© 1999 Macmillan Magazines LtdNATURE|VOL 397|11 FEBRUARY 1999|www.nature.com 491review articleChanging sources of nutrientsduring four million years ofecosystem developmentO. A. Chadwick, L. A. Derry, P. M. Vitousek, B. J. Huebert & L. O. Hedin........................................................................................................................................................................................................................................................As soils develop in humid environments, rock-derived elements are gradually lost, and under constant conditions itseems that ecosystems should reach a state of profound and irreversible nutrient depletion. We show here that inputsof elements from the atmosphere can sustain the productivity of Hawaiian rainforests on highly weathered soils.Cations are supplied in marine aerosols and phosphorus is deposited in dust from central Asia, which is over 6,000 kmaway.Terrestrial biogeochemists traditionally distinguish atmosphericallyderived from rock-derived elements1,2. Atmospherically derivedelements such as carbon and nitrogen have an important gasphase; they enter terrestrial ecosystems either through biologicalprocesses such as photosynthesis and biological N2®xation, orthrough deposition from the atmosphere (for example, dissolved inprecipitation or by dry deposition of particles and gases). Rock-derived elements such as calcium, magnesium, potassium andphosphorus are important constituents of minerals; they enterterrestrial ecosystems through partial or complete dissolution ofthese minerals (chemical weathering). A central conceptual modelfor the development of soils and ecosystems is built around thisdistinction3; new substrates that are laid down by volcanic erup-tions, glacial recessions, or other processes that initiate the forma-tion of wholly new soils and ecosystems, generally lackatmospherically derived elements, particularly nitrogen andcarbon, but they are rich in minerals that can supply rock-derivedelements. Consequently, rates of plant production and other eco-system processes are often constrained by the supply of nitrogen inyoung developing ecosystems, or such systems are dominated byplants with well developed N2-®xing symbioses, such as alder andvarious legumes4±6.This model assumes that, over time, the stock of weatherableminerals in soil is depleted and rock-derived elements are lostwithout replacement, whereas atmospherically derived elementscan be replenished continuously. Eventually, in the absence ofdisturbance of the substrate, ecosystems can reach a terminalsteady state of profound and irreversible depletion of rock-derivedelements. Phosphorus in particular is commonly implicated as themaster regulator of ecosystems in the long term3,7, although otherrock-derived elements (Ca, Mg, K) are more mobile than phos-phorus and should be depleted more rapidly.Several lines of evidence suggest that the sharpness of thedichotomy between rock- and atmospherically derived elementshas been overstated, with signi®cant consequences for our under-standing of biogeochemistry. These include: (1) studies of the massbalance of elements in watersheds show that the quantities ofputatively rock-derived elements dissolved in precipitation can bea substantial fraction of the weathering sources of those elements8,9;(2) large quantities of dust are entrained in the atmosphere in aridregions, and transported globally10±14; and (3) atmospheric deposi-tion of all elements may be understated by standard measurements.The dry deposition of particles and gases to vegetation surfaces canbe substantial15,16, and where cloudwater is deposited in forests it islikely to be a particularly important source of elements17,18.We evaluated sources of biologically signi®cant elements andtheir implications for ecosystem functioning across a developmen-tal sequence of sites in the Hawaiian islands. Although recentanalyses of ecosystem dynamics have focused on climate andclimate change, in Hawaii as elsewhere19±21, the quantity andavailability of nutrient elements in the soil also control plantproductivity and other aspects of ecosystem functioning22. More-over, terrestrial ecosystems are altered by anthropogenic changes innutrient supply23,24and by interactions between climate change andnutrient availability25at least as much as by climate change itself.Hawaii is a model ecosystemThe Hawaiian islands are formed by a stationary convective plume(hot spot) that taps lava from the upper mantle. The volcanoes growfor about 600,000 yr, drifting to the northwest on the Paci®c litho-spheric plate26. Each island (and each volcano within an island) isprogressively older along a transect from active volcanoes in thesoutheast to the oldest islands in the northwest27. Along the islandchain, we located six sites that differ markedly in the age of theirunderlying substrate, from 300 through to 2,100, 20,000, 150,000,and 1,400,000 to 4,100,000 years old (Fig. 1). The sites are similar intheir current climate, with a mean annual temperature of 16 8C, andabout 2,500 mm of precipitation annually. The substrate is basalticrock admixed with tephra and pumice, with an initial chemistrythat varies relatively little in space or time28. Over time, thetopography evolves from classical shield volcanoes with gentleslopes, to deeply dissected landforms29; however, we were able toFigure 1 Location of the study sites. The youngest two sites are on the still-activeKilauea volcano; the 20,000-yr-old site is on Mauna Kea, the 150,000-yr-old site ison Kohala, the 1,400,000-yr-old site is on East Molokai, and the 4,100,000-yr-old siteon Kauai.© 1999 Macmillan Magazines Ltdlocate residual shield surfaces on even the oldest sites. None of thesites has been cleared or systematically altered by people.Finally, the Hawaiian islands form the most remote archipelagoon Earth. Relatively few species reached Hawaii naturally (blownthere in storms or attached to migrating birds) in the tens ofmillions of years before human discovery30. Some of these speciesthen radiated to occupy a range of environments far broader thando continental species, resulting in an astonishing degree of bio-logical similarity among sites. Overall, this near constancy ingeology, topography, climate and biota, coupled with a very wideand well characterized range in substrate age, makes the Hawaiianislands an extraordinary resource for studies of soil and ecosystemdevelopment.We