Ecology

Selection of materials for joint repair of chromium-contaminated soilTable 1 shows the results of the orthogonal test. Range analysis was performed according to the results of Table 1. The range-analysis results are shown in Table 2.Table 1 Orthogonal design scheme and results.Full size tableTable 2 Orthogonal test results range analysis calculation table.Full size tableTable 2 shows that, from the perspective of unconfined compressive strength, the primary and secondary order of the 28 day strength, factors affecting the combined repair of chromium-contaminated soil were cement content → fly-ash synthetic zeolite content → CaS5 content. The best test ratio was: CaS5 content 3 times, synthetic zeolite content 15%, and cement content 20%. The unconfined compressive strength of the contaminated soil after remediation increased with the increase in cement content, but the relationship between the content of CaS5 and synthetic zeolite, and the unconfined compressive strength of the specimen was not very obvious. From the perspective of toxicity leaching, the primary and secondary order of factors affecting the total chromium leaching concentration of the combined remediation of chromium-contaminated soil were cement content → fly-ash synthetic zeolite content → CaS5 content. The primary and secondary order of factors affecting the leaching concentration of Cr(VI) in the combined remediation of contaminated soil were CaS5 content → cement content → fly-ash synthetic zeolite content. The best test ratios of the total chromium and Cr(VI) toxicity leaching test were: CaS5 content is 4 times, synthetic zeolite content 15%, and cement content 20%. Total chromium and Cr(VI) leaching concentration of the chromium-contaminated soil after joint remediation was negatively correlated with the content of CaS5, synthetic zeolite, and cement content. The change of total chromium leaching concentration was most significantly affected by cement content and synthetic zeolite. Second, the change of Cr(VI) leaching concentration was most significantly affected by CaS5 content. From the perspective of leaching concentration, when reducing agent CaS5, adsorbent synthetic zeolite, and curing agent cement were all at maximum, the leaching effect of total chromium and Cr(VI) was best. However, considering the actual engineering cost and dosage of the preparation should be reduced as much as possible for meeting the requirements. Therefore, comprehensive balance analysis determined the optimal ratio for joint repair of chromium-contaminated soil to be 3 times the dosage of CaS5, 15% synthetic zeolite, and cement amount 20%.Strength change of combined repair of chromium-contaminated soil under action of dry–wet cycleThe test compared the variation of unconfined compressive strength with the number of dry and wet cycles under different conditions of chromium content, combined to repair standard specimens of chromium-contaminated soil, and test results are shown in Fig. 1.Figure 1The relationship between unconfined compressive strength and the number of dry wet cycles.Full size imageFigure 1 shows that, in the beginning, the unconfined compressive strength of the combined repair of chromium-contaminated soil increased with the increase in the number of wet and dry cycles. After reaching the maximal value, it gradually decreased as the number of dry–wet cycles continued to increase. In the initial stage of the dry–wet cycle, the unconfined compressive strength of the combined repair of chromium-contaminated soil increased to varying degrees. For 1000 and 3000 mg/kg of chromium-contaminated soil, the peak of the unconfined compressive strength appeared at 2 times during the dry–wet cycle, and the peak of the unconfined compressive strength of 5000 mg/kg chromium-contaminated soil appeared at 4 dry–wet cycles. After that, unconfined compressive strength gradually decreased with the progress of dry–wet cycles, and the decrease rate became slower. From strength-loss analysis, the higher the chromium content was, the greater the change in strength loss. After 16 wet and dry cycles, the strength-loss rates of 1000, 3000, and 5000 mg/kg chromium-contaminated soil were 17.95%, 22.27%, and 28.73%, respectively, and strength loss was within 30%, showing better water stability21,22.From analysis of the strength-change process, after 28 days of curing for the joint repair of chromium-contaminated soil, the physical and chemical interaction between cement hydrate and soil in the repair preparation was still occurring, as was the strength increase and dry–wet cycle caused by its hydration products. The weakening effect on strength is a dynamic equilibrium process of mutual decline and growth, and the equilibrium state of the two reaction degrees directly affected the strength of solidified chromium-contaminated soil23. In the initial stage of the dry–wet cycle, the strength increase caused by the interaction between remediation agent and chromium-contaminated soil continued. At that time, the destructive effect of the dry–wet cycle on the joint repair of chromium-contaminated soil was not significant in comparison. As the number of dry–wet cycles increased, hydration products formed and became stable. Dry shrinkage and wet expansion cause internal stress in the joint repair of chromium-contaminated soil, and the soil has cracks due to internal stress changes. A dry–wet cycle has a relatively destructive effect that is gradually noticeable and resulting in a decrease in strength. After many instances of drying and wetting, the strength of repairing chromium-contaminated soil was decreased and stabilized.Figure 1 also shows that, compared with low-content chromium-contaminated soil, the high-content chromium-contaminated-soil solidified body strength peak appeared later, and the peak value was low. This is because the higher the chromium ion content was, the more serious the delay of the hydration reaction of the repair agent was, and the more obvious the weakening effect on the strength of the cured body was, which is not conducive to strength growth. The weakening effect of the dry–wet cycle on strength continued to exist, which led to the repaired contaminated soil with a high content of chromium having lower strength.Toxic-leaching changes of combined remediation of chromium-contaminated soil under dry–wet cycleThe experiment compared the variation of hexavalent chromium and total chromium leaching concentration with the number of dry–wet cycles in standard specimens of the combined repair of chromium-contaminated soil under different chromium-content conditions of the contaminated soil. Test results are shown in Fig. 2.Figure 2Effect of drying–wetting cycle timeson leaching concentration of Cr.Full size imageFigure 2 shows that the leaching concentration of Cr(VI) and total chromium decreased in the initial stage of the dry–wet cycle of the remediation of chromium-contaminated soil. After that, as the number of dry–wet cycles increased, leaching concentration also increased, but the content was low (1000 mg/kg). The medium content (3000 mg/kg) of chromium-contaminated soil Cr(VI) and total chromium leaching concentration fluctuated slightly, and the change was relatively stable, while the high content of chromium-contaminated soil (5000 mg/kg) Cr(VI) leaching the concentration fluctuated greatly, and total chromium increased significantly. Compared with the low-content chromium-contaminated soil, the leaching concentration of the solidified body of high-content chromium-contaminated soil was higher.In the beginning of the dry–wet cycle, the physical and chemical interaction between the cement hydrate and the soil in the repair preparation was still happening. The fly-ash synthetic zeolite had the adsorption effect of metal chromium ions and hydroxide precipitation in the alkaline environment. The formation of chromium ions could meet the requirements of curing/stabilizing chromium ions, and heavy-metal chromium ions are not easy to leach. With the increase in the number of dry–wet cycles, a series of evolutionary processes occurred, such as the expansion of local microcracks, the increase in macropores, the appearance of internal cracks in the contaminated soil, and the appearance of cracks and peeling phenomena on the outside of the contaminated-soil damage. At this time, the contact area between the heavy-metal ions in the contaminated soil and the external environment, especially water, increased, which reduced the ability of the repair agent to adsorb and wrap chromium ions, so that chromium ions were easily leached. In the leaching test, the use of the acidic leaching solution also destroyed the pH balance of the repaired chromium-contaminated soil, the hydrated gel was dissolved and desorbed, and the heavy metals changed, thereby accelerating the leaching of heavy-metal ions24.From analysis of the leaching law shown by the contaminated soil with different chromium content levels, when chromium content in the contaminated soil was low, the remediation agent could effectively solidify/stabilize most of the chromium ions in the soil Cr(VI) and low total chromium leaching. When the chromium content in the contaminated soil was high, the limited content of the repair agent showed an insufficient solidification/stabilization effect of the heavy-metal chromium ions. Because a higher concentration of chromium ions hindered the formation of hydration products of the repair agent, it weakened the adsorption and binding capacity of the hydrated gel. The heavy-metal chromium ions existed in the pores of the contaminated soil in a free state, making the repair agent solidify the chromium ions, the stabilization effect decreased, and the leaching of Cr(VI) and total chromium increased.Overall, the effect of the dry–wet cycle on the joint repair of chromium-contaminated soil was limited, and the joint repair of chromium-contaminated soil had strong resistance to dry–wet cycles, especially the low- and medium-content chromium-polluted soil.Combined repair of quality loss of chromium-contaminated soil under action of dry–wet cyclesThe cumulative mass-loss rate of the sample was calculated from Formula (1), and the result is shown in Fig. 3. With the increase in the number of wet and dry cycles, the cumulative mass-loss rate of the composite preparation to repair chromium-contaminated soil gradually increased; and the higher the chromium content of the contaminated soil was, the greater the cumulative mass-loss rate was. The cumulative mass-loss rate of 16 wet and dry cycles was less than 1%, which shows that the joint repair of chromium-contaminated soil had strong resistance to dry and wet cycles.Figure 3Change of cumulative mass loss rate during dry wet cycle.Full size imageFigure 4 is a photograph of the appearance change of a solidified 5000 mg/kg chromium-contaminated-soil sample after a dry–wet cycle. The soundness-evaluation results of the sample after each dry–wet cycle are shown in Fig. 5.Figure 4Appearance changes of cured chromium contaminated soil samples with dry and wet cycles at (a) 0 times; (b) 2 times; (c) 4 times; (d) 8 times; and (e) 16 times.Full size imageFigure 5Soundness evaluation results of cured chromium contaminated soil samples.Full size imageFigures 4 and 5 show that, after two dry–wet cycles of the joint repair of chromium-contaminated soil, the appearance of the sample did not significantly change, compared with 0 cycles, the surface changed from smooth to rough. Slight cracks appeared from the fourth cycle. Obvious cracks appeared in the sample at the end of the eighth cycle, and a small part of the sample fell off. The sample began to show obvious cracks from the end of the 15th dry–wet cycle, and large pieces of slack simultaneously appeared. The sample was subjected to 16 wet and dry cycles, and soundness was still not at e–h level, indicating that the joint repair of chromium-contaminated soil had strong resistance to dry and wet cycles.Combined repair of chromium-contaminated-soil microstructure changes under action of dry–wet cyclesAfter the joint repair of chromium-contaminated-soil specimens underwent a certain number of wet and dry cycles, the strength, leaching characteristics, and appearance of the specimens significantly changed. From the microstructure, there had to be corresponding changes. Therefore, scanning electron microscope (SEM) and X-ray diffraction (XRD) were used to further analyze the microstructure changes of specimens with different chromium content levels under the action of different wet and dry cycles, as shown in Figs. 6 and 7.Figure 6SEM images of 5000 mg/kg chromium contaminated soil specimens after different dry wet cycles at (a) 0 times; (b) 2 times; (c) 8 times; and (d) 16 times.Full size imageFigure 7XRD pattern of 5000 mg/kg chromium contaminated soil specimen after different dry wet cycles.Full size imageFigure 6 shows that the combined repair of chromium-contaminated soil after 28 days of curing had many pores in the specimen at 0 dry–wet cycles (standard sample), the physical and chemical interaction between the cement hydrate and the soil in the repair preparation still continued, and there were platelike calcium hydroxide crystals on the surface. After two dry–wet cycles, the contaminated soil was denser, and the overall structure was more complete than that in the samples without dry–wet cycles. The plate-shaped calcium hydroxide crystals were reduced, and a large number of fibrous and flocculent hydrated gels could be seen on the surface of the structure. This shows that the reaction between remediation agent and chromium-contaminated soil continued, which is consistent with the law that strength did not drop but rose during the two dry and wet cycles in the unconfined-compressive-strength test. After the test piece had undergone 8 dry–wet cycles, the surface of the test piece not only had a large increase in pores, but also had local cracks, indicating that the structure of the test piece was damaged under the action of the dry–wet cycle, which is consistent with the unconfined compressive strength found in the experiment, coinciding with a sharp drop. After 16 wet and dry cycles, the surface of the specimen not only showed a large number of pores and cracks, but also had obvious roughness. It showed that the dry–wet cycle effect caused the hydration products and cement materials in the soil to be destroyed and dissolved out, and the coupling and supporting forces between soil particles are weakened, and the strength of the soil is reduced accordingly, which was consistent with the macroscopic test results.Figure 7 shows that the main crystal phases of the chromium-contaminated soil were SiO2 and Al2O3 for the samples that did not undergo a dry–wet cycle. A small number of CSH, CAH, Ca(OH)2, and CaCO3 crystals could also be detected from the diffraction peaks. Cr3+ and Cr6+ formed hydroxide precipitates in a highly alkaline environment and wrapped them on the surface of cement, hindering their contact reaction with water. Compared with 0 cycles, SiO2 and Al2O3 in the second cycle were decreased, while the contents of CSH, CAH, Ca(OH)2, and CaCO3 significantly increased. This is because in the process of dry and wet cycles, the sample is fully exposed to moisture and air, so the hydration, depolymerization-cementation, pozzolanic, and carbonation reactions between composite preparation and chromium-contaminated soil continued. After two dry–wet cycles, more hydration products were generated than in the specimens without dry–wet cycles, which filled the pores between the particles of the solidified body, effectively blocking the permeability of the pores, and making the contaminated soil denser, and more structured and complete. At the same time, the full progress of the hydration reaction also delayed the damage rate of the water body to the soil in the dry–wet cycle, so that the soil could maintain a certain strength in the harsh environment, which is consistent with the above-mentioned growth trend of the soil strength. At the same time, the extension of a large amount of fibrous calcium silicate hydrate greatly increased the internal specific surface area of the soil. Free-state Cr3+ and Cr6+ were adsorbed or produced hydroxide precipitation and filled in the pores of the soil, and free ion concentration was also greatly reduced, which is consistent with the above ion-leaching test results. For the specimens with 8 dry and wet cycles, the content of hydration products such as CAH and CSH was reduced. This is due to a series of evolutionary processes such as the expansion of local microcracks, the increase in macropores, the appearance of internal cracks in the contaminated soil, and the appearance of cracks and peeling on the outside of the contaminated soil. Structural integrity was destroyed, and strength was accordingly reduced. By 16 wet and dry cycles, a large amount of fibrous CSH disappeared, which weakened the cementation between soil particles. At this time, the heavy-metal ions originally wrapped in the contaminated soil solidified the body and the external environment, the contact area with the water was increased, the pH value of the environment was decreased, hydrate CSH was decalcified, and Ca/Si ratio was decreased. This reduced the adsorption capacity of the compound formulation to chromium ions, so that chromium ions were dissolved out of the soil. More

