Ecology

Sampling sitesLongitudinal observation campaigns for I. ricinus nymph activity were carried out at 11 sampling sites in forest areas from seven different tick observatories across France. Tick observatories are located at the following French municipal areas, where the coordinates of the centre of each municipal area and the climatic types29 are also provided as: (1) La Tour de Salvagny (45° 48′ 50.6″ N 4° 42′ 53.2″ E; Mixed climates); (2) Saint-Genès-Champanelle (45° 43′ 23.8″ N 3°01′ 08.0″ E; Mountain climate); (3) Etiolles (48° 37′ 59.9″ N 2° 28′ 00.1″ E; Degraded oceanic climate); (4) Carquefou (47° 17′ 58.5″ N 1° 29′ 26.0″ W; Oceanic climate); (5) Gardouch (43°23′ 25.7″ N 1° 41′ 02.1″ E; South-West Basin climate); (6) Velaine-en-Haye (48° 42′ 13.4″ N 6° 01′ 16.1″ E; Semi-continental climate); (7) Les Bordes (47° 48′ 47.3″ N 2° 24′ 01.3″ E; Degraded-oceanic climate) (Fig. 1). The observation campaigns were carried out from April/June 2014 to May/June 2021 in most observatories, except for Les Bordes, which began in April 2018.Figure 1The map was created using QGIS version 3.8, Zanzibar (https://www.qgis.org). The climatic region types were previously classified by Joly et al.29.The distribution of tick observatories according to the climatic region types of continental France: (1) Etiolles (degraded oceanic); (2) Velaine-en-Haye (semi-continental); (3) Les Bordes (degraded oceanic); (4) Carquefou (oceanic); (5) La Tour de Salvagny (mixed); (6) Saint-Genès-Champanelle (mountain); (7) Gardouch (south-west basin). Phenological patterns observed at each observatory were also indicated.Full size imageEach tick observatory corresponds to one sampling site except La Tour de Salvagny, Gardouch, and Les Bordes (Table S1). In La Tour de Salvagny, we had to withdraw the observations at the original site (La Tour de Salvagny A) in September 2016 because the site became no longer accessible. In April 2017, we continued our observations at a nearby site, approximately 2 km apart (La Tour de Salvagny B). In Gardouch, the activity of questing nymphs was observed both inside and outside the enclosed area of an experimental station on roe deer (Capreolus capreolus), referred to as Gardouch Inside and Gardouch Outside, respectively. The estimated population density of roe deer in Gardouch Inside (50 individuals per 100 ha) was higher than Gardouch Outside (less than 20 individuals per 100 ha) (H. Verheiden, personal communication, 15th October 2021). Furthermore, three sampling sites in Les Bordes, approximately 1.2 km apart, were referred to as Les Bordes A, B, and C, respectively. Additional sampling sites of these observatories were considered and reported as distinct sampling sites in further analyses, resulting in a total to 11 sampling sites from 7 observatories. Furthermore, due to their geographical proximity, meteorological/climatic factors of different sampling sites from the same observatories were considered identical in subsequent statistical analyses, whereas land cover and topography factors could be varied.Field observation campaigns were planned and carried out by local investigators who had been trained on the sampling protocol. The locations of forests, sampling sites, and passages were chosen where their biotopes are known to be suitable for I. ricinus tick populations around each observatory at the time the field observation campaigns started30. The observations were never carried out during the daytime when the weather was highly unfavourable to questing ticks, e.g., heavy rain, snow, or snow cover.Sampling protocol for questing Ixodes ricinus nymphsActivity of questing I. ricinus nymphs was observed by a cloth-dragging sampling technique31. Within a 1-km radius, a 1 m × 1 m white cloth was dragged over 10 observation units of 10 m short-grass vegetative forest floors, called transects. For each transect, a repeated removal sampling design was used27. The cloth-dragging sampling process was successively repeated three times per sampling. All nymphs found on white cloth in each campaign were removed and collected in a vial for subsequent morphological identification32 by the same acarologists at the corresponding laboratories. As a result, the questing nymph activity of each sampling site was monitored as a total number of confirmed I. ricinus nymphs collected from three repeated sampling on 10 transects, equivalent to a surface area of 100 m2. This measure was considered as an indicator for tick abundance on the day of sampling. The same transects were repeatedly sampled throughout the study period at approximately 1-month intervals.Environmental dataWe tested 28 environmental variables to explain the observed I. ricinus nymph activity (Table 1). These variables could be categorized as: (1) Daytime duration and meteorological variables (time-dependent, 9 variables); (2) Land cover, topography, and bioclimatic variables (time-independent, 19 variables).Table 1 Environmental variables (meteorological, land cover, topography, and bioclimatic variables) used to explain I. ricinus nymph counts per 100 m2 in regression analysis.Full size tableDaytime duration and meteorological variablesDaytime duration ((daytime)) from January 2013 to June 2021 at each sampling site was obtained from the corresponding latitude using geosphere package33. Hourly meteorological data (2-m temperature and relative humidity) were recorded locally at each forest. Subsequently, daily mean, minimum, and maximum values of temperature (({T}_{M}), ({T}_{N}), and ({T}_{X}); in °C) and relative humidity (({U}_{M}), ({U}_{N}), and ({U}_{X}); in %) were derived from these hourly records. The meteorological seasons of the temperate area in northern hemisphere are defined as: (1) Spring, 1st March to 31st May; (2) Summer, 1st June to 31st August; (3) Autumn, 1st September to 30th November; (4) Winter, 1st December to 28th or 29th February.Missing values found on these local daily-level variables were imputed by the random forest algorithm in mice package34. External daily meteorological data, i.e., daily average temperature and relative humidity, derived from neighbouring weather stations (Météo-France or INRAE), as well as month and year information, were used as auxiliary variables (Table S2). As a result, the imputation process creates a total of 500 iterated values for each variable. The median values of 500 imputations were used to replace the missing values.The imputed daily meteorological data were subsequently used to calculate the averaged values in different lagged time intervals for further analysis, called interval-average variables15. The interval-average variables were generated to reduce the uncertainty that might arise during the imputation process and to capture the cumulative effects of the meteorological variables, which were mean temperature ({T}_{M}) and minimum relative humidity ({U}_{N}). The interval-average variables were defined as the average values of a meteorological variable (Min) {({T}_{M}), ({U}_{N})} during a period between ({t}_{1}) to ({t}_{2}) month(s) before the sampling, denoted as ({M}^{{t}_{1}:{t}_{2}}), where 1 month consists of 28 days. As temperature conditions affect several ecological processes of tick populations, particularly developmental and questing rates3, the mean temperature ({T}_{M}) was selected for further analysis to reflect the overall temperature effects. While the minimum relative humidity ({U}_{N}) was chosen for the following reasons: (1) the survival of I. ricinus is highly sensitive to desiccation conditions6,7,8. As a result, when compared to mean or maximum relative humidity, minimum relative humidity is a relatively strong indicator of the effects of desiccation stress; (2) the variation of minimum relative humidity among all sites was higher than that of the mean and maximum relative humidity. This high variation allowed us to better describe meteorological characteristics of each sampling site.Here, we hypothesized that interval-average meteorological conditions influence the dynamics of observed nymph activity at different time lags in different manners. Short-term lags may have an impact on immediate responses, such as the probability of questing. At the same time, long-term lags may influence the dynamics of nymph abundance, which is associated with development and survival rates. Therefore, we explored the impact of each meteorological variable at following time lags on the observed nymphs activity in subsequent regression analysis: (1) 1-month moving average condition, ({M}^{0:1}); (2) previous 3-to-6-month moving average condition, ({M}^{3:6}); (3) 6-month moving average condition, ({M}^{0:6}); (4) 12-month moving average condition, ({M}^{0:12}). For instance, ({T}_{M}^{0:1}) denotes 1-month moving average temperature, representing an average of temperature between 0 and 1 months (0–28 days) before the day of sampling.In addition to the interval-average variables, monthly and seasonal average values of mean temperature and minimum relative humidity during the observation period were also calculated to describe the characteristics of meteorological conditions of each sampling site.Land cover, topography, and bioclimatic variablesWe obtained land cover, topography, and bioclimatic data from a 1-km radius buffer area around the center of each sampling site to capture habitat characteristics across all 10 transects. All the variables were handled and obtained by using QGIS version 3.8.035. The digital elevation model (DEM) data derived from the Shuttle Radar Topography Mission (SRTM) database36 was used to describe the topographic features of sampling sites, which included the mean (({mean}_{elv})) and standard deviation (({sd}_{elv})) of the elevation (in m above sea level), the proportion of flat area (({p}_{flat}); defined by the slope ≤ 2.5%37), the proportion of area facing north (({p}_{north})), east (({p}_{east})), west (({p}_{west})), and south (({p}_{south})), and the catchment area ((catchment)) as a proxy variable for moisture. Bioclimatic variables for each site (historical average conditions during 1970–2000) were derived from the WorldClim database38, including the annual mean temperature (({BIO1}_{Temp}); in °C), the mean diurnal range (({BIO2}_{Diur}); in °C), the maximum temperature of the warmest month (({BIO5}_{maxTemp}); in °C), and the annual precipitation (({BIO12}_{Prec}); in mm). The land cover features of each sampling site were described using the CORINE Land Cover (CLC) 201839, while the characteristics of forests were explained by the BD forêt version 2 data40. The forest fragmentation was characterized by the percentage of forest-covering area (({p}_{Forest})), the forest edge density (({ED}_{Forest}); in m/km2), and the number of forest patches (({n}_{Forest})). While the diversities of the land cover types (level-1 and level-2 CLC) and the forest types were calculated by using the Shannon’s diversity index41 ((H)) as (H=sum_{i=1}^{S}{p}_{i}mathrm{ln}{p}_{i}), where (S) is the total number of land cover/forest types and ({p}_{i}) is the proportion of land cover/forest type (i) within the 1-km radius buffer area. The Shannon’s diversity index for level-1 CLC, level-2 CLC, and forest types were denoted as ({H}_{CLC1}), ({H}_{CLC2}), and ({H}_{Forest}), respectively. Finally, the soil pH data (({pH}_{soil})) was retrieved from the European Soil Data Centre (ESDC) database42.Statistical analysisAll the statistical analyses were carried out using the programming language R version 3.6.043. The variations of questing nymph population of each site were described by using (1) baseline annual nymph counts (spatial variation); (2) phenological patterns (seasonal variation). A baseline annual nymph count of site (i) (({{N}_{base}}_{i})) was defined as a summation of monthly median nymph counts ({varvec{tilde{N}}}_{i}={{tilde{N }}_{i,t}}) across all 12 months (tin left{mathrm{1,2},dots ,12right}) and expressed as: ({{N}_{base}}_{i}=sum_{t=1}^{12}{tilde{N }}_{i,t}). Subsequently, the monthly median nymph counts of each site ({varvec{tilde{N}}}_{i}) were transformed into normalized monthly median nymph counts ({varvec{tilde{N}}}_{i}^{*}={{tilde{N }}_{i,t}^{*}}) following Eq. (1) to have a range value of 0 to 1, which allows us to compare phenological patterns among all sites that have different annual baseline nymph counts.$${tilde{N }}_{i,t}^{*}=frac{{tilde{N }}_{i,t}}{mathrm{max}({stackrel{sim }{{varvec{N}}}}_{i})}$$
(1)
The term (mathrm{max}({stackrel{sim }{{varvec{N}}}}_{i})) denoted the maximum monthly median nymph counts. The normalized median nymph count ({tilde{N }}_{i,t}^{*}) of 1 indicates the maximum nymph activity (peak), while the value ({tilde{N }}_{i,t}^{*}) of 0 designates the absence of nymph activity. Afterwards, the phenological patterns were descriptively classified using the following criteria: (1) the season which the peaks of activity arrive; (2) evidence of reduced activity during winter (November–January); (3) the number of activity waves in a year, whether the pattern is unimodal or bimodal. After assigning phenological patterns to each site, the overall trends of different patterns were derived from medians of the normalized monthly median nymph count ({tilde{N }}_{i,t}^{*}) from all sites that belonged to each pattern. Furthermore, the directional changes in the maximum nymph counts were tested using a Spearman’s rank correlation coefficient, a p-value More

