The objective of this study was to use side-view digital photography to describe the structure and dynamics of natural forest stands in the context of the FunDiv Europe exploratory network. From April to December 2012, as well as April to December 2013 for some regions, time-series of digital images were recorded at each forest plot using commercially available digital camera units. Three regions of interest (ROI) representing the over-, mid- and under-storey layers of the forest plots were manually delineated for each image time-series. The image processing workflow produced a list of six key variables for each of the three ROIs though additional variables are also provided in this dataset. The six variables we suggest as key foci variables are: 1) Growing season length (days) from Green Cover, 2) Growing season onset from Green Cover (day of the year), 3) Growing season offset (day of the year) from Green Cover, 4) Mean Green Cover (%), 5) Mean ROI texture (Mean information gain; MIG), 6) Mean ROI anisotropy (MIG ratio).

The sampling design follows that of the exploratory platform. Each plot from each region was sampled though some plots are missing from the dataset due to loss of equipment. "Sampling" means the acquisition of images. Specifically, time-series of digital images were recorded at each forest plot using commercially available digital camera units (PlantCam ™ timelapse camera, Model WSCA04-00106. Wingscapes, Alabaster, AL, USA), which capture repeat side-view digital images at a programmed time interval. Within 5 m from the southern side of each forest plot, one camera was installed on a large (DBH > 10cm) and visually healthy tree. The camera was installed at a height of 2m on the tree side and oriented towards the magnetic North. The camera was fastened to the tree securely and the view angle adjusted parallel to the ground slope. Each camera was programmed to record a single image every 3 hours during daylight (9 am to 5 pm). The sampling period started in early April and images were acquired until late November - December.

This dataset covers all exploratory regions with the exception of Romania which is excluded due to loss of equipment and data. Data for one to six plots are missing from each region due to loss of equipment and data. Locations include all plots for the exploratory regions of Finland, Germany, Poland, Spain and Italy.

This data is derived from image analysis applied to a time series of images collected during the 2012 and 2013 growing seasosn (approx. April - December depending on region). For the regions of Germany, Poland, Spain and Italy images were acquired during the 2012 growing season, while Finland images were acquired during the 2013 growing season. The German plots were also re-sampled (i.e., images acquired) during the 2013 growing season and, therefor, Germany has two time-series or phenologies (2012 & 2013 seasons). For each plot and each acquisition year, Three images were captured daily for the entire growing season/sampling period. However, the data here represents analysis performed on only images acquired at the 12:00 pm (site local time) acquisition interval. The analysis was performed January - May, 2014.

These data represent preliminary analysis. To date, no image correction or equalization was done across the image time-series. Images, and thus image metrics, can be influenced by light and weather conditions. These extternal influences particularly apply to the structural image metrics or variables. Image metrics based on greenness indices are less influenced by light conditions. Further analysis will explore image equalization techniques in an attempt to standardize image intensity across an image time-series. As mentioned some cameras were lost in certain plots and thus so too their image data. These plots are not included.

Three regions of interest (ROI) representing the over-, mid- and under-storey layers of the forest plots were manually delineated for each image time-series. For each ROI in an image time-series, growing season onset (green-up) and offset (green-down) dates were estimated using two image greenness indices that we term "Percent Green" and "Green Cover" (Meyer and Neto, 2008; Sonnentag et al., 2012). The two greenness metrics are somewhat redundant though operate to estimate greenness in slightly different manners. For further details on differentiating the two greenness metrics please see the povided references or contact the data authors. We suggest any users of this data first use the "Green Cover" metric though one can use and compare both. For both greenness metrics, dates of maximum greenness increase (green onset dates) and decrease (green offset dates) were obtained from the 1st and 2nd derivatives of the smoothed greenness time-series (e.g., Noormets, 2009). In addition, the minimum value, maximum value, mean and standard deviation of greenness values for the time-series were also calculated. Furthermore, differences in structural heterogeneity of light interception/reflection patterns for the different forest strata were quantified in the over-, mid- and under-storey image ROIs. Structural heterogeneity metrics of texture and anisotropy were calculated on grey-tone transformed images following the method proposed by Proulx et al. (2014). For each plot, structural heterogeneity metrics were calculated for all images taken during each growing season (i.e., the plot time series; typically May 1 to October 31 though some chronologies extend longer) and subsequently the maximum, minimum, mean and standard deviation values were calculated.Thus, the image processing workflow produced a list of 22 variables for each of the 3 ROIs, for a total of 66 image-derived values per plot. However, for a given ROI we suggest 6 key variables of focus for those wanting to use this dataset: 1) Green Cover Growing season length (days), 2) Green Cover Growing season onset (day of the year), 3) Green Cover Growing season offset (day of the year), 4) Mean Green Cover (%), 5) Mean ROI Texture (Mean information gain; MIG), 6) Mean ROI Anisotropy (MIG ratio). Image-derived values could not be calculated for all forest plots in the FunDiv exploratory network since nearly all cameras were stolen in one bioclimatic region (Romania) and a few other cameras failed in the other regions. Finally, the growing season length, onset and offset dates could not be calculated for time-series that did not present a clear seasonal pattern (i.e., sharp green-up and green-down trends). This situation occurred in ~5% of the forest plots depending on the ROI.