Documentation Revision Date: 2017-03-16
Data Set Version: V1
Summary
From a temporal stack of >240 Landsat TM and ETM+ scenes (30-m resolution) covering the 1984-2013 period, the fraction of photosynthetic vegetation (fPV) was quantified. The fPV data were used to parameterize a dual sigmoid logistic growth curve representing seasonal patterns of fPV from which phenological parameters including spring onset and autumn offset were derived for each of the four parks.
Based on similar phenology and tree species in common, 113 sites were selected across the four parks. In total, 222 trees of three deciduous species (Liriodendron tulipifera, Quercus rubra, and Quercus alba) were cored. Individual rings were separated and annual wood production, carbon and nitrogen content, and δ13C, and δ15N isotope ratios were measured.
This data set includes 42 data files. There are 40 files in GeoTIFF (.tif) format; with ten files per park. These GeoTIFF files are maps of the Landsat and model-derived average phenology parameters that were used to stratify and randomly select sites for tree coring within each park. There are also two files in comma-separated (.csv) format: 1) results of individual tree ring measurements and analyses for each site; and 2) site locations and physical characteristics, and site averages of phenology parameters and tree ring analysis results.
Figure 1. Location map of sites across the four study regions spanning the Appalachian Oak mesophytic forest of the eastern United States. The background maps for the park sites are the 30-year average phenology parameter, spring onset of greenness (spring_onset), as provided with this data set.
Citation
Elmore, A.J., D. Nelson, S.M. Guinn, and R. Paulman. 2017. Landsat-based Phenology and Tree Ring Characterization, Eastern US Forests, 1984-2013. ORNL DAAC, Oak Ridge, Tennessee, USA. http://dx.doi.org/10.3334/ORNLDAAC/1369
Table of Contents
- Data Set Overview
- Data Characteristics
- Application and Derivation
- Quality Assessment
- Data Acquisition, Materials, and Methods
- Data Access
- References
Data Set Overview
This data set provides a 30-year record of Landsat TM and ETM+ derived forest phenology and the results of tree ring analyses for annual wood production and nitrogen and carbon isotopic composition at 113 selected forested sites in the eastern United States. The sites are located in four national parks: Prince William Forest Park (PRWI), Harpers Ferry National Historical Park (HAFE), Catoctin Mountain Park (CATO), and Great Smoky Mountains National Park (GRSM). Phenology and tree ring data cover 1984-2013.
From a temporal stack of >240 Landsat TM and ETM+ scenes (30-m resolution) covering the 1984-2013 period, the fraction of photosynthetic vegetation (fPV) was quantified. The fPV data were used to parameterize a dual sigmoid logistic growth curve representing seasonal patterns of fPV from which phenological parameters including spring onset and autumn offset were derived for each of the four parks.
Based on similar phenology and tree species in common, 113 sites were selected across the four parks. In total, 222 trees of three deciduous species (Liriodendron tulipifera, Quercus rubra, and Quercus alba) were cored. Individual rings were separated and annual wood production, carbon and nitrogen content, and δ13C, and δ15N isotope ratios were measured.
Project: North American Carbon Program (NACP)
The North American Carbon Program (NACP) is a multidisciplinary research program designed to improve scientific understanding of North America's carbon sources and sinks and of changes in carbon stocks needed to meet societal concerns and to provide tools for decision makers.
Related Publication:
Elmore, A. J., Nelson, D.M., and Craine, J.M. Earlier springs are causing reduced nitrogen availability in North American eastern deciduous forests, Nature Plants (2016). http://dx.doi.org/10.1038/nplants.2016.133
Acknowledgements:
This project was supported by the National Aeronautics and Space Administration's Terrestial Ecology Program and North American Carbon Program NNX12AK17G.
Data Characteristics
Spatial Coverage
The four study areas are located in the eastern US and are separated by more than 500 km: Prince William Forest Park, Harpers Ferry Historical Park, Catoctin Mountain Park, and the Great Smoky Mountains Park.
Spatial Resolution
Landsat imagery and the phenology parameter maps are at 30-m resolution. The sampling sites were randomly selected within common phenology classes at this same 30-m resolution, within the four national parks.