Wildlife Conservation Research Unit, Recanati-Kaplan Centre, Zoology, University of Oxford, Oxford, UKHans Bauer & Claudio Sillero-ZubiriEvolutionary Ecology Group, Biology, University of Antwerp, Antwerp, BelgiumHans BauerLaboratory for Applied Ecology, Natural Resource Conservation, University of Abomey-Calavi, Cotonou, BeninAristide Comlan TehouDepartment of HydroSciences and Environment, University Iba Der Thiam, Thiès, SénégalMallé GueyeDirection de la Faune, de la Chasse et des Parcs et Réserves, Ministère de l’Environnement de la Salubrité Urbaine et du Développement Durable, Niamey, NigerHamissou GarbaDirection de la Faune et des Chasses, Ministère de l’Environnement et du Développement Durable, Ouagadougou, Burkina FasoBenoit DoambaNational Parks Directorate, Ministry of Environment and Sustainable Development, Dakar, SenegalDjibril DiouckThe Born Free Foundation, Horsham, UKClaudio Sillero-Zubiri More

1.Wetzel, R. G. In Limnology 3rd edn (ed. Wetzel, R. G.), Ch. 9, 151–168 (Academic Press, 2001).2.Schindler, D. Warmer climate squeezes aquatic predators out of their preferred habitat. Proc. Natl Acad. Sci. USA 114, 9764–9765 (2017).CAS 
Article 
ADS 

Google Scholar 
3.North, R. P., North, R. L., Livingstone, D. M., Köster, O. & Kipfer, R. Long-term changes in hypoxia and soluble reactive phosphorus in the hypolimnion of a large temperate lake: consequences of a climate regime shift. Glob. Change Biol. 20, 811–823 (2014).Article 
ADS 

Google Scholar 
4.Fernández, J. E., Peeters, F. & Hofmann, H. Importance of the autumn overturn and anoxic conditions in the hypolimnion for the annual methane emissions from a temperate lake. Environ. Sci. Technol. 48, 7297–7304 (2014).Article 
ADS 

Google Scholar 
5.Michalak, A. M. et al. Record-setting algal bloom in Lake Erie caused by agricultural and meteorological trends consistent with expected future conditions. Proc. Natl Acad. Sci. USA 110, 6448–6452 (2013).CAS 
Article 
ADS 

Google Scholar 
6.Schmidtko, S., Stramma, L. & Visbeck, M. Decline in global oceanic oxygen content during the past five decades. Nature 542, 335–339 (2017).CAS 
Article 
ADS 

Google Scholar 
7.Breitburg, D. et al. Declining oxygen in the global ocean and coastal waters. Science 359, (2018).8.Jankowski, J., Livingstone, D. M., Bührer, H., Forster, R. & Niederhauser, P. Consequences of the 2003 European heat wave for lake temperature profiles, thermal stability, and hypolimnetic oxygen depletion: implications for a warmer world. Limnol. Oceanogr. 51, 815–819 (2006).Article 
ADS 

Google Scholar 
9.Yvon-Durocher, G., Jones, J. I., Trimmer, M., Woodward, G. & Montoya, J. M. Warming alters the metabolic balance of ecosystems. Phil. Trans. R. Soc. B 365, 2117–2126 (2010).Article 

Google Scholar 
10.Seki, H., Takahashi, Y., Hara, Y. & Ichimura, S. Dynamics of dissolved oxygen during algal bloom in Lake Kasumigaura, Japan. Water Res. 14, 179–183 (1980).CAS 
Article 

Google Scholar 
11.Jacobson, P. C., Stefan, H. G. & Pereira, D. L. Coldwater fish oxythermal habitat in Minnesota lakes: influence of total phosphorus, July air temperature, and relative depth. Can. J. Fish. Aquat. Sci. 67, 2002–2013 (2010).CAS 
Article 

Google Scholar 
12.Harke, M. J. et al. A review of the global ecology, genomics, and biogeography of the toxic cyanobacterium, Microcystis spp. Harmful Algae 54, 4–20 (2016).Article 

Google Scholar 
13.Vaquer-Sunyer, R. & Duarte, C. M. Thresholds of hypoxia for marine biodiversity. Proc. Natl Acad. Sci. USA 105, 15452–15457 (2008).CAS 
Article 
ADS 

Google Scholar 
14.Woolway, R. I. & Merchant, C. J. Worldwide alteration of lake mixing regimes in response to climate change. Nat. Geosci. 12, 271–276 (2019).CAS 
Article 
ADS 

Google Scholar 
15.Livingstone, D. M. Impact of secular climate change on the thermal structure of a large temperate central European lake. Clim. Change 57, 205–225 (2003).Article 

Google Scholar 
16.Zhang, Y. et al. Dissolved oxygen stratification and response to thermal structure and long-term climate change in a large and deep subtropical reservoir (Lake Qiandaohu, China). Water Res. 75, 249–258 (2015).CAS 
Article 

Google Scholar 
17.Bouffard, D., Ackerman, J. D. & Boegman, L. Factors affecting the development and dynamics of hypoxia in a large shallow stratified lake: hourly to seasonal patterns. Wat. Resour. Res. 49, 2380–2394 (2013).CAS 
Article 
ADS 

Google Scholar 
18.O’Reilly, C. M. et al. Rapid and highly variable warming of lake surface waters around the globe. Geophys. Res. Lett. 42, 10773–10781 (2015).ADS 

Google Scholar 
19.Nürnberg, G. K. Trophic state of clear and colored, soft- and hardwater lakes with special consideration of nutrients, anoxia, phytoplankton and fish. Lake Reserv. Manage. 12, 432–447 (1996).Article 

Google Scholar 
20.Ho, J. C., Michalak, A. M. & Pahlevan, N. Widespread global increase in intense lake phytoplankton blooms since the 1980s. Nature 574, 667–670 (2019).CAS 
Article 
ADS 

Google Scholar 
21.Kosten, S. et al. Warmer climates boost cyanobacterial dominance in shallow lakes. Glob. Change Biol. 18, 118–126 (2012).Article 
ADS 

Google Scholar 
22.Müller, B., Bryant, L. D., Matzinger, A. & Wüest, A. Hypolimnetic oxygen depletion in eutrophic lakes. Environ. Sci. Technol. 46, 9964–9971 (2012).PubMed 

Google Scholar 
23.Winslow, L. A., Leach, T. A. & Rose, K. C. Global lake response to the recent warming hiatus. Environ. Res. Lett. 13, 054005 (2018).Article 
ADS 

Google Scholar 
24.Livingstone, D. M. An example of the simultaneous occurrence of climate-driven “sawtooth” deep-water warming/cooling episodes in several Swiss lakes. Verh. Int. Ver. Limnol. 26, 822–828 (1997).
Google Scholar 
25.Williamson, C. E. et al. Ecological consequences of long-term browning in lakes. Sci. Rep. 5, (2015).26.Rose, K. C., Winslow, L. A., Read, J. S. & Hansen, G. J. A. Climate-induced warming of lakes can be either amplified or suppressed by trends in water clarity. Limnol. Oceanogr. Lett. 1, 44–53 (2016).Article 

Google Scholar 
27.Woolway, R. I. et al. Northern hemisphere atmospheric stilling accelerates lake thermal responses to a warming world. Geophys. Res. Lett. 46, 11983–11992 (2019).Article 
ADS 

Google Scholar 
28.Carpenter, S. R., Stanley, E. H. & Vander Zanden, M. J. State of the world’s freshwater ecosystems: physical, chemical, and biological changes. Annu. Rev. Environ. Resour. 36, 75–99 (2011). Article 

Google Scholar 
29.R Core Team. R: A Language and Environment for Statistical Computing. http://www.R-project.org/ (R Foundation for Statistical Computing, Vienna, 2017).30.Borchers, H. W. pracma: Practical Numerical Math Functions. R package version 2.1.5 https://CRAN.R-project.org/package=pracma (2018).31.Winslow, L. A. et al. rLakeAnalyzer: Lake Physics Tools. R package version 1.11.4. https://CRAN.R-project.org/package=rLakeAnalyzer (2017).32.Winslow, L. A. et al. LakeMetabolizer: an R package for estimating lake metabolism from free-water oxygen using diverse statistical models. Inland Waters 6, 622–636 (2016).CAS 
Article 

Google Scholar 
33.Carslaw, D. C. & Ropkins, K. Openair – an R package for air quality data analysis. Environ. Model. Softw. 27-28, 52–61 (2012).Article 

Google Scholar 
34.Moran, P. A. P. The interpretation of statistical maps. J. R. Stat. Soc. B 10, 243–251 (1948).MathSciNet 
MATH 

