Climate change exposure and vulnerability of the global protected area estate from an international perspective

Protected areas are essential to conserve biodiversity and ecosystem benefits to society under increasing human pressures of the Anthropocene. Anthropogenic climate change, however, threatens the enduring effectiveness of protected areas in conserving biodiversity and providing ecosystem services, because it modifies and redistributes biodiversity with unknown consequences for ecosystem functioning within protected areas. Here, we assess (a) the climate change exposure of the global terrestrial protected area estate and (b) the climate change vulnerability of national protected area estates.


| INTRODUC TI ON
Protected areas (PAs) are effective in conserving biodiversity, ecosystem functioning and services under increasing human pressures of the Anthropocene. Local biodiversity is generally higher inside than outside PAs (Gray et al., 2016). PAs preserve species and populations better than other conservation measures (Geldmann et al., 2013). For global biodiversity conservation, PAs are particularly effective when they are located in biodiversity hotspots (Joppa, Visconti, Jenkins, & Pimm, 2013), actively managed and funded . PAs cannot stop but decelerate the global biodiversity loss (Geldmann, Manica, Burgess, Coad, & Balmford, 2019). Further, PAs safeguard ecosystem services such as climate change mitigation and adaptation (MacKinnon, Dudley, & Sandwith, 2011); natural catastrophe control and the provision of natural resources (Xu et al., 2017); tourism and recreation (Balmford et al., 2009); and poverty reduction (Andam, Ferraro, Sims, Healy, & Holland, 2010). They are consequently considered crucial tools to meet the Sustainable Development Goals (SDG) and Aichi Biodiversity Targets (Mace et al., 2018). Conservationists perceive PAs as the most important policy for biodiversity conservation in the face of climate change (Hagerman & Satterfield, 2014).
Already in the 1980s scientists have warned of climate change being an inevitable threat to PA effectiveness (Peters & Darling, 1985). PAs are exposed to various direct and indirect climate change effects, for example increasing temperatures, melting of snow and ice, more severe droughts and storms, seasonal shifts, rising sea level and increased environmental acidification (Gross, Woodley, Welling, & Watson, 2017). Climate change is predicted to cause gains (Berteaux et al., 2018) and losses of biodiversity within PAs (Velazco, Villalobos, Galvão, & De Marco Júnior, 2019). In any case, the risk of PA downgrading, downsizing and degazettement (PADDD) increases for PAs that lose the biodiversity they were meant to protect (Thomas & Gillingham, 2015). Climate change modifies and redistributes biodiversity and thus forms novel ecosystems whose functioning and contributions to human well-being are unclear (Pecl et al., 2017). Climate change additionally co-occurs with other threats to biodiversity, such as human land use, implying interactive effects (Schulze et al., 2018). Therefore, the future effectiveness of PAs in preserving biodiversity and ecosystem services under climate change is uncertain.
Predicting the future climate inside PAs is required to inform conservation management and policymakers of potential climate change impacts on PAs (Rannow et al., 2014). Conservation management and policy are mainly adopted at the national to local scale. However, global studies about climate change impact neither address national authorities nor represent the local extent of PAs Beaumont et al., 2011;García-López & Allué, 2013;Bellard et al., 2014;Garcia et al., 2014;Ordonez et al., 2016;Li, Wu, et al., 2018;Li, Kou, et al., 2018); and the climate change research that focuses on PAs comprises a limited geographical extent only, for example North America (Batllori, Parisien, Parks, Moritz, & Miller, 2017) or Europe (Nila, Beierkuhnlein, Jaeschke, Hoffmann, & Hossain, 2019). A recent biogeographical investigation predicting climate shifts within PAs worldwide does not contemplate the governmental level either (Hoffmann, Irl, & Beierkuhnlein, 2019). A national view of the local climate change impact on individual PAs worldwide is missing but vital to support local to national conservation policy and management in reaching global conservation goals beyond 2020 despite climate change .
