Ecology

Here, we examine infection dynamics in the four regions of focus to learn more about how various control interventions performed in each region. Since the first documented cases emerged in these regions, the virus was able to spread freely across much of the first and second phases with a gradual increase in the control interventions. In the four regions of Makkah, Madinah, Eastern, and Riyadh, cases peaked on 12th May (6397 cases; 95% CI 5960–9697), 15th May (1967 cases; 95% CI 1625–2308), 23rd June (10367; 95% CI 8948–11785) and 11th June (11273 cases; 95% CI 11068–12491) respectively according to the fitted model shown in Fig. 2. As the epidemic progressed, more measures were adopted to contain the disease, and the disease’s infectiousness sharply decreased after the third period. There are a few factors responsible for the sudden declining trend: first our model is dependent on official data on cases that have been documented, and these data will only ever reflect a portion of the overall number of cases. Second, different regions developed different testing strategies, and some locations altered their approach to testing during the course of the time period that was investigated. It is possible that the beginning of the Hajj term (period 4) was a contributing factor in the decrease in the number of documented cases. Additionally, previous to this time period, the government indicated that it would be increasing the size of its local testing in order to detect new cases. It is possible that the efficacy of interventions would be reduced if increases are found to be occurring during the falling phase of an epidemic. This may result in measures being kept in place for a longer period of time than they would have been had more accurate data been provided.Figure 2With the use of the Delay Rejection Adaptive Metropolis method, the relevant parameters were estimated for each of the four areas of interest by fitting the data from 13th March until 25th September.Full size imageWe estimate the effective reproduction number (R_t) as an indicator of SARS-CoV-2 transmission before and after the interventions. Figure 3 depicts the dramatic shift in the rate of SARS-CoV-2 transmission as a result of decreased social contact and other control measures. At the beginning of the pandemic, (R_t) for SARS-CoV-2 in Saudi regions was between 4 and 6 as illustrated in Tables 6 and 7. In other words, on average each case spread to between four and six others. Considering that each new generation of SARS-CoV-2 cases occurs every five days, it is evident that this pandemic was rapidly expanding out of control. Moreover, we assumed that the transmission rate and the documented infection rate did not change during the first two periods since interventions were carried out gradually until a complete lockdown took place. As more measures were introduced, the spread of the disease began to decrease. Therefore, our data were based on the weekly reported number of documented SARS-CoV-2 cases broken down by region. As a result, it became clear that the reliability of the (R_t) value was relatively high for transmission.Figure 3Distribution of Rt estimates derived from 10000 MCMC samples for Makkah, Madinah, Eastern, and Riyadh, respectively. The black dot in the centre of each violin plot denotes the median, the thick bar in the plot denotes the interquartile range, and the thin bar in the plot denotes the lowest and maximum values. The mean and the credible interval for 95%, which is shown in parentheses, are labelled below or above, respectively.Full size imageThe effects of the events and interventions on the dynamics of SARS-CoV-2 in the regions of interest are considered. First, if the controls remained in phase four in Makkah, our model projects that the total number of documented cases would increase to 81047 (95% CI 79421–82672). In Al-Madinah, the cumulative number of documented cases would have increased to 22997 (95% CI 19578–26415). The number of cumulative documented cases may have reached 80520 (95% CI 78335–82704) if controls stayed steady in the Eastern region at the level they were at in phase four. If the pattern shown during the fourth period is taken into account, we estimate that there would have been 67150 (95% CI 63731–70568) documented infections in the Riyadh region. Figure 4 illustrates these findings.Figure 4The relevant parameters were estimated for each of the four regions of interest by first fitting the data of each region, and then predicting using the parameters from period 4. This was done with each of the four regions of interest separately.Full size imageWe now explore the impact of controls remaining in place at the same level as that implemented in phase three. In that case, the number of documented cases in Makkah would have increased to 116641 (95% CI 105015–128266). Similarly, the total number of documented cases in Al-Madinah would have increased to 53877 (95% CI 50458-57295) if the outbreak had been allowed to continue at the same level. If the controls had remained unchanged from how they were in phase three in the Eastern region, the total number of documented cases would have been 310459 (95% CI 298362–334981). Finally, in Riyadh this would have resulted in 665241 documented cases (95% CI 651822 to 678659). Figure 5 highlights these findings.Figure 5For each of the four areas of interest, the relevant parameters were estimated by first fitting the data of each region and then predicting using the parameters from period 3. This was carried out for each of the four areas of interest independently.Full size imageWe now investigate the impact of second-period controls remaining in place. In that case, the number of documented cases would increase to 1236642 (95% 1218314–1251626), 442865 (95% CI 439446–456283), 454031 (95% CI 441846–466215), and 2322624 (95% CI 1919206-3026042) in the regions of Makkah, Madinah, Eastern, and Riyadh, respectively (see Fig. 6).Figure 6For each of the four regions of interest, the relevant variables were identified by first fitting the data from each area and then making predictions using the parameters from period 2. Each of the four regions of interest was done separately.Full size imageThe efficacy of NPIs is dependent on when they are adopted, with earlier adoption resulting in greater success in lowering transmission rates of infectious diseases. In the early stages of COVID-19, Saudi regions made the decision to gradually implement measures in order to understand the severity of the disease and reduce the economic and social costs of lockdowns, as well as the political costs. In Fig. 6, if the government were to rely on the interventions of the second phase, then the number of cases of infection would considerably rise owing to the ineffectiveness of the measures. In the third period as in Fig. 5, the government made it possible to relax some of the control measures, but it is ultimately up to each area to decide whether they will maintain the same level of control or whether they will increase or decrease it. In comparison to the control measures carried out during the third and fourth periods, this led to significantly improved outcomes. The reason that these time periods were chosen is that there was no stiffening of the NPI response in most Saudi regions during the first two periods and control interventions were improved later on.Significant undetected infections resulted in the fast spread of new coronaviruses (SARS-CoV-2) which is illustrated in Fig. 7. The proportion of undocumented infections, including asymptomatic cases and undocumented symptomatic individuals who did not seek medical treatment or be tested for mild symptoms, was greater than that of Wuhan at the onset of the pandemic26, which may be a result of the following factors: first, the medical configuration was not optimal and public awareness was limited during the onset of the pandemic while the undocumented rate progressively increased; Second, contact tracing procedures employed in Saudi regions may have become overwhelmed if the number of early-stage cases in Saudi regions rises substantially. The discrepancy between the predicted proportions of asymptomatic (undocumented) cases may be attributable to the difficulty in the un-identifiability of parameters in epidemiological models. There were a substantial number of asymptomatic infected individuals with high infectivity in Saudi regions, where the epidemic situation escalated rapidly. Our research emphasises the frequency of asymptomatic SARS-CoV-2 cases and their role in transmission in order to increase people’s knowledge of asymptomatic cases and to serve as a guide for the prevention and control of SARS-CoV-2.Table 3 Estimated transmission rate in Saudi Regions.Full size tableTable 4 Estimated ascertainable infection rate.Full size tableIn this model, we fitted dynamic transmission rates because of varied preventable measures by the Saudi government at the level of the country or region. After a series of actions taken by the government, regions and cities went into lockdown, resulting in a decrease in the transmission rate as in Table 3. Before the interventions were introduced, in the first two periods of our study, we assumed the transmission rate did not change since individual and community responses had not effectively taken place. After severe interventions were implemented, the transmission rates were allowed to vary in later periods and reduced gradually due to the control measures that reduced the spread of disease27. Estimates of documented infection rates are presented in Table 4. Our model estimates show the documented infection rate has continued to decrease in the last two periods. Thus, the parameters we fit across periods are a measure of how effective the lockdown was in bringing down the documented infection rate28.Risk of resurgenceThe risk of resurgence in Saudi Arabia’s four regions has been examined in this section after the relaxation of intervention measures. There will be a rise in disease activity if control measures are relaxed without taking into account increases in the number of cases being detected, isolated, and/or traced. We predict the first week of no new cases of infection and the week when all current infections in Saudi Arabia will be eradicated.In the Makkah region, had the trend continued into the fourth period, the number of documented infections would have dropped to zero on average by the 6th September (23rd August to 27th September), and all infections would have been eradicated by the 26th of October (7th October to 14th November). On the 28th June, the number of weekly active infections (including presymptomatic, symptomatic, and asymptomatic cases) reached its highest point of 230,230 (95% CI 226811–234364), and on 8th September, that number dropped to 44023 (95% CI 40604–47441).Therefore, the number of documented infections would have reached zero in Al-Madinah region on average on 6th November (23rd October to 22nd November), and all infections would have been eliminated by 1st December (27th November to 14th December). On 23rd June, weekly active infections (including presymptomatic, symptomatic, and asymptomatic cases) peaked at 130,134 (95% CI 126715–133552) and then declined to 60023 (95% CI 58604–63441) on 25th September.If the trend had continued as it did in the fourth period in the Eastern region, the average number of documented infections would have reached zero on 2nd November (from 23rd October to 18th November), and the total eradication of infections would have happened on 1st December (26th November to 22nd December). The number of weekly active infections (including presymptomatic, symptomatic, and asymptomatic cases) peaked at 65000 (95% CI 61581–68418) during the week of July 23rd and subsequently decreased to 800 (95% CI 765–834) on 8th of September.Lastly, the model predicted that the number of weekly active infections in the Riyadh region (including presymptomatic, symptomatic, and asymptomatic infections) peaked on 28th June at 562332 (95% CI 513379–619542) and then decreased to 188215 (95% CI 174796–191633) on 18th September. On average, we expected that the number of documented infections would have decreased to zero on 18th October (7th October to 14th November) and that the total number of infections would have been eliminated on 1st December if the trend continued as it did in the fourth period (20th November to 23rd December). Figure 7 illustrates these findings. We found that if control measures were lifted 30 days following the first day of zero documented cases.Figure 7The estimated number of infected cases that were active (presymptomatic, symptomatic, and asymptomatic) during the research period in the areas of Makkah, Madinah, Eastern, and Riyadh respectively.Full size imageThe probability of resurgence, which we define as the number of active documented cases greater than 100 could be as high as 0.96 in Eastern, 0.95 in Madinah, 0.97 in Makkah, and 0.96 in Riyadh. If we adopt more stringent conditions of lifting controls after observing no confirmed cases for a continuous period of 30 days, the probability of resurgence decreases to 0.31, 0.28, 0.30, and 0.30, with probable resurgence occurring on 13th February, 7th February, 2nd January, and 8th January for Eastern, Makkah, Madinah and Riyadh, respectively (Fig. 8). Despite the use of a simplified model, these results emphasize the hazards of ignoring undetermined occurrences when modifying intervention techniques.Figure 8Figure demonstrating the effect of relaxing all control measures in all four regions 30 days following the first day without confirmed cases.Full size imageSensitivity analysisFor the purpose of testing the robustness of our research results, we conducted a series of sensitivity analyses by varying the durations of the latent and infectious periods, the ratio of transmissibility in asymptomatic (undocumented) cases to symptomatic (documented) cases, and the initial documented infection rate. We conduct eight sensitivity analyses (S1 to S8) within each model for each region of Saudi Arabia to assess the robustness of our model results. For instance, the sensitivity analysis performed for S1 was based on the changes of the latent period and pre-symptomatic infectious period, respectively, and other parameters remain the same. These modifications were carried out with the help of reference15,29, and the same approaches were used for the other parts of the sensitivity analysis, which is summarised in Table 5.Table 5 Description of essential model parameters that were not fitted in the MCMC, where (D_e) refers to the latent period, (D_p) refers to the pre-symptomatic infectious period, (D_i) refers to the symptomatic infectious period, (gamma _0) refers to the initial ascertain rate and (alpha) refers to the ratio of the transmission rate for P and A to I.Full size tableIn particular, for (S1), we raised the incubation period to 7 days (upper 95% CI based on ref15) and the pre-symptomatic infectious period to 3 days (upper 95% CI based on ref29). Therefore we set (D_e = 4) and (D_p=3), and modified (E_0) and (P_0) as needed. The transmissibility of the undocumented cases was assumed to be 0.46 (lower 95 % CI according to ref.31) of the infection cases for (S2); for (S3), the transmissibility of the asymptomatic (undocumented) cases was assumed to be 0.62 (upper 95 % CI according to ref31). We assumed that in (S4), the initial documented infection rate was (gamma _0) = 0.14 (lower 95 % CI according to ref13) and adjusted (A_0), (P_0) and E(0) accordingly. Similarly for (S5) we assumed the initial documented infection rate was (gamma _0) = 0.42 (upper 95 % CI according to ref13) and adjusted (P_0), (A_0), and (E_0) accordingly. In (S6) we set the variables (D_ e=3) and (D_p=1.1), and altered the values of (P_0) and (E_ 0) as necessary in accordance with13. In (S7) we assumed that the transmission rate of asymptomatic (undocumented) cases was half that of documented cases by setting 0.5. Finally, in (S8) we assumed that the infectious period ((D_i)) was double that of symptomatic cases by setting 6 days. Both (S7) and (S8) were based on30. The results of our sensitivity analysis are summarised in Tables 6 and 7. We note that the variation in the model predictions of (R_t) varies from setting to setting. However, these variations appear to be fairly small, proposing the robustness of the results to the specification of associated values in fairly realistic ranges13,13. Our sensitivity analysis provides information about the importance of each parameter to the model representing the transmission of SARS-CoV-2. An increase (or decrease) in parameter values, while other parameters’ values remain the same, contributes to an increase (or decrease) in effective reproduction numbers. For example, an increase in infectious period would result in a higher effective reproduction number at the beginning of the epidemic and a longer time required to clear all infections in Saudi regions32. Our sensitivity analysis indicates that almost all model parameters may have an important role in spreading this virus among susceptible people. In particular, the contact rate from person-to-person and the transition rate of asymptomatic (undetected) individuals play a significant role in disease spread. Our important findings, of a significant decrease in (R_t) after interventions and the existence of a substantial number of presymptomatic and asymptomatic cases, were found to be robust. This highlights that Saudi authorities should pay attention to intervention strategies in the event of a resurgence of cases and quarantining those who were in contact with active cases can effectively reduce the disease33. In Tables 6 and 7 we show the estimated effective reproduction number (R_t) associated with 95% CIs obtained from those eight sensitivity analyses for all four regions and all five time periods.Table 6 Sensitivity analysis of the effective reproduction number (R_t) for Eastern and Madinah.Full size tableTable 7 Sensitivity analysis of estimated effective reproduction number (R_t) for Makkah and Riyadh.Full size table More