Google Scholar 
35.Kalogirou, S. lctools: Local Correlation, Spatial Inequalities, Geographically Weighted Regression and Other Tools. R package version 0.2-7. https://CRAN.R-project.org/package=lctools (2019).36.Copernicus Climate Change Service (C3S). ERA5: Climate Data Store (CDS) https://cds.climate.copernicus.eu/cdsapp#!/home (accessed 1 October 2019).37.Gelman, G. & Hill, J. Data Analysis Using Regression and Multilevel/Hierarchical Models (Cambridge Univ. Press, 2007).38.Quinn, G. P. & Keough, M. J. Experimental Design and Data Analysis for Biologists (Cambridge Univ. Press, 2002).39.Lumley, T. leaps: Regression Subset Selection. R package version 3.1. https://CRAN.R-project.org/package=leaps (2020).40.Wood, S. N. Generalized Additive Models: An Introduction with R 2nd edn (CRC Press, 2017).Book 

Google Scholar 
41.Wood, S. & Scheipl, F. gamm4: Generalized Additive Mixed Models using ‘mgcv’ and ‘lme4’. R package version 0.2-5. https://CRAN.R-project.org/package=gamm4 (2017).42.Pinheiro, J. C. & Bates, D. M. Mixed Effects Models in S and S-Plus (Springer, 2000).43.Burnham, K. P., Anderson, D. R. & Huyvaert, K. P. AIC model selection and multimodel inference in behavioral ecology: some background, observations, and comparisons. Behav. Ecol. Sociobiol. 65, 23–35 (2011).Article 

Google Scholar 
44.Hosmer, D. W. & Lemeshow, S. Applied Logistic Regression 2nd edn (John Wiley and Sons, Inc., 2000).45.Homer, C. G. et al. Completion of the 2011 National Land Cover Database for the conterminous United States – Representing a decade of land cover change information. Photogramm. Eng. Remote Sensing 81, 345–354 (2015).
Google Scholar 
46.Lele, S. R., Keim, J. L. & Solymos, P. ResourceSelection: Resource Selection (Probability) Functions for Use-Availability Data. R package version 0.3-2. https://CRAN.R-project.org/package=ResourceSelection (2017).47.Cutler, D. R. et al. Random forests for classification in ecology. Ecology 88, 2783–2792 (2007).Article 

Google Scholar 
48.Liaw, A. & Wiener, M. Classification and regression by randomForest. R News 2, 18–22 (2002).
Google Scholar 
49.Messager, M. L., Lehner, B., Grill, G., Nedeva, I. & Schmitt, O. Estimating the volume and age of water stored in global lakes using a geo-statistical approach. Nat. Commun. 7, 13603 (2016).CAS 
Article 
ADS 

Google Scholar 
50.Stetler, J. T., Jane, S. F., Mincer, J. L., Sanders, M. N. & Rose, K. C. Long-term lake dissolved oxygen and temperature data, 1941–2018 ver 2. Environmental Data Initiative https://doi.org/10.6073/pasta/841f0472e19853b0676729221aedfb56 (2021).51.Adrian, R., Jane, S. F., & Rose, K. C. Widespread deoxygenation of temperate lakes – Müggelsee data. IGB Leibniz-Institute of Freshwater Ecology and Inland Fisheries dataset. https://doi.org/10.18728/568.0 (2021).52.Jenny, J.-P. Time series dataset of dissolved oxygen, water temperature and Secchi depths profiles in Lakes Annecy and Geneva. Portail Data INRAE V1, https://doi.org/10.15454/BUJUSX (2021). More

A review of existing despiking proceduresAmong despiking algorithms for raw, high-frequency, EC data, a popular approach was developed by Vickers and Mahrt6 (hereinafter VM97). The method consists in estimating the sample mean and standard deviation in overlapping temporal windows whose width in time is 5 min. The temporal window slides point by point, and any data point whose value exceeds (pm 3.5 sigma) (sample standard deviation) is flagged as a spike. The method is highly sensitive to the masking effect (where less extreme spikes go undetected because of the existence of the most extreme spikes), a reason for which the procedure is iterated increasing by 0.1 the threshold value at each pass, until no more spikes are detected.A revised version of the VM97 procedure was proposed by Metzger et al.14 (hereinafter M12), who suggested replacing the mean and standard deviation by more robust estimates, such as the median and the median absolute deviation (MAD), respectively. The authors found that this method reliably removed spikes that were not detected by VM97, showing a superior performance.To reduce the high-computational burden attributable to the windowed computations prescribed by the VM97 algorithm, Mauder et al.7 (hereinafter M13) proposed to estimate median and MAD over the whole flux averaging period (usually 30 or 60 min). M13 suggested to consider as spike those observations exceeding (pm 7cdot)MAD. Such an approach was selected as candidate method in the data processing scheme at the ICOS ecosystem stations15.Starkenburg et al9 recommended the approach developed by Brock16 (hereinafter BR86) as the best method for despiking EC data. This algorithm is currently implemented in the processing pipeline adopted by the National Ecological Observatory Network (NEON, https://www.neonscience.org). It is based on a two-stage procedure, where the first step consists in extracting the signal by means of a rolling third-order median filter which replaces the center value in the window with the median value of all the points within the window; the second step aims at identifying spikes by analyzing the histogram of the differences between the raw signal and the median filtered signal. Specifically, the differences are initially binned into 25 classes. Then, the first bins with zero counts on either side of the histogram are identified and points in the original signal that exceed the empty bins are flagged as spikes. If no bin with zero counts is found, then the number of bins is doubled (for example from 25 to 51, with one bin added ensuring to retain an odd number because the mean of differences, which is expected to be close to zero, should fall into the central bin of the histogram). The procedure is iterated by increasing the number of bins until the bin width is not less than the acquiring instrument resolution.The proposed despiking algorithmFigure 1Flowchart of the proposed despiking algorithm.Full size image
In order to define a modeling framework suitable for the representation of a sequence ((x_t)_{t in Z}) of observed raw EC data indexed by time t and contaminated by spikes, we assume a component model as follows:$$begin{aligned} left. begin{aligned} x_t&= mu _t + v_t + s_t,\ end{aligned}right. end{aligned}$$
(1)
where (mu _t) denotes the low frequency component (signal); (v_t) the deviations from the signal level (residuals) whose variability ((sigma _t^2)) is allowed to change slowly over time; and (s_t) the spike generating mechanism which is zero most of time but occasionally generates large absolute values.To achieve unbiased estimates of both the signal and the scale parameter ((sigma _t)) when data are contaminated by errors, the use of robust estimators is required. One of the most popular measures of robustness of a statistical procedure is the breakdown point, which represents the proportion of outlying data points an estimator can resist before giving a biased result. The maximum breakdown point is 50%, since, if more than half of the observations are contaminated, it is not possible to distinguish between the distribution of good data and the distribution of outlying data. Described in these terms, the arithmetic mean has a breakdown point of 0% (i.e. we can make the mean arbitrarily large just by changing any of the data point), whereas the median has a breakdown point of 50% (i.e. it becomes biased only when 50% or more of the data are large outliers).The proposed despiking procedure (hereinafter RobF) makes use of robust functionals whose breakdown point is 50% and consists in three stages (see Fig. 1). In the first step the signal ((mu _t)) extraction is carried out by means of the repeated median (RM) regression technique10,17. The second step involves the estimation of the time-varying scale parameter (sigma _t) by means of the (Q_n) estimator12. A detailed description of the robust functionals will be provided in the following sections. Spikes are detected in the third step, through the examination of outlier scores calculated as:$$begin{aligned} z_t=frac{x_t-mu _t}{sigma _t}. end{aligned}$$
(2)
Any values of (|z_t|) exceeding a pre-fixed threshold value ((z_{th})) is considered as spike. The choice of the threshold value should be based on the outlier scores data distribution which can vary across time. In this work (z_{th}) was set equal to 5 which means that for Normal- and Laplace-distributed data there is a 1 in 3.5 million and 1 in 300 chance, respectively, that an anomalous value is the result of a statistical fluctuation over the spectrum of plausible values. Once detected, spikes are removed and replaced by (mu _t) estimates obtained by the RM filter.Repeated median filterThe idea underlying moving time window based approaches is that of approximating the signal underlying observed data by means of local estimates that approximate the level of data in the center of the window.To this end, we fit a local linear trend11 of the form$$begin{aligned} mu _{t+i}=mu _t+ibeta _t, quad i=-k,ldots ,k, quad mathrm {to} quad {x_{t-k},ldots ,x_{t+k}}, end{aligned}$$
(3)
where k is the parameter defining the time window of length (n=2k+1), whereas (mu _t) and (beta _t) are estimated by means of the RM filter10 as$$begin{aligned} left. begin{aligned} tilde{mu }_t^{RM}&=medbigl (x_{t-k}+ktilde{beta }_t,ldots ,x_{t+k}-ktilde{beta }_tbigr ),\ tilde{beta }_t^{RM}&=med_{i=-k,ldots ,k} Bigl (med_{j=-k,ldots k,j ne i} frac{x_{t+i}-x_{t+j}}{i-j}Bigr ). end{aligned}right. end{aligned}$$
(4)
The only parameter required for the application of the RM filter is k, which controls how many neighbouring points are included in the estimation of (mu _t). Its choice depends not only on the time series characteristics, but also on the situations a procedure needs to handle. For despiking purposes, k has to be chosen as a trade-off problem between the duration of periods in which trends can be assumed to be approximately linear and the maximum number of consecutive outliers the estimator allows to resist before returning biased results.Results of previous studies18 for the evaluation of the RM filter performance in the removal of patches of impulsive noise showed that the RM resists up to 30% subsequent outliers without being substantially affected. Therefore, the minimal window width should be larger than at least three times the maximal length of outlier patches to be removed.To this end, the optimal time window width selection is carried out through a preliminary analysis of the data distribution. Specifically, the time series is subject to a preliminary de-trending procedure, where trend is approximated by a 5-degree polynomial function whose parameters are estimated via iterated re-weighted least squares (IWLS) regression. The optimal window width is then set equal to 4 times the maximum number of values exceeding (pm 3cdot s_g) in 30 s intervals, where (s_g) is the (global) standard deviation estimated by the (Q_n) estimator on de-trended data. To prevent cases where few or no data exceed the threshold values, a minimum window width of 5 s is imposed (i.e. 51 time steps for data sampled at 10 Hz acquisition frequency).
({{Q}}_n) scale estimatorBeyond the ability of the filter adopted for signal extraction, the effectiveness of a despiking strategy depends also on the robustness of the scale parameter, (sigma _t), which is of fundamental importance for the outlier scores derivation. Raw EC time series cannot be assumed to be identically distributed as variability may vary over time as the effect of changes in turbulence regimes and heterogeneity of the flux footprint area. In such situations, global estimates of the scale parameter are unrepresentative of the local variability. Consequently, the spike detection procedure becomes ineffective. To cope with this feature, the scale parameter (sigma _t) was estimated in rolling time windows whose width was set equal to those adopted for the signal extraction. As a robust estimates of (sigma _t), we used the (Q_n) estimator12$$begin{aligned} Q_n=2.2219{|x_i-x_j|;i0), then the process (X_t) is said to be integrated of order d, meaning that (X_t) needs to be differenced d times to achieve stationarity. To allow heteroskedasticity, we assume that (varepsilon _t= sigma _t e_t), where (e_t) is a sequence of independently and identically distributed variables with mean 0 and variance 1 and (sigma _t^2) is the conditional variance allowed to vary with time.The latter was simulated by means of a CGARCH process, which can be written as:$$begin{aligned} left. begin{aligned} sigma _t^2&=q_t + sum _{i=1}^r alpha _i (varepsilon _{t-i}^2 – q_{t-i}) + sum _{j=1}^s beta _j (sigma _{t-j}^2 -q_{t-j})\ q_t&=omega + eta _{11} q_{t-1} + eta _{21} (varepsilon _{t-1}^2 – sigma _{t-1}^2), end{aligned}right. end{aligned}$$
(7)
where (omega), (alpha _i), (beta _j), (eta _{11}), (eta _{21}) are strictly positive coefficients; (q_t) is the permanent (long-run) component of the conditional variance allowed to vary with time following first order autoregressive type dynamics. The difference between the conditional variance and its trend, (sigma _{t}^2 – q_{t}), is the transitory (short-run) component of the conditional variance. The conditions for the non-negativity estimation of the conditional variance23 are related to the stationary conditions that (alpha _i + beta _j < 1) and that (eta _{11} < 1) (such quantities provide a measure of the persistence of the transitory and permanent components, respectively).Model order specification and parameter estimation were performed by analyzing real EC data (more detail are provided in the “Results and discussion” section). With this modelling framework, we simulated 18,000 values as in EC raw data sampled at 10 Hz scanning frequency within a 30-min interval. Simulations were executed in the R v.4.0.2 programming environment by using the tools implemented in the rugarch package24.Once simulated, synthetic time series were intentionally corrupted with 180 spiky data points (1% for a sample size of 18000). Two macro-scenarios were considered. In the first scenario (S1), isolated or consecutive spike events of short duration were generated. In particular, 180 spike locations were randomly selected in such a way to obtain 30 single spikes, 30 spikes as double (consecutive) events, and 30 spikes as triple (consecutive) events. In the second scenario (S2), instead, time series were contaminated by impulsive peaks of longer duration. To this end, spike locations were carried out by randomly selecting five blocks of 50 consecutive data points. Once located, spikes were generated by multiplying the corresponding time series values (after mean removal) for a factor 10 in such a way to have magnitude similar to those commonly encountered on real, observed EC data. To simulate consecutive spike events as imposed by S2 scenario, generated spiky data points were taken in absolute term. Each scenario was permuted 99 times.MetricsThe ability of the despiking algorithms was assessed by comparing the number of artificial spikes inserted into the time series with the number of spikes identified by the method. More particularly, by referring to the (2times 2) confusion matrix as reported in Table 1, a valid despiking procedure maximizes decisions of type true positive (TP) while, at the same time, keeping decisions of the types false negative (FN) and false positive (FP) at the lowest levels possible. This trade-off can be measured in terms of Precision and Recall, which are commonly used for measuring the effectiveness of set-based retrieval25. For any given threshold value, the Precision is defined as the fraction of reported spikes that truly turn out to be spikes:$$begin{aligned} text {Precision}=frac{text {TP}}{text {TP}+text {FP}}, end{aligned}$$ (8) while the Recall is correspondingly defined as the fraction of ground-truth spikes that have been reported as spikes:$$begin{aligned} text {Recall}=frac{text {TP}}{text {TP}+text {FN}}. end{aligned}$$ (9) Table 1 Confusion matrix.Full size tableAs a measure that combines Precision and Recall, we consider the balanced F1-Score, which is the harmonic mean of the two indices above-mentioned, and given by:$$begin{aligned} text {F1-Score}=2 cdot frac{text {Precision} cdot text {Recall}}{text {Precision} + text {Recall}}. end{aligned}$$ (10) We have (0le text {F1-Score} le 1) where 0 implies that no spikes are detected and 1 indicates that all, and only, the spikes are detected. The closer to 1 the F1-Score index, the greater the effectiveness of the despiking method.In addition to the previous outlined metrics, a comparison between variances of (simulated) uncorrupted time series and the one estimated after the application of the despiking procedure has been performed.For an overall evaluation of the performance of the despiking algorithms, the Friedman test26 using a significance level (alpha =0.05), followed by a post-hoc test based on the procedure introduced in Nemenyi27 was applied. The Friedman test is a non-parametric statistical test, equivalent to repeated-measures ANOVA, which can be used to compare the performances of several algorithms28. The null hypothesis of the Friedman test is that there are no significant differences between performances of all the considered algorithms. Provided that significant differences were detected by the Friedman test (that is the null hypothesis is rejected) the Nemenyi test can be used for pairwise multiple comparisons of the considered algorithms. Nemenyi test is similar to the post-hoc Tukey test for ANOVA, and its output consists of a critical difference (CD) threshold. In order to do that, ranks are assigned to algorithms. For each data set, the algorithm with the best performance gets the lowest (best) average rank. The mean performance of two despiking algorithms is judged to be signifycantly different if the corresponding average ranks differ by at least the critical difference (the graphical output of Nemenyi test was implemented using tools provided in the R package tsutils (https://CRAN.R-project.org/package=tsutils)). More