Here, we approach this research gap by assessing the climate change exposure of the terrestrial PAs worldwide at the highest spatial resolution for which global climate data is available, that is approximately 1km. In a first step, we assessed the climate change exposure of PAs as the climate anomalies predicted for the year 2070 within each grid cell covered by a PA. In a second step, we summarized the climate anomalies by each PA and present the PAs' median climate anomalies by country and management category. In a third step, we calculated country-specific correlations between median climate anomalies and other PA characteristics to provide additional information about the climate change vulnerability of national PA estates. In a fourth step, we compare the median climate anomalies and other PA characteristics between national PA estates via a principal component analysis (PCA). The outcomes inform proactive management that can compensate for negative impacts of climate change on PA effectiveness (Game, Lipsett-Moore, Saxon, Peterson, & Sheppard, 2011). Our work sets out to support climate-smart policy and management of PAs, particularly at the national to local level. The area and IUCN management category of each PA was retrieved from the WDPA. We consider PA area as a proxy for the amount of available resources for biodiversity to adapt to climate change within PAs. The IUCN management categories I to IV mean stricter protection, while categories V and VI allow for the sustainable use of natural resources, for example via silviculture and agriculture (Dudley, 2008). We applied the Terrain Ruggedness index (TR) as a measure of topographic heterogeneity. The TR index has also a resolution of 30 arc seconds (Amatulli et al., 2018). Planar area has a TR of 0m, whereas mountain areas can have a TR of up to 2,000 m in the Himalayas (Amatulli et al., 2018). The median of the TR values inside PAs was used to represent the topographic heterogeneity of each PA. The median is more robust against extreme values than the mean. The human footprint index 2009 is the most recent global indicator of human pressure and involves eight indicators of human land use (Venter et al., 2016a): population density, buildings, electric infrastructure, roads, railways, navigable waterways, cropland and pasture. We calculated the median human footprint of each PA by taking the median of the raster cell values that fall within each PA polygon. The irreplaceability index provided by Le Saout et al. (2013) reflects the conservation value of PAs in terms of the species diversity covered by PAs (Hoffmann, Beierkuhnlein, Field, Provenzale, & Chiarucci, 2018). This irreplaceability index represents the degree of overlap between each PA included in the WDPA (version October 2012) and the ranges of species on the IUCN Red List (Le Saout et al., 2013). The index involves ranges of 21,296 species: 6,240 amphibians, 9,793 birds and 5,263 mammals.

| Climate data
We used the WorldClim global climate data provided by Hijmans et al. (2005) including 19 bioclimatic variables with a resolution of 30 arc seconds. The 19 bioclimatic variables cover the full climate spectrum relevant for biodiversity, from annual trends (e.g., mean annual temperature and annual precipitation) to seasonal trends (e.g., annual range in temperature and precipitation) and extreme conditions (e.g., temperature of the coldest and warmest month, and precipitation of the wettest and driest quarters of the year). The 19 bioclimatic variables are listed in  MPI-ESM-LR and MRI-CGCM3. We selected the ten GCMs based on data availability (Hijmans et al., 2005) and dissimilarity between GCM outputs (Knutti, Masson, & Gettelman, 2013) to represent a diversity of predictive skills for different geographical regions worldwide.
Since WorldClim does not provide a monthly time series of mean climate variables for the period 1960-1990, we used the monthly time series provided by Abatzoglou et al. (2018). These data represent locally observed interannual climate variability (ICV), that is the standard deviation of mean monthly climate data from 1960 to 1990. The ICV data have a resolution of 2.5 arc minutes, which is coarser than the resolution of WorldClim's mean climate data of current and future conditions (30 arc seconds). To assign the ICV data to the mean climate data of current and future conditions, we disaggregated the ICV data to the resolution of 30 arc seconds. Matching both datasets in this way is appropriate, because the WorldClim data were used as input data for the calculation of the ICV data by Abatzoglou et al. (2018). For the ICV data, we calculated the 19 bioclimatic parameters via the biovars function of R package dismo (Hijmans, Phillips, Leathwick, & Elith, 2017).

| Calculating climate change exposure
Climate change exposure can be measured by a variety of climate change metrics (Bellard et al., 2014;Dawson, Jackson, House, Prentice, & Mace, 2011;Li, Wu, et al., 2018). Climate change metrics are either calculated for a single locality, that is at the local level, or for a set of localities, that is at the regional level (Garcia et al., 2014).
