Effects of human depopulation and warming climate on bird populations in Japan

Quantifying biodiversity trends in economically developed countries, where depopulation, associated secondary succession, and climate warming are ongoing, provides insights for global biodiversity conservation in the 21st century. However, few studies have assessed the impacts of secondary succession and climate warming on species’ population trends at a national scale. We estimated the population trends of common breeding bird species in Japan and examined the associations between the overall population trend and species traits with the nationwide bird count data on 47 species collected from 2009 to 2020. The overall population trend varied among species. Four species populations increased moderately, 18 were stable, and 11 declined moderately. Population trends for 13 species were uncertain. The difference in overall trends among the species was associated with their habitat group and temperature niche. Species with relatively low‐temperature niches experienced more pronounced declines. Multispecies indicators showed a moderate increase in forest specialists and moderate declines in forest generalists (species that use both forests and open habitats) and open‐habitat specialists. Forest generalists and open‐habitat specialists also declined more rapidly at sites with more abandoned farmland. All species groups showed an accelerated decline or decelerated increase after 2015. These results suggest that common breeding birds in Japan are facing deteriorating trends as a result of nationwide changes in land use and climate. Future land‐use planning and policies should consider the benefits of passive rewilding for forest specialists and active restoration measures (e.g., low‐intensive forestry and agriculture) for nonforest specialists to effectively conserve biodiversity in the era of human depopulation and climate warming.