Our results show that the effects of the forest disturbance drivers on biodiversity are likely to be different from those simply expected from the baseline proportions of the forest disturbance drivers if we take into account the threatened species’ distributions. The amount of forest habitat is a primary factor for species diversity of many taxa, including mammals, amphibians, reptiles, birds, insects, and plants18. Indeed, our results revealed that threatened forest species have been exposed to a disproportional decrease in their habitat amount globally (i.e., lower proportions of forest with no or minor loss in all regions when species ranges were considered). Although this finding may be intuitive as population size and/or species range are part of the criteria in the IUCN assessment19, the detected pattern supports the validity of our approach of combining a forest disturbance map and species ranges for evaluating the impact of forest disturbances on threatened species. Moreover, we found that the dominant drivers differ among regions: the proportion of forestry, for example, increased in northern regions such as North America and Europe, whereas that of shifting agriculture increased in tropical regions when threatened species’ distributions were considered. These facts indicate although several influential international schemes for conservation have been implemented for regulating forestry20,21, different mechanisms aiming to directly tackle the over land use for local agriculture may increase their importance when we consider conservation in tropical regions. Our findings suggest that the social and economic drivers underlying the forest disturbance that impacts biodiversity differ among regions or nations, and it is important to establish specific conservation strategies in order to be effective.Based on the findings, we further emphasize that the combinations of multiple interacting drivers are likely to vary among regions. For example, the frequency and extent of stand-replacing natural disturbances such as wildfires have clearly been magnified by climate change, particularly in the Northern Hemisphere (e.g.,22). After such natural disturbances, societal demand for timber and/or pest reduction compels forest managers to ‘salvage’ timber by logging before it deteriorates, a common practice even in locations otherwise exempt from conventional green-tree harvesting, such as national parks or wilderness areas23. Thus, salvage logging clearly mediates the interaction between disturbances by forestry and wildfires and is likely to further affect biodiversity under climate change. Especially in regions where infrastructure (e.g., irrigation systems) has not been well developed, unpredictable changes in precipitation due to climate change was reported to increase forest disturbance by unregulated increases of agricultural land use24. Such regions largely overlapped with regions where shifting agriculture was identified as a dominant disturbance driver for threatened species in this study. Moreover, species themselves shift their ranges in response to climate change25, which would also shift major disturbance drivers and influential interactions of drivers to which the species are exposed, given the region-specific driver patterns. These examples clearly suggest the necessity to understand both the region-specific interrelations among multiple drivers and species’ responses for better prediction of land-use change and thus its effects on biodiversity.Shifting agriculture was the most dominant driver in all tropical regions corresponding to the recent estimates suggesting that the cover of regenerating secondary forest is increasing worldwide26. We demonstrated that this tendency is more drastic especially within the range of threatened species. The effect of shifting agriculture per unit area might be more limited than that of commodity-driven deforestation, which permanently alters forests into other land uses, since habitat structure might recover as the forest vegetation regenerates to a secondary state following the abandonment of the small clearings. However, ample evidence shows that many types of agricultural activities significantly degrade the conservation value of primary forest, especially in the tropics27, which often recovers very slowly if ever28 with the loss of irreplaceable conservation values. Therefore, given the wide areas of dominance of shifting agriculture across all tropical regions, its effect is likely to be pervasive. Consistently, our results show that species extinction risk (i.e., IUCN Red List status) is positively related to the proportional coverage of shifting agriculture (Fig. 2). In addition, as expected, a larger current proportion of shifting agriculture within a species range worsens the change rate in IUCN Red List status of the species (Fig. 4b). Furthermore, the effect is anticipated to be magnified for forest specialists because they are exposed to larger proportions of shifting agriculture than are forest generalist (Fig. 2), and they are also reported to recover more slowly than do forest habitat generalists27,28.A guideline for forest restoration suggested that appropriately sized landscapes should contain ≥40% forest cover (higher percentages are likely needed in the tropics), with about 10% in a very large forest patch and the remaining 30% in many evenly dispersed smaller patches and semi-natural wooded elements (e.g., vegetation corridors)29. Importantly, the guideline also suggests that the patches should be embedded in a high-quality matrix. Although younger secondary forest cannot be a substitute for pristine forest until 50 years or more after a disturbance, it can help to improve the quality of matrix in agricultural landscapes30. Indeed, we show that the negative impacts of shifting agriculture and forestry on IUCN status change have improved over time (Fig. 4b, c), presumably corresponding to the forest regenerating and recovery process. In contrast, the pattern of commodity-driven deforestation, a land use accompanied with permanent forest loss, showed a prolonged negative impact on IUCN status change (Fig. 4a). Notably, whether regenerating forests can move towards a highly diverse and structurally complex state or towards a state of low to intermediate levels of biodiversity and structural complexity depends on the amount of remaining intact mature forest in the landscape29. Therefore, a promising direction for future research would be to develop our analysis further to include spatiotemporal relationships among mature forest remnants, secondary forests, disturbance drivers, and threatened species populations.For conserving the core patches of mature forests, the establishment of protected areas (PAs) is one of the most effective legal measures that has been widely used to regulate land use for biodiversity31. On the other hand, for improving matrix quality, balancing conservation and use of the ecosystem would be critically important; shifting agriculture, for example, causes forest degradation, but it also contributes to food supply chains sourced from smallholder farmers and to food security of local communities8. In fact, establishing mechanisms for managing biodiversity-friendly landscapes has been intensively discussed recently, given the large potential influence of these landscapes on conservation32. These mechanisms include setting an international target on OECMs15. Our finding of a disproportional decrease in forest proportions with minor or no loss within species ranges supports the urgency of the discussion. At the same time, our results highlight an opportunity because large portions of the disturbed forests for threatened species are dominated by shifting agriculture at the global scale, especially in the tropics. As suggested above, if manged properly, such landscapes can still retain or improve functions as essential habitats and/or matrix for a variety of forest-dwelling species. Our analytical method provides a tool set to identify and prioritize areas where such attempts are urgently needed.Global demands for natural resources and ecosystem services drive land use in forests33 and thus affect biodiversity. Therefore, connecting the supply chains to the five major drivers of forest disturbance and their spatial overlaps with biodiversity is essential to inform how we should regulate and design material flows from forest ecosystems to keep them sustainable by minimizing the effects on biodiversity. Existing studies examining the impacts of resource consumption on biodiversity through supply chains of various sectors have often been assessed at the country scale (e.g.,12), partly because the availability of statistics needed to estimate material flows in supply chains is usually limited at finer (i.e., subnational) scales (but see34). We believe that our study provides the first basis for filling the resolution gap between trade statistics and local biodiversity effects by identifying patterns of the local co-occurrence of biodiversity and the forest disturbance drivers that can be directly linked to resource production at the national scale. Note, however, that downscaling a remotely sensed global data set into finer scales inevitably propagates errors and biases which include both those in the original maps and those in the processed data produced by analyses. Thus, preparation of more high-resolution data sets is essential, especially for disturbance drivers and threatened species’ distributions in our case, to keep the errors and biases at a reasonable level at focal spatial scales.The effectiveness of area-based conservation measures to regulate land use for conservation including PAs and OECMs also depends strongly on social and ecosystem conditions. For example, a few studies show that the effectiveness of PAs in halting or slowing forest disturbances depends on PA characteristics such as size and history, as well as on the management entities such as subnational governments or indigenous peoples35,36,37. Moreover, there has been no attempt to elucidate whether PAs and OECMs are effective at regulating supply chains as a supply-side measure by balancing resource production, ecosystem services for local communities, and biodiversity conservation; to tackle this issue, it will be necessary to conduct extensive analyses integrating spatial and temporal patterns of biodiversity, forest loss, its drivers, and material flows in global food supply chains. Though it is challenging and beyond the scope of this paper, solving this issue is urgent and raises a promising opportunity for future research. More