Wilkinson, D. M. Dispersal biogeography. Encyclopedia of Life Science. (Nature Publishing Group, 2001).Jan, P.L. et al. Range expansion is associated with increased survival and fecundity in a long-lived bat species. Proc. R. Soc. B. 286, 1–9. https://doi.org/10.1098/rspb.2019.0384 (2019).Article 

Google Scholar 
IUCN. IUCN Guidelines for the Prevention of Biodiversity Loss Caused by Alien Invasive Species. https://portals.iucn.org/library/node/12413(2000).McKinney, M. L. Urbanization as a major cause of biotic homogenization. Biological Conservation 127, 247–260. https://doi.org/10.1016/j.biocon.2005.09.005 (2006).
Google Scholar 
Galko, J. et al. Invázne a nepôvodné druhy v lesoch Slovenska: hmyz—huby—rastliny. (Národné lesnícke centrum, 2018).Lockwood, J. L., Hoopes, M. F. & Marchetti, M. P. Invasion ecology_draft_2ed. (John Wiley & Sons, 2013).Colautti, R. I. & MacIsaac, H. J. A neutral terminology to define ‘invasive’ species: Defining invasive species. Divers. Distrib. 10, 135–141. https://doi.org/10.1111/j.1366-9516.2004.00061.x (2004).Article 

Google Scholar 
Ambros, M., Dudich, A., MIKLÓS, P., Stollmann, A. & Žiak, D. Ryšavka tmavopása (Apodemus agrarius)–novỳ druh cicavca Podunajskej roviny (Rodentia: Muridae). Lynx, series nova 41, (2010).Bradley, B. A. et al. Disentangling the abundance—impact relationship for invasive species. Proc. Natl. Acad. Sci. 116, 9919–9924. https://doi.org/10.5281/zenodo.2605254 (2019).Article 
ADS 
CAS 

Google Scholar 
Nackley, L. L., West, A. G., Skowno, A. L. & Bond, W. J. The nebulous ecology of native invasions. Trends Ecol. Evol. 32, 814–824. https://doi.org/10.1016/j.tree.2017.08.003 (2017).Article 

Google Scholar 
Archer, S. Tree-grass dynamics in a Prosopis-thornscrub savanna parkland: reconstructing the past and predicting the future. Ecoscience 2, 83–99. https://doi.org/10.1080/11956860.1995.11682272 (1995).Article 

Google Scholar 
Wigley, B. J., Bond, W. J. & Hoffman, M. T. Thicket expansion in a South African savanna under divergent land use: local vs. global drivers? Global Change Biol. 16, 964–976. https://doi.org/10.1111/j.1365-2486.2009.02030.x (2010).Spiegel, C. S., Hart, P. J., Woodworth, B. L., Tweed, E. J. & LeBrun, J. J. Distribution and abundance of forest birds in low-altitude habitat on Hawai’i Island: Evidence for range expansion of native species. Bird Conserv. Int. 16, 175–185. https://doi.org/10.1017/S0959270906000244 (2006).Article 

Google Scholar 
Livezey, K. B. Range expansion of Barred Owls, part II: facilitating ecological changes. Am. Midland Nat. 323–349 (2009).Veech, J. A., Small, M. F. & Baccus, J. T. The effect of habitat on the range expansion of a native and an introduced bird species. J. Biogeogr. 38, 69–77. https://doi.org/10.1111/j.1365-2699.2010.02397.x (2011).Article 

Google Scholar 
Krupa, J. J. & Haskins, K. E. Invasion of the meadow vole (Microtus pennsylvanicus) in southeastern Kentucky and its possible impact on the southern bog lemming (Synaptomys cooperi). American Midland Naturalist 14–22 (1996).Jareño, D. et al. Factors associated with the colonization of agricultural areas by common voles Microtus arvalis in NW Spain. Biol. Invasions 17, 2315–2327. https://doi.org/10.1007/s10530-015-08774 (2015).Article 

Google Scholar 
Malygin, V. M., Baskevich, M. I. & Khlyap, L. A. Invasions of the common vole sibling species. Russ. J. Biol. Invas. 11, 47–65 (2020).Article 

Google Scholar 
Tong, X., Wang, R. & Chen, X.-Y. Expansion or invasion? A Response to Nackley et al. Trends Ecol. Evol. 33, 234–235. https://doi.org/10.1016/j.tree.2018.01.008 (2018).Thompson, K. & Davis, M. A. Why research on traits of invasive plants tell us very little. Trends Ecol. Evol. 24,115–116. https://doi.org/10.1016/j.tree.2011.01.007 (2011).Davis, M. A. & Thompson, K. Eight ways to be a colonizer; two ways to be an invader: A proposed nomenclature scheme for invasion ecology. Bull. Ecol. Soc. Am. 81, 226–230 (2000).
Google Scholar 
Davis, M. A., Thompson, K., Grime, J. P. & Charles, S. Elton and the dissociation of invasion ecology from the rest of ecology. Diversity and Distribution 7, 97–102. https://doi.org/10.1046/j.1472-4642.2001.00099.x (2001).Article 

Google Scholar 
Davis, M. A. Invasion biology. (Oxford University Press on Demand, 2009).Davis, M. A. et al. Don’t judge species on their origins. Nature 474, 153–154 (2011).Article 
CAS 

Google Scholar 
Klempa, B. et al. Complex evolution and epidemiology of Dobrava-Belgrade hantavirus: Definition of genotypes and their characteristics. Adv. Virol. 158, 521–529. https://doi.org/10.1007/s00705-012-1514-5 (2013).CAS 

Google Scholar 
Kraljik, J. et al. Genetic diversity of Bartonella genotypes found in the striped field mouse (Apodemus agrarius) in Central Europe. Parasitology 143, 1437–1442. https://doi.org/10.1017/S0031182016000962 (2016).Article 

Google Scholar 
Latinne, A. et al. Phylogeography of the striped field mouse, Apodemus agrarius (Rodentia: Muridae), throughout its distribution range in the Palaearctic region. Mamm. Biol. 100, 19–31. https://doi.org/10.1007/s42991-019-00001-0 (2020).Article 

Google Scholar 
Dudich, A. & Szabó, I. Über die Verbreitung der Hystrichopsylla Taschenberg, 1880 (Siphonaptera) in Ungarn. Folia Entomol. Hung 45, 27–32 (1984).
Google Scholar 
Dudich, A. Dynamika areálu ryšavky tmavopásej (Apodemus agrarius Pall.)–expanzia či invázia. Pp.: 53–62. Invázie a invázne organizmy. SEKOS pre SNK SCOPE, Nitra (1997).Karaseva, E. V., Tikhonova, G. N. & Bogomolov, P. L. Distribution of the Striped field mouse (Apodemus agrarius) and peculiarities of its ecology in different parts of its range. Zoologičeskij žurnal 71, 106–115 (1992).
Google Scholar 
Polechová, J. & Graciasová, R. Návrat myšice temnopásé, Apodemus agrarius (Rodentia: Muridae) na jižní Moravu. Lynx (Praha), ns 31, 153–155 (2000).Bryja, J. & Řehák, Z. Další doklady současné expanze areálu myšice temnopásé (Apodemus agrarius) na Moravě. Lynx (Praha), ns (2002).Herzig-Straschil, B., Bihari, Z. & Spitzenberger, F. Recent changes in the distribution of the field mouse (Apodemus agrarius) in the western part of the Carpathian basin. Annalen des Naturhistorischen Museums in Wien. Serie B für Botanik und Zoologie 421–428 (2003).Dudich, A., Ambros, M., Stollmann, A., Uhrin, M. & Urban, P. Ryšavka tmavopása Apodemus agrarius (Pallas) v Novohrade. Príroda okresu Veľký Krtíš – 15 rokov od celoslovenského tábora ochrancov prírody 110–115 (2003).Horáček, I. & Ložek, V. Biostratigraphic investigation in the Hámorská cave (Slovak karst). Pp.: 49–60. Krasové sedimenty. Fosilní záznam klimatickỳch oscilací a změn prostředí. Knihovna České speleologické společnosti, Svazek 21, (1993).Pazonyi, P. & Kordos, L. Late Eemian (Late Pleistocene) vertebrate fauna from the Horváti-lik (Uppony, NE Hungary). Fragmenta Palaeontologica Hungarica 22, 107–117 (2004).
Google Scholar 
Obuch, J., Danko, S. & Noga, M. Recent and subrecent diet of the barn owl (Tyto alba) in Slovakia. Slovak Raptor J. 10, 1–50 . https://doi.org/10.1515/srj-2016-0003 (2016).Article 

Google Scholar 
Obuch, J. Temporal changes in proportions of small mammals in the diet of the mammalian and avian predators in Slovakia. Lynx, n. s 55, 86–106. https://doi.org/10.37520/lynx.2021.007 (2022).Niethammer, J. Die Verbreitung der Brandmaus (Apodemus agrarius) in der Bundesrepublik Deutschland. Acta Sc. Nat. Brno 10, 43–55 (1976).
Google Scholar 
von Lehmann, E. Die Brandmaus in Hessen als Beispiel für die Problematik der Verbreitungsgrenzen vieler Säugetierarten. Natur und Museum 106, 112–117.Farský, O. Úlovky myšice temnopásé, Apodemus agrarius (Pallas), na Moravě a ve Slezsku v letech 1920 až 1940 [Caughts of the striped field mouse, Apodemus agrarius (Pallas), in Moravia and Silesia in the years 1920–1940]. Lynx, n. s 5, 11–18 (1965).Heroldová, M., Homolka, M. & Zejda, J. Některé nepublikované nálezy Apodemus agrarius v Čechách a na Moravě v návaznosti na současnỳ stav znalostí o jejím rozšíření (Rodentia: Muridae). Lynx, series nova 44, (2013).Ambros, M., Dudich, A. & Stollmann, A. Fauna drobných hmyzožravcov a hlodavcov (Insectivora, Rodentia) vybraných mokraďných biotopov južného Slovenska. Rosalia (Nitra) 14, 195–202 (1999).
Google Scholar 
Balát, F. Potrava sovy pálené na jižní Moravě a na jižním Slovensku [Food of Barn owl on the south Moravia and south Slovakia]. Zoologické listy 5, 237–256 (1956).
Google Scholar 
Baláž, I., Stollmann, A., Ambros, M. & Dudich, A. Drobné cicavce rezervácie Lohótsky močiar. Chránené územia Slovenska 58, 27–29 (2003).
Google Scholar 
Bridišová, Z., Baláž, I. & Ambros, M. Drobné cicavce prírodnej rezervácie Alúvium Žitavy [Small mammals of Alúvium Žitavy natural reservation]. Chránené územia Slovenska 69, 7–9 (2006).
Google Scholar 
Dudich, A., Lysỳ, J. & Štollmann, A. Súčasné poznatky o rozšírení drobnỳch zemnỳch cicavcov (Insectivora, Rodentia) južnej časti Podunajskej nížiny. Spravodaj Oblastného múzea v Komárne, Prírodné vedy 5, 157–186 (1985).
Google Scholar 
Kristofik, J. Small mammals in floodplain forests. Folia Zoologica (Czech Republic) (1999).Méhely, L. Két új poczokfaj a magyar faunában. Állattani közlemények 7, 3–14 (1908).
Google Scholar 
Noga, M. The wintering and food ecology of Long-eared Owl in South-Western part of Slovakia (Comenius University in Bratislava, 2007).
Google Scholar 
Pachinger, K., Novackỳ, M., Facuna, V. & Ambruš, B. Dynamika a zloženie synúzií mikromammálií na izolovanỳch ostrovoch vnútozemskej delty Dunaja v oblasti vodného diela Gabčíkovo. Acta Environ. Universitatis Comenianae 9, 71–77 (1997).
Google Scholar 
Poláčiková, Z. Small terrestrial mammals’ (Eulipotyphla, Rodentia) synusia of selected localities in western Slovakia. Ekológia (Bratislava) 29, 131–139. https://doi.org/10.4149/ekol_2010_02_131 (2010).
Google Scholar 
Reiterová, K. et al. Úloha drobnỳch cicavcov–dôležitỳch rezervoárov v cirkulácii larválnej toxoplazmózy. Slovenskỳ Veterinárny Časopis 4, 217–222 (2010).
Google Scholar 
Stanko, M., Mošanský, L. & Fričová, J. Small mammal communities (Eulipotyphla, Rodentia) of the middle part of alluvium Ipeľ river (Lučenská and Ipeľská basins). Ochrana prírody 26, 43–52 (2010).
Google Scholar 
Spitzenberger, F. & Engelberg, S. A new look at the dynamic western distribution border of Apodemus agrarius in Central Europe (Rodentia: Muridae). Lynx, series nova 45, (2014).Sládkovičová, V. H., Žiak, D. & Miklós, P. Synúzie drobných zemných cicavccov mokradných biotopov Podunajskej roviny. Folia faunistica Slovaca 18, 13–19 (2013).
Google Scholar 
Tulis, F. et al. Expansion of the Striped field mouse (Apodemus agrarius) in the south-western Slovakia during 2010–2015. Folia Oecologica 43, 64–73 (2016).
Google Scholar 
Ambros, M. et al. Zmeny v rozšírení ryšavky tmavopásej (Apodemus agrarius) na Slovensku. in Zborník príspevkov z vedeckého kongresu ‘Zoológia 2022’ 10 (2022).Dudich, A. Ektoparazitofauna cicavcov a vtákov južnej časti Podunajskej nížiny so zreteľom na Žitný ostrov. 1. Siphonaptera. Žitnoostrovské múzeum Dunajská Streda—Spravodaj múzea 9, 61–96 (1986).Dudich, A. Príspevok k poznaniu drobných zemných cicavcov (Insectivora, Rodentia) a ich ektoparazitov (Acarina, Anoplura, Siphonaptera) okolia ŠPR Čenkovská lesostep (Podunajská nížina). Iuxta Danubium (Komárno) 10, 186–191 (1993).
Google Scholar 
Cyprich, D., Krumpál, M. & Dúha, J. Blchy (Siphonaptera) cicavcov (Mammalia) Štátnej prírodnej rezervácii Šúr. Ochrana prírody 8, 241–289 (1987).
Google Scholar 
Lapin, M., Faško, P., Melo, M., Štastný, P. & Tomlain, J. Climate zones. in Landscape Atlas of the Slovak Republic (Harmanec: VKÚ, 2002).Faško, P. & Štastný, P. Average annual precipitation. in Landscape Atlas of the Slovak Republic (2002).Russell, J. C., Stjernman, M., Lindström, Å. & Smith, H. G. Community occupancy before-after-control-impact (CO-BACI) analysis of Hurricane Gudrun on Swedish forest birds. Ecol. Appl. 25, 685–694.  https://doi.org/10.1890/14-0645.1 (2015).Article 

