Varying Ecological Successions in Lakes Subdivided by Volcanic Eruption at Akan Caldera, Japan


 It is difficult to continuously observe ecological succession processes within lakes occurring over long-time spans. Thus, the process is generally shown as “lake types” or "hydrarch succession" reflected by trophic levels or differing aquatic vegetation, based on inductive inference by comparison of many lakes. Alternatively, long-term changes are simulated via microcosms or mesocosms in experimental systems, or lake history can be reconstructed by sediment analysis. Here, we try to demonstrate lake ecological succession processes over thousands of years by showing an example of lakes with diverse trophic levels and aquatic vegetation which were formed by segmentation inside Akan Caldera in eastern Hokkaido, Japan, due to volcanic eruptions. We found oligotrophic, mesotrophic, eutrophic and dystrophic lake systems in the caldera, despite similar ages and process of origin. Total water phosphorus concentration, defining trophic level, was significantly correlated with the ratio of accumulated watershed area to lake area and volume. Twenty-one species of aquatic macrophytes were classified into five groups clearly corresponding to respective or combinations of trophic level. This study is the first to visualize lake serial stages by documenting a series of trophic levels and associated aquatic vegetation groups as a result of differing eutrophication rates over time.


Introduction
Ecological succession within lakes is not only a major component of freshwater ecology and limnology, but also a platform for diagnosis and management of worldwide deteriorating aquatic environments resulting from human activity since the 20th century [1][2][3][4][5][6][7] . In general, lake succession proceeds thusly: when a lake is rst formed the water is oligotrophic with a paucity of biota, but the trophic level subsequently increases, as nutrients and minerals in ow from the watershed, and the lake basin shallows from accumulation of sediments. The diversity of biota and biomass increase or uctuate throughout this process, and eventually the lake attains the state of bog or marsh 1,2,4,5,8 . Historically, Forel, the "father" of limnology 5 , explained this transition as analogous to human ageing 9 , and this concept has been widely accepted due to both intuitive and experiential evidence 1,2,5,10 . Lake ecological succession is mostly characterized as "lake succession" or "lake type" based on differences in water quality, including trophic condition, and as "hydrarch succession" according to changes in aquatic ora 2,6,10,11 . Although the process and rate of succession varies with climatic zones, lake basin size, and initial water quality, in temperate regions lakes and hydrarch succession generally progress as oligotrophic-mesotrophic-eutrophic lake types, and submerged-oating leaved-emergent plants, respectively 2,4,6,12,13 . However, in any case, as Forel 9 pointed out when he compared the succession process to human aging, it is impossible to follow a normal lake's evolution, because of the immense time-scales required for a lake to disappear by sedimentation. Therefore, this general picture of long-term succession has been indirectly formed by comparing many lakes with different trophic conditions 10,12,14−18 , historical reconstruction of sedimentation processes 6,8,19 , follow-up survey of small-scale dams or reservoirs 2,20 , and experimental microcosms and mesocosms 5,21−25 . Despite these efforts, the perception of ecological succession in large lakes still relies on many assumptions, due to their long durations of biological ageings 2,21,26,27 . For example, the rate of eutrophication under natural conditions, a leading factor in succession, is thought to vary depending on lake size. However, few studies have successfully demonstrated that lake trophic conditions are related to lake size relative to watershed area 1,2,4,5 . The main reason for this missing information is how the load of trophic substances and sediments from the watershed is in uenced by indigenous variables, such as land form, soil, degree of weathering and erosion, local climate (temperature and rainfall), vegetation, and land use (farmland, factory and urbanization) in addition to the time effect 1,2,4,5,10 . Accordingly, a set of different size lakes of the same age and with similar watershed environmental conditions is needed to examine if lake trophic condition is related to the lake size and watershed area. In this paper, we offer insight into this problem by comparing a group of lakes in a large caldera. The lakes of Akan Caldera, Hokkaido, Japan, were formed by volcanic eruptions thousands of years ago, which divided a huge caldera lake into several lakes of varying size [28][29][30] . The watersheds of these lakes share similar environmental variables, making it possible to test the relationships between trophic conditions and aquatic ora, and lake size relative to watershed area.