Systematic paleontologyOrder Rodentia Bowdich, 182131Family Ischyromyidae Alston, 187632Genus Asiomys Qi, 198733Asiomys dawsoni Qi, 198733Figure 3A–EMaterial. Fragment of right calcaneus (IVPP V24417), early Middle Eocene, Huheboerhe, Irdin Manha Formation, Erlian Basin, China.Description. The bone is damaged and most probably that of a juvenile as it shows loss of the tissue in the extremities of the bone such as the calcaneal tuber and calcaneal eminence, which are usually less calcified in juveniles. The bone is relatively large (Table 1), with an elongated calcaneal tuber and a relatively short body (Fig. 3A–D). The sustentaculum tali is partly damaged; it has a subcircular articulation facet, which was probably more extended craniocaudally than mediolaterally. The caudal margin of the sustentaculum tali is inclined cranially, similar to the condition seen in species A and more than in species B (Fig. 3A). The sustentacular facet overlaps about one-half of the craniocaudal reach of the ectal facet. The groove for the ‘spring ligament’ (sensu Szalay and Decker34), which runs along the medial edge of the sustentaculum tali, is poorly pronounced. Likewise, the calcaneal groove for the tendon of the flexor fibularis muscle is shallow and poorly marked, most probably due to poor preservation. The ectal facet is relatively wide and similarly shaped as in species B (below). The peroneal process is completely damaged.Table 1 Measurements (in mm) of ischyromyid calcanei from the early middle Eocene of the Erlian Basin, Nei Mongol, China.Full size tableFigure 2Linear measurements of the calcaneus. Abbreviations: AEW, ectal facet anterior width; BL, calcaneal body length; BW, calcaneal body width; CCL, calcaneocuboid facet length; CCW, calcaneocuboid facet width; CL, calcaneus length; CMT, calcaneus maximum thickness; CW, calcaneal width; EL, ectal facet length; TEW, ectal facet total width; TL, tuber calcanei length; TT, tuber calcanei thickness; TW, tuber calcanei width; TWM, tuber calcanei width in mid-length. (Figure created in Corel Draw X4 (v.14.0.0.567) by Łucja Fostowicz-Frelik).Full size imageFigure 3Ischyromyid calcanei from the early middle Eocene of the Erlian Basin, Nei Mongol, China. (A–E), Asiomys dawsoni (IVPP V24417), right calcaneus, juvenile?; (F–K), species A (IVPP V24416), right calcaneus, adult; (L–Q), species B (IVPP V24418), right calcaneus, adult. In: A, F and L, dorsal; B, G and M, medial; C, H and N, lateral; D, I and O plantar; J and P caudal; E, K and Q, cranial views. Explanatory line drawings (right side) show important morphological features. Note sustentacular facet marked pale yellow. Scale bar equals 10 mm. (Photographs taken by Łucja Fostowicz-Frelik; drawings created in Corel Draw X4 (v.14.0.0.567) by Łucja Fostowicz-Frelik).Full size imageThe calcaneal tuber is strongly compressed, but it resembles in shape those of species A and B. A long groove for the calcaneofibular ligament is impressed on its lateral side.The anterior plantar tubercle is large and swollen, similar to that in species A, and touches the brim of the calcaneocuboid surface. The latter, only slightly damaged laterally, is round in outline, without a distinct pit, and inclined about 20–30°.Systematic remark: The fossil was associated with Asiomys dentition found in the same spot. We attribute specimen IVPP V24417 to Asiomys dawsoni, based on this fact and its distinctive size (Asiomys being the largest rodent in the assemblage). Asiomys is the only ischyromyid rodent known from the basal strata of the Irdin Manha Formation of Huheboerhe.Genus indet.Species AFigure 3F–KMaterial. Right calcaneus (IVPP V24416), early Middle Eocene, Irdin Manha Escarpment, Irdin Manha Formation, Erlian Basin, China.Description. The right almost complete calcaneus of an adult specimen is relatively large (Table 1), comparable in length to the calcaneus of a coypu (Myocastor coypus) or Asiatic brush-tailed porcupine (Atherurus macrourus). The bone has a characteristically elongated calcaneal tuber and rather short body (Fig. 3F–I). The calcaneal tuber is quite slender in comparison with the structure found in the coypu and porcupines. The shape of the bone resembles most closely the calcaneus of Paramys wortmani (see35: Fig. 12B), although in Paramys the calcaneal tuber is more compressed mediolaterally.The sustentaculum tali is large and eminent, reaching far medially and tapering, although its medial end forms a blunt edge parallel to the long axis of the bone. This medial edge also bears a well-marked but not deep groove of the calcaneonavicular (or ‘spring’) ligament (Fig. 3G). The sustentacular facet (facies articularis talaris media in Fostowicz-Frelik36: Fig. 12B2) is round, with only slight anteroposterior compression. It occupies almost the whole dorsal surface of the sustentaculum, encroaching slightly onto the calcaneal body. In that it differs from Notoparamys and Paramys wortmani, which both have a much more medially placed sustentacular facet, which does not encroach on the calcaneal body. The range of the sustentacular facet overlaps less than one-third of the ectal facet (posterior facies articularis talaris in Fostowicz-Frelik36: Fig. 12B2) on its anterior and medial sides. The calcaneal eminence is slightly longer than that in Marmota and Sciurus, in proportions closer to that of porcupines and of similar size as in Paramys wortmani. The ectal facet is wide, long, and has a distinctly helical course, even more strongly marked than in North American ischyromyids (see Rose and Chinnery35: Fig. 12A). It is, however, inclined more strongly mediolaterally than in Notoparamys and Paramys, and faces strongly medially. On the dorsal side of the calcaneal eminence, posterolateral to the ectal facet, there is a flattened rough area (finely pitted), marking the place of attachment of the lateral collateral ligaments binding the distal fibula and the astragalus with the calcaneus and stabilizing the astragalocalcaneal joint.A calcaneal body is short and stocky with poorly marked tendon ridges at the dorsal surface. A large peroneal process is partly damaged at its lateral margin. The process is placed closer to the cuboid surface than the sustentaculum tali. The position of the sustentaculum tali and the proportions of the calcaneal body of specimen IVPP V24416 resemble rather closely the calcaneus of Paramys wortmani (see35).The calcaneal tuber is not ‘pinched’ at its dorsal side but moderately compressed, thus there is no coracoid ridge posterior to the ectal facet. At the lateral side of the tuber, there is a long groove for the calcaneofibular ligament running askew, towards the dorsal surface of the calcaneal tuber. The groove for the calcaneofibular ligament is more weakly expressed than in the North American paramyines and arboreal sciurids, but similar to that of Marmota.The caudal surface of the calcaneal tuber is subcircular (only slightly more extended dorsoplantarly than mediolaterally, see Fig. 3 and Table 1). The groove for the calcaneal tendon (= Achilles tendon) is deep and placed asymmetrically at the surface (Fig. 3J). Also, the medial process of the calcaneal tuber is much better developed and extending medially.The plantar surface of the bone is almost straight with a delicate flexure cranially to a well-developed plantar heel process (Fig. 3G). The anterior plantar tubercle is relatively large, swollen, but shifted medially, towards the sustentaculum tali. It is placed very close to the cuboid surface, almost touching its margin; such location and the medial shifting resembles the condition in some ground squirrels, e.g., Cynomys (see Fostowicz-Frelik et al.8: Fig. 3D–F). The anterior plantar tubercle is also somewhat flattened and inclined medially and forms a well-marked calcaneal groove for the tendon of the flexor fibularis muscle.The calcaneocuboid articular surface is semicircular, slightly wider mediolaterally than long dorsoplantarly, which distinguishes species A from Marmota and paramyines (see35). It is almost transversally positioned, not inclined, as in most of the rodent taxa (coypu and porcupines included), and gently concave; it is also confluent and level with the cuboid pit, forming one round surface at the cranial end of the bone.Genus indet.Species BFigure 3L–QMaterial Right calcaneus (IVPP V24418), early Middle Eocene, Daoteyin Obo, Irdin Manha Formation, Erlian Basin, China.Description The bone is complete, slightly larger than in species A (Table 1), matching in length the calcaneus of the coypu. Its overall structure is very similar to the calcaneus of Paramys (either P. wortmani or P. taurus, see Rose and Chinnery35: Fig. 12B, C). It has a long and strong calcaneal tuber and a relatively strong but short calcaneal body (Fig. 3L). The tuber is more compressed mediolaterally than in species A; thus, the caudal surface of the tuber is extended more dorsoplantarly than mediolaterally (Fig. 3P). The attachment for the calcaneal tendon forms a rounded concavity at the caudal side of the tuber, and is more horizontally and symmetrically located at the surface than in species A. The lateral surface of the calcaneal tuber bears a marked scar from the calcaneofibular ligament, although the scar is convex, not concave as in species A and in other compared taxa (e.g., Cynomys).The sustentaculum tali is large and round; it is located relatively close to the calcaneal body, not extending as far medially as in the North American paramyines (see35). It is slightly longer anteroposteriorly and located more caudally (closer to the ectal facet) than in species A. Thus, the sustentacular surface overlaps ca. one-half of the cranial part of the ectal facet. The medial edge