We selected eleven major wheat production provinces of China for the study area, which comprise the largest winter wheat-sowing fraction: Henan, Shandong, Anhui, Jiangsu, Hebei, Hubei, Shanxi, Shaanxi, Sichuan, Xinjiang, and Gansu (Fig. 1). The wheat planting area is about 22 million ha in these provinces, accounting for more than 93% of the total wheat planting area. The total wheat production in these regions contributes more than 96% of the total wheat production in China, with more than 128 million tons in 201933.We developed a methodological framework for high-resolution AGB mapping. It mainly includes three parts: (1) Data acquisition and processing. (2) The WOFOST model parameterization and calibration. (3) Data assimilation (Fig. 1). Each part is explained in more detail below.Data acquisition and processingMeteorological dataChina Meteorological Forcing Dataset34,35 is used as weather driving data for the WOFOST model. The dataset is based on the internationally existing Princeton reanalysis data, Global Land Data Assimilation System data, Global Energy and Water Cycle Experiment-Surface Radiation Budget radiation data, and Tropical Rainfall Measuring Mission precipitation data. It is made by fusing the conventional meteorological observation data of the China Meteorological Administration. It includes seven elements: near-surface air temperature, air pressure, near-surface total humidity, wind speed, ground downward shortwave radiation, ground downward longwave radiation, and ground precipitation rate. The meteorological drive elements required for WOFOST are daily radiation, minimum temperature, maximum temperature, water vapor pressure, average wind speed, and precipitation. Details of these variables that participated in the WOFOST model can be referred to in Table S1.Soil characteristics measurements and phenology observationsSoil and phenology data were collected at 177 agricultural meteorological stations (AMS) from 2007 to 2015 (Fig. 1). Soil characteristics include soil moisture content at wilting points, field capacity, and saturation. To be consistent with the corresponding units in the crop model, the original data in weight water content was converted into volume water content through the corresponding soil bulk density measurements. Winter wheat phenology observations include the date of emergence (more than 50% of the wheat seedlings in the field show the first green leaves and reached about 2 cm), anthesis (the inner and outer glumes of the middle and upper florets of more than 50% of the wheat ears in the whole field are open, and the anthers loose powder), and maturity (more than 80% of the wheat grains turn yellow, the glumes and stems turn yellow, and only the upper first and second nodes are still slightly green). In most cases, the phenological stage “anthesis” is missing. The anthesis date was calculated by adding seven days to the observed heading date (when more than 50% of the wheat in the whole field exposes the tip of the ear from the sheath of the flag leaf).County-level yield statistics dataThe county-level yield data was collected from city statistical yearbooks of the study area from 2007 to 2015. Since most statistical yearbooks do not directly record per-unit yield data, the county-level yield was obtained by dividing the total yield and planting area. It is worth noting that all yields were calculated in units of metric kilograms per cultivated hectares (kg·ha−1).The winter wheat land cover dataWe used a winter wheat land cover product from a 1 km resolution dataset named ChinaCropArea1km36. This data was derived from GLASS leaf area index products and crop phenology from 2000 to 2015. This dataset is the base map of our data production.The MODIS LAI dataWe used the improved 8-days MODIS LAI products (i.e., 1 km) generated by Yuan et al.32 to assimilate the WOFOST model. The products applied the modified temporal-spatial filter and Savitzky-Golay filter to overcome the spatial-temporal discontinuity and inconsistence of raw MODIS LAI products, which makes them more applicable for the realm of land surface and climate modeling. The products can be accessed via the Land-Atmosphere Interaction Research Group website at Sun Yat-sen University (http://globalchange.bnu.edu.cn/research/lai).The WOFOST model parameterization and calibrationThe WOFOST model introductionThe WOFOST model was initially developed as a crop growth simulation model to evaluate the yield potential of various crops in tropical countries37. In this study, we chose the WOFOST model because the model reaches a trade-off of the complexity of the crop model and is suitable for large-scale simulations3. The WOFOST model is a typical crop growth model that explains crop growth based on underlying processes such as photosynthesis and respiration and their response to changing environmental conditions38. The WOFOST model estimates phenology, LAI, aboveground biomass, and storage organ biomass (i.e., grain yield) at a daily time step39 (Fig. 2).Fig. 2Schematic overview of the major processes implemented in WOFOST. The Astronomical module calculates day length, some variables relating to solar elevation, and the fraction of diffuse radiation.Full size imageZonal parameterizationWe first divided the study area covered by AMS into seamless Thiessen polygon zones. Each Thiessen polygon contains only a single AMS. These zones represent the whole areas where any location is closer to its associated AMS point than any other AMS point. Then, we assigned parameters to the entire zone based on the AMS data, including crop calendar (date of emergence) and soil water retention parameters (soil moisture content at wilting point, field capacity, and saturation). Besides, we also optimized two main crop parameters for controlling phenological development stages, namely TSUM1 (accumulated temperature required from emergence to anthesis) and TSUM2 (accumulated temperature required from anthesis to maturity), by minimizing the cost function of the observational and simulated date corresponding to anthesis and maturity.Parameter calibration within a single zoneWe implemented the calibration of parameters within every single zone, as illustrated in Fig. 3. We calculated the average statistical yield of each county within every single zone from 2007 to 2015, then ranked the counties in descending order and divided them into three groups, namely high, medium, and low-level yield counties, by the 33% quantile and 67% quantile of the average statistical yield. The three counties corresponding to 17% quantile, 50% quantile, and 83% quantile would be used for subsequent calibration and represent the corresponding three yield level groups. We used the statistical yields (converted to dry matter mass based on the standard moisture content of 12.5%) of the three counties for multiple years and a harvest index for each province to convert the county-level yield to AGB for calibration. The harvest index of each province was mainly estimated from the AMS’s dynamic growth records on the biomass composition of the dominant winter wheat varieties of the province and a published literature40. Besides, we collected the maximum LAI observations on all agrometeorological stations in all years in the study area, according to its histogram. We found that the histogram follows a normal distribution with a mean of 6.5 and a standard deviation of 1.5. Finally, we calibrated three sets of parameters corresponding to three yield level groups in each single zone according to the three selected counties.Fig. 3Flow chart of parameter calibration within a single zone.Full size imageWe designed a three-step calibration strategy for a specific yield level group. Firstly, as winter wheat varieties did not change significantly according to information recorded by agrometeorological stations from 2007 to 2015, we assumed the crop parameters of winter wheat remain unchanged every three years to combine three years of observational data to calibrate the parameters of the WOFOST model better. We maximized a log-likelihood function based on the maximum LAI statistics and every three-year county-level yield and AGB data mentioned to optimize selected crop parameters (see Table S2 in the Supplement Materials).The log-likelihood function was constructed as follows:$$log;{{rm{L}}}_{{rm{LAI}}}=-frac{1}{2}left[dlogleft(2pi right)+logleft(left|{Sigma }_{{rm{LAI}}}right|right)+{rm{MD}}{left({{bf{x}}}_{{rm{LAI}}};{mu }_{{rm{LAI}}},{Sigma }_{{rm{LAI}}}right)}^{2}right]$$
(1)
$$log;{{rm{L}}}_{{rm{TWSO}}}=-frac{1}{2}left[dlog(2pi )+logleft(left|{{boldsymbol{Sigma }}}_{{rm{TWSO}}}right|right)+{rm{MD}}{left({{bf{x}}}_{{rm{TWSO}}};{{boldsymbol{mu }}}_{{rm{TWSO}}},{{boldsymbol{Sigma }}}_{{rm{TWSO}}}right)}^{2}right]$$
(2)
$$log;{{rm{L}}}_{{rm{AGB}}}=-frac{1}{2}left[dlog(2pi )+logleft(left|{{boldsymbol{Sigma }}}_{{rm{AGB}}}right|right)+{rm{MD}}{left({{bf{x}}}_{{rm{AGB}}};{{boldsymbol{mu }}}_{{rm{AGB}}},{{boldsymbol{Sigma }}}_{{rm{AGB}}}right)}^{2}right]$$
(3)
$$log;{rm{L}}=log;{L}_{{rm{LAI}}}+log;{L}_{{rm{TWSO}}}+log;{L}_{{rm{AGB}}}$$
(4)
Where log L is the natural logarithm of the likelihood function, d is the dimension, that is, the number of years of joint calibration, which is set to 3 in this study xLAI is the vector composed of the maximum value of the 3-year LAI simulated by WOFOST, μLAI and ΣLAI are the mean vector and error covariance matrix of maximum LAI based on observation statistics. The annual maximum LAI was assumed to be independent, and the mean and standard deviation for each year was set the same as the result of Fig. 3. Similarly, xTWSO and xAGB are the yield vector and AGB vector at maturity of 3 years simulated by WOFOST, and μTWSO, μAGB are their corresponding county-level statistic vector, ΣTWSO and ΣAGB are their corresponding error covariance matrix. In this study, we assumed that the annual yield or AGB was independent, and their corresponding standard deviation was 10% of their statistical value. |Σ| is the determinant of Σ. The expression ({rm{MD}}{({bf{x}};{boldsymbol{mu }},{boldsymbol{Sigma }})}^{2}={({bf{x}}-{boldsymbol{mu }})}^{{rm{T}}}{{boldsymbol{Sigma }}}^{-1}({bf{x}}-{boldsymbol{mu }})), where MD is the Mahalanobis distance between the point x and the mean vector μ.Secondly, we optimized the inter-annual irrigation. We optimized two parameters every year: the critical value of soil moisture (denoted as SMc) and the amount of irrigation (denoted as V). When the soil moisture simulated by WOFOST is lower than SMc, an irrigation event will be triggered, and the irrigation amount is V cm. In this study, we combined three-year data for calibration with six parameters for optimization. The optimization strategy is the same as the previous step by maximizing the log-likelihood function. Finally, we fixed the optimized irrigation parameters and repeated the first step to calibrate the selected crop parameters and obtain the final optimal parameters.Data assimilationConsidering that MODIS LAI is relatively low compared to the actual LAI of winter wheat41, we select a weak-constraint cost function based on the least square of normalized observational and simulated LAI as shown in Eq. (5), which is assimilating the trend information of MODIS LAI into the crop growth model.$$J={sum }_{{rm{t}}=1}^{{rm{n}}}{left(frac{{{rm{LAI}}}_{{rm{MODIS}}}^{{rm{t}}}-{{rm{LAI}}}_{{rm{MODIS}}}^{min}}{{{rm{LAI}}}_{{rm{MODIS}}}^{max}-{{rm{LAI}}}_{{rm{MODIS}}}^{min}}-frac{{{rm{LAI}}}_{{rm{WOFOS}}}^{{rm{t}}}-{{rm{LAI}}}_{{rm{WOFOS}}}^{min}}{{{rm{LAI}}}_{{rm{WOFOS}}}^{max}-{{rm{LAI}}}_{{rm{WOFOS}}}^{min}}right)}^{2}$$
(5)
Where ({{rm{LAI}}}_{{rm{MODIS}}}^{{rm{t}}}) and .. are MODIS LAI and WOFOST simulated LAI of time t. ({{rm{LAI}}}_{{rm{MODIS}}}^{max}) and ({{rm{LAI}}}_{{rm{WOFOS}}}^{max}) are maximum of MODIS LAI and WOFOST simulated LAI. ({{rm{LAI}}}_{{rm{MODIS}}}^{min}) and ({{rm{LAI}}}_{{rm{WOFOS}}}^{min}) are minimum of MODIS LAI and WOFOST simulated LAI. J is the value of the cost function.We reinitialize the day of emergence (IDEM), the life span of leaves growing at 35 °C (SPAN), and thermal time from emergence to anthesis (TSUM1) in the WOFOST model on each 1 km winter wheat pixel according to cost function between WOFOST LAI and MODIS LAI. Besides, we applied the Subplex algorithm from the NLOPT library (https://github.com/stevengj/nlopt) for parameter optimization. More