Results And Discussion
Study area Akan Caldera is situated at the southern end of the Akan-Shiretoko Volcanic Chain, a volcanic region in eastern Hokkaido, Japan 30 . The outer shape of the caldera is oblong (24 km east-west × 13 km north-south) 30 , and a "central cone", Mt. Oakan, rises at the center of its inner basin (Fig. 1a, b). Within the caldera there are lakes and marshes of various sizes surrounding Mt. O-akan (Fig. 1a, Supplementary Table S1 online). Lake watersheds are isolated from external input by the caldera-wall 31,32 , and the water systems, connected by rivers or underground ow 28,32,33 , are roughly divided north-south and join at the southern foot of Mt. O-akan, where they then discharge through Akan River, the notch in the south caldera-wall (Fig. 1a).
Large lakes are distributed from west to northeast of Mt. O-akan and smaller lakes are localized from south to east (Fig. 1a). This peculiar lake arrangement is a result of the formation history of the caldera and Mt. O-akan. The Akan region has witnessed more than ten large volcanic eruptions in the last 1.5 million years, and the present oblong-shaped caldera was formed by the largest eruption of 200,000 years ago 30 . After the last large eruption (150,000 years ago), a huge lake, "Koakanko" (ancient Lake Akan), was generated in the caldera [28][29][30] . The lake was narrowed by post-caldera volcanic activity in the southwest part of the caldera, and by 110,000 years ago landform of the inside caldera-wall was almost completed [28][29][30] . A minor eruption of Mt. O-akan occurred 13,000 years ago slightly southeast of the center of the caldera, and it stopped when the lava ow reached the caldera-wall, so that Ko-akanko was separated into large and small basins [28][29][30] .
The developmental history of Mt. O-akan also shows that Pond Hyotan and Pond Junsai of the southern water system were rst divided from Ko-akanko, and other lakes of the northern water system were formed 5,000 to 2,500 years ago [28][29][30] . The lake depth charts of the large lakes, Akan, Panke and Penke, show the remains of valleys on the bottom of the inside of the caldera-wall extending as far as the base of Mt. O-akan 28,31,33 . This suggests that the water level of Ko-akanko was extremely low or the basin was exposed by discharge through the notch before the eruption of Mt. O-akan 28 . Therefore, the lakes of the northern water system are thought have formed by re-ooding after damming by Mt. O-akan, but the timing of lake formation and the developmental processes are not fully understood.
Akan Caldera occupies a part of Akan-Mashu National Park (designated in 1934 34 ), and is mostly covered with subarctic forests except for a town on the south side of Lake Akan, Akanko-onsen (Fig. 1a, b). Only Lake Akan has been developed as a sightseeing area due to the presence of Marimo, Aegagropila linnaei, a ball-shaped green alga designated a Japanese natural treasure 34,35 . Since the 1950s, the increase in tourism has resulted in eutrophication from sewage discharge. This continued until the 1980s when public sewage treatment service was provided 36 .

Water quality Topography
Shoreline development AWA AWA-LA ratio AWA-LV ratio  Table S4 online). Strong (|r| ≥ 0.7) and signi cant (p < 0.05, by test of no correlation) correlation coe cients were found between: TP and accumulated watershed area (AWA), TP and AWA-lake volume (LV) ratio (AWA/LV), electrical conductivity (EC) and AWA, EC and AWA-lake area (LA) ratio (AWA/LA), EC and AWA/LV, pH and shoreline development, and dissolved oxygen (DO) and AWA/LV (Table 1). Although correlation coe cients between TP and AWA/LA, and DO and AWA/LA did not reach 0.7 and − 0.7, respectively, both were statistically signi cant (Table 1).
When two-dimensional plots were drawn on these nine combinations, Lake Akan was discriminated as an outlier in TP and AWA/LA, TP and AWA/LV, EC and AWA/LA, and EC and AWA/LV (Table 1, Fig. 2b). Lake Akan was subject to anthropogenic eutrophication in the second half of the 20th century as previously explained 36 . We thus estimated phosphorus concentration before eutrophication. The oldest P 2 O 5 data, 0.010 mg l − 1 , measured in Lake Akan in 1931 37  Finally, the regression lines of DO and AWA/LA, and DO and AWA/LV were drawn with a slight negative slope due to the speci c low DO data for Lake Jiro ( Supplementary Fig. S3 online), and may not indicate an environmental gradient. Lake Jiro has no in ow and out ow rivers (Fig. 1a), and the water appears to be supplied through underground ow from the upstream Lake Akan 39 . However, in addition to DO, Chl-a, DOC, and chemical oxygen demand (COD) in Lake Jiro were the lowest concentrations among the Akan Caldera Lakes, and TP was the highest (Supplementary Table S2 online).
Furthermore, a portion of the lake surface of Lake Jiro does not freeze in winter 39 . These results suggest that other water sources, such as groundwater, may be involved in water formation in Lake Jiro