INTRODUCTION
The expansion and intensification of agriculture to support an increasing global population has been a major threat to global biodiversity over the past few centuries, and this trend is expected to continue into the 21st century (Leclère et al., 2020;Tilman et al., 2017;Rigal et al., 2023).In contrast to the global trend of human population growth, 61 countries or areas are projected to lose more than 1% of their population between 2022 and 2050 (United Nations, 2022).Additionally, the percentage of the world population over 65 years old is projected to increase from 10% in 2022 to 16% in 2050 (United Nations, 2022).Rural depopulation and aging will likely lead to a decline in agriculture and forestry activities, thereby mitigating the pressure of anthropogenic disturbances on the land (Li & Li, 2017;Schuh et al., 2020).The regrowth of grasslands and forests in such abandoned lands is expanding globally.By the end of the 21st century, they are projected to reach 3.7−4.5 billion ha, corresponding to 6−13% of the total land area (Hurtt et al., 2020).Therefore, the questions of whether and how the reduction in human activities provides new opportunities for global biodiversity conservation are important (Leclère et al., 2020;Tilman et al., 2017;Daskalova & Kamp, 2023).
Secondary succession may restore degraded ecosystems through reforestation (Ceaușu et al., 2015;Fayet et al., 2022;Paul et al., 2016).However, the habitat value of these regenerated areas may not reach that of natural habitats, even after several decades, or may be ephemeral because of recultivation (Atkinson et al., 2022;Crawford et al., 2022;Isbell et al., 2019).Furthermore, forest transition after agricultural abandonment does not always have positive effects on biodiversity and ecosystem services (Queiroz et al., 2014;Quintas-Soriano et al., 2022).This is particularly evident in countries with traditional agricultural landscapes, where low-intensive forestry and agriculture have supported rich species diversity over long periods, often for centuries (Benton et al., 2003;Katoh et al., 2009).Largescale and long-term secondary succession has led to the loss of landscape heterogeneity, resulting in a decrease in the number of available habitats and a subsequent reduction in species diversity at the national scale (Otero et al., 2015;Sugimoto et al., 2022).
Secondary succession and climate change may have additive or synergistic effects on biodiversity (Raven & Wagner, 2021;Daskalova & Kamp, 2023;Rigal et al., 2023).To quantify the large-scale biodiversity changes associated with these environmental changes, previous researchers derived population indicators from composite trends of species abundance in regions characterized by progressive declines in human activity and rising temperatures (Sirami et al., 2008;Herrando et al., 2014;Stephens et al., 2016).Species' population trends do not solely reveal the causes of population changes (Krebs, 1991).However, when combined with information on species' characteristics, such as habitat preferences and climatic niche positions, these trends can identify associations that may underlie the observed trends, including land-use changes and climate warming (Gregory et al., 2019;Soykan et al., 2016).Multispecies indicators (MSIs) of the population trends can then be used to aggregate trends for groups of species that are affected differently by specific changes (Gregory et al., 2019;Stephens et al., 2016).Therefore, nationwide biodiversity indicators can provide valuable information for national land-use planning to effectively conserve biodiversity under depopulation and climate warming scenarios (Sugimoto et al., 2022).
In previous studies, biodiversity indicators highlight the importance of habitat preference in explaining population trends under secondary succession, showing more positive trends for forest species than for nonforest species (Herrando et al., 2016;Reif et al., 2008;Sirami et al., 2008).Other studies show the importance of species' temperature niches, with warm-adapted species increasing in abundance and coldadapted species declining (Gregory et al., 2009;Stephens et al., 2016).However, there is a lack of comprehensive large-scale analyses examining the combined impacts of secondary succession and climate warming on biodiversity trends (but see Herrando et al., 2014).Moreover, while the biodiversity impacts of secondary succession have been investigated primarily in several European counties, there is limited evidence from other regions that have distinct biological communities, farmland types, climates, and views on the relationship between farmland abandonment and biodiversity (Queiroz et al., 2014).For example, East Asian countries, such as China, South Korea, and Japan, have also experienced forest expansion and maturation following the abandonment of forestry and agriculture, and the surface area of such abandoned lands may further increase in the future (Baek et al., 2022;Li & Li, 2017).Despite the ongoing land-use changes, few studies have investigated the biodiversity trends in East Asian countries, except for migratory shorebirds using the East Asian-Australasian flyway (Amano et al., 2010;Wang et al., 2022).Addressing this knowledge gap would improve the accuracy of predictions regarding changes in global biodiversity associated with future increases in abandoned land (Leclère et al., 2020;Crawford et al., 2022).
We aimed to address these gaps by assessing, for the first time, the population trends of common breeding bird species in Japan over 12 years (from 2009 to 2020).Japan is one of the world's most depopulated countries (Ohashi et al., 2019), and it has traditional agricultural landscapes, called Satoyama.These landscapes are characterized by mosaics of secondary forests and grasslands, farmlands, ponds, rivers, and rural settlements (Jiao et al., 2019;Kadoya & Washitani, 2011;Katoh et al., 2009).Agriculture and forestry are being increasingly abandoned in Japan as the human population and primary industry workers decline and age (Figure 1).Consequently, seminatural grasslands and farmlands are being lost and forests are maturing, resulting in the loss of landscape mosaics (Katoh et al., 2009).Therefore, we expected contrasting population trends between forest (positive trends) and nonforest specialists (negative trends).Additionally, considering the increase in annual mean temperature over the past decade (Figure 1), we also expected warm-adapted species to show more positive trends than cold-adapted species.We present Japan as a pioneer economically developed country with ongoing depopulation and climate challenges and sought to provide insights into land management for other countries around the world subject to similar change (Hori et al., 2021).

