ReadMe for Seasonality of soil moisture mediates responses of ecosystem phenology to elevated
CO2 and warming in a semi-arid grassland. Authors: Zelikova, Tamara, Williams, David,
Hoenigman, Rhonda, Blumenthal, Dana, Morgan, Jack, Pendall, Elise
2015-6-15
This readme file describes the data files accompanying the above publication. Detailed methods
describing the experimental design and data collection can be found in the published paper, but
please contact tzelikov@uwyo.edu with any further questions.
The following files are included:
1) PHACE_Greenness_Dryad.xlsx
This data file contains greenness data for each experimental plot from 2006-2013. Briefly, repeat
digital plot photos taken at an overhead distance to measure 1m2 of the PHACE (Prairie Heating
and CO2 Enrichment) experimental plots from 2006-2013. The same exact location was
photographed each time. Greenness from digital photos was as the number of pixels that fall
within a range of values in the green spectrum divided by the total number of pixels in the photo
and multiplied by 100 to yield percent green (Matlab Code readme file
PHACEGreennessMatlabCode.txt and also included as a supplement with the publication).
Aboveground plant biomass in each PHACE plot was harvested mid-July annually. Photos taken
after the biomass harvest in July were masked using Photoshop (masking the non-harvested areas
and quantifying the remaining portions of the photo where biomass was harvested). To calculate
greenness in the post-harvest photos, we subtracted PercentGreenness(HarvestedOnly)/0.5m2
greenness from the total photo greenness (PercentGreenness/m2 column) and multiplied by 2 to
quantify what the greenness would have been in the entire photo if biomass was not harvested.
In two instances, at the very end of the growing season, the estimate of greenness in the
PercentGreenness(HarvestedOnly)/0.5m2 column was slightly greater than the overall greenness,
yielding a negative result when the values were subtracted. The slight difference in greenness
estimates is likely the result of running the Matlab code for the complete photo and the photo
with the non-harvested areas masked separately, with pixels aligning slightly differently on each
run. In those two instances, we replaced the negative PercentGreenness(NonHarvestedOnly)/m2
values with zeros. There were two instances when we did not have a photo for a particular plot on
a particular date. We did not include those dates in the analysis and the missing values indicate
“No Photo”.
Data Column Headings
PhotoDate – the date the photos were taken
DOY – calendar day of year
SampYr – year
PlotID – plot number
Block – experimental block
CO2 – “c” ambient, “C” elevated
Temperature - “t” ambient, “T” heated
TreatmentCode – combination of CO2 and Temperature treatments
PercentGreenness/m2 – percent green in the entire plot
PercentGreenness(HarvestedOnly)/0.5m2 - post biomass harvest in July, non-harvested areas of
the plot were masked using Photoshop and we quantified greenness in the remaining portions of
the photo where biomass was harvested.
PercentGreenness(NonHarvestedOnly)/m2 - subtracted PercentGreenness(HarvestedOnly)/0.5m2
greenness from the total photo greenness (PercentGreenness/m2 column) and multiplied by 2 to
quantify what the greenness would have been in the entire photo if biomass was not harvested.
2) PHACEGreennessMatlabCode.txt
This is the Matlab code used to quantify greenness from digital photographs. An explanation of
the purpose of the analysis and the method used is included in the publication “Seasonality of soil
moisture mediates responses of ecosystem phenology to elevated CO2 and warming in a semi-arid
grassland”. Authors: Zelikova, Tamara, Williams, David, Hoenigman, Rhonda, Blumenthal,
Dana, Morgan, Jack, Pendall, Elise.
Briefly, we converted each image into a data matrix using Matlab R2011a (The Math-Works,
Natick, Massachusetts) and the imread() command. Each element in the matrix represented a
pixel with a value for red, green, and blue weight of the pixel between 0 and 255. These values
were converted to an HSV (hue, saturation, value) scale for classification using the rgb2hsv()
command, which creates a value for hue, saturation, and value between 0 and 1 for each pixel. In
the Matlab HSV scale, the greenest hues have a value of approximately 0.25, in contrast to brown
hues with a value of approximately 0.09. The saturation HSV captures how much of the hue
exists in the color. In pixels with a high saturation, the visible appearance of the pixel will be the
hue, while pixels with a low saturation will appear grey. The HSV value parameter describes the
brightness of the pixel. High values indicate that the pixel is dark and low values indicate a light
pixel, such that a pixel with a value of 0 is white and a pixel with a value of 1 is black.
Using the HSV scale, we defined the upper and lower boundaries of HSV values for
“green” for a subset of images and applied these boundaries to the other images. All other pixels
were classified as not green. The HSV range for green was 0.16 - 0.50 for Hue, 0.075 - 1.0 for
Saturation, and 0.09 - 0.91 for Value. The percentage of greenness for each image was the
number of pixels identified as “green” divided by the total number of pixels in the image.
3) PHACE_SamplePoint_Dryad.xlsx
This data file contains plant cover data estimated using Sample Point (Booth et al. 2006), a free
software package that superimposes a grid of crosshairs over each image to facilitate manual user
classification of 225 focal image pixels. Using this method, we quantified the live and dead cover
of litter, soil, the dominant perennial graminoids, a sub-shrub, perennial C3 forbs, and other
grasses not already identified by species. When the crosshair intersected a shadow in the photo
and we could not reliably distinguish plant cover type, we marked it as “Shadow”. Plants that did
not fall into the above categories were identified as “Unknown”. Because biomass was clipped
annually in July, we present cover by species from the entire photo area through July of each year
and from August until October, the cover is scaled to include only non-harvested portions of the
plot. Blank cells indicate missing values and in 2012, we did not separately identify FORBdead
and GRASSdead.
Data Column Headings
PhotoDate – the date the photos were taken
DOY – calendar day of year
SampYr – year
PlotID – plot number
Block – experimental block
CO2 – “c” ambient, “C” elevated
Temperature - “t” ambient, “T” heated
TreatmentCode – combination of CO2 and Temperature treatments
Unknown - plants that did not clearly delineate into species or cover types
Litter – unidentifiable plant litter
Shadow - when the crosshair intersected a shadow in the photo and we could not reliably
distinguish plant cover type
PASM – Pascopyrum smithii
BOGR – Bouteloua gracilis
HECO - Hesperostipa comata
CAEL - Carex eleocharis
ARFR - Artemisia frigida
SPCO - Sphaeralcea coccinea
FORB – all other forbs not identified by species
GRASS – all other graminoids not identified by species
PASMdead - Pascopyrum smithii dead biomass
BOGRdead - Bouteloua gracilis dead biomas
HECOdead - Hesperostipa comata dead biomass
CAELdead - Carex eleocharis dead biomass
ARFRdead - Artemisia frigida dead biomass
SPCOdead - Sphaeralcea coccinea dead biomass
FORBdead – all other forbs not identified by species, dead biomass
GRASSdead – all other graminoids not identified by species, dead biomass