Google Scholar 
Desrosiers, M., Planas, D. & Mucci, A. Short-term responses to watershed logging on biomass mercury and methylmercury accumulation by periphyton in boreal lakes. Can. J. Fish. Aquat. Sci. 63, 1734–1745. https://doi.org/10.1139/f06-077 (2006).Article 
CAS 

Google Scholar 
Hanisch, J. R., Tonn, W. M., Paszkoswki, C. A. & Scrimgeour, G. J. Stocked trout have minimal effects on littoral invertebrate assemblages of productive fish-bearing lakes: A whole-lake BACI study: Stocked trout have minimal effects on littoral invertebrates. Freshw. Biol. 58, 895–907. https://doi.org/10.1111/fwb.12095 (2013).Article 

Google Scholar 
Louhi, P., Mäki-Petäys, A., Erkinaro, J., Paasivaara, A. & Muotka, T. Impacts of forest drainage improvement on stream biota: A multisite BACI-experiment. For. Ecol. Manage. 260, 1315–1323. https://doi.org/10.1016/j.foreco.2010.07.024 (2010).Article 

Google Scholar 
Conner, M. M., Saunders, W. C., Bouwes, N. & Jordan, C. Evaluating impacts using a BACI design, ratios, and a Bayesian approach with a focus on restoration. Environ Monit Assess 188, 555. https://doi.org/10.1007/s10661-016-5526-6
(2016).Article 

Google Scholar 
Popescu, V. D., de Valpine, P., Tempel, D. & Peery, M. Z. Estimating population impacts via dynamic occupancy analysis of Before-After Control–Impact studies. Ecol. Appl. 22, 1389–1404. https://doi.org/10.1890/11-1669.1 (2012).Article 

Google Scholar 
Horváth, G. F. & Herczeg, R. Site occupancy response to natural and anthropogenic disturbances of root vole: Conservation problem of a vulnerable relict subspecies. J. Nat. Conserv. 21, 350–358. https://doi.org/10.1016/j.jnc.2013.03.004 (2013).Article 

Google Scholar 
Pounder, K. C. et al. Novel Hantavirus in Wildlife, United Kingdom. Emerg. Infect. Dis. 19, 673–675. https://doi.org/10.3201/eid1904.121057 (2013).Article 

Google Scholar 
Kreisinger, J., Bastien, G., Hauffe, H. C., Marchesi, J. & Perkins, S. E. Interactions between multiple helminths and the gut microbiota in wild rodents. Phil. Trans. R. Soc. B 370, 20140295. https://doi.org/10.1098/rstb.2014.0295 (2015).Article 

Google Scholar 
Kim, H.-C. et al. Hantavirus surveillance and genetic diversity targeting small mammals at Camp Humphreys, a US military installation and new expansion site Republic of Korea. PLoS ONE 12, e0176514. https://doi.org/10.1371/journal.pone.0176514 (2017).Article 

Google Scholar 
Burnham, K. P., Anderson, D. R. & Burnham, K. P. Model selection and multimodel inference: a practical information-theoretic approach. (Springer, 2002).Sugiura, N. Further analysis of the data by Akaike’s information criterion and the finite corrections: Further analysis of the data by akaike’ s. Commun. Stat. Theory Methods 7, 13–26 (1978).Article 
MATH 

Google Scholar 
Hurvich, C. M. & Tsai, C.-L. Regression and time series model selection in small samples. Biometrika 76, 297–307 (1989).Article 
MathSciNet 
MATH 

Google Scholar 
Morris, E. K. et al. Choosing and using diversity indices: insights for ecological applications from the German Biodiversity Exploratories. Ecol. Evol. 4, 3514–3524. https://doi.org/10.1002/ece3.1155 (2014).Article 

Google Scholar 
Ingram, J. C. Berger–Parker Index. in Encyclopedia of Ecology 332–334 (Elsevier, 2008). https://doi.org/10.1016/B978-008045405-4.00091-4.Bürkner, P.-C. brms : An R package for bayesian multilevel models using Stan. J. Stat. Soft. 80, (2017).Wang, Y., Naumann, U., Eddelbuettel, D., Wilshire, J. & Warton, D. mvabund: Statistical Methods for Analysing Multivariate Abundance Data. (2022).Warton, D. I., Thibaut, L. & Wang, Y. A. The PIT-trap—A “model-free” bootstrap procedure for inference about regression models with discrete, multivariate responses. PLoS ONE 12, e0181790. https://doi.org/10.1371/journal.pone.0181790 (2017).Article 

Google Scholar 
Oksanen, J. et al. vegan: Community Ecology Package. (2022).R Core Team. R: A Language and Environment for Statistical Computing. (R Foundation for Statistical Computing, 2022).Percie du Sert, N. et al. Reporting animal research: Explanation and elaboration for the ARRIVE guidelines 2.0. PLoS Biol. 18, e3000411. https://doi.org/10.1371/journal.pbio.3000411 (2020).Szacki, J. & Liro, A. Movements of small mammals in the heterogeneous landscape. Landsc. Ecol. 5, 219–224 (1991).Article 

Google Scholar 
Szacki, J., Babinska-Werka, J. & Liro, A. The influence of landscape spatial structure on small mammal movements. Acta Theriol. 2, (1993).Pimentel, D., Pimentel, M. & Wilson, A. Plant, animal, and microbe invasive species in the United States and World. Biol. Invasions 193 (2007).Valéry, L., Hervé, F., Lefeuvre, J. C. & Simberloff, D. Invasive species can also be native. Trends Ecol. Evol. https://doi.org/10.1016/j.tree.2009.07.003 (2009).Article 

Google Scholar 
Deinet, S. et al. Wildlife comeback in Europe: The recovery of selected mammal and bird species. (2013).Gompper, M. Top Carnivores in the Suburbs? Ecological and Conservation Issues Raised by Colonization of North-eastern North America by Coyotes: The expansion of the coyote’s geographical range may broadly influence community structure, and rising coyote densities in the suburbs may alter how the general public views wildlife. BioScience 52, 185–190. https://doi.org/10.1641/0006-3568 (2002).Dalecky, A. et al. Range expansion of the invasive house mouse Mus musculus domesticus in Senegal, West Africa: A synthesis of trapping data over three decades, 1983–2014. Mammal. Rev. 45, 176–190. https://doi.org/10.1111/mam.12043 (2015).Article 

Google Scholar 
Konečnỳ, A. et al. Invasion genetics of the introduced black rat (Rattus rattus) in Senegal West Africa. Mol. Ecol. 22, 286–300. https://doi.org/10.1111/mec.12112 (2013).Article 

Google Scholar 
Bramley, G. N. Home ranges and interactions of kiore (Rattus exulans) and Norway rats (R. norvegicus) on Kapiti Island, New Zealand. New Zeal. J. Ecol. 328–334 (2014).O’Rourke, R. L., Anson, J. R., Saul, A. M. & Banks, P. B. Limits to alien black rats (Rattus rattus) acting as equivalent pollinators to extinct native small mammals: The influence of stem width on mammal activity at native Banksia ericifolia inflorescences. Biol. Invasions 22, 329–338. https://doi.org/10.1007/s10530-019-02090-x (2020).Article 

Google Scholar 
White, T. A. et al. Range expansion in an invasive small mammal: influence of life-history and habitat quality. Biol. Invasions 14, 2203–2215. https://doi.org/10.1007/s10530-012-0225-x (2012).Article 

Google Scholar 
McDevitt, A. D. et al. Invading and expanding: Range dynamics and ecological consequences of the greater white-toothed shrew (Crocidura russula) invasion in Ireland. PLoS ONE 9, e100403. https://doi.org/10.1371/journal.pone.0100403 (2014).Article 
ADS 
CAS 

Google Scholar 
Aguilar, J.-P., Pélissié, T., Sigé, B. & Michaux, J. Occurrence of the stripe field mouse lineage (Apodemus agrarius Pallas 1771; Rodentia; Mammalia) in the Late Pleistocene of southwestern France. C.R. Palevol 7, 217–225 (2008).Article 

Google Scholar 
Kordos, L. Historico-zoogeographical and ecological investigation of the subfossil vertebrate fauna of the Aggtelek Karst. Vert. Hung. 18, 85–100 (1978).
Google Scholar 
Hairston, N. G., Smith, F. E. & Slobodkin, L. B. Community structure, population control, and competition. Am. Nat. 94, 421–425 (1960).Article 

Google Scholar 
Agnew, P., Hide, M., Sidobre, C. & Michalakis, Y. A minimalist approach to the effects of density-dependent competition on insect life-history traits. Ecol. Entomol. 27, 396–402 (2002).Article 

Google Scholar 
Krebs, C. J. Beyond population regulation and limitation. Wildl. Res. 29, 1–10 (2002).Article 

Google Scholar 
Huitu, O., Norrdahl, K. & Korpimäki, E. Competition, predation and interspecific synchrony in cyclic small mammal communities. Ecography 27, 197–206.  https://doi.org/10.1111/j.0906-7590.2003.03684.x (2004).Article 

Google Scholar 
Lofgren, O. Niche expansion and increased maturation rate of Clethrionomys glareolus in the absence of competitors. J. Mammal. 76, 1100–1112 (1995).Article 

Google Scholar 
Hansson, L. Competition between Rodents in Successional Stages of Taiga Forests: Microtus agrestis vs. Clethrionomys glareolus. Oikos 40, 258 (1983).Article 

Google Scholar 
Eccard, J. A. & Ylönen, H. Direct interference or indirect exploitation? An experimental study of fitness costs of interspecific competition in voles. Oikos 99, 580–590. https://doi.org/10.1034/j.1600-0706.2002.11833.x (2002).Article 

Google Scholar 
Gliwicz, J. Competition among forest rodents: Effects of Apodemus flavicollis and Clethrionomys glareolus on A. agrarius. Acta Zool. Fennica 1984. (1984).Neet, C. R. & Hausser, J. Habitat selection in zones of parapatric contact between the common shrew Sorex araneus and Millet’s shrew S. coronatus. J. Anim. Ecol. 235–250 (1990).Zub, K., Jędrzejewska, B., Jędrzejewski, W. & Bartoń, K. A. Cyclic voles and shrews and non-cyclic mice in a marginal grassland within European temperate forest. Acta Theriol. 57, 205–216 (2012).Article 
CAS 

Google Scholar 
Henttonen, H. et al. Long-term population dynamics of the common shrew Sorex araneus in Finland. in Annales Zoologici Fennici 349–355 (JSTOR, 1989).Gębczyńska, Z., Gębczyński, M., Morzuch, K. & Zielińska, D. M. Food eaten by four species of rodents in polluted forests. Acta Theriol. 34, 465–477 (1989).Article 