Here, we aim at analysing future climate change at the local level of PAs. Hence, we applied a local climate change metric. The most fundamental local climate change index is the climate anomaly metric, which is a measure of the magnitude of climate change at a given location indicating demographic population changes, particularly of species close to their climatic tolerance limits and with low adaptation capacity (Garcia et al., 2014). We refer to this local climate change index as climate change exposure.
We calculated the climate anomaly of each climate cell covered by a PA as the standardized Euclidean distance (SED) between independent climate variables of mean current  and mean future (2061-2080) climate conditions relative to the current ICV . The SED is a widely applied metric to estimate future climate anomaly (Bellard et al., 2014;Garcia et al., 2014;Mahony, Cannon, Wang, & Aitken, 2017;Ordonez et al., 2016;. The standardization of climate distance by the ICV makes the SED robust against distance inflation, which occurs when interannual climate variability is high but not considered by the distance metric. The SED between the mean current and mean future climate at a given location will be lower under high ICV than under low ICV, all else being constant.
We computed the SED based on independent climate variables.
By applying the SED to independent climate variables, we avoid variance inflation resulting from intercorrelated climate variables.
To produce independent climate variables, we projected the mean current, mean future and ICV climate data onto the first five principal components of the ICV data. In other words, the axes of the PCA represent the spatial variation of 1960-1990 interannual climate variability. Thus, the PCA well reflects the entire spatial and temporal variation of current climate conditions as a reference to measure future climate anomaly. We log10-transformed the precipitation variables before we conducted the PCA to correct for nonlinearity.
We thus reduced the 19 bioclimatic variables to five independent climate variables that were computationally practicable for us. The first five PCA axes account for 92% of the variation in the ICV data.
The PC loadings are shown in Table S.1. The PC space was built on the ICV data of all climate cells covered by a PA (n = 26,038,594). In the PCA space with axes scaled to have unit variance, the SED equals the Mahalanobis distance (Mahony et al., 2017).
We defined the following parameters to calculate the SED: The fewer climate variables are considered in measuring climate distance, the lower is the risk of Type I inference error (i.e., overestimating climate distance) and the higher is the risk of Type II inference error (i.e., underestimating climate distance). Because five variables are relatively few to represent all dimensions of the climate space, our results may underestimate the climate change impact in regions of low climate anomaly (Mahony et al., 2017).

| Estimating climate change vulnerability of national PA estates
We consider the PA characteristics "area," "elevation" and "terrain ruggedness" as indicators for the PAs' capacity to buffer climate change impact. The larger the PA area, the more and more diverse resources are likely provided for species to adapt to climate change via migration and adaptation. High resource diversity is also found in PAs of mountain regions, that is of higher elevation and terrain ruggedness. Terrain ruggedness is a proxy for climate and habitat diversity, and thus of resource availability and the adaptation capacity of PAs' biodiversity to impacts of climate change Lawler et al., 2015). We further assume that an increasing human footprint decreases the adaptive and buffer capacity of PAs because high human footprints indicate landscape fragmentation and human land use, lowering habitat extent, connectivity and resource availability, and hindering species adapting and migrating to track suitable climate conditions (Di Marco, Venter, Possingham, & Watson, 2018;Venter et al., 2016b). "Irreplaceability" represents the PAs' ecological importance for the conservation of globally threatened species (Le Saout et al., 2013).
We summarized the cell-based climate anomalies by individual PAs using the median, grouped the resulting median climate anomalies of each PA by country and correlated the median anomalies of PAs to other PA characteristics (see Section 2.1). We tested for correlations by using Pearson's correlation coefficient r and a modified t-test accounting for spatial autocorrelation (Dutilleul, Clifford, Richardson, & Hemon, 1993). The countrywide correlations be- Please note that the r and p-values cannot be compared between countries and do not represent the degree of climate change vulnerability of nationwide PA estates. However, the presence or absence of a significant correlation adds information about the climate change vulnerability of national PA estates. For example, a negative correlation between climate change exposure and PA area means that smaller PAs are more exposed to climate anomalies in the country, which is a valuable information for national conservation policy.
To compare the median climate anomalies and other PA characteristics of national PA estates, we additionally performed a PCA. We used the PAs' median climate anomaly, latitude, area, elevation, terrain ruggedness, the human footprint and irreplaceability as input data for the PCA (n = 84,032). We then calculated group centroids of the PAs' seven PC scores using countries as grouping factors and show these centroids along the first two PCA axes. The first two PCA axes account for 48% of the variation in the PA data for RCP 4.5 and 8.5. The data on PAs' median climate anomalies and characteristics are supplied under https://doi.org/10.5061/dryad.f4qrf j6tf and linked to the WDPA via the WDPA ID.