Bird abundance data
We obtained bird data from the Monitoring Sites 1000 Project, a nationwide monitoring survey of biodiversity across Japan's terrestrial and aquatic ecosystems (Biodiversity Center of Japan, 2022a).We used data sets from forest and grassland sites and forest-agriculture sites (48 and 71 sites, respectively) (Appendix S2).These sites were selected nonrandomly, but efforts were made to evenly distribute all the forest and grassland sites and 18 forest-agriculture sites across the country (Biodiversity Center of Japan, 2019, 2022b).Fifty-three forest-agriculture sites were selected from those proposed by volunteer groups or individuals (Biodiversity Center of Japan, 2019).Nonrandom selection of survey sites can introduce bias in trend estimates when spatial variation in population trends exists (Husby et al., 2021), so we also tested the sensitivity of the estimated trends to the choice of survey sites (details in Appendix S3 and "Statistical Analyses").
At each site, surveys were conducted every 1−5 years.At the forest and grassland sites, skilled volunteers visited each site up to four times during the breeding season (April-July) and recorded all birds detected within a 50-m radius of five pointcount locations, which were spaced >100 m apart along a fixed 1-km transect (Kawamura et al., 2019;Spake et al., 2020).At the forest-agriculture sites, skilled volunteers visited each site up to six times during the breeding season and recorded all birds detected within 50 m along a fixed 1-km transect (Katayama et al., 2014).Both types of sites were usually surveyed on clear or cloudy days (from 04:00 to 09:00 h) without rain or strong winds to minimize variations in detection probability (Katayama et al., 2014;Spake et al., 2020).The same observers usually visited each site across the survey years and their identities were not disclosed to protect their personal information.
We used the bird data from 2009 to 2020 because a different survey method was used at the forest and grassland sites before 2009.We took the maximum number of individuals observed for each species at each site each year from the repeated surveys following the approach of Spake et al. (2020).We assumed that the abundance was generally underestimated by point or linetransect counts and, therefore, that the maximum number of birds detected in any visit represented the minimum number at that location (Spake et al., 2020).We concluded that the imperfect detection of birds would not affect our findings (further discussion in Appendix S4).
Our analyses focused on common bird species observed at 20 or more sites (total of over 300 individuals).For less common species, it was difficult to obtain reliable population trends because their 95% confidence intervals were much larger than those for the common species in our preliminary analysis (results not shown).We also excluded colonial breeding species (i.e., swallows, egrets, and herons) because most of the survey sites did not include their breeding colonies, making it difficult to count their abundance accurately.Therefore, we used 47 bird species for the analyses (Appendix S5).After excluding sites with survey periods of fewer than 5 years, we included 119 sites in the analyses (Appendix S3).We expected that the 5-year criterion would at least accurately determine the direction of population trends (positive or negative), even though it may not precisely match the overall population trend (Wauchope et al., 2019).

Species traits
We selected six species traits to explain the differences in the population trends among the 47 species (Table 1).We classified habitat preference into three groups (forest specialists, forest generalists, and open-habitat specialists) to test our prediction on the impact of land-use changes on bird population trends.We defined forest specialists as species that only use forest in the breeding season, forest generalists as species that use both forests and open habitats (including farmlands, grasslands, wetlands, ponds, rivers, and lakes) in the breeding season, and open-habitat specialists as species that use only open habitats in the breeding season.We obtained information on the habitat use of each species from a comprehensive database of the life history, ecology, and morphology of birds in Japan (Takagawa et al., 2011).The database provides binary data (0, no; 1, yes) indicating the use of five habitat types for each species: forests; farmlands, grasslands, including bare grounds; ponds, rivers, and lakes; and wetlands.This binary information is primarily based on a comprehensive bird catalog in Japan.We supplemented this with information derived from several peer-reviewed papers and expert knowledge (Takagawa et al., 2011).Although this classification is qualitative and based partly on expert knowledge, the quantitative classifications of bird species' habitat preference, based on empirical data, align with qualitative classifications based on expert opinion (Lin et al., 2023).Therefore, expert opinion is considered effective, at least for broad assessments of habitat classifications.
We used the median and range of annual mean temperatures in the breeding range of each species, based on Ueta et al. (2022), as measures of temperature niche position and breadth, respectively.Because the temperature niche position and breadth were positively correlated (r = 0.82), we used only the temperature niche position in our analyses to determine whether species with relatively low-temperature positions experienced more significant declines, considering the recent warming observed in Japan (Figure 1).The temperature position was standardized before the analyses.Further details on the temperature niches are in Appendix S6.
We also used five additional species traits (use of understory, feeding guild, migratory status, annual productivity, and annual productivity) to test the potential impacts of other environmental changes on bird population trends.Information on these traits came from the comprehensive bird database in Japan (Takagawa et al., 2011) and other studies (see below).We included use or not of understory vegetation because recent studies report declines in the populations of several understory bird species in some areas in Japan, presumably owing to herbivory by the sika deer (Cervus nippon), whose population has increased more than seven-fold over recent decades (Seki et al., 2014;Ueta et al., 2014).We used feeding guild to determine whether insectivorous birds are declining, as they are in Europe and North America because of agricultural intensification and loss of natural habitats (Bowler et al., 2019;Stanton et al., 2018).We used migratory status to determine whether the decline in tropical migratory birds in the 1970s to the 1990s, which was possibly a consequence of habitat destruction and excessive hunting in their wintering areas of Southeast Asia (Higuchi & Morishita, 1999;Tamada, 2006;Yamaura et al., 2009), is ongoing.We also included annual productivity, which could be associated with the vulnerability of the population to habitat loss and fragmentation (Amano & Yamaura, 2007).The annual productivity was log-transformed and then standardized before analyses.