Virus propagation and enumerationEchovirus-11 (E11, Gregory strain, ATCC VR737) and Coxsackievirus-A9 (CVA9, environmental strain from sewage, kindly provided by the Finnish National Institute for Health and Welfare) stocks were produced by infecting sub-confluent monolayers of BGMK cells as described previously [7]. Viruses were released from infected cells by freezing and thawing the culture flasks three times. To eliminate cell debris, the suspensions were centrifuged at 3000 × g for 5 min. Each stock solution was stored at −20 °C until use. Infectious virus concentrations were enumerated by a most probable number (MPN) infectivity assay as described in the Supplementary Information. The assay limit of detection (LoD), defined as the concentration corresponding to one positive cytopathic effect in the lowest dilution of the MPN assay under the experimental conditions used, corresponding to 2 MPN/mL.Inactivation of enteroviruses by bacterial consortia from lake waterTo study the inactivation of CVA9 and E11 by a bacterial consortium from lake water, four surface water samples were collected from Lake Geneva (Ecublens, Switzerland) during the summer 2021. Each sampling event was conducted on warm and sunny days, to minimize biological variation. Immediately after sampling, large particles of the sample were removed by filtering 500 mL of water on a 8 μm nitrocellulose filter membrane (Merck Millipore, Cork, Ireland). The sample was then filtered through a 0.8 μm nitrocellulose filter membrane (Merck Millipore) to remove large microorganisms such as protists. The resulting water sample corresponds to the bacterial fraction used to study virus inactivation.For inactivation experiments, each virus was spiked into individual 1 mL aliquots of fractionated lake water to a final concentration of 106 MPN/mL, and samples were incubated for 48 h at 30 °C without shaking. Duplicate experiments were conducted for each virus and each lake water sample. Experiments to control for thermal inactivation were conducted using the same procedure but by replacing the fractionated lake water with sterile milliQ water. Viral infectivity at times 0 h and 48 h was determined by MPN as described above. Virus decay was calculated as log10 (C/C0), where C is the residual titer after 48 h of incubation, and C0 is the initial titer. The experimental LoD was approximately 5-log10.These same experiments were conducted for three new water samples in the presence of four protease inhibitors with the following final concentrations: E64—10 μM (E3132, Sigma–Aldrich, Saint-Louis, MO, USA), GM6001—4 μM (CC1010, Sigma–Aldrich), Chymostatin—100 μM (C7268, Sigma–Aldrich), and PMSF—100 μM (P7626, Sigma–Aldrich). Each inhibitor was added to 1 mL of fractionated lake water, vortexed for 30 seconds, and incubated at room temperature for 15 min, before adding the two viral strains under the same conditions as described above.Bacterial isolation, cultivation, and storageBacteria were isolated from two water samples from Lake Geneva’s Ecublens beach, taken in November 2019 (Fall, 89 isolates) and May 2020 (Spring, 47 isolates). Bacteria recovery was performed on R2A agar plate (BD Difco, Franklin Lakes, NJ, USA) as described previously [15]. Briefly, successive dilutions from 10−1 to 10−5 were carried out in sterile water for each sample. For each dilution, a volume of 1 mL was deposited on three separate R2A plates, before being incubated at 22, 30, and 37 °C. After 5 days of incubation, each colony was picked and enriched on a new R2A plate. To ensure purity, each isolate was successively plated five times on R2A plate and incubated at the same temperature as the initial isolation. Each purified isolate was cryopreserved in R2A / 20% glycerol at −80 °C. The isolates were named based on the water body (Lake (L)), isolation temperature, and the isolation order (L-T°C-number).Bacterial identificationThe identification of each isolate was performed by 16 S rRNA gene sequencing using the pair of primers 27 F (5’- AGA GTT TGA TCM TGG CTC AG- 3’, Microsynth AG, Balgach, Switzerland) / 786 R (5’- CTA CCA GGG TAT CTA ATC – 3’, Microsynth AG), following a methodology previously described [15]. The thermocycling conditions and the purification of PCR products are described in the Supplementary Information. The complete list of isolated bacteria and associated accession numbers is given in Supplementary Table 1.Phylogenetic inference and metadata visualizationThe consensus from 16 S rRNA gene sequences of the 136 isolates was aligned using the MUSCLE algorithm [16]. The phylogenetic analysis of 566 bp aligned sequences from the V2-V4 16 S rRNA gene regions (Positions: 152–717) was performed using Molecular Evolutionary Genetics Analysis X software [17]. Phylogeny was inferred by maximum likelihood, with 1000 bootstrap iterations to test the robustness of the nodes. The resulting tree was uploaded and formatted using iTOL [18].Virus incubation with bacterial isolatesFor the preparation of the bacteria before co-incubation, each one was first cultured on R2A agar for 48 h at their initial isolation temperature. Overnight suspensions of each bacterial isolate were grown in R2A broth at room temperature under constant agitation (180 rpm). For co-incubation experiments, 200 μL of each bacterial suspension were mixed with 100 μL of a 105 MPN/mL stock of E11 or CVA9. Then, each condition was supplemented with 600 μL of R2A broth. Incubation was carried out for 96 h at room temperature, without shaking. At the end of the co-incubation, each tube was centrifuged for 15 min at 9000 × g (4 °C) to eliminate bacteria, and the residual infectious viral titer was enumerated by MPN assay as described above [7]. Each co-incubation experiment was carried out in triplicate. Control experiments were performed under the same conditions but using sterile R2A. Virus decay was quantified as log10 (Cexp/Cctrl), where Cexp is the residual titer after a co-incubation for 96 h, and Cctrl is the titer after incubation of the virus in sterile R2A for 96 h. The experimental LoD was 3-log10.Protease activity measurement using casein and gelatin agar platesCasein agar was prepared as follows: 20 g of skim milk (BD Difco), supplemented with 1 g glucose were reconstituted with 200 mL of distilled water. Likewise, a 10% bacteriological agar solution was prepared in a final volume of 200 mL. Finally, a solution consisting of 0.8% NaCl, 0.02% KCl, 0.144% Na2HPO4, and 0.024% KH2PO4 was reconstituted in 600 mL of water. All solutions were autoclaved for 15 min at 110 °C. The solutions were mixed, and 25 mL were poured into each Petri dish. Gelatin agar was composed of 0.4% peptone, 0.1% yeast extract, 1.5% gelatin and 1.5% bacteriological agar. The mixture was autoclaved 15 min at 120 °C, and 25 mL of medium was poured into each Petri dish.For each isolate, an overnight suspension was performed in R2A broth at room temperature, before spotting 15 μL of each suspension at the center of both gelatin and casein agar plates. Each plate was incubated at 22, 30, or 37 °C for 72 h, depending on the initial isolation temperature of the bacteria. Casein-degrading activity (cas), which is exerted by many different protease classes, and gelatin-degrading activity (gel), which is mostly caused by MMPs, were revealed by a hydrolysis halo around the producing bacteria. Hydrolysis diameters were measured in millimeters (mm) to report the extent of the proteolytic effect of each strain on both substrates.Protease activity quantification in cell-free supernatantUsing the same bacterial suspensions as for bacterial/virus co-incubation, 200 μL of each suspension was inoculated into 600 μL of R2A broth and incubated without shaking for 96 h at room temperature. Each culture was centrifuged for 15 min at 9000 × g at 4 °C. The resulting cell-free supernatants (CFS) were stored at −20 °C until use. For each CFS, protease activity was measured using the Protease Activity Assay Kit (ab112152, Abcam, Cambridge, UK), which measures general protease activity (pgen) except MMPs, and the MMP Activity Assay Kit (ab112146, Abcam), which selectively measures MMP activity (mmp). Briefly, for the Protease Activity Assay kit, 50 μL of the substrate was added into each well of a dark-bottom plate containing 50 μL of each CFS. Standard trypsin provided by the kit was used as a positive control. For the MMP Activity Assay kit, 50 μL of each CFS was incubated with 50 μL of 2 mM APMA for 3 h at 37 °C, prior to the activity test. Collagenase I (C0130, Sigma–Aldrich) was used as a positive control. R2A broth was used as a negative control for each assay. Protease activity was measured at time 0 and after 60 min, using a Synergy MX fluorescence reader (BioTek). The excitation and emission wavelengths were set to 485 and 530 nm, respectively. The emitted fluorescence, generated by proteolytic cleavage of the substrate of each kit, was calculated as follows: ∆RFU = RFU (60 min) − RFU (0 min). Proteolytic activity was calculated in mmol/min/μL based on the emitted fluorescence measured for trypsin and collagenase I at known proteolytic activities.Data analysisStatistical analyses to compare inactivation data were performed by one-way t-test or one-way ANOVA with Dunnett’s post-hoc test in GraphPad Prism v.9. An alpha value of 0.05 was used as a threshold for statistical significance. For each dataset we confirmed that data were normally distributed.To analyze a potential correlation between protease activity and viral decay, the decay values for each virus strain was related to the four protease activity tests of this study using a scatterplot combined with a Kernel density estimation. The analyses were performed with R v.3.6.1 using the SmoothScatter function of the R Base package.A Left-Censored Tobit model (CTM) with mixed effects was chosen to investigate interactions between protease activity and the decay measured for each virus strain. Briefly, the CTM with mixed effect was chosen for three reasons: (1) The protocol used to measure viral decay had a limit of quantification of −3-log10, and 152 measurement points reached the detection limit, requiring the use of this value as the left-censored value of the model; (2) The two virus strains used in the study showed distinct responses after exposure to environmental bacteria, preventing the use of a multiple linear regression model; (3) Among biological replicates of co-incubation experiments, inactivation variability was observed, suggesting the concomitant action of random biological effects (e.g., production of other compounds than proteases by bacteria, or differences in protease production rate between replicates for each bacterial isolate). The resulting statistical model was then formulated as follows:$$log left( {frac{{C_{{{{{{mathrm{exp}}}}}}}}}{{C_{{{{{{mathrm{ctrl}}}}}}}}}} right) = ; beta _0 + beta _1;{rm I}_{{{{{{{{mathrm{virus}}}}}}}}_i = 2} + beta _2sqrt {left[ {pgen} right]_i} + beta _3sqrt {left[ {mmp} right]_i} + beta _4sqrt {left[ {cas} right]_i} \ + beta _5sqrt {left[ {gel} right]_i} + beta _6I_{{{{{{{{mathrm{virus}}}}}}}}_i = 2}sqrt {left[ {pgen} right]_i} + beta _7I_{{{{{{{{mathrm{virus}}}}}}}}_i = 2}sqrt {left[ {mmp} right]_i} \ + beta _8I_{{{{{{{{mathrm{virus}}}}}}}}_i = 2}sqrt {left[ {cas} right]_i} + beta _9I_{{{{{{{{mathrm{virus}}}}}}}}_i = 2}sqrt {left[ {gel} right]_i} + alpha _{{{{{{{{mathrm{id}}}}}}}}_i} + varepsilon _i$$$${{{mbox{where}}}}; log left( {frac{{C_{{{{{{mathrm{exp}}}}}}}}}{{C_{{{{{{mathrm{ctrl}}}}}}}}}} right) = left{ {begin{array}{*{20}{c}} { – 3} & {{{{{{{{mathrm{if}}}}}}}};{{{{{{{mathrm{log}}}}}}}}left( {frac{{C_{{{{{{mathrm{exp}}}}}}}}}{{C_{{{{{{mathrm{ctrl}}}}}}}}}} right) le – 3} \ {{{{{{{{mathrm{log}}}}}}}}left( {frac{{C_{{{{{{mathrm{exp}}}}}}}}}{{C_{{{{{{mathrm{ctrl}}}}}}}}}} right)} & {{{{{{{{mathrm{otherwise}}}}}}}}} end{array}} right.$$$$alpha _{{{{{{{{mathrm{id}}}}}}}}_i}sim {{{{{{{mathrm{i}}}}}}}}.{{{{{{{mathrm{i}}}}}}}}.;{{{{{{{mathrm{d}}}}}}}}.;{rm N}left( {0,;sigma _{{{{{{{{mathrm{id}}}}}}}}}^2} right)$$$${{{{{{{mathrm{for}}}}}}}};i in left{ {1,2, ldots } right}$$for which β0 defines the model intercept, (beta _1{rm I}_{{{{{{{{mathrm{virus}}}}}}}}_i = 2}) corresponds to the main effect of the virus factor on the viral decay, (beta _2,;beta _3,;beta _4,;{{{{{{{mathrm{and}}}}}}}};beta _5) corresponds to the main effects of the different protease activity measurements on viral decay, (beta _6I_{{{{{{{{mathrm{virus}}}}}}}}_i = 2},;beta _7I_{{{{{{{{mathrm{virus}}}}}}}}_i = 2},;beta _8I_{{{{{{{{mathrm{virus}}}}}}}}_i = 2},{{{{{{{mathrm{and}}}}}}}};beta _9I_{{{{{{{{mathrm{virus}}}}}}}}_i = 2}) corresponds to the interaction effects between each of these variables and the viral decay, (alpha _{{{{{{{{mathrm{id}}}}}}}}_i}) corresponds to the mixed effect of the model and (varepsilon _i) corresponds to the error term of the model. The selection of the model is further detailed in the Supplementary Information (Supplementary Material and Figs. S1 and S2).The full dataset included in the correlation analysis and the CTM is provided in Supplementary Table 2. A description of the variables used is given in the Supplementary Information. The dataset was analyzed using the censReg package in R [19]. The R code is given in the Supplementary Information. More

