Soil development over millennial timescales - a comparison of soil chronosequences of different climates and lithologies

This paper reports soil development over time in different climates, on time-scales ranging from a few thousand to several hundred thousand years. Changes in soil properties over time, underlying soil-forming processes and their rates are presented. The paper is based on six soil chronosequences, i.e. sequences of soils of different age that are supposed to have developed under the similar conditions with regard to climate, vegetation and other living organisms, relief and parent material. The six soil chronosequences are from humid-temperate, Mediterranean and semi-arid climates. They are compared with regard to soil thickness increase, changes in soil pH, formation of pedogenic iron oxides (expressed as Fed/Fet ratios), clay formation, dust influx (both reflected in clay/silt ratios), and silicate weathering and leaching of base cations(expressed as (Ca+Mg+K+Na)/Al molar ratios) over time. This comparison reveals that the increase of solum thickness with time can be best described by logarithmic equations in all three types of climates. Fed/Fet ratios (proportion of pedogeniciron Fed compared to total iron Fet) reflects the transformation of iron in primary minerals into pedogeniciron. This ratio usually increases with time, except for regions, where the influx of dust (having low Fed/Fet ratios) prevails over the process of pedogeniciron oxide formation, which is the case in the Patagonian chronosequences. Dust influx has also a substantial influence on the time courses of clay/silt ratios and on element indices of silicate weathering. Using the example of a 730 kasoil chronosequence from southern Italy, the fact that soils of long chronosequences inevitably experienced major environmental changes is demonstrated, and, consequentially a modified definition of requirements for soil chronosequences is suggested. Moreover, pedogenic thresholds, feedback systems and progressive versus regressive processes identified in the soil chronosequences are discussed.

In the Mediterranean region of Europe, soil chronosequences were investigated mainly in southern Spain and Italy. Important studies were done by Alonso et al. [11] and Dorronsoro and Alonso [12] on soils on Holocene to Pleistocene fluvial terraces in Spain, and by Scarciglia et al. [13] on soils on Quaternary marine terraces in Calabria (southern Italy).
Soil chronosequences in temperate-and cool-humid climates were studied both in North America and Europe. Studies in North America include those of Singleton and Lavkulich [14] who investigated soils on beach sand on Vancouver Island; Barrett and Schaetzl [15] and Barrett [16] who examined podzolisation in sandy terraces and beach ridges at Lake Michigan; Alexander and Burt [17], and Burt and Alexander [18] who analysed soils on moraines of Mendenhall Glacier, SE Alaska. European soil chronosequence studies comprise those of Arduino et al. [19] who analysed iron oxides and clay minerals in soils on fluvial terraces in northern Italy; Bain et al. [20] who investigated soils on river terraces ranging in age from 80-13,000 years in Scotland; and Mokma et al. [21] who studied a Holocene Podzol chronosequence in Finland. Sequences of moraine ridges in mountain regions, e.g. in the Sierra Nevada (California) and in the European Alps, have also been used for studying soil chronosequences.Birkeland and Burke [22] investigated soil catena chronosequences on moraine ridges in the eastern Sierra Nevada, California, and Egli et al. [23,24,25] who analysed Holocene soil chronosequences in the Swiss Alps.
In contrast to the above-mentioned Mediterranean and temperate to cool climates, where numerous soil chronosequences have been studied over the last four decades, soil chronosequence studies in tropical regions and some extreme environments such as cold deserts are rare. The few existing studies include those of Pillans [26] who investigated soils on basaltic lava flows with ages ranging from 10 ka to 5.59 Ma in the tropical climate of northern Queensland (Australia), and Muhs [27] who studied soils on Pleistocene reef terraces of Barbados. These soils formed in Sahara dust, volcanic ash from the Lesser Antilles island arc, and detrital carbonate from the underlying reef limestone. A soil chronosequence in a cold desert environment was analysed by Bockheim [28] who studied soils on moraines ranging from ca. 6 ka to 250 ka in age in the Transantarctic Mountains.
This paper compares the direction and rates of soil-forming processes in humid-temperate, Mediterranean and semi-arid climates. It aims at contributing to the knowledge on natural soil-forming processes that is required for the three main purposes mentioned in the beginning of the introduction. For this purpose, six soil chronosequences are presented. In addition to the comparison of three climatically different regions, also two diverse pathways of soil formation in different parent materials in the same region in humid-temperate climate are included ( Figure 1). The sites for these soil chronosequences were selected trying to avoid considerable human influence. Man has become a major soil-forming factor [29]. This factor is excluded here as far as possible, in order to ensure a good understanding of (semi-) natural soil formation first, as a base for unambiguous identification of the effect of the soil-forming factor man, separately. However, a certain extent of human influence could not be avoided, especially in the Mediterranean soil chronosequence where the soils are largely influenced by long-term agricultural use and related erosion.
Similarities and differences between the soil chronosequences in terms of identified soil-forming processes and their rates are discussed. Using the example of the Mediterranean soil chronosequence from southern Italy, comprising ca. 730 ka, the fact that soils of long chronosequences inevitably experienced major environmental changes is demonstrated, and, consequentially a modified definition of requirements for soil chronosequences is suggested.