Temporal Coverage
1984-01-01 through 2013-12-31
Temporal Resolution
Annual
Site boundaries: (All latitude and longitude given in decimal degrees, datum: WGS84)
| Site (Region) | Westernmost Longitude | Easternmost Longitude | Northernmost Latitude | Southernmost Latitude |
|---|---|---|---|---|
| Prince William Forest Park | -77.441 | -77.322 | 38.6439 | 38.5592 |
| Harpers Ferry Historical Park | -77.796 | -77.627 | 39.3558 | 39.2881 |
| Catoctin Mountain Park | -77.525 | -77.407 | 39.6947 | 39.5422 |
| Great Smoky Mountains National Park | -84.034 | -82.983 | 35.8981 | 35.3389 |
Data File Information
There are 42 data files in this data set. This includes two files in comma-separated (.csv) format and 40 files in GeoTIFF (.tif) format.
| File name | Description |
|---|---|
| dendrophenology_rings.csv | This file provides individual tree ring width and wood analyses of C, N, 13C, 15N, tree diameter at breast height (DBH), and calculated Basal Area Increment (BAI) |
| dendrophenology_sites.csv | This file provides site location and physical characteristics, species occurrence, and site averages of phenology parameters and tree ring measurement results. |
| XXXX_phenology_pre2013_parameter.tif | 40 GeoTIFF files, 10 parameter files per park, of the Landsat and model-derived average phenology maps used to stratify and randomly select sites for tree coring within each park. File naming syntax described below. |
Table 1. Variables in the file dendrophenology_rings.csv
| Variable ID | Units | Description |
| site_id | Four letter code indicating the park unit followed by a number indicating the site number | |
| region | Harpers Ferry Historical Park (HAFE); Catoctin Mountain Park (CATO); Prince William Forest Park (PRWI); Great Smokey Mountains Park (GRSM) | |
| elevation | meters | Elevation |
| slope | degrees | Topographic slope |
| aspect | degrees | Topographic aspect |
| tree_id | The site_id, followed by the tree number. There were two trees at each site. | |
| core_id | The tree id, followed by a core id. Each tree was cored two times but only one core was measured for isotopes. | |
| sp_code | A four letter code indicating the tree species. QUAL = Quercus alba; QURU = Quercus rubra; LITU = Liriodendron tulipifera | |
| DBH | cm | Diameter at breast height (DBH) |
| ring_width | mm | Ring width |
| BAI | mm^2 | Basal Area Increment (BAI) |
| Year | YYYY | Year the ring grew on the tree |
| dC | per mil | delta 13C of carbon in the tree ring |
| C_wood | % | Percent carbon in ring wood |
| CO2_air | ppm | Atmospheric concentration of CO2. Data are from the Scripps CO2 Program, Scripps Institution of Oceanography |
| d13C_CO2_air | per mil | delta 13C of CO2 in air. Data are from the Scripps CO2 Program, Scripps Institution of Oceanography |
| d13C_wood | per mil | Carbon isotope discrimination measured in wood |
| Ci | ppm | The internal CO2 concentration |
| Ci_Ca | unitless | The seasonally integrated ratio of internal (Ci) to atmospheric (Ca) CO2 concentration. |
| Ca_minus_Ci | ppm | Ca minus Ci (used to calculate iWUE) |
| iWUE | ratio | iWUE equal to the ratio of photosynthesis (A) to stomatal conductance of water vapor (gw) |
| early_sum_ann | fPV | Mean residual fraction of photosynthetic vegetation (fPV) in the early summer. Early summer is defined as starting 20 days after the spring onset and ending on DOY 221. |
| late_sum_ann | fPV | Mean residual fraction of photosynthetic vegetation (fPV) in the late summer. Late summer is defined as starting on DOY 221 and ending 20 days before the autumn offset of greeness |
| spring_ann | days | Mean offset in the DOY direction of fPV relative to the mean fPV on that DOY across all years |
| autumn_ann | days | |