F I G U R E 1 Predicted climate anomalies within the terrestrial PA estate for the year 2070 under the moderate emission scenario RCP 4.5 and the high emission scenario RCP 8.5. The climate anomaly represents the magnitude of future climate change at a given location. Climate anomalies were calculated for each grid cell of approximately 1km resolution, using the standardized Euclidean distance between the current and future climate conditions. Here, we show the mean and standard deviation (SD) of climate anomalies resulting from future climate projections of ten global climate models. The SD is a measure of the variation among future climate predictions.   Figure S.1 for results of RCP 8.5. We summarized the mean climate anomalies (Figure 1) for each PA using the median. The IUCN management categories from I to VI cover a gradient of human integration, from strict human exclosure to sustainable human land use, respectively. The black numbers represent the number of PAs within the countries and IUCN management categories. "NA" means no management category was available. The boxplots were ordered by decreasing median. The limits of the grey box show the lower and upper quartiles, that is the interquartile range. The whiskers extend to the lowest and highest values within 1.5 times the interquartile range. The black dots indicate outliers beyond the whiskers. The alpha-3 country codes are given (i.e., ISO 3166). "Global" composes all PAs, while "Trans" refers to transboundary PAs  1  298  37  3  29  1  11  29  1  4282  1  9  3  9  8  57  4  1  15  3  202  364  6592  517  1456  2  4  2  1  6  49  15  2  23  346  1  1  6  13  3  859  192  54  5376  112  12  129  1384  10  3  345  1  3  42523  2  2  1  125  1  99  20  310  5  38  3  74  7  7  88  139  2  23  48  16  3  62  1  1  123  192  12  2  1489  3  1415  20  1  234  72  1  79  52  1  3  2284  752  547  322  525  9  141  2  781  227  179  52  15  10  15  640  10  2  938  566  30  11  341  47  33  2  2  6  2  1  314  861  2  11  12  4  2389  589  53  43  23  24  464  30  29  2  5  45  8  116  3  3  1  5  1  12  2  380  140  3  10  500  10  16  102  24  54  1  3  23  2   I  II  III  IV  V ARE  TLS  BMU  MRT  DNK  NOR  CUW  CHL  NCL  LBN  MLT  CUB  SWE  MUS  BHS  URY  KAZ  JAM  CYP  HKG  STP  ALA  BES  NLD  LVA  FIN  LTU  EST  TKM  TCA  GRL  SYC  VUT  ECU  MDG  DJI  BLR  USA  GIB  TWN  EGY  KHM  SWZ  GBR  AUS  ARG  DEU  RUS  KOR  NAM  POL  UKR  LAO  IRL  DMA  GMB  Global  FJI  BWA  AZE  CAN  SSD  SEN  LKA  BEL  TCD  IRN  ARM  PAK  NER  PRY  TUN  NZL  KGZ  BEN  VNM  JPN  THA  LUX  MTQ  LCA  PRT  MAR  MNG  PRI  FRA  ISL  ZAF  IND  BRB  GHA  BFA  COM  ISR  TGO  GEO  GNB  ESP  CZE  HUN  BGR  SVK  PAN  ZWE  GRD  BRA  CIV  KEN  CHN  PHL  MMR  Trans  TZA  CRI  TTO  NGA  ROU  MOZ  TUR  GRC  CHE  PNG  GNQ  AND  DZA  COL  TJK  AUT  SVN  BIH  CMR  IDN  NPL  ITA  HRV  SLV  HND  GAB  SRB  ZMB  NIC  SLE  BLZ  IRQ  MWI  JOR  PER  LBR  MKD  ALB  MLI  BTN  VEN  SUR  MEX  BOL  ETH  CAF  UGA  COG  MYS  GIN  PSE  GTM  MNE  BDI CUW  DNK  NCL  MLT  CYP  MUS  REU  CHL  BES  NOR  BHS  CUB  LVA  SWE  NLD  FIN  LKA  BLR  URY  LTU  AUS  CXR  GMB  VNM  GBR  TWN  RUS  VGB  HTI  KHM  ARG  FJI  DMA  WSM  SWZ  PAK  ISR  SEN  UZB  KOR  USA  CYM  NAM  DEU  Global  MDG  CAN  NER  NZL  POL  ISL  IRN  CZE  KAZ  MNG  TCD  KEN  AZE  LUX  THA  PRY  CAF  BRA  ESP  MAR  BFA  IND  JPN  BEN  SDN  TJK  LCA  BGD  BLZ  GRD  UKR  IDN  TTO  PRT  PAN  ETH  AGO  HUN  GEO  ZWE  DOM  GIN  KGZ  ARM  SVN  SVK  HND  GHA  BGR  PHL  BWA  TZA  CIV  TUN  CRI  MOZ  KWT  GNQ  SSD  NGA  ROU  AFG  AUT  DZA  GRC  ITA  SRB  FRA  ZMB  NIC  MMR  TGO  COL  VEN  MWI  BRN  JOR  GUF  SLV  BIH  BTN  MLI  IRQ  MKD  CMR  MEX  MYS  NPL  SUR  GTM  SLE  ALB  COD  UGA  GUY  