Google Scholar 
Babinska-Werka, J. Response of rodents to an increased and quantitatively diverse food base. Acta theriologica 35, (1990).Margaletic, J., Glavaš, M. & Bäumler, W. The development of mice and voles in an oak forest with a surplus of acorns. J. Pest Sci. 75, 95–98 (2002).Holisova, V. The food of Apodemus agrarius (Pall.). Zoologické listy 16, 1–14 (1967).Obrtel, R. & Hološová, V. The trophic niche of Apodemus agrarius in northern Moravia. Folia Zool. 30, 125–138 (1981).
Google Scholar 
Grant, P. R. Interspecific competition among rodents. Annu. Rev. Ecol. Syst. 3, 79–106 (1972).Article 

Google Scholar 
Redfield, J. A., Krebs, C. J. & Taitt, M. J. Competition between Peromyscus maniculatus and Microtus townsendii in grasslands of coastal British Columbia. J. Anim. Ecol. 607–616 (1977).Kincaid, W. B. & Cameron, G. N. Effects of species removal on resource utilization in a Texas rodent community. J. Mammal. 63, 229–235 (1982).Article 

Google Scholar 
Yurkonis, K. A., Meiners, S. J. & Wachholder, B. E. Invasion impacts diversity through altered community dynamics. J. Ecol. 93, 1053–1061 (2005).Article 

Google Scholar 
Powell, K. I., Chase, J. M. & Knight, T. M. A synthesis of plant invasion effects on biodiversity across spatial scales. Am. J. Bot. 98, 539–548. https://doi.org/10.3732/ajb.1000402 (2011).Article 

Google Scholar 
Jaksic, F. M. Vertebrate invaders and their ecological impacts in Chile. Biodivers. Conserv. 7, 1427–1445 (1998).Article 

Google Scholar 
Richter-Boix, A. et al. Effects of the non-native amphibian species Discoglossus pictus on the recipient amphibian community: Niche overlap, competition and community organization. Biol. Invasions 15, 799–815. https://doi.org/10.1007/s10530-012-0328-4 (2013).Article 

Google Scholar 
Kumschick, S., Bacher, S. & Blackburn, T. M. What determines the impact of alien birds and mammals in Europe? Biol. Invasions 15, 785–797. https://doi.org/10.1007/s10530-012-0326-6 (2013).Article 

Google Scholar 
Tedeschi, L., Biancolini, D., Capinha, C., Rondinini, C. & Essl, F. Introduction, spread, and impacts of invasive alien mammal species in Europe. Mam. Rev. Mam. 12277. https://doi.org/10.1111/mam.12277 (2021).Harris, D. B. Review of negative effects of introduced rodents on small mammals on islands. Biol. Invasions 11, 1611–1630. https://doi.org/10.1007/s10530-008-9393-0 (2009).Article 

Google Scholar 
Traveset, A. et al. A review on the effects of alien rodents in the Balearic (Western Mediterranean Sea) and Canary Islands (Eastern Atlantic Ocean). Biol. Invasions 11, 1653–1670. https://doi.org/10.1007/s10530-008-9395-y (2009).Article 

Google Scholar 
Jung, T. S., Nagorsen, D. W., Kukka, P. M. & Barker, O. E. Alien invaders: recent establishment of an urban population of house mice (Mus musculus) in the Yukon. Northwest. Nat. 93, 240–242 (2012).Article 

Google Scholar 
McKinney, M. L. & Lockwood, J. L. Biotic homogenization: A few winners replacing many losers in the next mass extinction. Trends Ecol. Evol. 14, 450–453. https://doi.org/10.1016/S0169-5347(99)01679-1 (1999).Article 
CAS 

Google Scholar 
MacGregor-Fors, I., Morales-Pérez, L., Quesada, J. & Schondube, J. E. Relationship between the presence of House Sparrows (Passer domesticus) and Neotropical bird community structure and diversity. Biol. Invasions 12, 87–96. https://doi.org/10.1007/s10530-009-9432-5 (2010).Article 

Google Scholar 
Phillips, B. L. Behaviour on Invasion Fronts, and the Behaviour of Invasion Fronts. in Biological Invasions and Animal Behaviour (eds. Weis, J. S. & Sol, D.) 82–95 (Cambridge University Press, 2016). https://doi.org/10.1017/CBO9781139939492.007.Pietrek, A. G., Goheen, J. R., Riginos, C., Maiyo, N. J. & Palmer, T. M. Density dependence and the spread of invasive big-headed ants (Pheidole megacephala) in an East African savanna. Oecologia 195, 667–676. https://doi.org/10.1007/s00442-021-04859-1 (2021).Article 
ADS 

Google Scholar 
Stanko, M. Ryšavka tmavopása (Apodemus agrarius, Rodentia) na Slovensku. (Parazitologický ústav SAV, 2014).Thompson, K., Hodgson, J. G. & Rich, T. C. Native and alien invasive plants: more of the same?. Ecography 18, 390–402 (1995).Article 

Google Scholar  More

Conservationist Jim Groombridge in Hawaii (standing) performing a ‘heli-hook-up’, in which a net full of equipment is hooked up to the hovering helicopter, to save it needing to land.Credit: Jim Groombridge/Maui Forest Bird Recovery Project

Since his undergraduate degree, Jim Groombridge has been part of several teams that work with critically endangered animals, including the Mauritius kestrel (Falco punctatus), which was brought back from the brink of extinction. But he has also experienced the devastation of some species being lost forever, despite all possible interventions. After receiving his PhD from Queen Mary University of London in 2000, he worked as a project coordinator at the Maui Forest Bird Recovery Project in Makawao, Hawaii. Conservation science spans many topics including climate change, working with local communities, epidemiology, genomics and designing protected areas. Projects can range from single-species conservation to ecosystem-level or landscape conservation, such as restoring whole islands. Now a professor in biodiversity conservation at the University of Kent’s Durrell Institute of Conservation and Ecology in Canterbury, UK, Groombridge teaches bachelor’s and master’s students about leadership of conservation teams and how to motivate them in the face of setbacks.What is special about leading conservation teams?Conservation field teams are slightly quirky, and those quirks can define what makes a team work well or not. One is that team leaders are rarely trained in management tasks, such as overseeing a budget, interacting with project partners and local governments, dealing with team members who feel passionate about what they do and facing the high stakes involved. Team members are enthusiastic, passionate and seldom motivated by money.Another quirk is that, in a small conservation team of four to six people, there is often a mix of skill sets and experience. You can have highly experienced specialists in a particular area, such as screening parrots for diseases, or reintroduction biology, and you might also have volunteers with only passion and enthusiasm to offer.How do you lead a team with such variable experience?Even with those different levels of expertise, you still need to meet high standards for specimen and data collection. At the moment, for example, I’m sequencing the genome of the pink pigeon (Nesoenas mayeri), using samples collected in the 1990s. There’s a sense of responsibility, especially if you’re working with species that are rare, because if you mess it up, they could go extinct. It’s not unusual to have volunteers with only two or three weeks’ worth of experience handling extremely rare samples or working with valuable data sets. Their learning curve is pretty steep. As a leader, you need to make sure that you understand the details — ranging from tasks such as collecting data and monitoring and recording invasive species to, for example, knowing how to trap a mongoose — so that you can make sure that everyone is collecting the data in the same way.

Jim Groombridge (far left), who studies biodiversity conservation at the University of Kent, UK, with one of the field crews involved in an operation to translocate a bird called the po‘ouli in Hawaii.Credit: Jim Groombridge/Maui Forest Bird Recovery Project

What do team members tend to have in common?They often share a passion for nature. They want to save the environment, they want to save a species from going extinct, they want to make a difference. That level of emotion is important. It creates an energy, which needs to be channelled proactively and positively into the project to make it a success.In 2002, for example, I was leading a team working to save a bird called the po‘ouli (Melamprosops phaeosoma) on the island of Maui, part of the Hawaiian archipelago. We were trying to translocate one of the last known birds into the range of another one to give them the opportunity to breed. There was huge excitement, but after four weeks of failing to catch the bird, there was also a lot of frustration.How do you manage a team with such strong emotions?Morale is really important. So is being able to deal with difficulties when they arise. That’s what gets small teams through tough times. With the po‘ouli, I had to make sure that the team had fun, and that people genuinely enjoyed themselves. That meant taking time out with the team in the evenings and ensuring that everyone had a bit of a laugh, so it wasn’t deadly serious all the time. Also, I made sure that team members got to perform the aspects of the job that they were good at, to increase their confidence and well-being. We eventually trapped the po‘ouli and moved it, but even though the birds were in the same territory, they didn’t breed.How do you manage expectations amid failure?I had to remind the team about the broader picture of what we had achieved. This was the first time anyone had followed the po‘ouli in the forest for ten days. I think we learnt more about the ecology of that species in that time than anyone had learnt in 30 years. We held the translocated bird for about two hours before we released it, and it took food items from us, which showed that the birds could be kept in captivity if necessary. We learnt a huge amount that could be applied to another project.
Treading carefully: saving frankincense trees in Yemen
You have to manage people’s expectations and have goals that are achievable. If you are starting a project on a species with fewer than ten individuals left in the wild, and your goal is to have thousands, that’s a difficult leap of imagination. Instead, perhaps start with finding a food that a species would eat in captivity. People need to remain connected with what’s achievable. There’s a delicate balance between being aspirational and being pragmatic.As a team member, what do you wish more conservation leaders knew?Often, there is too much emphasis placed on the command structure. Innovation in a conservation team is undersold, and easily quashed by a type of line-manager approach. The hierarchy in a team is important because people know what to do and who to report to, but you also have to encourage team members to use their initiative and ask questions. I remember when my team and I were in the cloud forests, tropical mountain regions covered by clouds for most of the year in Hawaii, we were struggling with baiting rats, which prey on eggs and fledglings of native birds. It’s one of the wettest places on Earth, and the rat poison basically turns to cottage cheese. However, one of my colleagues designed a bait box, which kept the bait dry for many weeks. When you’re working with critically endangered species and in field conditions, ingenuity is crucial.
This interview has been edited for length and clarity. More