| N_wood | % | Percent nitrogen in wood |
| d15_N | per mil | delta 15N of nitrogen in wood |
Table 2. Variables in the file dendrophenology_sites.csv
| Variable ID | Units | Description |
|---|---|---|
| site_id | Four letter code indicating the park unit followed by a number indicating the site number | |
| region | Harpers Ferry Historical Park (HAFE); Catoctin Mountain Park (CATO); Prince William Forest Park (PRWI); Great Smokey Mountains Park (GRSM) | |
| easting | meters | Spatial location east of the meridian of the UTM zone identified in the srid |
| northing | meters | Spatial location north of the equator |
| Datum_srid | EPSG code which corresponds to the WGS 84 / UTM zone 18N or 17N. The values are 32617 or 32618 | |
| slope | degrees | Topographic slope |
| aspect | degrees | Topographic aspect |
| elevation | meters | Elevation |
| LITU | binary | Indicates presence of Liriodendron tulipifera among the tree species cored at the site. |
| QURU | binary | Indicates presence of Quercus rubra among the tree species cored at the site. |
| QUAL | binary | Indicates presence of Quercus alba among the tree species cored at the site. |
| av_min_veg_cover | fraction | The average minimum vegetation cover. fPV is the fraction of photosynthetically active vegetation |
| av_cover_amplitude | fraction | The vegetation cover amplitude, achievable only when the rate of greendown during the summer growing season equals 0. |
| spring_inflection_point | days | The day of year (DOY) of the spring inflection point (i.e., days since Jan 1). |
| greenness_increase_max_greeness | days | Proportional to the length of the spring greenup period. Approximately equal to 1/3 of the length of time (in days) between the onset of greenness increase and the onset of greenness maximum. |
| autumn_inflection_point | days | The day of year of the autumn inflection point |
| autumn_greendown | days | Proportional to the length of the spring greenup period |
| greendown_fPV | fPV/day | Rate of greendown during the summer growing season |
| d15N_slope | per mil/yr | Trend over time in delta 15N |
| d15N_p_value | P-value for the trend over time in delta 15N | |
| d15N_mean | per mil | Site mean delta 15N |
| D13C_slope | per mil/yr | Trend over time in carbon isotope discrimination (Delta 13C) |
| D13C_p_value | P-value for the trend over time in Delta 13C | |
| D13C_mean | per mil | Site mean Delta 13C |
| BAI_slope | mm^2/yr | Trend over time in Basal Area Increment (BAI) |
| BAI_p_value | P-value for the trend over time in BAI | |
| BAI_mean | mm^2 | Site mean BAI |
| iWUE_slope | unitless ratio/yr | Trend over time in intrinsic Water Use Efficiency (iWUE) |
| iWUE_p_value | P-value for the trend over time in iWUE | |
| iWUE_mean | unitless ratio | Site mean iWUE |
Note: The Elmore et al., 2012 study did not include the Great Smoky Mountains National Park site.
GeoTIFF Files
There are 40 GeoTIFF files with this data set. The files are Landsat and model-derived average phenology parameter maps used to stratify and randomly select sites for tree coring within each park.
The files are named XXXX_phenology_pre2013_"parameter".tif
Where:
XXXX = park [ Harpers Ferry Historical Park (HAFE); Catoctin Mountain Park (CATO); Prince William Forest Park (PRWI); Great Smokey Mountains Park (GRSM) ]; and
"parameter"= phenology parameters.
Example file name: cato_phenology_pre2013_autumn_offset.tif
[ The "_pre2013_" element denotes that the files contain data from 1984 to 2013. ]
Table 3. Descriptions of parameters provided in the GeoTIFF files. There are 10 files/parameters per park. Note cross references to corresponding variables in the *_sites.csv data file and bootstrapping parameter numbers in Section 4.