Foster, G. & Rahmstorf, S. Global temperature evolution 1979–2010. Environ. Res. Lett. 6, 044022 (2011).ADS 
Article 

Google Scholar 
Cahill, N., Rahmstorf, S. & Parnell, A. C. Change points of global temperature. Environ. Res. Lett. 10, 084002 (2015).ADS 
Article 

Google Scholar 
Yan, X. H. et al. The global warming hiatus: Slowdown or redistribution?. Earth’s Future 4, 472–482 (2016).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Karl, T. R. et al. Possible artifacts of data biases in the recent global surface warming hiatus. Science 348, 5632 (2015).Article 

Google Scholar 
Cohen, J. L., Furtado, J. C., Barlow, M., Alexeev, V. A. & Cherry, J. E. Asymmetric seasonal temperature trends. Geophys. Res. Lett. 39, 04705. https://doi.org/10.1029/2011GL050582 (2012).ADS 
Article 

Google Scholar 
Pepin, N. C. & Lundquist, J. D. Temperature trends at high elevations: patterns across the globe. Geophys. Res. Lett. 35, 14 (2008).Article 

Google Scholar 
Rangwala, I. & Miller, J. R. Climate change in mountains: a review of elevation-dependent warming and its possible causes. Clim. Change 114, 527–547 (2012).ADS 
Article 

Google Scholar 
Wang, Q., Fan, X. & Wang, M. Recent warming amplification over high elevation regions across the globe. Clim. Dyn. 43, 87–101 (2014).CAS 
Article 

Google Scholar 
Fan, X., Wang, Q., Wang, M. & Jiménez, C. V. Warming amplification of minimum and maximum temperatures over high-elevation regions across the globe. PLoS ONE 10, e0140213 (2015).PubMed 
PubMed Central 
Article 

Google Scholar 
Pepin, N. et al. Elevation-dependent warming in mountain regions of the world. Nat. Clim. Chang. 5, 424 (2015).ADS 
Article 

Google Scholar 
Piccarreta, M., Lazzari, M. & Pasini, A. Trends in daily temperature extremes over the Basilicata region (southern Italy) from 1951 to 2010 in a Mediterranean climatic context. Int. J. Climatol. 35, 1964–1975 (2015).Article 

Google Scholar 
Gonzalez-Hidalgo, J. C., Peña-Angulo, D., Brunetti, M. & Cortesi, N. Recent trend in temperature evolution in Spanish mainland (1951–2010): from warming to hiatus. Int. J. Climatol. 36, 2405–2416 (2016).Article 

Google Scholar 
McCullough, I. M. et al. High and dry: high elevations disproportionately exposed to regional climate change in Mediterranean-climate landscapes. Landsc. Ecol. 31, 1063–1075 (2016).Article 

Google Scholar 
Sanz-Elorza, M., Dana, E. D., González, A. & Sobrino, E. Changes in the high-mountain vegetation of the central Iberian Peninsula as a probable sign of global warming. Ann. Bot. 92, 273–280 (2003).PubMed 
PubMed Central 
Article 

Google Scholar 
Peñuelas, J. & Boada, M. A global change induced biome shift in the Montseny mountains (NE Spain). Glob. Change Biol. 9, 131–140 (2003).ADS 
Article 

Google Scholar 
Linares, J. C. & Tíscar, P. A. Climate change impacts and vulnerability of the southern populations of Pinus nigra subsp. salzmannii. Tree Physiol. 30, 795–806 (2010).PubMed 
Article 

Google Scholar 
Giorgi, F., Hurrell, J. W., Marinucci, M. R. & Beniston, M. Elevation dependency of the surface climate change signal: a model study. J. Clim. 10, 288–296 (1997).ADS 
Article 

Google Scholar 
Palazzi, E., Mortarini, L., Terzago, S. & Von Hardenberg, J. Elevation-dependent warming in global climate model simulations at high spatial resolution. Clim. Dyn. 52, 2685–2702 (2019).Article 

Google Scholar 
Poyatos, R., Latron, J. & Llorens, P. Land use and land cover change after agricultural abandonment. Mt. Res. Dev. 23, 362–368 (2003).Article 

Google Scholar 
Mouillot, F., Ratte, J. P., Joffre, R., Mouillot, D. & Rambal, S. Long-term forest dynamic after land abandonment in a fire prone Mediterranean landscape (central Corsica, France). Landsc. Ecol. 20, 101–112 (2005).Article 

Google Scholar 
Zellweger, F. et al. Forest microclimate dynamics drive plant responses to warming. Science 368, 772–775 (2020).ADS 
CAS 
PubMed 
Article 

Google Scholar 
Ríos-Cornejo, D., Penas, Á., Álvarez-Esteban, R. & Del Río, S. Links between teleconnection patterns and mean temperature in Spain. Theor. Appl. Climatol. 122, 1–18 (2015).ADS 
Article 

Google Scholar 
Nogués-Bravo, D., Araújo, M. B., Errea, M. P. & Martinez-Rica, J. P. Exposure of global mountain systems to climate warming during the 21st Century. Glob. Environ. Chang. 17, 420–428 (2007).Article 

Google Scholar 
Vicente-Serrano, S. M., Beguería, S., López-Moreno, J. I., El Kenawy, A. M. & Angulo-Martínez, M. Daily atmospheric circulation events and extreme precipitation risk in northeast Spain: Role of the North Atlantic Oscillation, the Western Mediterranean Oscillation, and the Mediterranean Oscillation. J. Geophys. Res. Atmos. 114, D8 (2009).Article 

Google Scholar 
Guzman-Morales, J. & Gershunov, A. Climate change suppresses Santa Ana winds of southern California and sharpens their seasonality. Geophys. Res. Lett. 46, 2772–2780. https://doi.org/10.1029/2018GL080261 (2019).ADS 
Article 

Google Scholar 
Yu, M. & Ruggieri, E. Change point analysis of global temperature records. Int. J. Climatol. 39, 3679–3688 (2019).Article 

Google Scholar 
Giorgi, F. Climate change hot-spots. Geophys. Res. Lett. 33, 08707. https://doi.org/10.1029/2006GL025734 (2006).ADS 
Article 

Google Scholar 
García, M. J. L. Recent warming in the Balearic Sea and Spanish Mediterranean coast: Towards an earlier and longer summer. Atmósfera 28, 149–160 (2015).Article 

Google Scholar 
Toreti, A., Desiato, F., Fioravanti, G. & Perconti, W. Seasonal temperatures over Italy and their relationship with low-frequency atmospheric circulation patterns. Clim. Change 99, 211–227 (2010).ADS 
Article 

Google Scholar 
Scorzini, A. R. & Leopardi, M. Precipitation and temperature trends over central Italy (Abruzzo Region): 1951–2012. Theor. Appl. Climatol. 135, 959–977 (2019).ADS 
Article 

Google Scholar 
Lee, X. et al. Observed increase in local cooling effect of deforestation at higher latitudes. Nature 479, 384–387 (2011).ADS 
CAS 
PubMed 
Article 

Google Scholar 
Juang, J.-Y., Katul, G., Siqueira, M., Stoy, P. & Novick, K. Separating the effects of albedo from eco-physiological changes on surface temperature along a successional chronosequence in the southeastern United States. Geophys. Res. Lett. 34, 21408. https://doi.org/10.1029/2007.GL03129 (2007).ADS 
Article 