Sequence analysisProtein sequence searches were conducted in the GenBank nr database with BLASTP42 using default parameters, including automatically adjusting parameters for short input sequences (Table S1). Conserved domain searches were run against the GenBank Conserved Domain Database (CDD)43. Sequence alignments were conducted in CLC Main Workbench 20.0.4 (Qiagen) with the fast alignment algorithm, gap open cost = 10, and gap extension cost = 1. Biochemical properties of the fish putative thiaminase I protein sequences were predicted with the Create Sequence Statistics function in CLC Main Workbench 20.0.4 (Qiagen, Hilden, Germany). The molecular weights were calculated from the sum of the amino acids in the sequence, and the isoelectric points (pIs) were calculated from the pKa values for the individual amino acids in the sequence.Bacteria culturePure cultures of P. thiaminolyticus strain 818822 were cultured at 37 °C in Terrific Broth (MO BIO Laboratories, Carlsbad, CA) in either a shaking incubator or in a beveled flask with a stir bar and were harvested after 48–80 h of culture. Upon harvest, cultures were processed immediately or frozen whole in 50 mL Falcon tubes at − 80 °C. Fresh or thawed cultures were spun at 14,000×g, and culture supernatant was concentrated using Amicon-ultra 10 kDa molecular weight cut-off (MWCO) filters (EMD Millipore, Billerica, MA).The zebrafish and alewife candidate thiaminase I genes were cloned and overexpressed in E. coli to determine whether they produced functional thiaminases. The recombinant thiaminase I gene from P. thiaminolyticus was overexpressed in E. coli as a positive control. Candidate and control genes were synthesized (Integrated DNA Technologies, Inc., Coralville, Iowa) and placed into the pET52b vector (EMD Millipore). Insert sequences are provided in Supplementary Figs. S10–S13. The empty pET52b vector was used as a negative control. The plasmid was transformed into E. coli (Rosetta 2(DE3)pLysS Singles Competent Cells, EMD Millipore) according to the manufacturer’s instructions, and expression of candidate genes was induced by the addition of IPTG. Cells were lysed in 1X BugBuster (Millipore) according to the manufacturer’s instructions in the presence of benzonase nuclease, and soluble and insoluble fractions were separated by centrifugation.Tissue collectionsAdult common carp were captured from Lake Erie using short-set gill nets. Adult alewife and quagga mussels (Dreissena bugensis) were collected from Sturgeon Bay, Lake Michigan using bottom trawls. Fish collections were completed during July 2007. Sex of sampled fish was not identified. Upon collection, unanesthetized animals were immediately euthanized by flash freezing between slabs of dry ice and stored at − 80 °C. Fish were harvested by the Great Lakes Science Center, U.S. Geological Survey (USGS). Laboratory use of frozen animal tissues and wild type and recombinant bacteria was in accordance with institutional guidelines and biosafety procedures at Oregon State University and USGS. Animal care and use procedures were approved by the Great Lakes Science Center, USGS. All USGS sampling and handling of fish during research are carried out in accordance with guidelines for the care and use of fishes by the American Fisheries Society44. All methods are reported in accordance with applicable ARRIVE guidelines (https://arriveguidelines.org). Zebrafish from OSU’s zebrafish facility were anesthetized and euthanized by overdose with waterborne 200 ppm ethyl 3-aminobenzoate methanesulfonate (MS-222, Sigma-Aldrich, St. Louis, MO) following protocols approved by the OSU Animal Institutional Care and Use Committee and were frozen at − 80 °C after euthanization. Gills, liver, spleen, and the intestinal tract were dissected, and gill tissue was homogenized separately from liver, spleen, and gut, which were homogenized together and designated “viscera.” Homogenization and protein preparation procedures were the same as that for alewife. Zebrafish from Columbia Environmental Research Center (CERC), USGS cultures were anesthetized and euthanized by overdose with 200 ppm ethyl 3-aminobenzoate methanesulfonate (MS-222, Sigma-Aldrich, St. Louis, MO) in water following protocols approved by CERC Institutional Animal Care and Use Committee (IACUC). Whole fish (0.2–0.6 g) were homogenized in 10 mL cold phosphate buffer, pH 6.5. Whole common carp and alewife were thawed until they could just be dissected. Preliminary trial extractions on alewife stomach and intestines, spleen, and gills revealed similar results and revealed that gills and spleen tissue produced the cleanest protein preparations. Therefore, subsequent extractions for common carp and alewife used gill tissue. Samples were pooled from 3 to 5 individual fish, haphazardly chosen from the sampled fish without exclusions. Quagga mussels were thawed just sufficiently to be husked from their shell and were used whole. Researchers were aware of the species and tissue designation of each sample throughout the experiments. Animal tissues were placed in ice-cold (4 °C) beakers containing cold extraction buffer (16 mM K3HPO4, 84 mM KH2PO4, 100 mM NaCl, pH 6.5 with 1 mM DTT, 2 mM EDTA, 3 mM Pepstatin, 1X Protease inhibitor cocktail (Sigma), and 1 mM AEBSF). All extractions were carried out at 4 °C in pre-chilled glassware. Samples were mechanically homogenized using a rotor–stator tissue grinder. Samples were stirred gently for several hours to overnight at 4 °C, centrifuged at 14,000×g to remove debris, and strained through cheesecloth to remove any insoluble lipids. Extracts were then subjected to 30–75% ammonium sulfate precipitation. Pellets from the precipitation were resuspended in buffer (83 mM KH2PO4, 17 mM K2HPO4, and 100 mM NaCl), centrifuged to remove any remaining debris, and stored in 30% glycerol at − 20 °C.Protein electrophoresisNative PAGE was run using either pre-cast TGX gels (BioRad, Hercules, California) of varying percentage (7.5% to 12% or 8–16% gradient gels) or on hand-cast gels (TGX FastCast, BioRad) made according to the manufacturer’s instructions.Blue-native PAGE was used to estimate the mass of thiaminases in their native conformation. Blue-native PAGE45 gels were run using the NativePage Novex Bis–Tris system (Life Technologies) or hand-cast equivalents46. Light blue cathode buffer was used to facilitate visualization of the activity stain.Standard denaturing SDS-PAGE was used to estimate the molecular mass of thiaminases after denaturation. Denaturing SDS-PAGE was run using one of three relatively equivalent methods: pre-cast TGX gels (BioRad) according to the manufacturer’s instructions, hand-cast Tris–HCl gels using standard Laemmli chemistry47 with an operating pH of approximately 9.5, or hand-cast Bis–Tris gels (MOPS buffer) with an operating pH of approximately 7. For all denaturing and non-denaturing SDS-PAGE applications, standard Laemmli sample buffer was used, and samples were heated to 75 °C for 15 min to facilitate denaturation followed by brief centrifugation to eliminate any precipitated debris.Non-denaturing PAGE was used as an alternative to denaturing PAGE for the common carp thiaminase that could not be renatured (i.e., activity could not be recovered) following a denaturing SDS-PAGE. Non-denaturing PAGE was conducted using any of the three aforementioned gel chemistries with SDS-containing running buffers including reductant (DTT), but samples were not heated prior to application to the gel. Samples for non-denaturing PAGE were allowed to incubate in sample buffer at room temperature for 30 min prior to gel loading. This preserves the charge-shift induced by SDS but does not result in protein denaturation, facilitating in-gel analysis of thiaminase I activity after separation.To visualize proteins following electrophoresis, gels were stained with Coomassie stain (CBR-250 at 1 g/L in methanol/acetic acid/water (4:5:1) and destained with methanol/acetic acid/water (1.7:1:11.5). Mini-gels were run on BioRad’s mini-protean gel rigs. Midi-gels (16 cm length) were run on Hoefer’s SE660, and large-format gels (32 cm length) were run on a BioRad’s Protean Slab Cell. Mini-gels were generally run at room temperature, and midi- and large-format gels were run at 4 °C. Blue-native PAGE was always run at 4 °C.Two-dimensional electrophoresis (2DE) separated proteins in the first dimension based on pI and in the second dimension based on mass (either native or denatured). 2DE was performed by combining in-gel IEF with either denaturing SDS-PAGE, non-denaturing SDS-PAGE, or native PAGE. IPG strips were incubated in TRIS-buffered equilibration solution48 either with 6 M urea, SDS, and iodacetamide (denaturing) or without urea, SDS, and iodacetamide (non-denaturing) for 20 min. Low melting point agarose was used to solidify IGP strips in place. Agarose was cooled to just above the gelling temperature, as hot agarose inactivated thiaminase I activity.Isoelectric focusingIsoelectric focusing (IEF) was conducted both in-gel and in-liquid. In-gel IEF was conducted in immobilized pH gradient (IPG) strips using a Multifor II (GE Healthcare Life Sciences). Prior to rehydration, all protein preparations were desalted in low-salt (~ 5 to 10 mM) sodium or potassium phosphate buffer (pH 6.5) using 10 kDA MWCO filters. All samples were applied using sample volumes and protein concentrations recommended by the manufacturer. For standard denaturing in-gel IEF, rehydration solution consisted of 8 M urea, 2% CHAPS, 2% IPG buffer of the appropriate pH-range, 1% bromophenol blue, and 18 mM DTT. The IEF was conducted at maximum of 2 mA total current and 5 W total power, with an EPS3500 XL power supply in gradient mode. Voltage gradients were based on standard protocols recommended by the manufacturer. In-gel IEF was also performed under native conditions to allow thiaminase I activity staining of IPG strips. Protocols were essentially the same as those for denaturing conditions, with the following exceptions: (1) urea was eliminated and the CHAPS concentration was reduced to 0.5% in the rehydration solution; (2) rehydration was conducted at 14 °C; and (3) the water in the cooling tray was cooled to 4 °C.In-liquid IEF was conducted using a Rotofor (BioRad) according to the manufacturer’s instructions. Non-denaturing in-liquid IEF was also conducted using a focusing solution including no urea, 2% pH 3–10 biolyte, 0.5% CHAPS, 20% glycerol, and 5 mM DTT. The addition of glycerol helped retain activity but also increased focusing times. The Rotofor was run at a constant 15 W with a maximum current of 20 mA and voltage set for a maximum of 2000 V. Samples containing 8 M urea were cooled to 14 °C during focusing to avoid urea precipitation, whereas samples lacking urea were cooled to 4 °C during focusing. Protein extracts in salt solutions greater than 10 mM were desalted directly in focusing solution using a 10 kDA MWCO filter. Focusing runs were allowed to proceed until the voltage stabilized and fractions were harvested with the needle array and vacuum pump. Ampholytes were removed by addition of NaCl to 1 M and then samples were desalted into phosphate buffer using a 10kD MWCO filter.Thiaminase I activity measurementsFor quantitative measurements of thiaminase I activity, we conducted a radiometric assay at CERC as previously described49. Zebrafish homogenates were diluted 1:8, 1:16, or 1:32 in cold phosphate buffer, pH 6.5. Two replicates per dilution were assayed. Activity was calculated from the greatest dilution that gave activity within the linear range of the assay and was reported as pmol thiamine consumed per g tissue (wet weight) per minute (pmol/g/min).Thiaminase I activity stainingAfter electrophoresis, gels were stained for thiaminase I activity using a previously described diazo-coupling reaction19,50. Briefly, gels were washed 3 times in water, twice in 25 mM sodium phosphate buffer with 1 mM DTT, and once in 25 mM sodium phosphate buffer without DTT. Gels were then incubated in 0.89 mM thiamine-HCl and co-substrate (1.45 mM pyridoxine, 24 mM nicotinic acid, or 20 mM pyridine) in 25 mM sodium phosphate buffer for 10 min. Gels were briefly rinsed in water and placed in a lidded container and incubated at 37 °C for 30 min to allow thiamine degradation by any thiaminases in the gel. The diazo stain19,50 was then applied to detect remaining thiamine in the gel for five minutes with gentle agitation. Stained gels were rinsed with water and photographed, and further stained with Coomassie to visualize proteins. More