of the sustentacular shelf bears a deep groove for the ‘spring ligament’.The ectal facet is large, equally wide throughout its length, long and helical, although its course is straighter along the proximodistal direction than in species A. The ectal surface faces mediodorsally, with a slightly weaker medial component than in species A. The dorsal surface of the tuber, just caudal to the ectal facet, is not typically ‘pinched’ into a sagittally oriented crest, but it is, nevertheless, more mediolaterally compressed than in the species A, similar to Marmota.The calcaneal body forms about one-third of the bone length. Its dorsal surface is carved by deep longitudinal marks indicating the position of the extensor digitorum brevis muscle (Fig. 3). A middle-size peroneal process is located cranially at the calcaneal body. It is strong and long anteroposteriorly, reaching almost the edge of the calcaneocuboid surface. Its lateral edge shows a deep groove for the tendon of the peroneus longus muscle, while its dorsal surface forms a groove for the peroneus brevis muscle tendon (Fig. 3). Species B differs from the ground squirrels in the shape and location of the peroneal process, which is less extended laterally in species B than e.g., in marmots, although it is relatively much larger than in the coypu and porcupines.The anterior plantar tubercle looks less swollen than in species A; it is located at the very margin of the calcaneocuboid surface and as in species A is shifted medially (Fig. 3O, Q). The calcaneocuboid surface is slightly inclined (ca. 25°) anteromedially, which distinguishes the bone from species A, Marmota, and Notoparamys, which all have the calcaneocuboid facet almost transversal and perpendicular to the long axis of the calcaneus. In this respect, the calcaneocuboid surface resembles more closely the calcaneus of Paramys taurus (Rose and Chinnery35: Fig. 12C). The calcaneocuboid surface is almost round, slightly wider mediolaterally, resembling that of species A. A relatively small calcaneal pit (extending only to a half of the anterior plantar tubercle base, see Fig. 3Q), smaller but deeper than in species A, forms a shallow sink at the medial side of the surface, cranially to the sustentaculum tali.PCA analysisA Principal Component Analysis (PCA) was performed based on 14 measurements of the calcaneus. The analysis included the calcaneal measurements of five ischyromyid species (two described here as species A and B, and three comparative species from North America) and 16 extant large rodent species (Supplementary Table S1). The extant taxa represent six basic types of locomotor adaptations found in rodents: ambulatorial (terrestrial generalists), amphibious (swimming), arboreal (tree climbing), cursorial (four-pedal runners), ricochetal (bipedal jumpers), and semi-fossorial (burrowing).Principal Components 1 and 2 (PC1 and PC2) represent 87.48% and 5.75% of the variance, respectively, whereas Principal Components 3–4 represent further 4% of the variance (Supplementary Table S2). All the variables are positively correlated with PC1 and their loadings are very balanced (Fig. 4). Thus, it implies that the PC1 represents a proxy for the size of the bone. PC2 is most strongly correlated with the length of the calcaneal body, BL (-0.86) and more weakly correlated with the width of the cuboid facet (CCW) and anterior width of the ectal facet (AEW), 0.31 and 0.21, respectively (Fig. 4). The correlation with the length of the calcaneal body is an especially important factor for estimating an animal’s vertical jumping ability; the species with elongated calcaneal bodies are generally better jumpers (see8,36). The strong negative correlation of the length of the calcaneal body in the second component is illustrated by grouping the species with a strong jumping locomotor repertoire (e.g., squirrels and chinchillas) towards the left side of the plot (Fig. 4). Incidentally, this phenomenon does not concern the calcanei of ricochetal species (see the position of Pedetes versus that of Sciurus and Chinchilla: Fig. 4), where the mechanics of a jump are differently realized, and the stabilisation and relative stiffness of the ankle joint plays the most important role (thus, the calcaneal body and calcaneal tuber are more similar in size).Figure 4Principal component analysis of 14 metric parameters of rodent calcanei. The morphospace including paramyid calcanei from Nei Mongol in yellow circle. Lines connecting all data points represent a minimum spanning tree (MST) based on a Euclidean distance matrix. The loadings of the Components 1 and 2 shown at the corresponding axes. Strictly fossil taxa marked in red and pink, extant in black. (Figure created in Corel Draw X4 (v.14.0.0.567) by Łucja Fostowicz-Frelik).Full size imageIn the plot of PC1 against PC2, ischyromyids do not cluster together. Instead, the PCA morphospace is divided into two (or even three) broad groups of ischyromyid locomotor adaptations: the ambulatorial species and those with more pronounced jumping or cursorial ability. Chinese taxa fall among typically large ambulatorial rodents, such as the coypu (Myocastor) and porcupines (Atherurus and Hystrix). Closest to them there is the North American ischyromyid Quadratomus, which is somewhat shifted towards the cursorial species and can be thus distinguished as differently specialized (more cursorial). Two other North American ischyromyids, Ischyromys and Reithroparamys, are grouped with Chinchilla and Ondatra, respectively, which may imply some jumping and slightly scansorial locomotor adaptations for Ischyromys and those of typical agile generalist species for Reithroparamys.Although the sample is limited, the results of the PCA analysis point to general differences in the structure of the calcaneus, and thus, locomotor specialisation, between Asian and North American ischyromyid species. Moreover, Asian species seem to differ less from each other than the North American ones do, reflecting the overall greater species diversity and coverage of a wider niche spectrum of the North American ischyromyids.Functional and paleoecological implicationsThe studied calcanei add to our knowledge on the functional aspects of locomotion of ischyromyid rodents. Proximal tarsal morphology has been recently used to interpret the locomotor behavior of some extinct rodents (see e.g.,8,37,38,39). In the scheme of locomotor categories of Samuels and Van Valkenburgh40, attributions proposed for early ischyromyids fit into generally terrestrial41, arboreal42 or a mixture of those two35.A relatively short calcaneal body, widely spread sustentaculum tali, and a large peroneal process observed in most ischyromyid species (including these studied herein) indicate rather poor cursoriality. Instead, their ankle joint structure allows for a large freedom of foot movements in different planes. A medially extended sustentaculum tali together with a long and helically twisted ectal facet indicate a large degree of sliding between the calcaneus and astragalus along their articular facets, which makes possible a great degree of foot torsion resulting in foot eversion and inversion. This effect is further enhanced by an extended calcaneocuboid facet that is gently concave and oriented perpendicularly to the long axis of the calcaneus in species A.Such adaptations are helpful for both clinging to branches and adjusting to uneven or inclined substrate during climbing. A great degree of freedom of movement may be helpful also during burrowing, when the hind legs are used to push forward loose soil out of a burrow or an animal is forced to maintain a crouched posture, when it digs with its forelegs and head. Nevertheless, as much as the calcaneal structure may suggest some burrowing ability in ischyromyids (see Rose and Chinnery35), the rest of the postcranial skeleton known from the more complete specimens of North American representatives41 does not support fossorial adaptations. In particular, a long tail in the pre-Oligocene North American (see e.g., Paramys or Reithroparamys in Wood41: figs. 8 and 44, respectively) suggests some arboreal adaptations or at least occasional climbing, as such a tail greatly enhances balancing on uneven terrain. In contrast, typically fossorial mammals have reduced tails43.The overall morphology of dental and mandibular remains16,18 of Asian ischyromyids is similar to that of their North American counterparts16,19. As complete or even partial postcranial skeletons are unknown for the Asian ischyromyids, we can surmise their general locomotor adaptations based on calcaneal morphology which, although not in striking contrast with their North American counterparts, shows some differences.Overall, the calcaneal morphology of Chinese ischyromyids is closest to that of ground squirrels and especially porcupines (both Atherurus and Hystrix) and the coypu; the similarity to the last one is supported also by the PCA analysis. The calcaneal morphology and proportions may therefore reflect their locomotion behavior as generalized terrestrials, with a somewhat limited ability to climb (a rare but observed behavior in Hystrix) and to dig burrows (as does Atherurus43). A transverse and gently concave calcaneocuboid facet of species A facilitates foot rotation along the long axis, useful on an uneven, rocky terrain or while traversing branches, when an animal needs a flexible foot for a better grip (see Chester et al.44). On the other hand, the lack of both a characteristically bent calcaneal tuber and posteriorly located peroneal process in all ischyromyids (except for Notoparamys, see Rose and Chinnery35) argues against the arboreal adaptations characteristic of tree squirrels. More