Material and methods
Three soil chronosequences were studied in southern Norway. Two of them were on marine loamy sediments in the provinces Vestfold and Østfold, on both sides of the Oslo Fjord, and one was on beach sand in Vestfold (Tables 1-3). These areas have been subject to continuous glacio-isostatic upward movement, since the ice sheet in the southern Oslo Fjord region melted between ~13000 and 12000 calendar years ago. As a consequence of the steady uplift since the weight of the ice had gone, All soils were described according to FAO [38] and classified according to WRB [39] (Tables 1-6).

Soil thickness increase
Soil thickness was calculated according to [12] as thickness of A horizons + thickness of B horizons + ½ thickness of transitional AC/CA/BC/CB horizons. Soil thickness increase over time can be best described by logarithmic functions in all studied soil chronosequences ( Figure 2). The rates of increase in soil thickness are clearly influenced by climate and parent material. The highest rate is observed for the Mediterranean climate ( Figure 2; green triangles). The humid-temperate climate of southern Norway allows for medium rates of soil thickness increase over time, whereby rates in the loamy marine sediments (Figure 2; red squares) are higher than those in the sandy beach deposits (Figure 2; orange squares). The lowest rates of soil thickness increase are found in the semi-arid climate of Patagonia ( Figure 2; green diamonds), where the mean annual temperature (12.6 °C) is in between those of the two other regions, but the low amount of precipitation (MAP = 287 mm) limits the rates of weathering and primary plant and animal productivity, and increase in soil thickness.
In general, it was attempted to establish all soil profiles comprised in the chronosequences in flat positions, showing minimum erosion. This approach worked well in the Holocene chronosequences in Norway and Patagonia, whereas in the Holocene-Pleistocene chronosequence in southern Italy, spanning ~730 ka, it was impossible to include only soils that were not or only slightly eroded because erosion was ever-present. A theoretical function of increase in soil thickness over time was obtained by including only maximum soil thicknesses ( Figure 3). However, very few of the soils included in Because of the xeric moisture regime and variable re-carbonatisation through dust influx, the general pH trend with time in the Metaponto chronosequence shows a slow decrease with some scatter and even increased pH values in the older soils of the sequence ( Figure 6). Pleistocene alternations between Mediterranean forest (during interglacial periods) and steppe environments (during glacial periods) in southern Italy [41,42] must have been accompanied by alternating leaching and nonleaching or slight leaching conditions. However, the influence of paleo-environmental changes on the time courses of soil chemical parameters is difficult to reconstruct. It is likely that the rates of decarbonatisation and pH decrease in the upper 50 cm were considerably decreased and even temporarily reversed during glacial periods with steppe environments and enhanced dust influx rates. Also the environmental conditions within glacial periods were not constant but included, for example shifts between Artemisia steppe and forest steppe, which must have been related to different soil moisture regimes and hence different carbonate dynamics. The Patagonian soil chronosequences are located in the driest environment included in this work. However the gravelly beach ridges, on which the soil chronosequences in Patagonia were established, are highly water permeable. Therefore, decarbonatisation is taking place in the upper 20 to 40 cm of these soils, and pH decreases towards the soil surface, indicating that some leaching processes are active. Both chronosequences show a slow but distinct pH decrease with time, whereby a clear difference in pH levels is observed between the soil chronosequence on beach ridges with > 10 % fine earth, i.e. the Regosol sequence (Figure 4; dark blue diamonds), and the sequence on beach ridges with ≤ 10 % fine earth in the upper 75 cm, i.e. the Leptosol sequence (Figure 4; light blue diamonds). The pH values of the Regosols are 0.8 to 0.9 pH units higher than those of the Leptosols. The higher amounts of fine earth in the Regosols point to greater dust accumulation at these sites, compared to the Leptosol sites; it can thus be concluded that the higher pH level of the Regosol sequence is also due to the higher input of calcareous dust.    The s sedimen source a to textur on marin also sho observed ratios, bu distinct l high rate semi-arid main soi of the Pa Figure 9 upper 50 soil chronose ts. The pare and thus simi re. This varia ne loamy sed ow some sc d (Figure 9,      50 cm show decreasing trends in all climates and parent materials, whereby the general level of pH is higher in drier climates and in areas with substantial influx of calcareous dust than in humid climates. Formation of pedogenic iron oxides proceeds with time, usually leading to increasing Fe d /Fe t ratios, except for regions, where the influx of dust (having low Fe d /Fe t ratios) prevails over the process of pedogenic iron oxide formation, which is the case in the Patagonian chronosequences presented here. Dust influx has also a substantial influence on the time courses of clay/silt ratios and on element indices for silicate weathering. Variable dust influx and sediment inhomogeneity of pedons may lead to large scatter of data so that chronofunctions are difficult to obtain. Five of the six soil chronosequences presented here are Holocene soil chronosequences, comprising time-spans of ca. 7 ka (Patagonia) to ca. 11 ka (Norway). These soils experienced some climatic shifts such as the mid-Holocene climatic optimum, and the Little Ice Age, but this climatic variation was within a range that appears to have no measureable effect on the soil chronofunctions.