Evapotranspiration measurements and above-ground biomassFigure 1 shows daily evapotranspiration (ET, mm day−1) of each CC tested before mowing (DOY, day of the year, 184) and at 2, 8, 17 and 25 days after mowing (DOY 190, 196, 205 and 213); bare soil was also included as a reference. Before mowing, ET rates showed significant differences between and within the three groups. CR plants had a mean ET of 8.1 mm day−1, which was lower, compared to the other two groups (10.6 and 18.6 mm day−1 for GR and LE, respectively) and the bare soil control (8.5 mm day−1). On DOY 184, values as high as 9.4 (Glechoma hederacea L., GH) and 9.8 mm day−1 (Trifolium subterraneum L. cv. Denmark, TS) were found (Fig. 1), while ranging around 7 mm day-1, Dichondra repens J.R.Forst. & G.Forst. (DR), Hieracium pilosella L. (HP), and Sagina subulata (Swartz) C. Presl (SS) ET were lower than soil evaporation itself.Figure 1Vertical bars represent the daily water use as referred to unit of soil (ET, mm day−1) for the bare soil (yellow) and all the cover crop species as divided into creeping plants (shades of blue), legumes (shades of green) and grasses (shades of orange). Evapotranspiration was measured though a gravimetric method before (i.e. − 4) and at 2, 8, 17 and 25 days after mowing. ET data are mean values ± SE (n = 4).Full size imageOn the same day, a large ET variation was recorded within the GR group as Festuca arundinacea Schreb. cv. Thor (FA) scored the highest daily ET values (13.4 mm day−1), whereas in Festuca ovina L. cv. Ridu (FO), water loss was reduced by 45% (7.5 mm day−1). Within the 15 CCs, LE registered the highest pre-mowing ET with Trifolium michelianum Savi cv. Bolta (TM) peaking at 22.6 mm day−1. However, within LE, Medicago polymorpha L. cv. Scimitar (MP) showed ET values as low as 12.1 mm day−1 (Fig. 1).Two days after mowing, all tested CCs recorded ET values lower than 9 mm day−1 (Fig. 1). Moreover, water use reduction among LE ranged between 56% (M. polymorpha, MP) and 73% (T. michelianum, TM), such that T. michelianum (TM, 6.1 mm day−1), Medicago truncatula Gaertn. cv. Paraggio (MT, 5.6 mm day−1) and M. polymorpha (MP, 5.2 mm day−1) registered ET values lower than the bare soil (7.0 mm day−1). Even though registering a consistent ET reduction after mowing, GR retained ET rates slightly higher than bare soil, except for F. ovina (FO), which recorded the lowest at 6.3 mm day−1. Subsequent samplings showed that most of the CCs had a progressive recovery in water use (Fig. 1) and data taken 17 days after mowing confirmed that Lotus corniculatus L. cv. Leo (LC) and all GR fetched pre-mowing ET rates. Medicago lupulina L. cv. Virgo (ML) registered a partial recovery with similar rates (about 13 mm day−1) at 17 and 25 days after the mowing event. F. ovina and all remaining LE stayed below 10 mm day−1 with ET values close to the control until the end of the trial. At 17 days from grass cutting, under a quite high exceeding-the-pot biomass, both G. hederacea (GH) and T. subterraneum (TS) reached ET values as high as 12.0 and 11.4 mm day−1, respectively. On the other hand, D. repens (DR), H. pilosella (HP), and S. subulata (SS) even though with slightly higher ET values than those registered at the beginning of the trial (DOY 184), remained close to the soil evaporation rates until DOY 213.Aboveground dry clipped biomass at the first mowing date (ADW_MW1, DOY 188) showed large differences among groups, as represented in Table 1. ADW_MW1 within LE was quite variable, as values ranged between 274.3 g m−2 (M. polymorpha, MP) and 750.0 g m−2 (T. michelianum, TM). With a mean value of 565.9 g m−2, LE aboveground biomass was 80% higher than the mean GR ADW_MW1 (110.2 g m-2). F. ovina (FO) scored the lowest value at 48.4 g m−2 among grasses, while within the creeping group, G. hederacea (GH) and T. subterraneum (TS) had biomass development outside the pot edges totalling 89.6 g m−2 and 23.2 g m−2, respectively.Table 1 Aboveground dry biomass clipped at the first mowing event (ADW _MW1), the corresponding leaf area surface index (LAI) and water use per leaf area unit (ETLEAF) of all cover crops tested.Full size tableLeaf area index (LAI, m2 m−2) at mowing showed the highest values in LE with LAI peaking at 12.4 (Table 1). Among GR, LAI did not show significant differences, being around 1.2. Concerning CR, LAI was assessed at 0.2 and 0.8 for T. subterraneum (TS) and G. hederacea (GH) respectively, while LAI estimated through photo analysis ranged between 1.3 (D. repens, DR) and 3.6 (T. subterraneum TS).Evapotranspiration per leaf area unit (ETLEAF) was notably higher in GR, ranging between 7.75 (F. ovina, FO) and 9.22 (Lolium perenne L. cv. Playfast, LP) mm m−2 day−1 (Table 1). In descending order, ETLEAF was the highest in D. repens (DR, 5.46 mm m−2 day−1). Similar ETLEAF was found when comparing some LE and CR species such as M. truncatula (MT, 3.40 mm m−2 day−1), M. lupulina (ML, 4.05 mm m−2 day−1), G. hederacea (GH, 3.68 mm m−2 day−1), H. pilosella (HP, 3.86 mm m-2 day-1) and T. subterraneum (TS, 2.74 mm m−2 day−1). T. michelianum (TM), with 1.81 mm m-2 day-1 scored the lowest ETLEAF of all species (Table 1).Plotting LAI versus the before-mowing ET yielded a significant quadratic relationship (R2  > 0.76) (Fig. 2a) which helped to distinguish two different data clouds. Till LAI values of about 6, the model was linear, having at its lower end all GR and CR species with the inclusion of M. polymorpha (MP) as a legume, while, at the other end, M. truncatula (MT), L. corniculatus (LC) and M. lupulina (ML) were grouped together. T. michelianum (TM) was isolated from all CCs at 22.56 mm day−1.Figure 2Panel (a): quadratic regression of leaf area index (LAI, m2 m−2) vs cover crop evapotranspiration per unit of soil (ET, mm day−1). Each data point is mean value ± SE (n = 4). The quadratic model equation is y = − 0.128×2 + 2.9968x + 5.4716, R2 = 0.76. Panel (b): the quadratic regression between LAI corresponding to the clipped biomass (m2 m−2) and cover crop ET reduction (%). Each data point is mean value ± SE (n = 4). Quadratic model equation is y = − 0.8985×2 + 16.503x + 5.1491, R2 = 0.94.Full size imageWhen regressing the fraction of ET reduction, compared to pre-mowing values vs LAI (Fig. 2b), the same quadratic model achieved a very close fit (R2 = 0.94, p  1 mm) root diameters as affected by soil cover.Full size tableThe highest values of diameter class length (DCL, mm cm−3) for very fine roots (DCL_VF,  1.0 mm) roots although, most notably, L. corniculatus roots showed the highest abundance for both DCL_M (23.08 cm cm−3) and DCL_C (0.54 cm cm−3).At the 10–20 cm soil depth, GR confirmed the highest values for both very fine and fine roots, with F. arundinacea reaching maximum DCL of 2.269 and 5.215 cm cm-3, respectively (Table 2). L. corniculatus largely outscored any other species for both medium and coarse root diameter (6.173 and 0.037 cm cm−3, respectively), with F. arundinacea ranking second (3.157 and 0.016 cm cm−3, respectively).The highest root dry weight (RDW, mg cm-3) within the topsoil layer was reached by L. corniculatus (8.7 mg cm−3) and F. arundinacea (7.6 mg cm-3). Notably, such values were significantly higher than those recorded on the remaining species, except for the F. arundinacea vs F. rubra commutata comparison (Table 2). At 10–20 depth, scant variation was recorded in RDW measured in grasses, whereas L. corniculatus held its supremacy within legumes (4.5 mg cm−3). Within the creeping type, D. repens (DR) and G. hederacea (GH) scored RDW values as high as those determined for grass species (namely F. arundinacea , P. pratensis and F. rubra commutata), whereas S. subulata (SS) essentially had no root development.Soil aggregates and mean weight diameter (MWD)Table 3 reports the proportional aggregate weight (g kg−1) for both 0–10 and 10–20 cm soil depths. Compared to bare soil, the largest increase in large macroaggregates (LM,  > 2000 µm) in the top 10 cm of soil was achieved by L. corniculatus with 461 g kg−1. L. corniculatus differed from the rest of the LE group, whose grand mean (90 g kg−1) was the lowest of the three tested groups. As a legume, T. subterraneum (TS, 122 g kg−1) recorded the lowest values compared to fellow CR species, ranging between 211 (D. repens, DR) and 316 g kg−1 (G. hederacea, GH). GR recorded LM values slightly lower than those of CR, with a mean value of 217 vs 224 g kg-1.Table 3 Proportional aggregate weight (g kg−1) of sand-free aggregate-size fractions acquired from wet sieving as affected by soil cover and mean weight diameter (MWD). Aggregate-size fraction divided as macroaggregates with large size ( > 2 mm, LM) and small size (2 mm—250 μm, sM), microaggregates (250 μm—53 μm, m), and silt and clay ( More

Crowther, T. W. et al. Mapping tree density at a global scale. Nature 525, 201–205 (2015).Article 
CAS 

Google Scholar 
Asner, G. P. et al. A universal airborne LiDAR approach for tropical forest carbon mapping. Oecologia 168, 1147–1160 (2012).Article 

Google Scholar 
Walker, A. P. et al. Predicting long‐term carbon sequestration in response to CO2 enrichment: How and why do current ecosystem models differ? Glob. Biogeochem. Cy. 29, 476–495 (2015).Article 
CAS 

Google Scholar 
Madrigal-González, J. et al. Climate reverses directionality in the richness–abundance relationship across the World’s main forest biomes. Nat. Commun. 11, 1–7 (2020).Article 

Google Scholar 
Stephenson, N. L. Climatic control of vegetation distribution: the role of the water balance. Am. Nat. 135, 649–670 (1990).Article 

Google Scholar 
Weiskittel, A. R., Maguire, D. A., Monserud, R. A. Development of a hybrid model for intensively managed Douglas-fir in the Pacific Northwest. In Forest Growth and Timber Quality, 49 (USDA, Portland, 2009).Paoli, G. D., Curran, L. M. & Slik, J. W. F. Soil nutrients affect spatial patterns of aboveground biomass and emergent tree density in southwestern Borneo. Oecologia 155, 287–299 (2008).Article 

Google Scholar 
Yoda, K., Kira, T., Ogawa, H. & Hozami, K. Self-thinning in overcrowded pure stands under cultivated and natural conditions. J. Biol. Osaka City Univ. 14, 107–129 (1963).
Google Scholar 
Westoby, M. The self-thinning rule. Adv. Ecol. Res. 14, 167–225 (1984).Article 

Google Scholar 
Weiner, J. & Freckleton, R. P. Constant final yield. Annu. Rev. Ecol. Evol. S. 41, 173–192 (2010).Article 

Google Scholar 
Pillet, M. et al. Disentangling competitive vs. climatic drivers of tropical forest mortality. J. Ecol. 106, 1165–1179 (2018).Article 

Google Scholar 
Schluter, D. Experimental evidence that competition promotes divergence in adaptive radiation. Science 266, 798–801 (1994).Article 
CAS 

Google Scholar 
Pacala, S.W. & Levin, S.A. Biologically generated spatial pattern and the coexistence of competing species. Spatial Ecology: The Role of Space in Population Dynamics and Interspecific Interactions. (Princeton University Press, Princeton, NJ, p. 204-232, 1997).Asefa, M., Cao, M., Zhang, G., Ci, X. & Li, J. Yang Environmental filtering structures tree functional traits combination and lineages across space in tropical tree assemblages. Sci. Rep. 7, 1–10 (2017).Article 
CAS 

Google Scholar 
Pretzsch, H. & Biber, P. Tree species mixing can increase maximum stand density. Can. J. For. Res. 46, 1179–1193 (2016).Article 

Google Scholar 
Escudero, A. et al. Every bit helps: The functional role of individuals in assembling any plant community, from the richest to monospecific ones. J. Veg. Sci. 32, e13059 (2021).Article 

Google Scholar 
Jump, A. S. et al. Structural overshoot of tree growth with climate variability and the global spectrum of drought-induced forest dieback. Glob. Chang. Biol. 23, 3742–3757 (2017).Article 

Google Scholar 
M. Takyu, Y. Kubota, S.I. Aiba, T. Seino, T. Nishimura. Pattern of changes in species diversity, structure and dynamics of forest ecosystems along latitudinal gradients in East Asia. In Forest Ecosystems and Environments (Springer, Tokyo, 2005), pp. 49–58.Rivoire, M. & Le, G. A. Moguedec, generalized self-thinning relationship for multi-species and mixed-size forests. Ann. Sci. 69, 207–219 (2012).Article 

Google Scholar 
Salas‐Eljatib, C. & Weiskittel, A. R. Evaluation of modelling strategies for assessing self‐thinning behaviour and carrying capacity. Ecol. Evol. 8, 10768–10779 (2018).Article 

Google Scholar 
Schietti, J. et al. Forest structure along a 600 km transect of natural disturbances and seasonality gradients in central‐southern Amazonia. J. Ecol. 104, 1335–1346 (2016).Article 

Google Scholar 
Vanclay, J. K. & Sands, P. J. Calibrating the self-thinning frontier. For. Ecol. Manag. 259, 81–85 (2009).Article 

Google Scholar 
Sapijanskas, J., Paquette, A., Potvin, C., Kunert, N. & Loreau, M. Tropical tree diversity enhances light capture through crown plasticity and spatial and temporal niche differences. Ecology 95, 2479–2492 (2014).Article 

Google Scholar 
Lieth, H. Modeling the primary productivity of the world. In H. Lieth & R. H. Whittaker, eds. Primary Productivity of the Biosphere (Springer-Verlag, New York, New York, USA, 1975), pp. 237–264.Grieser, J., Gommes, R., Cofield, S., Bernardi, M. World Maps of Climatological net Primary Production of Biomass, NPP. Food and Agriculture Organization of the United Nations. (GEONETWORK. FAO, Rome, Italy, 2006).Hengl, T. et al. SoilGrids250m: global gridded soil information based on machine learning. PLoS One 12, e0169748 (2017).Article 

Google Scholar 
Wood, S. N. Stable and efficient multiple smoothing parameter estimation for generalized additive models. J. Am. Stat. Assoc. 99, 673–686 (2004).Article 

Google Scholar 
J.B. Grace. Structural Equation Modeling and Natural Systems. (Cambridge University Press, Cambridge, 2006).Aho, K., Derryberry, D. & Peterson, T. Model selection for ecologists: the worldviews of AIC and BIC. Ecology 95, 631–636 (2014).Article 

Google Scholar 
R Core Team (2021). R: A language and environment for statistical computing. (R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/).Wood, S., Scheipl, F. & Wood, M. S. Package ‘gamm4’. Am. Stat. 45, 339 (2017).
Google Scholar 
Lefcheck, J. S. piecewiseSEM: Piecewise structural equation modelling in r for ecology, evolution, and systematics. Methods Ecol. Evol. 7, 573–579 (2016).Article 

Google Scholar 
Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D. R. Core Team nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-148 (2020).Latham J, Cumani R, Rosati I, Bloise M. FAO Global Land Cover (GLC-SHARE) Database Beta-Release 1.0, Land and Water Division. 2014. http://www.fao.org/uploads/media/glc-share-doc.pdf. More

Hsu-Kim H, Kucharzyk KH, Zhang T, Deshusses MA. Mechanisms regulating mercury bioavailability for methylating microorganisms in the aquatic environment: A critical review. Environ Sci Technol. 2013;47:2441–56.Article 
CAS 

Google Scholar 
Podar M, Gilmour CC, Brandt CC, Soren A, Brown SD, Crable BR, et al. Global prevalence and distribution of genes and microorganisms involved in mercury methylation. Sci Adv. 2015;1:e1500675.Article 

Google Scholar 
Liu YR, Johs A, Bi L, Lu X, Hu HW, Sun D, et al. Unraveling microbial communities associated with methylmercury production in paddy soils. Environ Sci Technol. 2018;52:13110–8.Article 
CAS 

Google Scholar 
Lee C-S, Fisher NS. Bioaccumulation of methylmercury in a marine copepod. Environ Toxicol Chem. 2017;36:1287–93.Article 
CAS 

Google Scholar 
Parks JM, Johs A, Podar M, Bridou R, Hurt RAJ, Smith SD, et al. The genetic basis for bacterial mercury methylation. Science 2013;339:1332–5.Article 
CAS 

Google Scholar 
McDaniel EA, Peterson BD, Stevens SLR, Tran PQ, Anantharaman K, McMahon KD. Expanded phylogenetic diversity and metabolic flexibility of mercury-methylating microorganisms. mSystems 2020;5:e00299–20.Article 
CAS 

Google Scholar 
Cooper CJ, Zheng K, Rush KW, Johs A, Sanders BC, Pavlopoulos GA, et al. Structure determination of the HgcAB complex using metagenome sequence data: Insights into microbial mercury methylation. Commun Biol. 2020;3:320.Article 
CAS 

Google Scholar 
Kerin EJ, Gilmour CC, Roden E, Suzuki MT, Coates JD, Mason RP. Mercury methylation by dissimilatory iron-reducing bacteria. Appl Environ Microbiol. 2006;72:7919–21.Article 
CAS 