Statistical analyses
We modeled log-linear population trends and estimated annual indices for each species separately with (SSG, using) TRIM (trends and indices for monitoring data) (Pannekoek & van Strien, 2005).TRIM accounts for overdispersion and serial correlation and interpolates missing observations with a Pois-son log-linear model (Pannekoek & van Strien, 2005).This method has been used in bird monitoring studies in many parts of the world, including Europe, Africa, and Asia (Khadka & Lamichhane, 2021;Lehikoinen et al., 2019;Wotton et al., 2020).Although census methods differed between the forest and grassland sites (point counts) and forest-agriculture sites (line transects), this is unlikely to introduce systematic bias to the derived trends because the model treats each site independently and assesses the trends in relative rather than absolute abundance (Pannekoek & van Strien, 2005).
We calculated the annual abundance indices and their standard errors for each species over the 12 years.We assumed that all years were possible change points in the population trend and employed a stepwise selection of these change points with Wald tests (model 2 in TRIM).We first ran the model with both overdispersion and serial correlation terms for all species.For the species whose dispersion parameter was estimated to be low (<1.0) or whose serial correlation was low (<0.2),we refitted the model without these corrections.We estimated the slope of a linear regression across log-transformed annual abundance indices of each species (additive slope in TRIM) as a measure of the overall population trend.
To model the relationship between the overall trends of the 47 bird species and their species traits, we used phylogenetic least square regression (PGLS).The response variable was the additive slope, and the explanatory variables were the six traits (Table 1).We used the maximum likelihood estimate of Pagel's λ (Pagel, 1999) to correct for phylogenetic nonindependence, accounting for phylogenetic uncertainty (Appendix S7).To account for the differences in the uncertainty of the overall trend parameter obtained from TRIM, we weighted the estimates based on the inverse of their squared standard errors.This means that more precise estimates have larger importance in the model fitting than less precise estimates (Senzaki et al., 2020).Generalized variance inflation factors were <2.2, indicating low multicollinearity among the predictors (Zuur et al., 2010).
Where significant (p <0.05) effects of the factors were observed, we combined individual species indices produced by TRIM into MSIs with the MSI tool (Soldaat et al., 2017) in R 4.1.2(R Core Team, 2021).This tool uses species-specific annual indices and their standard errors provided by TRIM to calculate annual MSIs and confidence intervals, accounting for sampling error, with the Monte Carlo simulation method.This method calculates a mean and SE from 1000 simulated MSIs, back-transforms them to an index scale, and repeats the process 10,000 times.These indicators are a measure of biodiversity change, where a reduction in the index value mean occurs if more species are declining than increasing and vice versa (Soldaat et al., 2017).Subsequently, we tested for differences in the MSIs between each combination of factor levels with Monte Carlo procedures and reported the average difference in the multiplicative trends with standard error and the significance of that difference.Similarly, we assessed changes in MSIs before and after 2015.We chose this year based on a preliminary visualization of the MSIs.
To gain more direct insight into the impact of secondary succession on bird population trends, we also used a Poisson log-linear model with a binary predictor from TRIM for the degree of farmland abandonment (not or partly abandoned, 0; mostly or fully abandoned, 1).We defined farmland labeled as mostly or fully abandoned as abandoned farmland, and farmland labeled as not or partly abandoned as nonabandoned farmland.We carried out this analysis with 63 agriculture-forest sites (nonabandoned at 27 sites and abandoned at 36 sites), where the information on abandoned farmland was obtained through a survey completed by volunteers at each site.Volunteers were asked the degree of farmland abandonment at the spatial scale of the catchment, which in Japan usually corresponds to 10−100 km 2 (Nakamura et al., 2005).We did not include forest specialists in this analysis because they were not observed at many agricultural-forest sites, which would lead to imprecise estimates because of such a reduced data set.We then used the MSI tool to evaluate the potential impact of the predictors on the population trends.
Similarly, we examined the impact of management activities for biodiversity conservation on bird population trends with the TRIM binary predictor for the presence of management activity (absent, 0; present, 1).Of the 63 sites, 36 were subject to various management activities in each catchment, including extensive mowing of grasslands and bamboo; creation or maintenance of ponds, ditches, and wetlands; and control of invasive non-native species for biodiversity conservation (T.F., personal observation).We pooled these activities due to the limited sample size.