In contrast, the Mediterranean soil chronosequence comprises ca. 730 ka, thus spanning a number of glacial-interglacial cycles. This means that the soil-forming factors climate and organisms were subject to major changes over the residence time of most of the soils comprised in the soil chronosequence. In the case of southern Italy, the environmental conditions alternated between Mediterranean conditions during interglacial periods and open Artemisia steppe to forest steppe environments during glacial periods [41,42]. It is evident that rates of base leaching and formation of pedogenic oxides must have been significantly lower under steppe conditions than under Mediterranean conditions. It seems most likely that existing Mediterranean soils of several meters depth did not show any further deepening during periods of steppe environments. Instead, it can be assumed that enhanced dust influx and increased accumulation of soil organic matter took place in the uppermost part of the existing Mediterranean soils, and steppe soils such as Phaeozems and Chernozems formed. Each following shift to an interglacial period with Mediterranean conditions led to soil organic matter decomposition and hence degradation of the steppe soils. Mediterranean soil formation that had been interrupted during glacial periods continued, characterised by carbonate leaching, clay migration, and rubification. These conclusions are in agreement with Scarciglia et al. [13] who studied chronosequences of typical Mediterranean soils on Early to Late Pleistocene marine terraces along the northwestern coast of Calabria (southern Italy). A combination of chemical, mineralogical and micromorphological analyses revealed that soil-forming processes in those soils did apparently not proceed continuously but polycyclically. The authors thus concluded that the soils mainly developed during interglacial periods.
The problem of major changes in climate and vegetation that is discussed here, using the example of the Italian soil chronosequences, holds true for all soil chronosequences that extend back to pre-Holocene periods. Hotchkiss et al. [13], for example, discuss the impact of Pleistocene environmental changes on soil and ecosystem development in Hawaii. In such long soil chronosequences, the factor climate per se cannot be kept constant over time, as demanded in the original definition of soil chronosequences, specifying soil chronosequences as sequences of soils of different age that are supposed to have developed under similar conditions with regard to climate, vegetation and other living organisms, relief and parent material (see introduction). We thus need to accept that climate and vegetation varied over the time-spans comprised in these long soil chronosequences and that (at least) all soils of pre-Holocene age are polygenetic [29]. Hence, the requirements for these long soil chronosequences need to be modified as follows: soil chronosequences are sequences of soils of different age that are supposed to have developed under similar environmental conditions, or have experienced similar environmental changes through time, e.g. through glacial-interglacial cycles, with regard to the soil-forming factors climate, vegetation and other living organisms, relief and parent material.
In addition to the inevitable environmental variability in time discussed above, it is important to consider that, even under constant environmental conditions pedogenesis would not proceed uniformly over long time-spans. Instead, pedogenic thresholds and feedback systems naturally occur in the course of soil development. Examples for major pedogenic thresholds that occur in the soil chronosequences comprised in this paper include the: threshold of acidification and availability of organic complexing agents that has to be reached before podzolisation starts in the soil chronosequence on beach sand in Norway; and the threshold of calcium leaching that has to be reached before clay illuviation starts in the soil chronosequences on loamy sediments in Norway and in southern Italy. Feedback systems are operating in the chronosequences involving clay illuviation (in Norway and southern Italy), since clay accumulation in the developing Bt horizons will lead to a progressively finer pore system in the Bt horizons, which in turn will favour further accumulation of clay. Feedback systems are also effective in the Norwegian soil chronosequence on beach sand, where the precipitation of the first metal-organic colloids that precipitate in the subsoil will attract other colloids and thus self-enhance the development of spodic horizons. Similarly, the first carbonate crystals, precipitating in the Patagonian soils will form nuclei, on which further carbonates will precipitate.
Besides progressive pedogenic processes such as silicate weathering and formation of clay minerals and pedogenic oxides, also regressive processes need to be taken into account. The main regressive process, particularly in the Mediterranean study area, is erosion. Johnson [44] explained soil thickness as a result of the interplay of: 1) deepening through weathering; 2) upbuilding (e.g. by dust influx); and 3) erosion. According to Johnson and Watson-Stegner [45], upbuilding can be regressive if unweathered fresh material accumulates at a rate that exceeds the weathering rate, or it can be progressive if weathering keeps pace with sedimentation and the sediments do not lead to profile rejuvenation or simplification. In the case of the Italian chronosequence, upbuilding by dust deposition represents a regressive process, because dust introduces fresh carbonates to previously decarbonated and clay illuviated soils. In the Patagonian chronosequences, upbuilding by dust deposition can be regarded as a regressive process with respect to Fe d /Fe t and clay/silt ratios because both ratios are decreased by dust accumulation, whereas they should generally increase with proceeding soil development including neoformation of clay minerals and pedogenic iron oxides. On the other hand, the dust that accumulated in the interstitial voids between the pebbles in the Patagonian beach ridges in many sites, represents the only fine earth at all. Dust accumulation in these cases is an essential process to increase the water holding capacity of the extremely gravelly soils and allow for the succession of higher plants. From this point of view, dust accumulation in this particular environment may be rather regarded as a progressive, than a regressive, process.