| "parameter" Phenology parameter in respective map file. |
Units | Description | Corresponding variable in *_sites.csv data file | Corresponding parameter number. See Section 4. |
|---|---|---|---|---|
| autumn_offset | days | Equals the day of year of the autumn inflection point. | autumn_inflection_point | m5 |
| autumn_offset_slope | days | Proportional to the length of the autumn greendown period. | autumn_greendown | m6 |
| min_veg_cover | fraction | The average minimum vegetation cover. fPV is the fraction of photosynthetically active vegetation. | av_min_veg_cover | m1 |
| model_iterations | The number of model iterations required to calculate the phenology parameters | Not applicable | NA | |
| model_misfit | Model misfit, calculated using Equation 8 in Elmore et al., 2012 | Not applicable | NA | |
| num_landsat_images | The number of Landsat images used for that pixel (the value changes across the image due to cloud cover that obscures some pixels but not all) | Not applicable | NA | |
| potential_amplitude | fraction | Equals the vegetation cover amplitude, achievable only when "summer_greendown" equals 0. | av_cover_amplitude | m2 |
| spring_onset | days | Equals the day of year (DOY) of the spring inflection point (i.e., days since Jan 1) | spring_inflection_point | m3 |
| spring_onset_slope | days | Proportional to the length of the spring greenup period. Approximately equal to 1/3 of the length of time (in days) between the onset of greenness increase and the onset of greenness maximum. | greenness_increase_max_greeness | m4 |
| summer_greendown | fPV/day | Rate of greendown during the summer growing season | greendown_fPV | m7 |
Attributes of the *.tif files
All files contain the following attributes:
| File type | File format | Map units | Resolution | Number of bands | Data type | Fill value |
| raster | GeoTiff | meter | 30 | 1 | Float32 | 3.4E+38 |
The table below provides file names and attributes of the individual files. Minimum and maximum values may reflect the values the model derived for non-forest land areas.The phenology parameter files are unfiltered.
Table 4. Attributes of the GeoTIFF files.
| Filename | crs_proj4 | West | East | South | North | Min value | Max value |
|---|---|---|---|---|---|---|---|
| cato_phenology_pre2013_autumn_offset.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | -75.7 | 648.9 |
| cato_phenology_pre2013_autumn_offset_slope.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | -33.7 | 48.9 |
| cato_phenology_pre2013_min_veg_cover.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | -0.1 | 0.6 |
| cato_phenology_pre2013_model_iterations.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | 2 | 101 |
| cato_phenology_pre2013_model_misfit.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | 21.8 | 779.8 |
| cato_phenology_pre2013_num_landsat_images.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | 32 | 161 |
| cato_phenology_pre2013_potential_amplitude.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | -2.6 | 6.3 |
| cato_phenology_pre2013_spring_onset.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | -241.9 | 596.3 |
| cato_phenology_pre2013_spring_onset_slope.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | -23.6 | 43.8 |
| cato_phenology_pre2013_summer_greendown.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.525 | -77.407 | 39.5422 | 39.6947 | -0.0174 | 0.033 |
| grsm_phenology_pre2013_autumn_offset.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | -508.9 | 795.3 |
| grsm_phenology_pre2013_autumn_offset_slope.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | -40.9 | 66.3 |
| grsm_phenology_pre2013_min_veg_cover.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | -0.8 | 0.8 |
| grsm_phenology_pre2013_model_iterations.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | 2 | 101 |
| grsm_phenology_pre2013_model_misfit.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | 0 | 3414.1 |
| grsm_phenology_pre2013_num_landsat_images.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | 1 | 177 |
| grsm_phenology_pre2013_potential_amplitude.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | -7 | 8.2 |
| grsm_phenology_pre2013_spring_onset.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | -578.1 | 817.6 |
| grsm_phenology_pre2013_spring_onset_slope.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | -69.6 | 59.8 |
| grsm_phenology_pre2013_summer_greendown.tif | +proj=utm +zone=17 +ellps=WGS84 +units=m +no_defs | -84.034 | -82.983 | 35.3389 | 35.8981 | -0.0494 | 0.04794 |
| hafe_phenology_pre2013_autumn_offset.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | -217.6 | 806.3 |
| hafe_phenology_pre2013_autumn_offset_slope.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | -33.8 | 48.7 |
| hafe_phenology_pre2013_min_veg_cover.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | -0.1 | 0.6 |
| hafe_phenology_pre2013_model_iterations.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | 2 | 101 |
| hafe_phenology_pre2013_model_misfit.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | 18.7 | 970.4 |