Google Scholar 
Boulant, N., Kunstler, G., Rambal, S. & Lepart, J. Seed supply, drought, and grazing determine spatio-temporal patterns of recruitment for native and introduced invasive pines in grasslands. Divers. Distrib. 14, 862–874 (2008).Article 

Google Scholar 
Améztegui, A. Land-use changes as major drivers of mountain pine (Pinus uncinata Ram.) expansion in the Pyrenees. Glob. Ecol. Biogeogr. 19, 632–641 (2010).
Google Scholar 
Rambal, S. Relations entre couverts végétaux des parcours et cycle de l’eau. In L’eau des troupeaux en alpages et sur parcours: une ressource à gérer, aménager, partager (ed. Lepart, J.) 25–37 (Association Française de Pastoralisme et Cardère éditeur, 2015).
Google Scholar 
Fonderflick, J., Lepart, J., Caplat, P., Debussche, M. & Marty, P. Managing agricultural change for biodiversity conservation in a Mediterranean upland. Biol. Conserv. 143, 737–746 (2010).Article 

Google Scholar 
Abadie, J. et al. Forest recovery since 1860 in a Mediterranean region: Drivers and implications for land use and land cover spatial distribution. Landsc. Ecol. 33, 289–305 (2018).Article 

Google Scholar 
Cervera, T., Pino, J., Marull, J., Padró, R. & Tello, E. Understanding the long-term dynamics of forest transition: From deforestation to afforestation in a Mediterranean landscape (Catalonia, 1868–2005). Land Use Policy 80, 318–331 (2019).Article 

Google Scholar 
Wolpert, F., Quintas-Soriano, C. & Plieninger, T. Exploring land-use histories of tree-crop landscapes: a cross-site comparison in the Mediterranean Basin. Sustain. Sci. 15, 1267–1283 (2020).Article 

Google Scholar 
Lasanta-Martínez, T., Vicente-Serrano, S. M. & Cuadrat-Prats, J. M. Mountain Mediterranean landscape evolution caused by the abandonment of traditional primary activities: A study of the Spanish Central Pyrenees. Appl. Geogr. 25, 47–65 (2005).Article 

Google Scholar 
Malandra, F., Vitali, A., Urbinati, C., Weisberg, P. J. & Garbarino, M. Patterns and drivers of forest landscape change in the Apennines range, Italy. Reg. Environ. Change 19, 1973–1985 (2019).Article 

Google Scholar 
Zhang, Q. et al. Reforestation and surface cooling in temperate zones: Mechanisms and implications. Glob. Change Biol. 26, 3384–3401 (2020).ADS 
Article 

Google Scholar 
Rambal, S., Lacaze, B. & Winkel, T. Testing an area-weighted model for albedo or surface temperature of mixed pixels in Mediterranean woodlands. Int. J. Remote Sens. 11, 1495–1499 (1990).Article 

Google Scholar 
Luyssaert, S. et al. Land management and land-cover change have impacts of similar magnitude on surface temperature. Nat. Clim. Change 4, 389–393. https://doi.org/10.1038/nclimate2196 (2014).ADS 
Article 

Google Scholar 
Novick, K. A. & Katul, G. G. The duality of reforestation impacts on surface and air temperature. J. Geophys. Res. Biogeosci. 125, e05543 (2020).Article 

Google Scholar 
Davy, R. & Esau, I. Differences in the efficacy of climate forcings explained by variations in atmospheric boundary layer depth. Nat. Commun. 7, 1–8 (2016).Article 

Google Scholar 
Serafin, S. et al. Exchange processes in the atmospheric boundary layer over mountainous terrain. Atmosphere 9, 102. https://doi.org/10.3390/atmos9030102 (2018).ADS 
CAS 
Article 

Google Scholar 
Perugini, L. et al. Biophysical effects on temperature and precipitation due to land cover change. Environ. Res. Lett. 12, 053002 (2017).ADS 
Article 

Google Scholar 
Visbeck, M. H., Hurrell, J. W., Polvani, L. & Cullen, H. M. The North Atlantic oscillation: Past, present, and future. Proc. Natl. Acad. Sci. 98, 12876–12877 (2001).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Hurrell, J. W. Decadal trends in the North Atlantic oscillation: Regional temperatures and precipitation. Science 269, 676–679 (1995).ADS 
CAS 
PubMed 
Article 

Google Scholar 
Martín, P., Sabatés, A., Lloret, J. & Martin-Vide, J. Climate modulation of fish populations: the role of the Western Mediterranean Oscillation (WeMO) in sardine (Sardina pilchardus) and anchovy (Engraulis encrasicolus) production in the north-western Mediterranean. Clim. Change 110, 925–939 (2012).ADS 
Article 

Google Scholar 
Schwingshackl, C., Hirschi, M. & Seneviratne, S. I. Global contributions of incoming radiation and land surface conditions to maximum near surface air temperature variability and trend. Geophys. Res. Lett. 45, 5034–5044 (2018).ADS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Philipona, R., Behrens, K. & Ruckstuhl, C. How declining aerosols and rising greenhouse gases forced rapid warming in Europe since the 1980s. Geophys. Res. Lett. 36, L02806. https://doi.org/10.1029/2008GL036350 (2009).ADS 
Article 

Google Scholar 
Schwarz, M., Folini, D., Yang, S., Allan, R. P. & Wild, M. Changes in atmospheric shortwave absorption as important driver of dimming and brightening. Nat. Geosci. 13, 110–115 (2020).ADS 
CAS 
Article 

Google Scholar 
Norris, J. R. & Wild, M. Trends in aerosol radiative effects over Europe inferred from observed cloud cover, solar “dimming”, and solar “brightening”. J. Geophys. Res. Atmos. 112, D08214. https://doi.org/10.1029/2006JD007794 (2007).ADS 
Article 

Google Scholar 
Mateos, D. et al. Quantifying the respective roles of aerosols and clouds in the strong brightening since the early 2000s over the Iberian Peninsula. J. Geophys. Res. Atmos. 119, 10–382 (2014).Article 

Google Scholar 
Sanchez-Lorenzo, A. et al. Reassessment and update of long-term trends in downward surface shortwave radiation over Europe (1939–2012). J. Geophys. Res. Atmos. 120, 9555–9569 (2015).ADS 
Article 

Google Scholar 
Kambezidis, H. D., Kaskaoutis, D. G., Kalliampakos, G. K., Rashki, A. & Wild, M. The solar dimming/brightening effect over the Mediterranean Basin in the period 1979–2012. J. Atmos. Solar Terr. Phys. 150, 31–46 (2016).ADS 
Article 

Google Scholar 
Chiacchio, M. & Wild, M. Influence of NAO and clouds on long-term seasonal variations of surface solar radiation in Europe. J. Geophys. Res. Atmos. 115, 0022. https://doi.org/10.1029/2009JD012182 (2010).Article 

Google Scholar 
Wild, M. Decadal changes in radiative fluxes at land and ocean surfaces and their relevance for global warming. Wiley Interdiscipl. Rev. Clim. Change 7, 91–107 (2016).Article 

Google Scholar 
Held, I. M. & Soden, B. J. Water vapor feedback and global warming. Annu. Rev. Energy Environ. 25, 441–475 (2000).Article 

Google Scholar 
Dessler, A. E. & Sherwood, S. C. A matter of humidity. Science 323, 1020–1021 (2009).CAS 
PubMed 
Article 

Google Scholar 
Ruckstuhl, C., Philipona, R., Morland, J. & Ohmura, A. Observed relationship between surface specific humidity, integrated water vapor, and longwave downward radiation at different altitudes. J. Geophys. Res. Atmos. 112(D03302), 2007. https://doi.org/10.1029/2006JD007850 (2007).Article 

Google Scholar 
Parras-Berrocal, I. M. et al. The climate change signal in the Mediterranean Sea in a regionally coupled atmosphere–ocean model. Ocean Sci. 16, 743–765. https://doi.org/10.5194/os-16-743-2020 (2020).ADS 
Article 

Google Scholar 
Reale, M. et al. The regional earth system model RegCM-ES: Evaluation of the Mediterranean climate and marine biogeochemistry. J. Adv. Model. Earth Syst. 12, e001812 (2020).Article 

Google Scholar 
Sen, P. K. Estimates of the regression coefficient based on Kendall’s tau. J. Am. Stat. Assoc. 63, 1379–1389 (1968).MathSciNet 
MATH 
Article 

Google Scholar 
Kelliher, F. M., Leuning, R. & Schulze, E. D. Evaporation and canopy characteristics of coniferous forests and grasslands. Oecologia 95, 153–163 (1993).ADS 
CAS 
PubMed 
Article 

Google Scholar 
Linacre, E. T. Simpler empirical expression for actual evapotranspiration rates-a discussion. Agric. Meteorol. 11, 451–452 (1973).Article 

Google Scholar 
Jones, P. D., Jónsson, T. & Wheeler, D. Extension to the North Atlantic Oscillation using early instrumental pressure observations from Gibraltar and south-west Iceland. Int. J. Climatol. 17, 1433–1450 (1997).Article 

Google Scholar 
Palutikof, J. P. Analysis of Mediterranean climate data: measured and modelled. In Mediterranean Climate: Variability and Trends (ed. Bolle, H. J.) (Springer, 2003).
Google Scholar 
Martin-Vide, J. & Lopez-Bustins, J. A. The western Mediterranean oscillation and rainfall in the Iberian Peninsula. Int. J. Climatol. 26, 1455–1475 (2006).Article 

Google Scholar  More

1500-year stable carbon and oxygen isotopes in larch tree-ring celluloseThe δ13Ccell (Fig. 1a, Fig. S2) and δ18Ocell (Fig. 1b, Fig. S3) records span 516–2016 CE, at annual resolution. The δ13Ccell timeseries shows mostly increasing trends during the first millennium of the Common Era (516–1120 CE), and similarly at the end of the last millennium (1720–2016 CE). The maximum δ13Ccell value occurs in 2016 CE (−19.6‰; + 3.2σ), while the minimum occurs in 686 CE (−24.7‰, −3.6σ) relative to the average for the period 516–2016 CE (−22.04‰) (Table S2, Fig. S2). The standard error (SE) for the whole analysed period is 0.02.Figure 1Annually resolved δ13Ccell (a) and δ18O cell (b) in Siberian larch tree-ring cellulose chronologies for the period from 516 to 2016 CE. Chronologies are smoothed by a 101-year Hamming window to highlight a centennial scale. The dotted and dashed lines indicate the number of trees analysed.Full size imageThe δ18Ocell timeseries (Fig. 1b, Fig. S3) showed two positive and one negative extreme over the past 1500 years, with the minimum value (19.9‰; −6.3σ), occurring in 536 CE, and maximum values (31.9‰; + 3.8σ and 32.2‰; + 4.4σ), occurring in 1266 and 2008 CE, respectively (Table S2, Fig. S3). The SE for the whole analysed period is 0.03. The δ18Ocell data has higher standard deviation (SD) (1.15) than δ13Ccell (0.75).Less than 1% of values in the δ18Ocell record are classified as extreme, with the standard deviation ≥  ± 3σ. The δ13Ccell and δ18Ocell records are significantly correlated (r = 0.1, p = 0.0001, n = 1500).Local climate signals preserved in δ13Ccell and δ18Ocell recordsWe used weather observations from the local Mugur-Aksy weather station (50°N, 90°E, 1850 m asl) (Table S1) to derive quantitative paleoclimatic reconstructions from our δ13Ccell and δ18Ocell timeseries. A multiple linear regression analysis revealed significant correlations between δ13Ccell and July precipitation (r = −0.58; p  More