Distribution and species composition of macrophytes
We recorded 21 species of macrophytes in total (excluding emergent plants and macro algae) in 7 lakes, while no macrophyte species were observed in 3 lakes without in ow rivers (Supplementary Table S3 online). The correlation matrix of the number of macrophyte species was strong and signi cant with the following items among the above-mentioned lake topographic characters (Supplementary Table S4 online): boundary length (r = 0.720, p < 0.05), shoreline development (r = 0.703, p < 0.05), maximum depth (r = 0.924 < 0.001), mean depth (Fig. 4, r = 0.928, p < 0.001) and residence time (r = 0.921, p < 0.001). Several previous studies on relatively shallow or small lakes and ponds reported that the number of macrophyte species is correlated with lake area and that MacArthur and Wilson's "the theory of island biogeography", which theorizes how larger islands have more species than smaller islands, is often applicable [40][41][42][43][44][45][46][47] . However, in our study no signi cant correlation was found between lake area and the number of macrophyte species (Supplementary Table S4  in the large lakes of Akan Caldera is presumed to be more closely related to the area of littoral zone than the lake area. The two-dimensional plots of the number of macrophyte species and the above ve topographic characters showed that Pond Junsai, a dystrophic lake, was designated an outlier (Fig. 4). Furthermore, when the individual species was classi ed as submerged, oating-leaved and free-oating, the correlation coe cient was signi cantly greater with the number of species of submerged plants (Supplementary Table S4 Table S4 online), with most of these species localized in Pond Junsai. Thus, we conducted a cluster analysis to understand species composition of macrophytes in each lake (Fig. 5). The cluster was rstly divided into two groups: 6 species of oating-leaved and free-oating plants of Pond Junsai, and 14 submerged and 1 oating-leaved plants in the other lakes.
As mentioned above, prior to anthropogenic eutrophication, the phosphorus level in the rst half of the 20th century in Lake Akan appears to have been much lower than it is currently (Fig. 3). Thus, the current aquatic ora was also compared with results from the oldest known vegetation survey conducted in 1897 50 . Of the ten species of macrophytes in Lake Akan observed in our study, nine species were also found in the 1897 list, except for Potamogeton crispus, classi ed as mesotrophic. However, the two oligotrophic species R. nipponicus var. submersus and I. asiatica in the old list were not con rmed. These results suggest that the species composition may have changed to a more eutrophic vegetation type