| hafe_phenology_pre2013_num_landsat_images.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | 25 | 171 |
| hafe_phenology_pre2013_potential_amplitude.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | -3.2 | 6.3 |
| hafe_phenology_pre2013_spring_onset.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | -626.8 | 532.3 |
| hafe_phenology_pre2013_spring_onset_slope.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | -30.3 | 48.4 |
| hafe_phenology_pre2013_summer_greendown.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.796 | -77.627 | 39.2881 | 39.3558 | -0.0208 | 0.0387 |
| prwi_phenology_pre2013_autumn_offset.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 456.8 |
| prwi_phenology_pre2013_autumn_offset_slope.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 38.6 |
| prwi_phenology_pre2013_min_veg_cover.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 0.6 |
| prwi_phenology_pre2013_model_iterations.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 101 |
| prwi_phenology_pre2013_model_misfit.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 699.2 |
| prwi_phenology_pre2013_num_landsat_images.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 189 |
| prwi_phenology_pre2013_potential_amplitude.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 1.4 |
| prwi_phenology_pre2013_spring_onset.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 555.5 |
| prwi_phenology_pre2013_spring_onset_slope.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 37.5 |
| prwi_phenology_pre2013_summer_greendown.tif | +proj=utm +zone=18 +datum=WGS84 +units=m +no_defs | -77.441 | -77.322 | 38.5592 | 38.6439 | -3.4 | 0.007 |
Application and Derivation
These data are useful for climate change studies. Advances in remote sensing and dendroecological methods present opportunities to acquire information retrospectively to advance understanding of how phenological changes and resource availability to trees have been affecting forest productivity. This study provides a 30 year record of remotely-sensed forest phenology (Elmore et al., 2016).
As described in Elmore et al. (2016), all data were organized in a PostgreSQL object-relational database and queried for statistical analysis in the R statistical programing language. Two types of models were used to test for the effects of spring and autumn timing on resource availability and wood production. The first type of model used all available observations across trees and time. These models tested for the effects of time, spring anomaly and autumn anomaly on BAI, δ15N and Δ13C from tree rings, and included the main effects of region (GRSM, PRWI, HAFE and CATO) and species (QUAL, QURU and LITU) as blocking factors. Because the different measurements were made at different frequencies, the sample size available for each model varied. Observations of spring (n = 2,143) and autumn (n = 2,911) anomaly were sparse because of low availability of appropriately timed Landsat data. Further, these observations did not always line up with measurements of δ15N, resulting in n = 967 for models comparing spring anomaly with δ15N and n= 1,229 for models comparing autumn anomaly with Δ13C.
In the second set of models, the rate of change (slope) of trends was calculated in tree-ring-based measurements over time and rates were compared across sites. These models aid the interpretation of causal relationships between correlated variables identified in the first set of models. Sample sizes for these models equaled the number of sites (n = 113). Region and three nominal variables were also included in these models indicating the presence or absence of each of the three tree species at the site.
To augment exploration of regression statistics in tabular form, the relationship between δ15N and spring anomaly was visualized in scatterplot form. The scatterplot was generated by first modeling the effects of region and species on δ15N, then modeling the effect of spring anomaly on the model residuals. From the data used to construct this model, the annual mean spring anomaly and annual mean δ15N were calculated (Elmore et al., 2016).
Figure 2. Relationships between mean δ15N, Δ13C, and spring anomaly: a, Years exhibiting an early spring phenology anomaly are associated with low δ15N (r2 = 0.21; P = 0.02), after accounting for region and species effects. Symbol sizes represent sample sizes of paired δ15N and spring anomaly observations in each year, which range from 2 to 93; b, Averaged biennially across all 222 trees, Δ13C has increased and δ15N has decreased over time (Elmore et al., 2016).
Quality Assessment
Uncertainty in the average phenology parameters
Bootstrapped resampling was used to calculate uncertainty in the average phenology observations. These were reported in Elmore et al. 2012. Others have compared the annual phenology anomaly with in situ observations (Melaas et al., 2013, Melaas et al., 2016). Analytical precision (1 sigma) of an internal wood standard analyzed alongside samples was 0.1 per mil for del 13C and 0.3 per mil for del 15N. Uncertainty estimates are referenced in Elmore et al. 2016.
