrometry.Duplicate samples were incorporated in each procedure.These measurements include some that reflect the forms of elements that are available to biota on relatively short time scales ,4x8ft rolling benches and others that represent the total pools of elements and/or the cumulative effects of weathering.Available forms of elements can be dynamic; measurements reflect what was in the soil at the time of sampling, but those pools can change on annual time scales and certainly are likely to have been influenced by human land use, both pre- and post-European contact.Total element pools are a more conservative measure; they include forms that are not immediately available to organisms, but they reflect the integrated outcome of additions and losses of elements playing out at time scales of decades or longer.Soils were collected across broad rainfall gradients in and around the two valleys—particularly in the uplands , where sample locations received from about 1450– 3420 mm y-1 of rain at Pololu¯ and from around 875 to 2200 mm y-1 at Halawa.Consequently, mean values of soil properties must be compared with caution.The distribution of soil properties with variation in rainfall provides a more direct measure of differences between valleys and among slope positions within valleys.These distributions are illustrated for base saturation and the percentage of P remaining in Figure 4; both of these measures are relatively stable indicators of soil fertility that correlate well with the boundaries of intensive rainfed agriculture in leeward Kohala.Both base saturation and P remaining decline with increasing rainfall in the uplands surrounding both valleys, consistent with numerous measurements of both 30 cm and deep soils along rainfall gradients in Hawai’i.Where rainfall ranges overlap between the valleys, Halawa soils are substantially lower in both base saturation and P remaining than are Pololu¯ soils—again consistent with numerous measurements that show declining soil fertility with increasing substrate age in Hawai’i.Alluvial and colluvial soils within both valleys had much higher base saturation and P remaining than did upland soils with similar rainfall.Although the uplands surrounding both Pololu¯ and Halawa Valleys are nutrient-depleted and infertile, with most samples falling well below the thresholds that bounded intensive Hawaiian rainfed systems , the slope and alluvial soils are well above these fertility thresholds.Not surprisingly, the remnants of intensive pre-contact agriculture were absent in the uplands and abundant in the alluvial areas of both valleys and the lower slopes of Halawa.Similar patterns of variation with rainfall and slope position were observed for most soil measurements, excluding only total C, N, and P.We summarize these comparisons in Tables 1 and 2, comparing slope and alluvial soils in each valley with upland soils that fall within a similar range of rainfall.Because Halawa Valley is both older and has much lower rainfall than Pololu¯ , we confine our statistical comparisons to slope positions within each valley and its surrounding uplands—although upland Halawa soils are systematically less fertile than Pololu¯ soils at comparable rainfall.Analyses of variance for most soil properties yielded significant differences with sample position in both valleys ; for these properties, in every case uplands were significantly lower in fertility/nutrient availability than were slope and alluvial positions, and slope did not differ significantly from alluvium.Total element pools displayed different patterns.There too analyses of variance yielded significant differences with sample position in both valleys.For C and N, upland soils had significantly higher concentrations than slope and alluvial soils in both valleys; slope and alluvium did not differ significantly.For total P, upland and alluvial soils had significantly higher concentrations than slopes at Pololu¯ , whereas upland soils had significantly lower total P than slope or alluvial soils at Halawa.We suggest that the greater total C and N pools of upland soils reflect their greater effective age, relative to slope and alluvial soils.