RWA  PER  COG  are the cornerstones of conservation effort, but extending our high-resolution approach to the entire terrestrial surface would be extremely useful for environmental management worldwide. We highly recommend to follow this future perspective, although the F I G U R E 3 Global and country-specific correlations of the PAs' median climate anomalies (2070, RCP 4.5) with PA characteristics; see Figure S.2 for results of RCP 8.5. The PA characteristics "area," "elevation" and "terrain ruggedness" indicate the PAs' capacity to buffer the climate change impact; "irreplaceability" represents the PAs' importance for the conservation of globally threatened species. By relating the predicted climate anomalies to the PA characteristics at the country level, we provide additional information about the climate change vulnerability of national PA estates. PAs are assumed to be particularly vulnerable to climate change when the predicted climate anomalies, the human footprint and irreplaceability are high, while the area, elevation and terrain ruggedness are low. Bars reflect Pearson's correlation coefficients r; red for positive and blue for negative coefficients. Asterisks represent the significance level considering spatial autocorrelation (*p ≤ .05, **p ≤ .01, ***p ≤ .001), while no asterisk means non-significant correlation (p > .05). The alpha-3 country codes are shown (i.e., ISO 3166). "Global" composes all PAs, while "Trans" refers to transboundary PAs F I G U R E 4 Principal components analysis of the PAs' median climate anomalies (2070, RCP 4.5) and other PA characteristics grouped by countries; see Figure S.3 for results of RCP 8.5. The PA characteristics "area," "elevation" and "terrain ruggedness" indicate the PAs' capacity to buffer the climate change impact; "irreplaceability" represents the PAs' importance for the conservation of globally threatened species.  (Ordonez et al., 2016). High local climate anomalies can lead to physiological, morphological and behavioural changes of individuals and demographic changes of populations (Peñuelas et al., 2013). Species living close to their climatic tolerance limits and having low adaptation capacity are most affected by climate anomalies (Garcia et al., 2014), potentially leading to population declines (Foden et al., 2007) and local extinctions (Sinervo et al., 2010). Local climate anomalies can also positively affect biodiversity. Rising temperatures cause increasing plant diversity in high latitudes (Hill & Henry, 2011) and elevations (Steinbauer et al., 2018). The fitness of mountain lizards can increase due to warming (Chamaille-Jammes, Massot, Aragon, & Clobert, 2006). High-latitude PAs are projected to gain biodiversity under global warming (Berteaux et al., 2018). In Kruger National Park, climate change is expected to increase plant productivity and thus elephant populations (Scheiter & Higgins, 2012).
Climate anomalies cause new, non-analogue communities, that is communities without current analogues, because species differ in their ability to respond to climate change via dispersal, range dynamics and biotic interactions . The functioning of such novel communities remains largely unknown (Hobbs et al., 2006). Impacts of recent climate change onto ecosystem functioning and services are manifold (Scheffers et al., 2016). Mascaro et al. (2012 show that non-native species led to increased productivity, carbon storage and nutrient cycling in lowland Hawaiian rain forests. In contrast, forest carbon storage is decreasing with increasing frequency and intensity of droughts, fires, windthrow and insect outbreaks (Holmgren, Hirota, van Nes, & Scheffer, 2013;Seidl, Schelhaas, & Lexer, 2011).