Table 3. Summary of bootstrapping results (Elmore et al., 2012). This study did not include the Great Smoky Mountains National Park site.
|
Parameter |
Description |
Inversion estimate using dobs
|
Mean bootstrap estimate
|
Bootstrap 95% lower bound
|
Bootstrap 95% upper bound
|
|
m1 |
Minimum vegetation cover |
0.1023 |
0.9977 |
0.08046 |
0.1210 |
|
m2 |
Potential amplitude |
0.8802 |
0.8779 |
0.6917 |
1.042 |
|
m3 |
Spring onset of greenness |
108.2 |
108.5 |
100.0 |
115.7 |
|
m4 |
Slope of spring onset |
7.596 |
8.600 |
6.182 |
11.45 |
|
m5 |
Autumn offset of greenness |
311.4 |
311.4 |
301.6 |
321.5 |
|
m6 |
Slope of autumn offset |
7.473 |
6.971 |
4.657 |
8.742 |
|
m7 |
Summer greendown |
0.002250 |
0.002231 |
0.001499 |
0.002898 |
Sensitivity analysis of effect size on core chronology
To evaluate the sensitivity of our results to the potential for missing or false tree rings, or any other sources of error in core chronology (e.g., the possibility for N translocation between tree rings), a sensitivity analysis was performed in which 1%, 5%, and 10% of the cores were sampled at random for artificial tampering. For each sampled core a ring was either added or taken away from the bark end of the core. This procedure changed the alignment between measurements of ring width, δ13C and δ15N in ring wood, and year and phenology for the cores sampled for tampering. Any of the results that are sensitive to the exact alignment of these measurements will be influenced by this sensitivity analysis. 100 Monte Carlo simulations were performed at each of the three random sampling levels, repeatedly modeled the effects of year, spring anomaly, region, and species on δ15N and constructed histograms of the resulting effects of year and spring anomaly on δ15N.
This sensitivity analysis revealed that the effects of year and spring anomaly are robust to a range of core tampering. As the fraction of cores sampled for tampering increases some random permutations result in either an increase or a decrease in the effect size (both effects). However, the trend of δ15N with time is always negative and the effect of an earlier spring is always positive.
Data Acquisition, Materials, and Methods
Site description
The study sites were located in four national parks: Prince William Forest Park, Harpers Ferry Historical Park, Catoctin Mountain Park, and Great Smoky Mountains Park. These parks, referred to as ‘regions’ in this study are within the Appalachian Oak sub-region of the mesophytic forest region extending from New England to Tennessee. The mean annual temperature is 11.76 degrees C and the mean annual precipitation is 1,595 mm across the sampling regions. Elevation ranges from 30 m to more than 1,400 m, which influences temperature and precipitation patterns (Elmore et al., 2016).
Methods
Remotely sensed phenology data
Landsat scenes (>240 depending on location; Thematic Mapper (TM) and Enhanced Thematic Mapper plus +, 30 m pixel, path15/row33) were processed to reflectance using the Landsat Ecosystem Disturbance Processing System (LEDAPS), and analyzed for the photosynthetic vegetation fraction, fPV, using spectral mixture analysis and image end members. Observations of fPV were organized by DOY and fit with two sigmoid logistic growth curves, one increasing in spring and a second decreasing in autumn.
- To evaluate changes in phenology, spring and autumn scenes were identified that were positioned within 20 days of the spring and autumn inflection points, greater than 20% of the site average annual minimum fPV, and less than 20% (spring) and 40% (autumn) of the maximum fPV.
- These spring- and autumn-timed Landsat observations of fPV were compared against the 30-year average date of spring onset and autumn offset, respectively, to calculate anomalies. For any given location, between 5 and 16 spring fPV observations and 7 and 18 autumn fPV observations were identified (mean of 9.7 and 13.4, respectively).
- Residual vegetation cover was also calculated after accounting for the mean vegetation phenology.
Site selection
Maps of average phenology were used to stratify and randomly select locations for tree coring within each park. An ISODATA clustering algorithm was used to identify 20 phenology classes that differed primarily in growing season length and the amplitude of annual fPV. The standard deviation in phenology parameters (3 × 3 pixel neighbor rule) was also calculated and used to further constrain the extent of each phenology class to areas of low spatial variability in phenology. Six randomly located sites were then identified in each phenology class. Alternative sites for each phenology class were also identified; if trees of the target species were not identified at any given site an alternative site was used. Similar methods were employed at all park units used in the study; however, at GRSM we applied two further constraints that (1) all sites must be located below 1,400 m to ensure access to trees of the target species, and (2) sites were located within 250 m of a designated trail to facilitate site access. One hundred and twenty sites were initially identified (60 at northern parks (PRWI, HAFE, and CATO) and 60 at GRSM), but because of access and the availability of the target species only 113 were ultimately included in the study.