Smale, D. A. et al. Marine heatwaves threaten global biodiversity and the provision of ecosystem services. Nat. Clim. Change https://doi.org/10.1038/s41558-019-0412-1 (2019).Article 

Google Scholar 
Smith, K. E. et al. Socioeconomic impacts of marine heatwaves: Global issues and opportunities. Science 374, eabj3593 (2021).CAS 
Article 

Google Scholar 
Frolicher, T. L., Fischer, E. M. & Gruber, N. Marine heatwaves under global warming. Nature 560, 360–364. https://doi.org/10.1038/s41586-018-0383-9 (2018).ADS 
CAS 
Article 
PubMed 

Google Scholar 
Oliver, et al. Longer and more frequent marine heatwaves over the past century. Nat. Commun. https://doi.org/10.1038/s41467-018-03732-9 (2018).Article 
PubMed 
PubMed Central 

Google Scholar 
Oliver, et al. Projected marine heatwaves in the 21st century and the potential for ecological impact. Front. Mar. Sci. https://doi.org/10.3389/fmars.2019.00734 (2019).Article 

Google Scholar 
Hobday, A. J. et al. A hierarchical approach to defining marine heatwaves. Prog. Oceanogr. 141, 227–238 (2016).ADS 
Article 

Google Scholar 
Banzon, V., Smith, T. M., Chin, T. M., Liu, C. & Hankins, W. A long-term record of blended satellite and in situ sea-surface temperature for climate monitoring, modeling and environmental studies. Earth Syst. Sci. Data 8, 165–176 (2016).ADS 
Article 

Google Scholar 
Wernberg, T. et al. An extreme climatic event alters marine ecosystem structure in a global biodiversity hotspot. Nat. Clim. Change 3, 78–82. https://doi.org/10.1038/nclimate1627 (2013).ADS 
Article 

Google Scholar 
Arias-Ortiz, A. et al. A marine heatwave drives massive losses from the world’s largest seagrass carbon stocks. Nat. Clim. Change https://doi.org/10.1038/s41558-018-0096-y (2018).Article 

Google Scholar 
Smale, D. A. Impacts of ocean warming on kelp forest ecosystems. New Phytol. 225, 1447–1454 (2020).Article 

Google Scholar 
Couch, C. S. et al. Mass coral bleaching due to unprecedented marine heatwave in Papahānaumokuākea Marine National Monument (Northwestern Hawaiian Islands). PLoS ONE 12, e0185121. https://doi.org/10.1371/journal.pone.0185121 (2017).CAS 
Article 
PubMed 
PubMed Central 

Google Scholar 
Oliver, E. C. J. et al. The unprecedented 2015/16 Tasman Sea marine heatwave. Nat. Commun. https://doi.org/10.1038/ncomms16101 (2017).Article 
PubMed 
PubMed Central 

Google Scholar 
Montie, S., Thomsen, M. S., Rack, W. & Broady, P. A. Extreme summer marine heatwaves increase chlorophyll a in the Southern Ocean. Antarct. Sci. 32, 508–509 (2020).ADS 
Article 

Google Scholar 
Gupta, A. S. et al. Drivers and impacts of the most extreme marine heatwaves events. Sci. Rep. 10, 1–15 (2020).ADS 
Article 

Google Scholar 
Holbrook, N. J. et al. A global assessment of marine heatwaves and their drivers. Nat. Commun. 10, 1–13 (2019).CAS 
Article 

Google Scholar 
La Sorte, F. A., Johnston, A. & Ault, T. R. Global trends in the frequency and duration of temperature extremes. Clim. Change 166, 1. https://doi.org/10.1007/s10584-021-03094-0 (2021).ADS 
Article 

Google Scholar 
Thomsen, et al. Local extinction of bull kelp (Durvillaea spp.) due to a marine heatwave. Front. Mar. Sci. https://doi.org/10.3389/fmars.2019.00084 (2019).Article 

Google Scholar 
Strydom, S. et al. Too hot to handle: Unprecedented seagrass death driven by marine heatwave in a World Heritage Area. Glob. Change Biol. 26, 3525–3538. https://doi.org/10.1111/gcb.15065 (2020).ADS 
Article 

Google Scholar 
Leggat, W. P. et al. Rapid coral decay is associated with marine heatwave mortality events on reefs. Curr. Biol. 29, 2723-2730.e2724. https://doi.org/10.1016/j.cub.2019.06.077 (2019).CAS 
Article 
PubMed 

Google Scholar 
Wernberg, T. et al. Climate-driven regime shift of a temperate marine ecosystem. Science 353, 169–172. https://doi.org/10.1126/science.aad8745 (2016).ADS 
CAS 
Article 
PubMed 

Google Scholar 
Thomsen, M. S. & McGlathery, K. Facilitation of macroalgae by the sedimentary tube forming polychaete Diopatra cuprea. Estuar. Coast. Shelf Sci. 62, 63–73. https://doi.org/10.1016/j.ecss.2004.08.007 (2005).ADS 
Article 

Google Scholar 
Spalding, M. D. et al. Marine ecoregions of the world: A bioregionalization of coastal and shelf areas. Bioscience 57, 573–583 (2007).Article 

Google Scholar 
Costello, M. J. & Chaudhary, C. Marine biodiversity, biogeography, deep-sea gradients, and conservation. Curr. Biol. 27, R511–R527. https://doi.org/10.1016/j.cub.2017.04.060 (2017).CAS 
Article 
PubMed 

Google Scholar 
Tait, L. W., Thoral, F., Pinkerton, M. H., Thomsen, M. S. & Schiel, D. R. Loss of giant kelp, Macrocystis pyrifera, driven by marine heatwaves and exacerbated by poor water clarity in New Zealand. Front. Mar. Sci. https://doi.org/10.3389/fmars.2021.721087 (2021).Article 

Google Scholar 
Marin, M., Feng, M., Phillips, H. E. & Bindoff, N. L. A global, multiproduct analysis of coastal marine heatwaves: Distribution, characteristics, and long-term trends. J. Geophys. Res. Oceans 126, e2020JC016708. https://doi.org/10.1029/2020JC016708 (2021).ADS 
Article 

Google Scholar 
Kain, J. M. The seasons in the subtidal. Brit. Phycol. J. 24, 203–215 (1989).Article 

Google Scholar 
Atkinson, J., King, N. G., Wilmes, S. B. & Moore, P. J. Summer and winter marine heatwaves favor an invasive over native seaweeds. J. Phycol. 56, 1591–1600. https://doi.org/10.1111/jpy.13051 (2020).CAS 
Article 
PubMed 

Google Scholar 
Salinger, M. J. et al. The unprecedented coupled ocean-atmosphere summer heatwave in the New Zealand region 2017/18: Drivers, mechanisms and impacts. Environ. Res. Lett. 14, 044023 (2019).ADS 
Article 

Google Scholar 
Amaya, D. J., Miller, A. J., Xie, S.-P. & Kosaka, Y. Physical drivers of the summer 2019 North Pacific marine heatwave. Nat. Commun. 11, 1–9 (2020).Article 

Google Scholar 
Di Lorenzo, E. & Mantua, N. Multi-year persistence of the 2014/15 North Pacific marine heatwave. Nat. Clim. Change 6, 1042–1047. https://doi.org/10.1038/nclimate3082 (2016).ADS 
Article 

Google Scholar 
Cayan, D. R. Large-scale relationships between sea surface temperature and surface air temperature. Mon. Weather Rev. 108, 1293–1301 (1980).ADS 
Article 

Google Scholar 
Hipel, K. W. & McLeod, A. I. Time Series Modelling of Water Resources and Environmental Systems (Elsevier, 1994).
Google Scholar 
trend: non-parametric trend tests and changepoint detection.–R package ver. 1.1. 2 (2020).Costanza, R. et al. The value of the world’s ecosystem services and natural capital. Nature 387, 253–260 (1997).ADS 
CAS 
Article 

Google Scholar 
Halpern, B. S. et al. A global map of human impact on marine ecosystems. Science 319, 948–952 (2008).ADS 
CAS 
Article 

Google Scholar 
Harley, C. D. et al. The impacts of climate change in coastal marine systems. Ecol. Lett. 9, 228–241. https://doi.org/10.1111/j.1461-0248.2005.00871.x (2006).ADS 
Article 
PubMed 

Google Scholar 
Thomsen, M. S. & South, P. M. Communities and attachment networks associated with primary, secondary and alternative foundation species; a case study of stressed and disturbed stands of southern bull kelp. Diversity 11, 56. https://doi.org/10.3390/d11040056 (2019).Article 

Google Scholar 
Smale, D. A. & Wernberg, T. Extreme climatic event drives range contraction of a habitat-forming species. Proc. R. Soc. B Biol. Sci. 280, 20122829 (2013).Article 

Google Scholar 
Thomsen, M. S. et al. Cascading impacts of earthquakes and extreme heatwaves have destroyed populations of an iconic marine foundation species. Divers. Distrib. (2021).Rogers-Bennett, L. & Catton, C. Marine heat wave and multiple stressors tip bull kelp forest to sea urchin barrens. Sci. Rep. 9, 1–9 (2019).CAS 
Article 

Google Scholar 
Filbee-Dexter, K. et al. Marine heatwaves and the collapse of marginal North Atlantic kelp forests. Sci. Rep. 10, 1–11 (2020).Article 

Google Scholar 
Thomson, J. A. et al. Extreme temperatures, foundation species, and abrupt ecosystem change: An example from an iconic seagrass ecosystem. Glob. Change Biol. 21, 1463–1474. https://doi.org/10.1111/gcb.12694 (2015).ADS 
Article 