Google Scholar 
Gilmour CC, Podar M, Bullock AL, Graham AM, Brown SD, Somenahally AC, et al. Mercury methylation by novel microorganisms from new environments. Environ Sci Technol. 2013;47:11810–20.Article 
CAS 

Google Scholar 
Capo E, Bravo AG, Soerensen AL, Bertilsson S, Pinhassi J, Feng C, et al. Deltaproteobacteria and Spirochaetes-like bacteria are abundant putative mercury methylators in oxygen-deficient water and marine particles in the Baltic Sea. Front Microbiol. 2020;11:574080.Article 

Google Scholar 
Gionfriddo CM, Tate MT, Wick RR, Schultz MB, Zemla A, Thelen MP, et al. Microbial mercury methylation in Antarctic sea ice. Nat Microbiol. 2016;1:16127.Article 
CAS 

Google Scholar 
Jones DS, Walker GM, Johnson NW, Mitchell CPJ, Coleman Wasik JK, Bailey JV. Molecular evidence for novel mercury methylating microorganisms in sulfate-impacted lakes. ISME J. 2019;13:1659–75.Article 
CAS 

Google Scholar 
Christensen GA, Gionfriddo CM, King AJ, Moberly JG, Miller CL, Somenahally AC, et al. Determining the reliability of measuring mercury cycling gene abundance with correlations with mercury and methylmercury concentrations. Environ Sci Technol. 2019;53:8649–63.Article 
CAS 

Google Scholar 
Villar E, Cabrol L, Heimburger-Boavida LE. Widespread microbial mercury methylation genes in the global ocean. Environ Microbiol Rep. 2020;12:277–87.Article 
CAS 

Google Scholar 
Lin H, Ascher DB, Myung Y, Lamborg CH, Hallam SJ, Gionfriddo CM, et al. Mercury methylation by metabolically versatile and cosmopolitan marine bacteria. ISME J. 2021;15:1810–25.Article 
CAS 

Google Scholar 
King JK, Kostka JE, Frischer ME, Saunders FM, Jahnke RA. A quantitative relationship that demonstrates mercury methylation rates in marine sediments are based on the community composition and activity of sulfate-reducing bacteria. Environ Sci Technol. 2001;35:2491–6.Article 
CAS 

Google Scholar 
Regnell O, Watras CJ. Microbial mercury methylation in aquatic environments: A critical review of published field and laboratory studies. Environ Sci Technol. 2019;53:4–19.Article 
CAS 

Google Scholar 
Xie R, Wang Y, Huang D, Hou J, Li L, Hu H, et al. Expanding Asgard members in the domain of Archaea sheds new light on the origin of eukaryotes. Sci China Life Sci. 2022;65:818–29.Article 
CAS 

Google Scholar 
Seitz KW, Dombrowski N, Eme L, Spang A, Lombard J, Sieber JR, et al. Asgard archaea capable of anaerobic hydrocarbon cycling. Nat Commun. 2019;10:1822.Article 

Google Scholar 
Zaremba-Niedzwiedzka K, Caceres EF, Saw JH, Backstrom D, Juzokaite L, Vancaester E, et al. Asgard archaea illuminate the origin of eukaryotic cellular complexity. Nature 2017;541:353–8.Article 
CAS 

Google Scholar 
Liu Y, Makarova KS, Huang W-C, Wolf YI, Nikolskaya AN, Zhang X, et al. Expanded diversity of Asgard archaea and their relationships with eukaryotes. Nature 2021;593:553–7.Article 
CAS 

Google Scholar 
Zhang JW, Dong HP, Hou LJ, Liu Y, Ou YF, Zheng YL, et al. Newly discovered Asgard archaea Hermodarchaeota potentially degrade alkanes and aromatics via alkyl/benzyl-succinate synthase and benzoyl-CoA pathway. ISME J. 2021;15:1826–43.Article 
CAS 

Google Scholar 
Cai M, Liu Y, Yin X, Zhou Z, Friedrich MW, Richter-Heitmann T, et al. Diverse Asgard archaea including the novel phylum Gerdarchaeota participate in organic matter degradation. Sci China Life Sci. 2020;63:886–97.Article 
CAS 

Google Scholar 
Baker BJ, De Anda V, Seitz KW, Dombrowski N, Santoro AE, Lloyd KG. Diversity, ecology and evolution of Archaea. Nat Microbiol. 2020;5:887–900.Article 
CAS 

Google Scholar 
Farag Ibrahim F, Zhao R, Biddle Jennifer F, Atomi H. “Sifarchaeota,” a novel Asgard phylum from Costa Rican sediment capable of polysaccharide degradation and anaerobic methylotrophy. Appl Environ Micro. 2021;87:e02584–20.
Google Scholar 
Adam PS, Borrel G, Brochier-Armanet C, Gribaldo S. The growing tree of Archaea: new perspectives on their diversity, evolution and ecology. ISME J. 2017;11:2407–25.Article 

Google Scholar 
Cai M, Richter-Heitmann T, Yin X, Huang W-C, Yang Y, Zhang C, et al. Ecological features and global distribution of Asgard archaea. Sci Total Environ. 2021;758:143581.Article 
CAS 

Google Scholar 
Zhang C-J, Chen Y-L, Sun Y-H, Pan J, Cai M-W, Li M. Diversity, metabolism and cultivation of archaea in mangrove ecosystems. Mar Life Sci Tech. 2020;3:252–62.Article 

Google Scholar 
Dai SS, Yang Z, Tong Y, Chen L, Liu SY, Pan R, et al. Global distribution and environmental drivers of methylmercury production in sediments. J Hazard Mater. 2021;407:124700.Article 
CAS 

Google Scholar 
Tang WL, Liu YR, Guan WY, Zhong H, Qu XM, Zhang T. Understanding mercury methylation in the changing environment: Recent advances in assessing microbial methylators and mercury bioavailability. Sci Total Environ. 2020;714:136827.Article 
CAS 

Google Scholar 
Tsui MTK, Finlay JC, Balogh SJ, Nollet YH. In situ production of methylmercury within a stream channel in northern California. Environ Sci Technol. 2010;44:6998–7004.Article 
CAS 

Google Scholar 
Liu Y, Zhou Z, Pan J, Baker BJ, Gu JD, Li M. Comparative genomic inference suggests mixotrophic lifestyle for Thorarchaeota. ISME J. 2018;12:1021–31.Article 
CAS 

Google Scholar 
Lei P, Zhong H, Duan D, Pan K. A review on mercury biogeochemistry in mangrove sediments: Hotspots of methylmercury production? Sci Total Environ. 2019;680:140–50.Article 
CAS 

Google Scholar 
Beckers F, Rinklebe J. Cycling of mercury in the environment: Sources, fate, and human health implications: A review. Crit Rev Env Sci Tec. 2017;47:693–794.Article 
CAS 

Google Scholar 
de Oliveira DC, Correia RR, Marinho CC, Guimaraes JR. Mercury methylation in sediments of a Brazilian mangrove under different vegetation covers and salinities. Chemosphere 2015;127:214–21.Article 

Google Scholar 
Li R, Xu H, Chai M, Qiu GY. Distribution and accumulation of mercury and copper in mangrove sediments in Shenzhen, the world’s most rapid urbanized city. Environ Moni Assess. 2016;188:87.Article 

Google Scholar 
O’Connor D, Hou D, Ok YS, Mulder J, Duan L, Wu Q, et al. Mercury speciation, transformation, and transportation in soils, atmospheric flux, and implications for risk management: A critical review. Environ Int. 2019;126:747–61.Article 

Google Scholar 
Obrist D, Kirk JL, Zhang L, Sunderland EM, Jiskra M, Selin NE. A review of global environmental mercury processes in response to human and natural perturbations: Changes of emissions, climate, and land use. Ambio 2018;47:116–40.Article 

Google Scholar 
Capo E, Peterson BD, Kim M, Jones DS, Acinas SG, Amyot M, et al. A consensus protocol for the recovery of mercury methylation genes from metagenomes. Mol Ecol Resour. 2022; https://doi.org/10.1111/1755-0998.13687.Gionfriddo CM, Wymore AM, Jones DS, Wilpiszeski RL, Lynes MM, Christensen GA, et al. An improved hgcAB primer set and direct high-throughput sequencing expand Hg-methylator diversity in nature. Front Microbiol. 2020;11:541554.Article 

Google Scholar 
Yu R-Q, Barkay T. Chapter two – microbial mercury transformations: Molecules, functions and organisms. Adv Appl Microbiol. 2022;118:31–90.Article 

Google Scholar 
Chételat J, Richardson MC, MacMillan GA, Amyot M, Poulain AJ. Ratio of methylmercury to dissolved organic carbon in water explains methylmercury bioaccumulation across a latitudinal gradient from north-temperate to arctic lakes. Environ Sci Technol. 2018;52:79–88.Article 

Google Scholar 
Liu Y-R, Dong J-X, Han L-L, Zheng Y-M, He J-Z. Influence of rice straw amendment on mercury methylation and nitrification in paddy soils. Environ Pollut. 2016;209:53–9.Article 
CAS 

Google Scholar 
Moreau JW, Gionfriddo CM, Krabbenhoft DP, Ogorek JM, DeWild JF, Aiken GR, et al. The effect of natural organic matter on mercury methylation by Desulfobulbus propionicus 1pr3. Front Microbiol. 2015;6:1389.Article 

Google Scholar 
Chen C-F, Ju Y-R, Chen C-W, Dong C-D. The distribution of methylmercury in estuary and harbor sediments. Sci Total Environ. 2019;691:55–63.Article 
CAS 

Google Scholar 
Bravo AG, Bouchet S, Guédron S, Amouroux D, Dominik J, Zopfi J. High methylmercury production under ferruginous conditions in sediments impacted by sewage treatment plant discharges. Water Res. 2015;80:245–55.Article 
CAS 

Google Scholar 
Wang H, Su J, Zheng T, Yang X. Insights into the role of plant on ammonia-oxidizing bacteria and archaea in the mangrove ecosystem. J Soil Sediment. 2015;15:1212–23.Article 
CAS 

Google Scholar 
Imachi H, Nobu MK, Nakahara N, Morono Y, Ogawara M, Takaki Y, et al. Isolation of an archaeon at the prokaryote–eukaryote interface. Nature 2020;577:519–25.Article 
CAS 

Google Scholar 
Zhou J, Riccardi D, Beste A, Smith JC, Parks JM. Mercury methylation by HgcA: Theory supports carbanion transfer to Hg(II). Inorg Chem. 2014;53:772–7.Article 
CAS 

Google Scholar 
Smith Steven D, Bridou R, Johs A, Parks Jerry M, Elias Dwayne A, Hurt Richard A, et al. Site-directed mutagenesis of HgcA and HgcB reveals amino acid residues important for mercury methylation. Appl Environ Micro. 2015;81:3205–17.Article 
CAS 

Google Scholar 
Sousa FL, Neukirchen S, Allen JF, Lane N, Martin WF. Lokiarchaeon is hydrogen dependent. Nat Microbiol. 2016;1:16034.Article 
CAS 

Google Scholar 
Schaefer JK, Rocks SS, Zheng W, Liang L, Gu B, Morel FMM. Active transport, substrate specificity, and methylation of Hg(II) in anaerobic bacteria. Proc Natl Acad Sci USA 2011;108:8714.Article 
CAS 

Google Scholar 
Sakai S, Imachi H, Hanada S, Ohashi A, Harada H, Kamagata Y. Methanocella paludicola gen. nov., sp. nov., a methane-producing archaeon, the first isolate of the lineage ‘Rice Cluster I’, and proposal of the new archaeal order Methanocellales ord. nov. Int J Syst Evol Microbiol. 2008;58:929–36.Article 

Google Scholar 
Dridi B, Fardeau ML, Ollivier B, Raoult D, Drancourt M. Methanomassiliicoccus luminyensis gen. nov., sp. nov., a methanogenic archaeon isolated from human faeces. Int J Syst Evol Microbiol. 2012;62:1902–7.Article 
CAS 

Google Scholar 
Dietz R, Sonne C, Basu N, Braune B, O’Hara T, Letcher RJ, et al. What are the toxicological effects of mercury in arctic biota? Sci Total Environ. 2013;443:775–90.Article 
CAS 

Google Scholar 
Gilmour Cynthia C, Bullock Allyson L, McBurney A, Podar M, Elias Dwayne A, Lovley Derek R. Robust mercury methylation across diverse methanogenic archaea. mBio 2018;9:e02403–17.
Google Scholar 
Pan J, Chen Y, Wang Y, Zhou Z, Li M. Vertical distribution of Bathyarchaeotal communities in mangrove wetlands suggests distinct niche preference of Bathyarchaeota subgroup 6. Micro Ecol. 2019;77:417–28.Article 

Google Scholar 
Zhang C-J, Pan J, Duan C-H, Wang Y-M, Liu Y, Sun J, et al. Prokaryotic diversity in mangrove sediments across southeastern China fundamentally differs from that in other biomes. mSystems 2019;4:e00442–19.Article 
CAS 

Google Scholar 
Peng Y, Leung HC, Yiu SM, Chin FY. IDBA-UD: a de novo assembler for single-cell and metagenomic sequencing data with highly uneven depth. Bioinformatics 2012;28:1420–8.Article 
CAS 