1.Jane, S. F. et al. Distance, flow and PCR inhibition: eDNA dynamics in two headwater streams. Mol. Ecol. Resour. 15, 216–227 (2015).CAS 
Article 

Google Scholar 
2.Thomsen, P. F. & Willerslev, E. Environmental DNA: An emerging tool in conservation for monitoring past and present biodiversity. Biol. Conserv. 183, 4–18 (2015).Article 

Google Scholar 
3.Valentini, A. et al. Next-generation monitoring of aquatic biodiversity using environmental DNA metabarcoding. Mol. Ecol. 25, 929–942 (2016).CAS 
Article 

Google Scholar 
4.Harper, L. R. et al. Needle in a haystack? A comparison of eDNA metabarcoding and targeted qPCR for detection of great crested newt (Triturus cristatus). Ecol. Evol. 8, 6330–6341 (2018).Article 

Google Scholar 
5.Ficetola, G. F. et al. Replication levels, false presences and the estimation of the presence/absence from eDNA metabarcoding data. Mol. Ecol. Resour. 15, 543–556 (2015).CAS 
Article 

Google Scholar 
6.Willoughby, J. R., Wijayawardena, B. K., Sundaram, M., Swihart, R. K. & DeWoody, J. A. The importance of including imperfect detection models in eDNA experimental design. Mol. Ecol. Resour. 16, 837–844 (2016).CAS 
Article 

Google Scholar 
7.Burian, A. et al. Improving the reliability of eDNA data interpretation. Mol. Ecol. Resour. March, 1–12 (2021).
Google Scholar 
8.Klymus, K. E., Richter, C. A., Chapman, D. C. & Paukert, C. Quantification of eDNA shedding rates from invasive bighead carp Hypophthalmichthys nobilis and silver carp Hypophthalmichthys molitrix. Biol. Conserv. 183, 77–84 (2015).Article 

Google Scholar 
9.Buxton, A. S., Groombridge, J. J., Zakaria, N. B. & Griffiths, R. A. Seasonal variation in environmental DNA in relation to population size and environmental factors. Sci. Rep. 7, 46294 (2017).ADS 
CAS 
Article 

Google Scholar 
10.Mächler, E., Deiner, K., Spahn, F. & Altermatt, F. Fishing in the water: Effect of sampled water volume on environmental DNA-based detection of macroinvertebrates. Environ. Sci. Technol. 50, 305–312 (2016).ADS 
Article 

Google Scholar 
11.Spens, J. et al. Comparison of capture and storage methods for aqueous macrobial eDNA using an optimized extraction protocol: Advantage of enclosed filter. Methods Ecol. Evol. 8, 635–645 (2016).Article 

Google Scholar 
12.Djurhuus, A. et al. Evaluation of filtration and DNA extraction methods for environmental DNA biodiversity assessments across multiple trophic levels. Front. Mar. Sci. 4, 314 (2017).Article 

Google Scholar 
13.Lugg, W. H., Griffiths, J., van Rooyen, A. R., Weeks, A. R. & Tingley, R. Optimal survey designs for environmental DNA sampling. Methods Ecol. Evol. 9, 1049–1059 (2017).14.Mauvisseau, Q. et al. Influence of accuracy, repeatability and detection probability in the reliability of species-specific eDNA based approaches. Sci. Rep. 9, 1–11 (2019).
Google Scholar 
15.Willoughby, J. R., Wijayawardena, B. K., Sundaram, M., Swihart, R. K. & DeWoody, J. A. The importance of including imperfect detection models in eDNA experimental design. Mol. Ecol. Resour. 16 , 837–844 (2016).16.Griffin, J. E., Matechou, E., Buxton, A. S., Bormpoudakis, D. & Griffiths, R. A. Modelling environmental DNA data; Bayesian variable selection accounting for false positive and false negative errors. J. R. Stat. Soc. Ser. C Appl. Stat. 69, 377–392 (2020).MathSciNet 
Article 

Google Scholar 
17.Lahoz-Monfort, J. J., Guillera-Arroita, G. & Tingley, R. Statistical approaches to account for false-positive errors in environmental DNA samples. Mol. Ecol. Resour. 16, 673–685 (2016).CAS 
Article 

Google Scholar 
18.Stratton, C., Sepulveda, A. J. & Hoegh, A. msocc: Fit and analyse computationally efficient multi-scale occupancy models in r. Methods Ecol. Evol. 11, 1113–1120 (2020).Article 

Google Scholar 
19.Tingley, R., Coleman, R., Gecse, N., van Rooyen, A. & Weeks, A. Accounting for false positive detections in occupancy studies based on environmental DNA: A case study of a threatened freshwater fish (Galaxiella pusilla). Environ. DNA 00, 1–10 (2020).
Google Scholar 
20.Schmidt, B. R., Kéry, M., Ursenbacher, S., Hyman, O. J. & Collins, J. P. Site occupancy models in the analysis of environmental DNA presence/absence surveys: A case study of an emerging amphibian pathogen. Methods Ecol. Evol. 4, 646–653 (2013).Article 

Google Scholar 
21.Vörös, J., Márton, O., Schmidt, B. R., Gál, J. T. & Jelić, D. Surveying Europe’s only cave-dwelling chordate species (Proteus anguinus) using environmental DNA. PLoS ONE 12, e0170945 (2017).Article 

Google Scholar 
22.Biggs, J. et al. Using eDNA to develop a national citizen science-based monitoring programme for the great crested newt (Triturus cristatus). Biol. Conserv. 183, 19–28 (2015).Article 

Google Scholar 
23.Cantera, I. et al. Optimizing environmental DNA sampling effort for fish inventories in tropical streams and rivers. Sci. Rep. 9, 3085 (2019).ADS 
Article 

Google Scholar 
24.Dejean, T. et al. Improved detection of an alien invasive species through environmental DNA barcoding: The example of the American bullfrog Lithobates catesbeianus. J. Appl. Ecol. 49, 953–959 (2012).Article 

Google Scholar 
25.Eiler, A., Löfgren, A., Hjerne, O., Nordén, S. & Saetre, P. Environmental DNA (eDNA) detects the pool frog (Pelophylax lessonae) at times when traditional monitoring methods are insensitive. Sci. Rep. 8, 5452 (2018).ADS 
Article 

Google Scholar 
26.Nakagawa, H. et al. Comparing local- and regional-scale estimations of the diversity of stream fish using eDNA metabarcoding and conventional observation methods. Freshw. Biol. 63, 569–580 (2018).CAS 
Article 

Google Scholar 
27.Royle, J. A. & Link, W. A. Generalized site occupancy models allowing for false positives and false negative errors. Ecology 87, 835–841 (2006).Article 

Google Scholar 
28.Mackenzie, D. I. & Kendall, W. L. How should detection probability be incorporated into estimates of relative abundance?. Ecology 83, 2387–2393 (2002).Article 

Google Scholar 
29.MacKenzie, D. D., Nichols, J. D., Hines, J. E., Knutson, M. G. & Franklin, A. B. Estimating site occupancy, colonization, and local extinction when a species is detected imperfectly. Ecology 84, 2200–2207 (2003).Article 

Google Scholar 
30.Tyre, A. J., Tenhumberg, B., Field, S. A., Niejalke, D. & Possingham, H. P. Improving precision and reducing bias in biological surveys: Estimating false-negative error rates. Ecol. Appl. 13, 1790–1801 (2003).Article 

Google Scholar 
31.Dorazio, R. M. & Erickson, R. A. EDNAOCCUPUANCY: An R package for multi-scale occupancy modeling of environmental DNA data. Mol. Ecol. Resour. 18, 368–380 (2018).CAS 
Article 

Google Scholar 
32.Guillera-Arroita, G., Lahoz-Monfort, J. J., van Rooyen, A. R., Weeks, A. R. & Tingley, R. Dealing with false positive and false negative errors about species occurrence at multiple levels. Methods Ecol. Evol. 8, 1081–1091 (2017).Article 

Google Scholar 
33.Cole, D. J. Parameter Redundancy and Identi Ability (CRC Press, Boca Raton, 2020).Book 

Google Scholar 
34.Diana, A., Matechou, E., Griffin, J. E., Buxtron, A. S. & Griffiths, R. A. An Rshiny app for modelling environmental DNA data: Accounting for false positve and false negative observation error. bioRxiv https://doi.org/10.1101/2020.12.09.417600 (2020).Article 

Google Scholar 
35.Biggs, J. et al. Analytical and methodological development for improved surveillance of the great crested newt. Defra Project WC1067. (2014).36.Sewell, D., Beebee, T. J. C. & Griffiths, R. A. Optimising biodiversity assessments by volunteers: The application of occupancy modelling to large-scale amphibian surveys. Biol. Conserv. 143, 2102–2110 (2010).Article 

Google Scholar 
37.Buxton, A. S., Tracey, H. & Downs, N. C. How reliable is the habitat suitability index as a predictor of great crested newt presence or absence?. Herpertological J. 31, 51–57 (2021).
Google Scholar 
38.R-Core Team. R: language and environment for statistical computing. (2020).39.Oldham, R. S., Keeble, J., Swan, M. J. S. & Jeffcote, M. Evaluating the suitability of habitat for the great crested newt (Triturus cristatus). Herpetol. J. 10, 143–155 (2000).
Google Scholar  More

1.Bryan, B. A. et al. China’s response to a national land-system sustainability emergency. Nature 559, 193–204 (2018).ADS 
CAS 
Article 

Google Scholar 
2.Zhang, W. J., Xue, X., Peng, F., You, Q. G. & Hao, A. H. Meta-analysis of the effects of grassland degradation on plant and soil properties in the alpine meadows of the Qinghai-Tibetan Plateau. Glob. Ecol. Conserv. 20, e00774 (2019).3.Pan, T., Zou, X. T., Liu, Y. J., Wu, S. H. & He, G. M. Contributions of climatic and non-climatic drivers to grassland variations on the Tibetan Plateau. Ecol. Eng. 108, 307–317 (2017).Article 

Google Scholar 
4.Shen, H. H., Wang, S. P. & Tang, Y. H. Grazing alters warming effects on leaf photosynthesis and respiration in Gentiana straminea, an alpine forb species. J. Plant. Ecol. 6, 418–427 (2013).Article 

Google Scholar 
5.Li, G. Y., Jiang, C. H., Cheng, T. & Bai, J. Grazing alters the phenology of alpine steppe by changing the surface physical environment on the northeast Qinghai-Tibet Plateau, China. J. Environ. Manage. 248, 109257 (2019).6.Li, Y. M. et al. Changes of soil microbial community under different degraded gradients of alpine meadow. Agric. Ecosyst. Environ. 222, 213–222 (2016).Article 

Google Scholar 
7.Guo, N. et al. Changes in vegetation parameters and soil nutrients along degradation and recovery successions on alpine grasslands of the Tibetan plateau. Agric. Ecosyst. Environ. 284, 106593 (2019).8.Lin, L. et al. Predicting parameters of degradation succession processes of Tibetan Kobresia grasslands. Solid Earth 6, 1237–1246 (2015).ADS 
Article 

Google Scholar 
9.Li, H. D. et al. Assessing revegetation effectiveness on an extremely degraded grassland, southern Qinghai-Tibetan Plateau, using terrestrial LiDAR and field data. Agric. Ecosyst. Environ. 282, 13–22 (2019).Article 

Google Scholar 
10.Wang, G. X., Qian, J., Cheng, G. D. & Lai, Y. M. Soil organic carbon pool of grassland soils on the Qinghai-Tibetan Plateau and its global implication. Sci. Total Environ. 291, 207–217. https://doi.org/10.1016/s0048-9697(01)01100-7 (2002).CAS 
Article 

Google Scholar 
11.Yuan, Z. Q. et al. Responses of soil organic carbon and nutrient stocks to human-induced grassland degradation in a Tibetan alpine meadow. CATENA 178, 40–48 (2019).CAS 
Article 

Google Scholar 
12.Askari, M. S. & Holden, N. M. Quantitative soil quality indexing of temperate arable management systems. Soil Till Res. 150, 57–67 (2015).Article 

Google Scholar 
13.Lima, A. C. R., Brussaard, L., Totola, M. R., Hoogmoed, W. B. & de Goede, R. G. M. A functional evaluation of three indicator sets for assessing soil quality. Appl. Soil Ecol. 64, 194–200 (2013).Article 

Google Scholar 
14.Masto, R. E., Chhonkar, P. K., Singh, D. & Patra, A. K. Alternative soil quality indices for evaluating the effect of intensive cropping, fertilisation and manuring for 31 years in the semi-arid soils of India. Environ. Monit. Assess 136, 419–435. https://doi.org/10.1007/s10661-007-9697-z (2008).CAS 
Article 
PubMed 