Population trends and species traits
We used the records for 93,338 individuals of 47 bird species at 119 sites over the 12-year study period (Appendix S5).The TRIM results showed variable overall population trends among the species over the study period: moderate increases (4 species), stable states (17 species), moderate declines (11 species), and uncertain states (15 species) (Appendix S8).
The PGLS showed that habitat group significantly explained the species-specific population trends.Forest generalists and open-habitat specialists exhibited greater declines compared with forest specialists (Table 2 & Figure 2a).There was no significant difference in the population trend of forest generalists and open-habitat specialists (Tukey post hoc test, p = 0.47).In the forest specialist group, the abundance of 4 out of the 21 species (19.0%) significantly increased, eight species (38.1%) were stable, another eight (38.1%) were uncertain, and in only one species (4.8%), abundance significantly decreased (Figure 2a).In the forest-generalist group, the abundance of 9 out of the 19 species (47.4%) significantly decreased, nine species (47.4%) were stable, one (5.3%)was uncertain, and no species had a significant increase in abundance (Figure 2a).In the open-habitat specialist group, the abundance of one species (14.3%) significantly decreased, and abundance of the other six species (85.7%) was uncertain.The PGSL model also showed a significant effect of temperature niche position (Table 2); bird species with relatively low-temperature niche positions declined more (Figure 2b).There was no significant interaction between habitat group and climate niche position (p = 0.12).No other traits were significantly associated with the population trends (Table 2).

Multispecies indicators
The MSI showed a moderate decline across the 47 common land bird species from 2009 to 2020 (annual change −0.8%, overall change −10.7%) (Figure 3a).The MSI for forest specialists showed a moderate increasing trend (annual change 0.7%, overall change = 5.9%) (Figure 3b).In contrast, the MSI for forest generalists showed a moderate decline (annual change −1.7%, overall change −18.9%) (Figure 3c).The MSI for openhabitat specialists showed a more severe decline (annual change −2.8%, overall change −30.1%) (Figure 3d).There were significant differences in the MSI trends between forest specialists and forest generalists and open-habitat specialists, but not between forest generalists and open-habitat specialists (Table 3).
Based on temperature niche position, we classified the 47 bird species as either warm-adapted (above the median of all species) or cold-adapted (below the median) (Appendix S5).Both groups showed moderate declines in their MSIs over the 12 years (Figure 3e,f).The warm-adapted species showed a less severe decline (annual change −0.5%, overall change −8.7%) than the cold-adapted species (annual change −1.1%, overall change −12.6%), but the difference was not significant (Table 3).
Moreover, the MSIs showed a significant shift in trends before and after 2015 for the 47 common bird species, forest specialists, forest generalists, open-habitat specialists, warmadapted species, and cold-adapted species (Table 3 & Figure 3).The MSI of forest specialists changed from a moderate increase before 2015 to a stable trend after 2015, and the MSIs of the other groups changed from stable to a moderate decline (Figure 3).

DISCUSSION
The common bird species showed an overall population decline over the study period, particularly after 2015 (Figure 3a), coinciding with forest maturation, loss of open habitats, and an increase in annual mean temperature (Figure 1).The variation in population trends among the species was explained by habitat preferences (Figure 2a), with forest specialists generally increasing and nonforest specialists declining (Figure 3b-d).Temperature niche position also explained the

TABLE 3
Differences in the multispecies indicators (MSIs) of the population trends between each combination of habitat type, year, and site management based on Monte Carlo procedures (1000 iterations from the MSI tool in Soldaat et al., 2017).