Google Scholar 
Li D, Liu C-M, Luo R, Sadakane K, Lam T-W. MEGAHIT: an ultra-fast single-node solution for large and complex metagenomics assembly via succinct de Bruijn graph. Bioinformatics 2015;31:1674–6.Article 
CAS 

Google Scholar 
Hyatt D, Chen G-L, LoCascio PF, Land ML, Larimer FW, Hauser LJ. Prodigal: prokaryotic gene recognition and translation initiation site identification. BMC Bioinforma. 2010;11:119.Article 

Google Scholar 
Zhang C-J, Pan J, Liu Y, Duan C-H, Li M. Genomic and transcriptomic insights into methanogenesis potential of novel methanogens from mangrove sediments. Microbiome. 2020;8:94.Article 
CAS 

Google Scholar 
Kang DD, Froula J, Egan R, Wang Z. MetaBAT, an efficient tool for accurately reconstructing single genomes from complex microbial communities. PeerJ 2015;3:e1165.Article 

Google Scholar 
Sieber CMK, Probst AJ, Sharrar A, Thomas BC, Hess M, Tringe SG, et al. Recovery of genomes from metagenomes via a dereplication, aggregation and scoring strategy. Nat Microbiol. 2018;3:836–43.Article 
CAS 

Google Scholar 
Parks DH, Imelfort M, Skennerton CT, Hugenholtz P, Tyson GW. CheckM: assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes. Genome Res. 2015;25:1043–55.Article 
CAS 

Google Scholar 
Olm MR, Brown CT, Brooks B, Banfield JF. dRep: A tool for fast and accurate genomic comparisons that enables improved genome recovery from metagenomes through de-replication. ISME J. 2017;11:2864–8.Article 
CAS 

Google Scholar 
Chaumeil PA, Mussig AJ, Hugenholtz P, Parks DH. GTDB-Tk: a toolkit to classify genomes with the Genome Taxonomy Database. Bioinformatics 2019;36:1925–7.
Google Scholar 
Kanehisa M, Sato Y, Morishima K. BlastKOALA and GhostKOALA: KEGG tools for functional characterization of genome and metagenome sequences. J Mol Biol. 2016;428:726–31.Article 
CAS 

Google Scholar 
Huerta-Cepas J, Forslund K, Szklarczyk D, Jensen LJ, von Mering C, Bork P. Fast genome-wide functional annotation through orthology assignment by eggNOG-mapper. Mol Biol Evol. 2017;34:2115–22.Article 
CAS 

Google Scholar 
Finn RD, Clements J, Eddy SR. HMMER web server: interactive sequence similarity searching. Nucleic Acids Res. 2011;39:W29–W37.Article 
CAS 

Google Scholar 
Eddy SR. Accelerated profile HMM searches. PLoS Comput Biol. 2011;7:e1002195.Article 
CAS 

Google Scholar 
Edgar RC. MUSCLE: Multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 2004;32:1792–7.Article 
CAS 

Google Scholar 
Capella-Gutiérrez S, Silla-Martínez JM, Gabaldón T. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 2009;25:1972–3.Article 

Google Scholar 
Nguyen L-T, Schmidt HA, von Haeseler A, Minh BQ. IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol. 2015;32:268–74.Article 
CAS 

Google Scholar 
Price MN, Dehal PS, Arkin AP. FastTree 2 – approximately maximum-likelihood trees for large alignments. Plos ONE. 2010;5:e9490.Article 

Google Scholar 
Letunic I, Bork P. Interactive tree of life (iTOL) v3: an online tool for the display and annotation of phylogenetic and other trees. Nucleic Acids Res. 2016;44:W242–5.Article 
CAS 

Google Scholar 
Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, et al. Highly accurate protein structure prediction with AlphaFold. Nature 2021;596:583–9.Article 
CAS 

Google Scholar 
Trott O, Olson AJ. AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem. 2010;31:455–61.CAS 

Google Scholar 
Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods. 2012;9:357–9.Article 
CAS 

Google Scholar 
Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinforma (Oxf, Engl). 2010;26:841–2.Article 
CAS 

Google Scholar 
Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 2009;25:1754–60.Article 
CAS 

Google Scholar  More

Study areaWe focused on hiking activity in the four main islands of Japan (Honshu, Hokkaido, Kyushu, and Shikoku) and nearby small islands connected to the main islands by a bridge (Fig. 1a). These islands lie between latitudes 31.0° and 45.5°N, and the total area is 361,000 km2. The islands are generally mountainous and tallest mountains in central Honshu exceed 3000 m a.s.l. (Fig. 1c). In Tokyo, mean monthly temperatures range between 5.2 °C in January and 26.4 °C in August, while they range between − 18.4 °C in January and 6.2 °C in August at the summit of the highest mountain, Mt. Fuji (3776 m a.s.l., Japan Meteorological Agency). In northern Honshu and Hokkaido, snow depth can exceed 1 m even at low elevations and high mountains are covered with snow even in southern Japan.Vegetation excluding farmland and pasture covers 70.9% of the study area and the 93.9% is forest. Plantations of mostly evergreen conifers such as Japanese cedar (Cryptomeria japonica) occupy 37.6% of the vegetation area (National Surveys on the Natural Environment by the Biodiversity Center of Japan 2nd–7th; http://www.biodic.go.jp/trialSystem/top_en.html). Secondary vegetation after past human disturbances occupies 39.4% of the total vegetation and the remaining 23.0% is primary vegetation. The typical primary vegetation types are, from north to south, boreal mixed forest, deciduous broad leaved forest, and evergreen broad leaved forest.Grid squaresRecords of hiking activity were summarized for 4244 secondary grid squares based on Standard Grid Square System, which was defined by the Minister’s Order of Administrative Management Agency in 1973. In the system, the secondary grid was defined as a grid of 5′ in latitude and 7′ 30″ in longitude, which roughly corresponds to a 10 km grid in the study area. This is the standard grid system of the government and we adopted the system for convenience in future application uses and communication with practitioners. The grids, which are defined by latitude and longitude, are different in the area up to 22% between the north and south ends. Therefore, area of each grid was included in a model as an offset term.Hiking activityAccording to a government survey in 2016, (the Survey on Time Use and Leisure Activities by the Statistics Bureau of Japan, http://www.stat.go.jp/english/data/shakai/index.htm), 10.0% (about 10.7 million people) of Japan’s population age 15 or over enjoyed hiking/mountaineering in the last year. The census showed also that hiking is more popular among urban residents in the metropolitan areas. Both multi-day expedition to high mountains and day trek to low mountains in suburban areas are popular. Because of the severe winter climate, unskilled hikers use the high mountains in summer and early autumn only. During a summer vacation, whose peak time in Japan is August, many hikers enjoy multi-day trips to distant mountains. Spring and autumn are also popular seasons because of the mild weather and the scenic beauty of the fresh green or autumn colors.Data collectionIn this study, we used number of hiking records accumulated on the most popular social networking service for hikers in Japan (Yamareco; https://www.yamareco.com) as a surrogate for flow of recreation service. For all the registered destinations in the study area, the number of hiking records for each month and the latitude and longitude of the destination were collected from the service in September 2016 with the rvest28 package in R software29. This service launched in October 2005 hosts records of the hiking route, photos, participants, and impressions of a hiking trip and facilitates communication among users. Although monthly number of records for each destination is always available on the site, the exact date of each hiking record is not always public information for privacy reasons; therefore, all of the records from the almost 11 years since the start of the service were lumped together in our analysis. Hikers may record multiple places in a single trip, so the total number of records must be larger than the number of unique trips. Users of the service sometime record a place that is not a destination, e.g. start points and stations of trails, parking areas, stations of transports, and bus stops. Such records were excluded before analyses as far as it can be judged from the name of the place. As a result, the total number of hiking records was 4,708,229 records for 16,179 destinations. Finally, these records were assigned to the 4244 grids based on the latitude and longitude of each destination and then total number of records for each grid was used as a surrogate of the recreation service flow in our analysis. Not only total number but also monthly number was used in our analysis to examine seasonal changes in associations between the service and vegetation. Total record number of the grids was strongly right-skewed; no record (handled as 0 in our analysis) was found in 2036 grids while mean and maximum record number were 1109 and 350,384, respectively.Explanation variablesFifty ecological, environmental, and social/infrastructural variables (Table S1) were prepared for each grid by using ArcGIS version 10.5 (ESRI, Redlands, CA, USA). For vegetation and land-use attributes, National Surveys on the Natural Environment by the Biodiversity Center of Japan (2nd–7th; http://www.biodic.go.jp/trialSystem/top_en.html) and National Land Numerical Information (http://nlftp.mlit.go.jp/ksj-e/index.html) were used. The proportion of sea, that of total vegetation cover (excluding agricultural land and pasture) to land area, that of agricultural land (including pasture) to land area, that of natural vegetation (vegetation excluding plantations) to total vegetated area, and that of primary vegetation (vegetation with no record or evidence of a disturbance) to natural vegetation were summarized at four spatial scales: a radius of 10 km, 20 km, 50 km, and 100 km from the center of each grid. Spatial patterns of the three vegetation variables in 10 km radius were summarized in Fig. 1d–f.Maximum elevation, minimum elevation, and ruggedness (index of topographic heterogeneity30) were summarized at the four spatial scales based on a digital elevation model (10-m resolution) provided by the Geospatial Information Authority of Japan (https://fgd.gsi.go.jp/download/menu.php). For climatic variables (annual and monthly mean temperature, annual and monthly precipitation, annual and monthly hours of sunshine, and annual maximum snow depth), the National Land Numeric Information provided by the Ministry of Land, Infrastructure, Transport and Tourism of Japan (http://nlftp.mlit.go.jp/ksj-e/index.html) was referenced. Densities of population and roads at the four spatial scales were prepared from population census data from the Statistics Bureau of Japan (http://e-stat.go.jp/SG2/eStatGIS/page/download.html) and the National Land Numeric Information. For calculation of these densities, the sea surface was excluded. In addition, latitude and longitude of center of each grid were also used as explanatory variables to average effects of spatial coordinates.Statistical analysisIn this study, we employed BRT, a machine-learning method based on regression trees31 for modeling the complex relationship between a CES flow and landscape attributes12. BRT is an ensemble learning method where multiple regression trees are sequentially combined to minimize the loss function by means of gradient descent. This technique has advantage in the development of a model with a high predictive performance, in which high-dimensional interactions among explanatory variables and nonlinear responses are fully accounted for. In ecology, BRT has been frequently used for modeling of a species distribution32.Total and monthly numbers of hiking records were modeled as a function of the 50 variables described above under the assumption of a Poisson response. For temperature, precipitation, and hours of sunshine, annual and monthly average were used for the analysis of total and monthly records, respectively. In modeling by BRT, parameters for building of each learner and assembly of the learners must be carefully chosen to maximize generalization ability of a model31. In our case, candidate parameters were 2, 5, and 10 for the maximum depth of variable interactions for each learner; 2, 5, 10, and 20 for the minimum number of observations in the terminal nodes for each learner; 0.5 and 0.75 for the proportion of training data used for building each learner; and 1000, 2000, 4000, 6000, 8000 and 10,000 for the total number of learners (Table S2). In the model assembling process, the value of 0.01 was used as a shrinkage parameter. Ten-fold cross validation was used to obtain the best suites of parameters. R2 based on sum of squares:$${R}^{2}=1-frac{{sum ({y}_{i}-widehat{{y}_{i}})}^{2}}{{sum ({y}_{i}-overline{{y }_{i}})}^{2}}$$
was used for evaluation of the model’s prediction performance. The importance of explanatory variables was evaluated as an increase of mean absolute error after 100-times permutation of a variable33.Effects of each explanatory variable (a landscape attribute) on the response variable (record number) and the context dependence were visually inspected by individual conditional expectation (ICE) plot34. ICE plot visualizes the effect of a given explanatory variable for each observation by connecting outcome of a model for shifting values of the focal explanatory variable throughout the range while keeping other explanatory variables as the original value. Predictions were performed in log-scale and each line was centered to be zero at the left end of the x-axis to show relative effects of explanatory variables (c-ICE plot sensu Goltstein et al.34). Each line in ICE plot can be colored based on value of the second explanatory variable to assist assessment of the interactive effects of the two predictors. Friedman’s H statistic35 was used to detect explanatory variables whose interaction with the vegetation variables are important and therefore should be used for color-coding of an ICE plot. Friedman’s H is defined as a proportion of variance of partial dependence estimates explained by interactive effects for arbitrary suites of explanatory variables.Then, expected impacts of 0.1 decrease in the three local vegetation variables were assessed by the trained model and mapped. Although vegetation variables were sometimes more important at larger spatial scales (see “Results”), we focused on vegetation at a local (10 km radius) scale because most changes in vegetation occur at the scale in Japan (National Surveys on the Natural Environment by the Biodiversity Center of Japan, https://www.biodic.go.jp/kiso/fnd_list_h.html).All statistical analyses were performed using the R software packag29. The gbm36 package was used for BRT, the iml37 package was used for calculation of Friedman’s H statistic, and the cv.models (Oguro, https://github.com/Marchen/cv.models) packages was used for cross validation and parameter tuning. More