PAs are assumed to be particularly vulnerable to climate change when the predicted climate anomalies, the human footprint and irreplaceability are high, while the area, elevation and terrain ruggedness are low . We revealed that increas-  (Belote et al., 2018), ecosystem intactness (Watson, Iwamura, & Butt, 2013), conservation targets (Belote et al., 2017), the conservation capacity of land (Gillson, Dawson, Jack, & McGeoch, 2013), the management resources available (Wintle et al., 2011) and the risks of management actions (Ando et al., 2018) differ between PAs. Climate-smart management guidelines generally aim at the persistence and resistance of present biodiversity despite climate change, or at the adaption of biodiversity to climate change (Gross et al., 2017). Reasonable management interventions can vary from low intensity, for example monitoring, to high intensity, for example assisted migration and restoration (Dawson et al., 2011;Gillson et al., 2013). Appropriate management practice may be conservative, innovative, flexible, reversible or experimental (Belote et al., 2018). Alternatively, "no-regret" strategies could be applied, which intend to achieve conservation benefits irrespective of climate change (Hallegatte, 2009). In any case, adaptive PA management is a promising tool to ensure the enduring effectiveness and efficiency of PAs in the light of uncertain future developments (Rannow et al., 2014).
Our methodological approach implies assumptions that limit the implications of our findings. The climate anomaly index is the most fundamental indicator of climate change exposure at the local level, but cannot reveal the entire complexity of biodiversity and ecosystem responses to climate change at the local and regional level (Garcia et al., 2014). This local indicator does, for instance, not reflect shifts in seasonal climate nor changes in climate extremes, which are both extremely relevant for biodiversity, ecosystem functioning and services (Pecl et al., 2017;Scheffers et al., 2016). However, seasonal climate shifts and changing climate extremes are hardly predictable for the local level but global extent. Ideally, future studies would incorporate multiple climate change metrics of the local and regional level (Garcia et al., 2014) to understand the full potential of climate change impact.
Further, the predictive skills of GCMs differ between geographical regions (Bring et al., 2019). For our global assessment including national perspectives, we selected ten GCMs based on dissimilarity between GCM outputs (Knutti et al., 2013) to represent a diversity, and thus complementarity, of predictive skills. We then estimated the variation among future climate projections, but an uncertainty of the predictions remains that is inherent in the climate models and practically incalculable. Especially in mountain regions with a low density of climate stations (Hijmans et al., 2005), the quality of the WorldClim data is poor (Bobrowski & Schickhoff, 2017). Accordingly, our results must be carefully interpreted for those regions. Moreover, the climate data resolution of approximately 1 km do not consider microclimate, which can buffer climate change impact (Suggitt et al., 2018). Interacting effects between climate change and other threats to biodiversity and ecosystem functioning (e.g., invasive species) are neglected as well. Also, the human footprint index from 2009 and the irreplaceability index from 2012 are out of date and newer version are not available. Nevertheless, given that human land cover (Venter et al., 2016b) and species loss (Johnson et al., 2017) are increasing globally, our application of the human footprint and irreplaceability index may even underestimate the climate change exposure of PAs.
In addition, the climate change vulnerability of biodiversity, ecosystem functioning and services depend not only on climate change exposure, adaptive capacity and ecological importance, but also on resistance (or sensitivity, i.e., ability to remain in the original state despite change) and resilience (i.e., ability to return to the original state after change) (Dawson et al., 2011;Li, Wu, et al., 2018 Such a comprehensive analysis could be the foundation of a globally coordinated and adaptive PA planning and management system in the future. We perceive the development and application of a global adaptive PA management system as a major future task to reach global conservation and sustainability targets, and safeguard human well-being of generations to come.

ACK N OWLED G EM ENTS
We acknowledge support from the ECOPOTENTIAL project-EU Horizon 2020 research and innovation programme, grant agreement No. 641762. Open access funding enabled and organized by Projekt DEAL.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interests.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13136.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used in this study are open. Data references are given in the main text. The data we produced are available online at https://doi. org/10.5061/dryad.f4qrf j6tf.