Google Scholar 
Hughes, T. P. et al. Global warming and recurrent mass bleaching of corals. Nature 543, 373–377. https://doi.org/10.1038/nature21707 (2017).ADS 
CAS 
Article 
PubMed 

Google Scholar 
Le Nohaïc, M. et al. Marine heatwave causes unprecedented regional mass bleaching of thermally resistant corals in northwestern Australia. Sci. Rep. 7, 14999. https://doi.org/10.1038/s41598-017-14794-y (2017).ADS 
CAS 
Article 
PubMed 
PubMed Central 

Google Scholar 
Smale, D. A., Wernberg, T. & Vanderklift, M. A. Regional-scale variability in the response of benthic macroinvertebrate assemblages to a marine heatwave. Mar. Ecol. Prog. Ser. 568, 17–30. https://doi.org/10.3354/meps12080 (2017).ADS 
Article 

Google Scholar 
Cavole, L. et al. Biological impacts of the 2013–2015 warm-water anomaly in the Northeast Pacific: Winners, losers, and the future. Oceanography (Washington D.C.) https://doi.org/10.5670/oceanog.2016.32 (2016).Article 

Google Scholar 
Coleman, M. A., Minne, A. J. P., Vranken, S. & Wernberg, T. Genetic tropicalisation following a marine heatwave. Sci. Rep. 10, 12726. https://doi.org/10.1038/s41598-020-69665-w (2020).ADS 
CAS 
Article 
PubMed 
PubMed Central 

Google Scholar 
Collette, B. B. in Reproduction and sexuality in marine fishes 21–64 (University of California Press, 2010).Yatsu, A. & Shimada, H. Distributions of Epipelagic Fishes, Squids, Marine Mammals. Bulletin 53 North Pacific Commission, 111–146.Hirst, A., Roff, J. & Lampitt, R. A synthesis of growth rates in marine epipelagic invertebrate zooplankton. Adv. Mar. Biol. 44, 1–142 (2003).CAS 
Article 

Google Scholar 
Smale, D. A. & Wernberg, T. Satellite-derived SST data as a proxy for water temperature in nearshore benthic ecology. Mar. Ecol. Prog. Ser. 387, 27–37 (2009).ADS 
Article 

Google Scholar 
Bernardello, R., Serrano, E., Coma, R., Ribes, M. & Bahamon, N. A comparison of remote-sensing SST and in situ seawater temperature in near-shore habitats in the western Mediterranean Sea. Mar. Ecol. Prog. Ser. 559, 21–34 (2016).ADS 
Article 

Google Scholar 
Brewin, R. J. et al. Evaluating operational AVHRR sea surface temperature data at the coastline using benthic temperature loggers. Remote Sens. 10, 925 (2018).ADS 
Article 

Google Scholar 
Smit, A. J. et al. A coastal seawater temperature dataset for biogeographical studies: Large biases between in situ and remotely-sensed data sets around the Coast of South Africa. PLoS ONE 8, e81944. https://doi.org/10.1371/journal.pone.0081944 (2013).ADS 
CAS 
Article 
PubMed 
PubMed Central 

Google Scholar 
Marin, M., Bindoff, N. L., Feng, M. & Phillips, H. E. Slower long-term coastal warming drives dampened trends in coastal marine heatwave exposure. J. Geophys. Res. Oceans https://doi.org/10.1029/2021jc017930 (2021).Article 

Google Scholar 
Lourenço, C. R. et al. Upwelling areas as climate change refugia for the distribution and genetic diversity of a marine macroalga. J. Biogeogr. 43, 1595–1607. https://doi.org/10.1111/jbi.12744 (2016).Article 

Google Scholar 
Riegl, B. & Piller, W. E. Possible refugia for reefs in times of environmental stress. Int. J. Earth Sci. 92, 520–531. https://doi.org/10.1007/s00531-003-0328-9 (2003).Article 

Google Scholar 
El Glynn, P. W. Nino-southern oscillation 1982–1983: Nearshore population, community, and ecosystem responses. Annu. Rev. Ecol. Syst. 19, 309–346. https://doi.org/10.1146/annurev.es.19.110188.001521 (1988).Article 

Google Scholar 
Glynn, P. W. & D’Croz, L. Experimental evidence for high temperature stress as the cause of El Niño-coincident coral mortality. Coral Reefs 8, 181–191. https://doi.org/10.1007/bf00265009 (1990).ADS 
Article 

Google Scholar 
Glynn, P. W., Maté, J. L., Baker, A. C. & Calderón, M. O. Coral bleaching and mortality in panama and ecuador during the 1997–1998 El Niño-Southern Oscillation Event: Spatial/temporal patterns and comparisons with the 1982–1983 event. Bull. Mar. Sci. 69, 79–109 (2001).
Google Scholar 
Podestá, G. P. & Glynn, P. W. The 1997–98 El Niño event in Panama and Galápagos: An update of thermal stress indices relative to coral bleaching. Bull. Mar. Sci. 69, 43–59 (2001).
Google Scholar 
Ladah, L. B. & Zertuche-Gonzalez, J. A. Giant kelp (Macrocystis pyrifera) survival in deep water (25–40 m) during El Nino of 1997–1998 in Baja California, Mexico. Bot. Marina 47, 367–372. https://doi.org/10.1515/bot.2004.054 (2004).Article 

Google Scholar 
Kayanne, H. Validation of degree heating weeks as a coral bleaching index in the northwestern Pacific. Coral Reefs 36, 63–70 (2017).ADS 
Article 

Google Scholar 
Le Nohaïc, M. et al. Marine heatwave causes unprecedented regional mass bleaching of thermally resistant corals in northwestern Australia. Sci. Rep. 7, 1–11 (2017).ADS 
Article 

Google Scholar 
Marba, N. & Duarte, C. M. Mediterranean warming triggers seagrass (Posidonia oceanica) shoot mortality. Glob. Change Biol. 16, 2366–2375. https://doi.org/10.1111/j.1365-2486.2009.02130.x (2010).ADS 
Article 

Google Scholar 
Bennett, S., Wernberg, T., Arackal Joy, B., de Bettignies, T. & Campbell, A. H. Central and rear-edge populations can be equally vulnerable to warming. Nat. Commun. 6, 10280. https://doi.org/10.1038/ncomms10280 (2015).ADS 
CAS 
Article 
PubMed 

Google Scholar 
Filbee-Dexter, K. et al. Marine heatwaves and the collapse of marginal North Atlantic kelp forests. Sci. Rep. 10, 13388. https://doi.org/10.1038/s41598-020-70273-x (2020).ADS 
CAS 
Article 
PubMed 
PubMed Central 

Google Scholar 
Snover, M. L. Ontogenetic habitat shifts in marine organisms: Influencing factors and the impact of climate variability. Bull. Mar. Sci. 83, 53–67 (2008).
Google Scholar 
Harley, C. D. Climate change, keystone predation, and biodiversity loss. Science 334, 1124–1127 (2011).ADS 
CAS 
Article 

Google Scholar 
Kelaher, B. P., Coleman, M. A. & Bishop, M. J. Ocean warming, but not acidification, accelerates seagrass decomposition under near-future climate scenarios. Mar. Ecol. Prog. Ser. 605, 103–110 (2018).ADS 
CAS 
Article 

Google Scholar 
De Senerpont Domis, L. N. et al. Plankton dynamics under different climatic conditions in space and time. Freshw. Biol. 58, 463–482 (2013).Article 

Google Scholar 
Fossheim, M. et al. Recent warming leads to a rapid borealization of fish communities in the Arctic. Nat. Clim. Change 5, 673–677. https://doi.org/10.1038/nclimate2647 (2015).ADS 
Article 

Google Scholar 
Morales-Nin, B. & Panfili, J. Seasonality in the deep sea and tropics revisited: What can otoliths tell us?. Mar. Freshw. Res. 56, 585–598 (2005).Article 

Google Scholar 
Alongi, D. M. Ecology of tropical soft-bottom benthos: A review with emphasis on emerging concepts. Rev. Biol. Trop. 37, 85–100 (1989).
Google Scholar 
Hobday, A. J., Spillman, C. M., Paige Eveson, J. & Hartog, J. R. Seasonal forecasting for decision support in marine fisheries and aquaculture. Fish. Oceanogr. 25, 45–56. https://doi.org/10.1111/fog.12083 (2016).Article 

Google Scholar 
Spillman, C. M., Smith, G. A., Hobday, A. J. & Hartog, J. R. Onset and decline rates of marine heatwaves: Global trends, seasonal forecasts and marine management. Front. Clim. https://doi.org/10.3389/fclim.2021.801217 (2021).Article 

Google Scholar 
Schlegel, R. W., Oliver, E. C. J., Wernberg, T. & Smit, A. J. Nearshore and offshore co-occurrence of marine heatwaves and cold-spells. Prog. Oceanogr. 151, 189–205. https://doi.org/10.1016/j.pocean.2017.01.004 (2017).ADS 
Article 

Google Scholar 
Huang, B. et al. Improvements of the daily optimum interpolation sea surface temperature (DOISST) Version 2.1. J. Clim. 34, 2923–2939. https://doi.org/10.1175/jcli-d-20-0166.1 (2021).ADS 
Article 

Google Scholar 
OBPG, N. & Stumpf, R. P. Distance to Nearest Coastline: 0.01-Degree Grid. Distributed by the Pacific Islands Ocean Observing System (PacIOOS). http://pacioos.org/metadata/dist2coast_1deg.html and https://data.noaa.gov/dataset/dataset/distance-to-nearest-coastline-0-01-degree-grid2http://www.pacioos.hawaii.edu/metadata/dist2coast_1deg.html (2012).Schlegel, R. W. & Smit, A. J. heatwaveR: A central algorithm for the detection of heatwaves and cold-spells. J. Open Source Softw. 3(27), 821 (2018).ADS 
Article 

Google Scholar 
Sakurai, T., Yukio, K. & Kuragano, T. in Proceedings. 2005 IEEE International Geoscience and Remote Sensing Symposium, 2005. IGARSS’05. 2606–2608 (IEEE). More

AffiliationsSanger Institute, Wellcome Trust Genome Campus, Hinxton, UKPhysilia Ying Shi ChuaLaboratory of Genomics and Molecular Medicine, Department of Biology, University of Copenhagen, Copenhagen, DenmarkJacob Agerbo RasmussenCenter for Evolutionary Hologenomics, Globe Institute, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, DenmarkJacob Agerbo RasmussenAuthorsPhysilia Ying Shi ChuaJacob Agerbo RasmussenCorresponding authorCorrespondence to
Physilia Ying Shi Chua. More