Comparison of multispecies indicators Mean difference a SE p
Overall trend (Figure 3) variation.Although warm-adapted and cold-adapted species showed declining trends, species with a low-temperature position declined more rapidly (Figures 2b & 3e,f).These results support our prediction regarding recent changes in bird populations associated with secondary succession and climate warming at a national scale, highlighting the negative and positive effects of secondary succession.

Drivers of population changes
The overall positive trend in forest specialist birds aligns with findings from European countries that have experienced either expansion of forested area or forest maturation (Bowler et al., 2021;Herrando et al., 2016;Reif et al., 2022).This pattern supports the idea that secondary succession provides an oppor-tunity for restoring forest biodiversity (Cunningham et al., 2015;Regos et al., 2016).In Japan, the forest area has remained relatively stable since the 1960s at approximately 250,000 km 2 (Yamaura et al., 2009).However, the volume of plantation forests and natural forests has increased by 24.8% and 8.6%, respectively, from 2008 to 2018 (calculated from Appendix S1).This forest maturation resulting from the cessation of most forestry activities appears to be driving the increase in forest specialists (Figure 1).Although plantation forests generally support lower bird diversity compared with natural forests, they can still serve as important habitats for forest birds, such as the narcissus flycatcher (Ficedula narcissina), depending on the planted tree species and tree ages (Castaño-Villa et al., 2019;Kawamura et al., 2021;Spake et al., 2019).Therefore, the maturation of natural and plantation forests seems to have contributed to the increase in the abundance of forest birds across the country.Furthermore, the expansion of native woody plants in long-abandoned farmland, particularly in hilly and mountainous areas, may have also benefitted certain forest bird species (Katayama et al., 2021;Broughton et al., 2021).
The decline in the abundance of open-habitat specialist birds (Figure 3c) corresponded with the loss of open habitats in Japan (Figure 1).Similarly, grassland and farmland birds in North America and Europe have shown rapid declines in recent decades, likely as a consequence of agricultural intensification and abandonment (Bowler et al., 2019;Inger et al., 2015;Rosenberg et al., 2019;Stanton et al., 2018).In Japan, the impacts of agricultural intensification do not appear to have increased significantly over the past 20 years.For example, the cumulative ecological risks associated with pesticides have continuously decreased (Nagai et al., 2022) and the area of farmland consolidation (e.g., the expansion and shaping of agricultural land) has increased only slightly (Ministry of Agriculture, Forestry & Fisheries, 2022a).Moreover, growing evidence suggests that abandoned fields have transitioned to tall grasslands or forests, causing a loss of early-successional habitats and associated species, including the skylark (Alauda arvensis) (Copland et al., 2012;Koshida & Katayama, 2018; but see Kitazawa et al., 2019).Furthermore, rural abandonment could cause a nationwide decline in open-land butterflies (Sugimoto et al., 2022).Our results, therefore, suggest that the loss of early-successional habitats drives a large-scale decline in open-habitat specialists.
With ongoing secondary succession (Figure 1), we also found an overall negative trend in forest generalist birds (Figure 3c); most species declined or were stable (Figure 2).Previous studies show stable or increasing trends in habitat generalists under increased anthropogenic disturbances (Clavel et al., 2011;Kormann et al., 2018).These results may support the theory that reduced disturbance could lead to a more specialized forest species assemblage (Clavel et al., 2011).In Japan, declining generalist species, such as the meadow bunting (Emberiza cioides), may prefer early-successional habitats, such as open forests, grasslands, and farmlands, more than the generalist species with a stable trend (Yamaura et al., 2009).Additionally, some species use multiple habitats during their life cycle, each stage for different vital functions at a certain stage (Arroyo-Rodríguez et al., 2020;Fahrig et al., 2019), and may be affected by the reduction in or loss of habitat heterogeneity (Katoh et al., 2009;Otero et al., 2015).
Furthermore, we found that although even warm-adapted species declined (Figure 3e), cold-adapted species experienced more pronounced declines (Figure 2b) with ongoing climate warming.Importantly, all species groups showed an accelerated decline or decelerated increase after 2015, coinciding with an acceleration in temperature increase (Figure 1).These results suggest that climate warming is another significant driver of the recent bird population changes in Japan, exerting negative impacts on many common species.For example, some species in Japan, such as cuckoos, are facing an increasing ecological mismatch with their food resources (plants and invertebrates), which are advancing their spring phenology more rapidly than birds (Gordo & Doi, 2012;Ogawa-Onishi & Berry, 2013).The combined impacts of land-use changes and climate warming appear to be responsible for the more negative trends observed across bird groups after 2015.
Many bird species show delayed responses to changes in land use and climate, which can vary from a few years to decades, depending on the species and the type of environmental changes (Cornford et al., 2023;Daskalova et al., 2020).Therefore, the bird population trends we estimated may reflect the impacts of land-use changes a few years to decades earlier, when forest maturation and the loss of open habitats were more rapid than the present study period (2009-2020) (Figure 1).Climate change may play an increasingly important role in driving bird population trends in Japan in the future.Future research should aim to disentangle the effects of these drivers and consider time-delayed biodiversity responses (Conford et al., 2023).
Although the PGLS model did not support the effects of other traits, some of these factors might also be important at different spatial scales.In particular, the rapid increase in deer populations has threatened understory birds in certain areas of Japan because of the loss of understory vegetation (Seki et al., 2014;Ueta et al., 2014).The forest area damaged by deer and other animals was estimated to be 4870−6364 ha from 2017 to 2021, which accounts for <0.05% of the total forest area (Ministry of Agriculture, Forestry & Fisheries, 2022b).However, the affected area may increase in the future if global warming reduces snowfall and decreases the mortality of deer in winter (Kaji et al., 2010).