Speci city of Akan Caldera Lakes
To summarize the results, the lakes of Akan Caldera having the same origin, have developed into a series of oligotrophic, mesotrophic, eutrophic and dystrophic lakes, after being divided by a volcanic eruption (Fig. 2a). The trophic level of each lake indicated by TP was closely related to the ratio of watershed area (AWA) to lake size (LA and LV) (Fig. 2b). These results strongly suggest the possibility that the rate of eutrophication was different among the lakes, and we see the various stages of lake succession in progress in this system. However, the observed TP of Lake Akan and its downstream neighbors, Lake Jiro and Lake Taro, are assumed to have been impacted by anthropogenic eutrophication in the past (Fig. 3a, b).
Furthermore, the formation history of the Akan Caldera Lakes suggests that Ponds Hyotan and Junsai are signi cantly older than the other lakes [28][29][30] , and an in uence of the time lag on TP among lakes must also be taken into account. To understand the loading rate of TP and its long-term uctuation, the correct time of formation and subsequent eutrophication history of each lake should be clari ed through research on lake sediment, etc. In addition, the linear regression of the relationship between TP and AWA/LA and AWA/LV (Fig. 2b) suggests that the indigenous environmental variables of the watersheds may vary little. However, in reality, the geology and vegetation within the Akan Caldera are not uniform 28,34 , and further research is needed to determine the actual phosphorus loading from the watersheds.
Macrophyte species composition varied among lakes, driven by lake trophic conditions (Fig. 5), and is indicative of aquatic plant succession, or "hydrarch succession". Many species of macrophytes are classi ed into a variety of "types" according to trophic level of the habitats 44,49,51−55 . Schneider and Melzer 55 , for example, proposed seven categories ranging from oligotrophic to eutrophic and even polytrophic types. Meanwhile, Lacoul and Freedman 49 simpli ed into three categories: oligotrophic, eutrophic, and general types, based on the opinion that many macrophytes generally have a broad ecological range, occurring over wide trophic levels, while other species have a narrower distribution. In our study, however, all of the observed species belonged to only one of the ve occurrence types, with a species-speci c range of trophic level (Fig. 6).
Thirteen species were distributed among oligotrophic, mesotrophic and dystrophic lake-types, respectively, and eight species occurred in the oligotrophic and mesotrophic, and the oligotrophic, mesotrophic and eutrophic type lakes with wider trophic ranges. These results explain how the species composition of macrophytes in the Akan Caldera Lakes with different trophic types is determined by the combination of species with different trophic requirements. Importantly, we noted vegetation changes in Lake Akan due to anthropogenic eutrophication: this was characterized by disappearance of oligotrophic type species, appearance of mesotrophic type species, and survival of oligo-mesotrophic and oligo-meso-eutrophic type species, when the trophic level of Lake Akan shifted from oligotrophic to mesotrophic as shown in Fig. 6. This might be the rst to clearly and simply illustrate the change of species composition in aquatic plant community caused by the change of trophic level. Further investigation is necessary to test whether this is a universal phenomenon in limnology.
Trophic levels are not the only factor known to affect macrophyte distribution and species composition. Within some relatively small areas, the following physical factors have been important in producing environmental gradients between or within lakes: topography, geological qualities, in ow waters as physical factors in the watershed, lake basin morphology (mainly depth and area), water temperature, light conditions, turbidity, current ow, substrate (sediment). Chemical factors in play include inorganic ions, salinity, organic matters, conductivity, alkalinity, pH and utrients 12,15,24,42,44,47,49,51,52,56−60 . As this itemization indicates, a comprehensive understanding of the mechanisms that determine the distribution and species composition of macrophytes is di cult, due to multiple in uencing factors. In the Akan Caldera Lakes, species distribution was related to trophic conditions (Figs. 5 and 6), while number of species was closely related to lake size (Fig. 4). Lakes Akan, Panke and Penke, all large and oligo/mesotrophic lakes with high species count, appear to offer a large variety of the physical and chemical factors noted above. The topography of capes, bays and islands in these lakes diversi es wave action and substrate via varying wind-wave parameters, and "fetch", i.e. the length of the lake surface over which wind blows 5,49 . In ow rivers locally alter substrate, current and water quality, greater water depth lowers water temperature and reduces substrate grain size, and oligo/mesotrophic water allows sunlight to penetrate deeper into the littoral zone, leading to a gradual gradient in the light environment 5,6,24,43,44,47,49,57,60 . This environmental variability may offer habitat for a number of macrophytes in these large lakes. On the other hand, Pond Junsai, dominated by oating-leaved and free-oating plants, is the only example of dystrophic water quality in Akan Caldera Lakes. Its brownish lake water, derived from high DOC containing abundant humic substances, suppresses submerged plant growth due to high light absorption 6,24,44,49,60 .
Humic substances are thought to originate from decomposing terrestrial and/or littoral plant material 6,25,38,49 . Thus, eutrophication in Pond Junsai may have undergone a different process than in the other lakes because of differing surrounding vegetation, even if the initial process was the same. To understand the factors affecting plant distribution the relationships between detailed macrophyte distribution and habitat micro-environments in each lake, including impacts of the surrounding vegetation, must be elucidated.