Google Scholar 
15.Zhou, H. et al. Changes in the soil microbial communities of alpine steppe at Qinghai-Tibetan Plateau under different degradation levels. Sci. Total Environ. 651, 2281–2291 (2019).ADS 
CAS 
Article 

Google Scholar 
16.Yang, C., Zhang, F. G., Liu, N., Hu, J. & Zhang, Y. J. Changes in soil bacterial communities in response to the fairy ring fungus Agaricus gennadii in the temperate steppes of China. Pedobiologia 69, 34–40 (2018).Article 

Google Scholar 
17.Li, J. J. & Yang, C. Inconsistent response of soil bacterial and fungal communities in aggregates to litter decomposition during short-term incubation. Peerj 7, e8078 (2019).18.Yang, C., Li, J. J., Liu, N. & Zhang, Y. J. Effects of fairy ring fungi on plants and soil in the alpine and temperate grasslands of China. Plant Soil 441, 499–510 (2019).CAS 
Article 

Google Scholar 
19.Yang, C., Liu, N. & Zhang, Y. J. Soil aggregates regulate the impact of soil bacterial and fungal communities on soil respiration. Geoderma 337, 444–452 (2019).ADS 
CAS 
Article 

Google Scholar 
20.Wardle, D. A. et al. Ecological linkages between aboveground and belowground biota. Science 304, 1629–1633 (2004).ADS 
CAS 
Article 

Google Scholar 
21.Wu, G.-L., Ren, G.-H., Dong, Q.-M., Shi, J.-J. & Wang, Y.-L. Above- and belowground response along degradation gradient in an alpine grassland of the Qinghai-Tibetan Plateau. Clean-Soil Air Water 42, 319–323. https://doi.org/10.1002/clen.201200084 (2014).CAS 
Article 

Google Scholar 
22.Che, R. X. et al. Degraded patch formation significantly changed microbial community composition in alpine meadow soils. Soil Till Res. 195, 104426 (2019).23.Aßhauer, K. P., Wemheuer, B., Daniel, R. & Meinicke, P. Tax4Fun: predicting functional profiles from metagenomic 16S rRNA data. Bioinformatics 31, 2882–2884 (2015).Article 

Google Scholar 
24.Harris, R. B. Rangeland degradation on the Qinghai-Tibetan plateau: a review of the evidence of its magnitude and causes. J. Arid Environ. 74, 1–12. https://doi.org/10.1016/j.jaridenv.2009.06.014 (2010).ADS 
CAS 
Article 

Google Scholar 
25.Ren, G., Shang, Z., Long, R., Hou, Y. & Deng, B. The relationship of vegetation and soil differentiation during the formation of black-soil-type degraded meadows in the headwater of the Qinghai-Tibetan Plateau China. Environ. Earth Sci. 69, 235–245. https://doi.org/10.1007/s12665-012-1951-1 (2013).Article 

Google Scholar 
26.Zhang, Y. et al. Diversity of nitrogen-fixing, ammonia-oxidizing, and denitrifying bacteria in biological soil crusts of a revegetation area in Horqin Sandy Land Northeast China. Ecol. Eng. 71, 71–79. https://doi.org/10.1016/j.ecoleng.2014.07.032 (2014).Article 

Google Scholar 
27.Wang, Y. et al. Effects of grassland degradation on ecological stoichiometry of soil ecosystems on the Qinghai-Tibet Plateau. Sci. Total Environ. 722, 137910. https://doi.org/10.1016/j.scitotenv.2020.137910 (2020).ADS 
CAS 
Article 
PubMed 

Google Scholar 
28.Zhang, Y. et al. Soil bacterial and fungal diversity differently correlated with soil biochemistry in alpine grassland ecosystems in response to environmental changes. Sci. Rep. 7, 43077. https://doi.org/10.1038/srep43077 (2017).ADS 
Article 
PubMed 
PubMed Central 

Google Scholar 
29.Hartmann, M. et al. Resistance and resilience of the forest soil microbiome to logging-associated compaction. ISME J. 8, 226–244. https://doi.org/10.1038/ismej.2013.141 (2014).CAS 
Article 
PubMed 

Google Scholar 
30.Liu, S. B., Zamanian, K., Schleuss, P. M., Zarebanadkouki, M. & Kuzyakov, Y. Degradation of tibetan grasslands: consequences for carbon and nutrient cycles. Agric. Ecosyst. Environ. 252, 93–104 (2018).CAS 
Article 

Google Scholar 
31.He, S. Y. & Richards, K. Impact of meadow degradation on soil water status and pasture managementA case study in tibet. Land Degrad. Dev. 26, 468–479. https://doi.org/10.1002/ldr.2358 (2015).Article 

Google Scholar 
32.Yergeau, E., Hogues, H., Whyte, L. G. & Greer, C. W. The functional potential of high Arctic permafrost revealed by metagenomic sequencing, qPCR and microarray analyses. ISME J. 4, 1206–1214. https://doi.org/10.1038/ismej.2010.41 (2010).CAS 
Article 
PubMed 

Google Scholar 
33.Eichorst, S. A. et al. Genomic insights into the Acidobacteria reveal strategies for their success in terrestrial environments. Environ. Microbiol. 20, 1041–1063 (2018).CAS 
Article 

Google Scholar 
34.Fang, D. X. et al. Microbial community structures and functions of wastewater treatment systems in plateau and cold regions. Bioresour. Technol. 249, 684–693 (2018).CAS 
Article 

Google Scholar 
35.Mukhopadhya, I., Hansen, R., El-Omar, E. M. & Hold, G. L. IBD—what role do proteobacteria play?. Nat. Rev. Gastroenterol. Hepatol. 9, 219–230. https://doi.org/10.1038/nrgastro.2012.14 (2012).CAS 
Article 
PubMed 

Google Scholar 
36.Kjoller, A. H. & Struwe, S. Fungal communities, succession, enzymes, and decomposition (2002).37.Poll, C., Brune, T., Begerow, D. & Kandeler, E. Small-scale diversity and succession of fungi in the detritusphere of rye residues. Microbial. Ecol. 59, 130–140. https://doi.org/10.1007/s00248-009-9541-9 (2010).Article 

Google Scholar 
38.Jangid, K. et al. Land-use history has a stronger impact on soil microbial community composition than aboveground vegetation and soil properties. Soil Biol. Biochem. 43, 2184–2193. https://doi.org/10.1016/j.soilbio.2011.06.022 (2011).CAS 
Article 

Google Scholar 
39.Cao, C. et al. Soil bacterial community responses to revegetation of moving sand dune in semi-arid grassland. Appl. Microbiol. Biotechnol. 101, 6217–6228. https://doi.org/10.1007/s00253-017-8336-z (2017).CAS 
Article 
PubMed 

Google Scholar 
40.Tripathi, B. M. et al. Tropical soil bacterial communities in Malaysia: pH dominates in the equatorial tropics too. Microbial. Ecol. 64, 474–484. https://doi.org/10.1007/s00248-012-0028-8 (2012).Article 

Google Scholar 
41.Chu, H. et al. Bacterial community dissimilarity between the surface and subsurface soils equals horizontal differences over several kilometers in the western Tibetan Plateau. Environ. Microbiol. 18, 1523–1533. https://doi.org/10.1111/1462-2920.13236 (2016).CAS 
Article 
PubMed 

Google Scholar 
42.Wu, X. et al. Bacterial communities in the upper soil layers in the permafrost regions on the Qinghai-Tibetan plateau. Appl. Soil Ecol. 120, 81–88. https://doi.org/10.1016/j.apsoil.2017.08.001 (2017).Article 

Google Scholar 
43.Yang, C. et al. Assessing the effect of soil salinization on soil microbial respiration and diversities under incubation conditions. Appl. Soil Ecol. https://doi.org/10.1016/j.apsoil.2020.103671 (2020).Article 

Google Scholar 
44.Langille, M. G. I. et al. Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences. Nat. Biotechnol. 31, 814. https://doi.org/10.1038/nbt.2676 (2013).CAS 
Article 
PubMed 
PubMed Central 

Google Scholar 
45.Mermin, J. et al. Reptiles, amphibians, and human Salmonella infection: a population-based, case-control study. Clin. Infect. Dis. 38, S253–S261. https://doi.org/10.1086/381594 (2004).Article 
PubMed 

Google Scholar 
46.Wang, J. et al. Plant community ecological strategy assembly response to yak grazing in an alpine meadow on the eastern Tibetan Plateau. Land Degrad. Dev. 29, 2920–2931. https://doi.org/10.1002/ldr.3050 (2018).Article 

Google Scholar 
47.Ji, S., Geng, Y., Li, D. & Wang, G. Plant coverage is more important than species richness in enhancing aboveground biomass in a premature grassland, northern China. Agric. Ecosyst. Environ. 129, 491–496. https://doi.org/10.1016/j.agee.2008.11.002 (2009).Article 

Google Scholar 
48.Caporaso, J. G. et al. QIIME allows analysis of high-throughput community sequencing data. Nat. Methods 7, 335–336. https://doi.org/10.1038/nmeth.f.303 (2010).CAS 
Article 
PubMed 
PubMed Central 

Google Scholar 
49.Caporaso, J. G. et al. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J. 6, 1621–1624. https://doi.org/10.1038/ismej.2012.8 (2012).CAS 
Article 
PubMed 
PubMed Central 

Google Scholar 
50.Chen, W. et al. Consistent responses of surface- and subsurface soil fungal diversity to N enrichment are mediated differently by acidification and plant community in a semi-arid grassland. Soil Biol. Biochem. 127, 110–119. https://doi.org/10.1016/j.soilbio.2018.09.020 (2018).CAS 
Article 

Google Scholar 
51.Kanehisa, M. et al. KEGG for linking genomes to life and the environment. Nucleic Acids Res. 36, D480–D484. https://doi.org/10.1093/nar/gkm882 (2008).CAS 
Article 
PubMed 