Site-level population trends and potential bias
Nonforest specialist birds declined more rapidly at abandoned sites or sites without conservation management (Figures 4 & 5).However, the difference in trends was not statistically significant (Table 3).This may be explained by the qualitative nature of our explanatory variables based on interviews with surveyors.More quantitative data may be needed to understand the nonlinear responses of biodiversity to farmland abandonment.Small to medium proportions of abandoned farmlands can have positive effects on bird diversity by increasing habitat heterogeneity (Katayama et al., 2021).Additionally, our variable of management activity included various practices that may have mixed biodiversity outcomes.Future research should consider these issues when assessing the nationwide impacts of secondary succession and conservation management on bird population trends.
There was no clear difference in the elevation or land-cover types between the survey sites and random sites (Appendix S3).However, we found that the proportion of management activities at the study sites was likely to be higher than the proportion of management activities at all agriculture-forest sites in the country, which can potentially cause bias in trend estimates, especially for nonforest specialists (Figure 5).Our additional analysis showed that, compared with our estimates (Figure 3), the actual declines in forest generalist and open-habitat specialist birds could be higher and lower, respectively, on a national scale.

Conservation implications
The contrasting population trends we found between forest specialists and other species have implications for land-use management for biodiversity conservation in the face of human depopulation and climate warming, which is projected to continue in Japan and many other economically developed countries (Ohashi et al., 2019;United Nations, 2022).Although secondary succession could provide new opportunities for the restoration of forest specialist birds, it also poses threats to species that depend on early-successional habitats.Additionally, climate warming continues to pose a significant threat to many bird species in Japan, particularly those with lower temperature niches.Therefore, future projections of global biodiversity changes associated with secondary succession and climate change should take these species' niches into account.Systematic approaches to identify the priority areas for restoration and restoration modes (e.g., passive rewilding vs. active restoration) are also crucial for improving population trends across biological communities, especially in areas like Japan where both habitat groups include species of conservation concern (Perino et al., 2022).Spatial niche modeling of biological communities can provide insights into the spatial allocation of these restoration strategies (Sugimoto et al., 2022).
Considering that many farmland and grassland species in Japan are threatened with local or regional extinction (Sugimoto et al., 2022;Tamada, 2006), greater efforts are needed to facilitate active restorations of early-successional habitats.These efforts include the maintenance of open forests and forestagriculture mosaics (Figure 4; Katayama et al., 2021;Katoh et al., 2009), creation of seminatural grasslands and ponds (Deguchi et al., 2020;Tscharntke et al., 2021), and organic and other low-intensive farming (Katayama et al., 2019).Future research should evaluate the effectiveness of these restoration strategies in the nationwide population trends of biodiversity under projected climate changes.Such studies should consider common and rare species that may show different population trends (Burns et al., 2021).