Tree cores
At each site, the two largest specimens (occupying the largest portion of the canopy) of three species were targeted for coring: Quercus alba (QUAL), Quercus rubra (QURU) or Liriodendron tulipifera (LITU). If only one of the three species was present, two trees of the same species were cored. Across all 113 sites, 222 trees were cored using a 5.15-mm diameter increment borer at ∼1.4 m (breast height) above the forest floor. Two cores positioned at right angles to each other were extracted and stored in paper straws. The species and diameter at breast height (DBH) were recorded.
Core analyses
The cores were dried at 60 degrees C for at least 48 hours, sanded, and then scanned at 472 dots per centimeter. The images were analyzed for ring width using Cybis CDendro Software (Saltsjobaden) and ages assigned to each ring following close visual examination of the increment core. To ensure data quality, cores were visually cross-dated from the same tree and trees from the same site using plots of normalized ring widths and, for a subset of cores, a second researcher collected ring coordinates. One core from each tree was sectioned into annual increments using a razor blade.
Approximately 1 mg of wood from each increment from each core was analyzed for δ13C using a Carlo Erba NC2500 elemental analyser (CE Instruments) interfaced with a ThermoFinnigan Delta V+ isotope ratio mass spectrometer (IRMS). For δ13C analysis, a MgClO4 trap was used to remove water vapor before the transfer of sample gases to the IRMS.
Approximately 10 mg of wood from alternate increments was analyzed for δ15N using the same instruments. Alternate increments were sampled to limit any effects of nitrogen mobility that might have occurred between adjacent tree rings. A Carbosorb trap was used to remove CO2 in advance of removing water vapor with MgClO4 before the transfer of sample gases to the IRMS.
The δ13C and δ15N data were normalized to the VPDB and AIR scales, respectively, using a two-point normalization curve with internal standards calibrated against USGS40 and USGS41. Analytical precision (1σ) of an internal wood standard analyzed alongside samples was 0.1‰ for δ13C and 0.3‰ for δ15N. Carbon isotopes were normalized for trends in atmospheric δ13C resulting from fossil fuel use to arrive at carbon isotope discrimination against 13C (Δ13C).
As a metric of wood production, the Basal area increment (BAI) was calculated from the measured DBH and annual ring widths. The DBH was initially calculated for each year over the past 30 years by subtracting annual ring widths (×2) from the DBH at the time of coring (DBHt). To account for bark thickness 1 cm was subtracted from the measured DBH. The BAIt was then calculated for each year as follows:
BAIt = π[DBHt/2]2 - π[DBH t-1/2] 2
Data Access
These data are available through the Oak Ridge National Laboratory (ORNL) Distributed Active Archive Center (DAAC).
Landsat-based Phenology and Tree Ring Characterization, Eastern US Forests, 1984-2013
Contact for Data Center Access Information:
- E-mail: uso@daac.ornl.gov
- Telephone: +1 (865) 241-3952
References
Elmore, A. J., Nelson, D.M., and Craine, J.M. Earlier springs are causing reduced nitrogen availability in North American eastern deciduous forests, Nature Plants (2016). http://dx.doi.org/10.1038/nplants.2016.133
Elmore, A. J., Guinn, S. M., Minsley, B. J. & Richardson, A. D. Landscape controls on the timing of spring, autumn, and growing season length in mid-Atlantic forests. Glob. Change Biol. (2012). 18, 656–674. http://dx.doi.org/10.1111/j.1365-2486.2011.02521.x
Melaas, E. K., Friedl, M. A. and Zhu, Z. Detecting interannual variation in deciduous broadleaf forest phenology using Landsat TM/ETM plus data. Remote Sens. Environ. (2013) 132, 176–185. http://dx.doi.org/10.1016/j.rse.2013.01.011
Melaas, E. K., Friedl, M. A. and Richardson, A.D. Multiscale modeling of spring phenology across Deciduous Forests in the Eastern United States. Global Change Biology (2016) 22, 792–805. http://dx.doi.org/10.1111/gcb.13122