Google Scholar  More

Present results showed that Hilsa shad had low nucleotide diversity (0.001809–0.008811) like most of the Clupeiforms, e.g., Elongate ilisha (0.001–0.010), Tapertail anchovy (0.0011–0.0029) in Yangtze river and Japanese anchovy (0.0014–0.0090)44,45,46. Sea fish population had higher genetic diversity than anadromous population within same species or among same group47. Although, Hilsa and Kelee shad belonged to the same subfamily Dorosomantinae but Hilsa shad is anadromous in nature and Kelee shad is exclusively marine48. Because of this habit, nucleotide diversity of Hilsa shad was lower than Kelee shad (Hilsa kelee) (0.010337–0.014690)49. Correspondingly, marine Pacific herring (0.020)50 also had higher nucleotide diversity than Hilsa shad. There were several researchers also reported low nucleotide diversity of Hilsa shad population in the Hoogli, the Ganges and the Brahmaputra river of India10,17,18. Low genetic diversity suggested that only small portion of the total population had the scope of successful spawning. That might be associated with their long anadromous breeding migration journey. At that time huge numbers of individuals were caught in their long migratory routes by the fishermen. Frequent changing of spawning pattern is another reason of unsuccessful spawning51. Therefore, Government of Bangladesh should place some safety and protection actions including, public conscious, restriction on fishing gear, Hilsa fisheries management activities and proper timing of the fishing ban period.Previous studies on genetic population structure of T. ilisha were mostly based on allozymes, allele frequencies, microsatellite DNA markers and mitochondrial DNA regions: Cytochrome b (CytB), ATPase 6&8 (ATPase), 12 s and 16 s rRNA10,15,16,17,18. However, genomic data is more powerful marker than previous markers to present the history, evolution, population status and phylogeny of a fish. Recently, A study discover the population genomics and structure of Hilsa shad in Bangladesh waters based on genomic data at NGS platform by NextRAD sequencing, however they mistakenly assigned samples collected from the confluent of the Meghna River as the north-eastern riverine group19,20. Our study was also based on genomic data at the NGS platform. Conversely, we collected sequence data of 4434 nuclear genes applying a cross-species gene enrichment method22, to examine the genetic diversity and population status of hilsa shad from the Bay of Bengal, its estuaries and all possible lotic and lentic waters and two migratory cohorts.. This study provided a solid estimation of the population status of Hilsa shad using genome-wide data and to infer its genetic diversity.Result of the maximum likelihood IQtree and the population structure suggested that the fresh, estuarine and marine water of Bangladesh have a single population of Hilsa shad. In-addition DAPC, dendrogram and network on SNP loci analysis also represented the same trend. In the phylogenetic tree, samples of all locations were mixed together without making any specific cluster. In the population structure analysis, a single population was present with some admixtured individuals bearing small portion of genes from other group. Pairwise FST value between most locations were poor with non-significant P value (P  > 0.05), that support the deprived local population differences and homogeneity of this fish population throughout our studied locations. The hilsa shad population in Bangladesh might retrieve from a collapsed population. Once upon a time (upto first half of 1990s), this fish was most available and cheap fish in Bangladesh. Because of overexploitation and lack of proper management, the fish population was collapsed more than one decade. After that period, because of fishing ban period and public consciousness (first imposed in 2011), the population started to increase. Hilsa fish production in Bangladesh has doubled in a decade from 2006–2007 (279,189 MT) to 2017–2018 (517,189 MT)4,64. This fact probably caused low genetic diversity and divergence among populations of hilsa shad in the Bangladesh waters.Bangladesh has diversified fresh water habitats for Hilsa shad migration including main river system, coastal and freshwater small rivers, hill stream rivers, haors etc. but anadromous migration of this shad starts from same marine water body, the Bay of Bengal, which is their living ground. Furthermore, this fish has highly migratory nature among marine, estuarine and fresh water bodies. Therefore, it is difficult to draw a conclusion that there is more than one population in this water system. Low variation among groups and among population within groups also did not support more than one population. FST value between most of the locations was poor with non-significant P value, which suggested that the population differences were not significant. Although in some cases, P value was significant but due to their poor FST value that did not provide strong support of local population differences. Here present findings of this study were supported by the findings of some previous researchers who represented the single gene pool or stock of this species in the Bay of Bengal with a substantial gene flow18,52,53.All of the spawning grounds of Hilsa shad were identified in the coastal areas of Bangladesh especially at the lower stretches of the Meghna, the Tetulia, the Ander Manik and the Shahabazpur River e.g., Hatia (Moulavir char) Sandwip (Kalir char) and Bhola (Dhal char and Monpura)6,21. However, migratory plan is mainly initiated during the spawning season, which is activated with follow of fresh water runoff from the inland rivers, and naturally it occurs with the commencement of the south-west monsoon and consequent flooding of all the major rivers draining down to the upper Bay of Bengal and there are no considerable differences in any context. Isolation of spawning ground is an important factor for population differentiation11. Due to presence of un-alienated spawning grounds, it is less feasible to draw population differences of Hilsa shad in the upper streams of different rivers and in their living ground, Bay of Bengal. Therefore, the unique spawning grounds and sole major migratory down-stream route strengthen the presence of single population in all over the Bangladesh water without any significant population clusters. Without specify exact spawning grounds for every cluster, it is unrealistic to draw several clusters in this population.Hilsa population studies in Indian part across the Hoogli, the Bhagirathi, the Ganges and the Brahmaputra Rivers also suggested single and genetically homogeneous population in Indian part10,17,18. Hilsa shad population of the Hoogli-Bhagirathi river system and Hilsa stock of Bangladesh water used same natal habitat, Bay of Bengal. Moreover, the River Ganges is the upstream of the Padma River (Bangladesh) and the Bhagirathi River (India) as well as the Brahmaputra is the upstream of the Jamuna River (Bangladesh). Most of the Hilsa shad of River Ganges comes from the Padma River and as the same way the Brahmaputra river has no other significant source of this fish except the Jamuna River. So genetic homogeneity and unique population across these rivers of Indian part also supported the Hilsa shad’s single population in the Bangladesh water.Nevertheless, Rahman and Naevdal (2000) based on allozymes and muscle proteins as well as Mazumder and Alam (2009) based on mitochondrial D-loop region figured out more than one Hilsa population in Bangladesh waters15,54. Rahman and Naevdal (2000) mentioned two populations: 1. Marine and 2. Estuary and fresh water but they processed without explaining how this highly migratory species was separated into two distinct cohorts. Mazumder and Alam (2009) divided the population into two clusters like previous study but poor pairwise FST value between two groups showed that there were no differences between fresh water and marine-estuarine locations. Recently Asaduzzaman et al. (2020) reported three clusters in the Hilsa population in Bangladesh waters, first one was in marine and estuarine waters and another two belonged to north–western riverine (turbid freshwater) and north-eastern riverine (clear freshwater) ecotypes20. Existing of a single population, the most likely assumption from the present research varied with their findings. Our result suggested that as a highly migratory species, Hilsa shad is incapable to belong to more than one population when sampled at different sections of their migration route. Our postulation is the presence of single cluster in the Bangladesh water because all water bodies are almost connected to each other, raising high rate of gene flow and created large population size. Western and eastern river systems of Bangladesh have immaterial dissimilar water quality (e.g., turbidity) but this is not enough to make population differences of Hilsa shad since they migrate and start their life from same spawning grounds and used almost same route across the lower stream and coastal estuaries during their breeding migration. Asaduzzaman et al. (2020) reported that samples of the Meghna river (MR) was included in the north-eastern riverine (clear freshwater) ecotypes by DAPC and neighbor-joining tree analysis20. However, their sample collection site (MR) was located in the common migratory route for north–western riverine (turbid freshwater) and north-eastern riverine (clear freshwater) ecotypes. Therefore, this site should be representing the samples of both ecotypes rather than specific one.If we draw several specific populations or clusters in the upper streams of Bangladesh that means we had the scope to find this shad in the freshwater all over the year round. However, in the freshwater of Bangladesh, this fish was available in the summer (June–October) and winter season (January-March) only; these were related to their summer and winter migration respectably55. If one or two groups of this fish, continue their complete lifecycle in the freshwater (Western/Eastern part of Bangladesh) that states the assurance of continuous supply of this fish almost year round. However, the original scenario does not support this hypothesis. Finally we can conclude that, only one population of this fish inhabit in the Bangladesh waters without any instance of different populations and clusters (2–4) but in some specific locations, they had some particular characteristics. The Bay of Bengal is their main living ground, at the time of their breeding they come to the freshwater upper streams, spawn in the estuaries and finally return to the sea. Therefore, using all the same ecosystems (sea, estuary and freshwater rivers) in a cyclic fashion is essential to support their life cycle, which certainly pushes all the individuals to belong a unique population.In the population structure analysis, only one population of Hilsa shad was identified with some admixtured individuals (32%) containing partial genes from other population in the water bodies of Bangladesh. The mentioned other population might not represent the Hilsa population of the Hoogly and Bhagirathi river system, India because, the Hilsa shads of both migratory routes of Bangladesh and India showed genetic homogeneity10,17. The Ganges and Brahmaputra rivers of Indian part are the upstream of the Padma and the Jamuna river of Bangladesh and might be belonged to the same population. However, Hilsa population of the Arabian Sea was genetically heterogeneous from the Bay of Bengal18 and those different population genes of admixtured individuals might come from the Arabian Sea by oceanographic dispersion. Once (almost 18,000 years ago) the Arabian Sea had a close connection with the Bay of Bengal through the Laccadive Sea, the Gulf of Mannar and the Palk Bay. Therefore, this likely was an easy way for oceanographic dispersion of Hilsa shad between these two water bodies. After that period, a bridge of limestone shoals, coral reefs and tombolo called as ‘Ram Bridge’ or ‘Adam’s Bridge’ (about 48 km) originated between Pamban Island off the south-eastern coast of Tamil Nadu, India, and Mannar Island, off the north-western coast of Sri Lanka 56,57. Sarker et al. (2020) also mentioned that type of oceanographic dispersion between these two water bodies for another Clupeid fish species, Hilsa kelee49. The Irrawaddy, the Naaf and the Sittang River of Myanmar were also regarded as another important route for Hilsa migration6,58. There is also a possibility of inflowing of these different genes of other population from such population. Still there is no population structure study was conducted in the Myanmar part. Therefore, there is no scope to compare those admixtured individuals with the Hilsa population of Myanmar. However, for completing the full scenario, the Hilsa population of Myanmar also claims research attention in population genomics field.In the present study, Samples of both migration cohorts (summer and winter) were studied. The maximum likelihood IQ tree, population structure and DAPC suggested that samples of both migration cohorts were homogenous. Similarly, Jhingran and Natarajan (1969) and Ramakrishnaiah (1972) also did not find any significant temporal population differences in their previous studies59,60. Dwivedi (2019) found morphometric variations between seasonal migrants of Hilsa shad from Hooghly estuary, India using geometric morphometrics (GM) data61. They explained that these morphotypes might be related to the food availability and temperature fluctuation of summer and winter season but they did not incorporate that to the genetic level of the population. Quddus et al. (1984) reported two seasonal migratory populations of Hilsa shad in Bangladesh water based on spawning, fecundity and sex ratio8. Based on our findings and previous studies we can conclude these mentioned seasonal cohorts might be associated with their food availability and breeding rather than genome level.Hill stream river and haor were two important and unique ecosystems for fish diversity in Bangladesh, they belong to the unique characteristics in the ecological factors as well as fish diversity62,63. Infrequently Hilsa shad use these two water bodies as their migratory routes. Samples were collected from the Shomeswari River and the Dingapota Haor, Mohanganj as the representatives of hill stream river and haor population respectively. However, Hilsa shad of these two exclusive water bodies were similar to the samples of the some other fresh water bodies (i.e., CM, CN and MG) as they were belonging to the Hilsa population without any admixtured individuals. Samples of SS do not have any significant P value with other locations whereas MO samples had significant P value with five other locations but having poor FST value with three locations (i.e., BL, PP, MG). MO samples had only mentionable FST value with MP (estuarine) and MK (marine), which might be the result of differences in water quality of these two water bodies. In DAPC, phylogenetic tree and in network, the samples of hill stream river and haor failed to make any unique cluster or monophyletic clade that represent they are also the part of single unique Hilsa population of Bangladesh waters.Main migration was occurred through the Meghna river estuary, which is connected to the Padma, Meghna and Jamuna river system. However, there are some other alternative routes through some small coastal rivers e.g., the Pashur, the Bishkhali, the Balaswar, the Kocha river, which are connected to the Padma river through the Modhumati and the Gorai river. These coastal rivers passed through or beside the world largest mangrove forest Sundarban. Thus, these two routes are ecologically different from each other. Samples of these two routes have some genetic differences, because most of the locations (MK, CF and BL with PP and KN) of these two estuarine routes had significant P value, but their FST value was not satisfactorily high to make population differences. Ecological differences of these two routes might be played an important role to create this type of slight differences among them. Therefore, these scenarios were not significant enough to describe noteworthy differences in the population level, but may make a sign of upcoming population differences. More