Our results also highlight the importance of nationwide biodiversity indicators in Japan, which should be a key consideration for the 2050 biodiversity goals for East Asia (Díaz et al., 2020;Perino et al., 2022).However, knowledge gaps remain concerning the migration ecology of land birds in the East Asian Flyway, some of which are severely declining and endangered because of hunting, habitat loss and degradation, and climate change (Yong et al., 2021).Without addressing these flyway-wide challenges, national land management for biodiversity conservation may not work well for migratory land birds.There is much potential for strengthened regional collaboration between countries to establish standardized monitoring programs for migratory land birds, including mainland China (Kamp et al., 2015), Republic of Korea (Kim et al., 2021), Taiwan (Lin & Pursner, 2020), and Japan.Recent developments of the East Asian Bird Monitoring Scheme (Chan, 2015) can be a major step toward land bird conservation in East Asia.

ACKNOWLEDGMENTS
This work was conducted as a part of the Monitoring Sites 1000 Project in Japan.We are grateful to 700 volunteer surveyors who contributed to the project (Appendix S9) and to S. Okubo for his assistance with the GIS analyses.We also thank the editor and three anonymous reviewers for constructive comments.N.K. was financially supported by the Environment Research and Technology Development Fund of the Environmental Restoration and Conservation Agency of Japan (grant 2-2302) and JSPS KAKENHI (grants 17K15057 and 21K05631).F.M. was financially supported by the Czech Science Foundation GA ČR (project 23-07103S).

FIGURE 1
FIGURE 1 Social and ecological changes (see Appendix S1 for data sources) and expected population trends of common land birds in Japan over the past decades (dashed line in annual mean temperature graph, year 2015).Bird illustrations from https://www.ac-illust.com/.

TABLE 1
Explanatory variables for the population trends of 47 species of common breeding land birds in Japan from 2009 to 2020., forest generalist (1), or open-habitat specialist (2) Temperature niche position continuous range: 12.1−26.1 (median of annual mean temperatures where each species was recorded, 12.8 (mean number of broods per year × mean clutch size) *Numbers in parentheses are categories for analyses.

FIGURE 2
FIGURE 2Relationships between estimated population trend for 47 species of common breeding birds in Japan by (a) habitat group (green, moderate increase; blue, stable; orange, moderate decrease; black, uncertain; different lowercase letters, significant differences in the growth rate [p< 0.05, Tukey post hoc test]) and (b) temperature niche position (point size, proportional to precision of the population trend; heavy black line, marginal effect; thin lines, 95% CIs).

FIGURE 4
FIGURE 4 Multispecies indicators (MSIs) of bird population trends for (a, b) forest generalists and (c, d) open-habitat specialists in (a, c) agriculture-forest sites with mostly abandoned area (27 sites) and (b, d) agriculture-forest sites with no or partly abandoned area (36 sites) (blue, stable population; vermillion, moderately deceasing trends; light gray, population trends of individual species).A vertical line indicates that population trends before and after 2015 are significantly different.

FIGURE 5
FIGURE 5 Multispecies indicators (MSIs) in Japan for (a, b) forest generalists (18 species) and (c, d) open-habitat specialists (six species) in (a, c) agriculture-forest sites without conservation management (36 sites) and (b, d) agriculture-forest sites with conservation management (27 sites) (blue, stable population; vermillion, moderately deceasing trends; light gray, population trends of individual species).A vertical line indicates that population trends before and after 2015 are significantly different.

TABLE 2
Results of the phylogenetic generalized least square model for the relationship between the overall population trend of the 47 bird species and their species traits.