Conclusion
The Akan Caldera Lakes developed environmental diversity due to a variety of lake and watershed sizes brought about by unique volcanic activity [28][29][30] . In this paper, this diversity is displayed as two orderly series of changes in trophic levels and aquatic vegetation among the lakes, possibly illustrating lake ecological succession in "real time", previously thought impossible to see 2,9,21,26 . Odum 18 called such observable examples of a coincidental series of ecological succession events in nature a "natural laboratory of succession". He further stated that "these areas not only have a priceless natural beauty, but they also constitute a natural 'teaching laboratory' in which the 'visual display' of ecological succession is dramatic". In that sense, the Akan Caldera Lakes can be seen as a massive experiment conducted by nature in "Akan Caldera Laboratory", or seen as catalog visualizing a freshwater ecosystem, focusing on lake and hydrarch succession. These fortunate coincidences have also lead to attainment of conservation as a national park, so that the lakes and watershed have escaped severe disturbance from human activities, with the exception of Lake Akan 34,36 . The well-preserved natural environments of the Akan Caldera Lakes will enable a new study approach: how lakes and aquatic organisms situated in different succession stages have responded to changes in the surrounding environment, including climate change, now seen as an issue of paramount importance 26 .

Topography of lakes and watersheds
Lake area and boundary length were calculated using ARCGIS10 (Esri Japan Co.) based on the 1/25,000 numerical map data of the Geographical Survey Institute, Japan. Land area includes island areas.
Land watershed area was computed using the DEM10m data of the Geographical Survey Institute. After altitude data were changed into raster (altitude grid), subtle undulations were removed by the Fill tool, and bearing azimuth of ow was computed by the Flow Direction tool (north; 64, northeast; 128, east; 1, southeast; 2, south; 4, southwest; 8, west; 16 and northwest; 32). Accumulation value (number of cells accumulated toward the direction of ow) computed by the Flow Accumulation tool was extracted at accumulation values more than 30000 (sl30000) and more than 200 (sl200) by the Reclass command, and each watershed (ws30000, ws200) was computed by being grouped by every feeder of sl30000 and sl200 by the Stream Link tool. Finally, the Ws raster was converted into the polygon, and Ws30000 and ws200 were manually divided as a watershed for each lake, observing the DEM.
Lake volume and mean depth were calculated based on lake charts. Depth sounding for chart drawing was performed in autumn of 2014 in eight lakes, excepting Lakes Akan, Panke and Penke which already have lake charts. The whole lake was uniformly measured by a GPS sh nder (Lowrance HDS-8, Navico) on a motor boat. Elevation of the lake surface was obtained by the GNSS survey. The depth-sounding data were converted into contour drawings by chart drawing software, Reefmaster (ReefMaster Software Ltd.). Residence time was calculated with annual rainfall at 1200 mm 34 .

Water quality
Measurements of physical and chemical variables and collection of lake water were performed in ten lakes of Akan Caldera from October to November 2013 and in July 2014. Electrical conductivity (EC) and pH were directly recorded using portable sensors at the center of each lake. Water was collected at the same point using 2 l polycarbonate bottles and taken immediately to the laboratory. Dissolved oxygen (DO) and chemical oxygen demand (COD) were measured by titration with standard sodium thiosulfate solution and potassium permanganate solution, respectively. Total nitrogen (TN) and total phosphorus (TP) were measured by an auto analyzer (AACS-II, Bran + Luebbe Ltd.). Additionally, an aliquot of the water sample was ltered onto Whatman GF/F glass ber lters, and suspended solids (SS) measured gravimetrically after drying at 110°C for two hours. Chlorophyll-a (Chl-a), concentrated onto a Whatman GF/F glass lter, was quanti ed with a spectrophotometer (UV-1600, Shimadzu Co. Ltd.) after extraction using methanol (100%). The ltrate was used for measuring dissolved organic carbon (DOC) with a TOC meter (TOC VC, Shimadzu Co. Ltd.).

Macrophyte survey
Macrophytes were surveyed and collected by SCUBA diving or snorkeling along the entire shorelines of Lake Akan (August A correlation matrix of these standardized data was made and variables with strong correlation coe cients (|r| ≥ 0.7) were investigated in detail. In lakes where macrophytes occurred, the presence/absence of individual species was converted into 1/0 data, and a cluster analysis was conducted by the Ward method using the Euclid distance. The BellCurbe (Social Survey Research Information), an add-in program for Excel, was used for the analysis.  Table S2 online). (b) relationship between ratio of accumulated watershed area to lake volume (AWA/LV) and TP. The AWA/LV shows signi cant correlation with TP thought to be a parameter of eutrophication rate, but only Lake Akan (red square) is discriminated as an outlier. Similar results are found for accumulated watershed area to lake area (Table 1, Supplementary   Table S1 online). High TP speci cally observed on Lake Akan is regarded as an effect of past anthropogenic eutrophication.