Transcript Cushing_Grasslands_QAQC - LTER Information Management

Grasslands ANPP Data Integration*
JRN, SEV, SGS
1. Project History & BG (ANPP for Grasslands)
2. Data Integration to date
Issues, Questions
3. Data Analysis to date
Issues, Questions
4. ASM Workshop Feedback
Next Steps (May – June, 2007)
5. QA/QC Wish List
* Judy Cushing, Ken Ramsey, Nicole Kaplan, Kristin Vanderbilt
Lee Zeman, Carri Le Roy, Anne Fiala
Judith Kruger, Alan Knapp, Dan Milchunas, Esteban Muldavin
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
1
Project History
Are DataBank concepts transferable beyond the canopy?
Can database components help the IMs?
1.
Luquilla (Eda) – data visualization
2.
Cross site analysis of NPP (JRN, SEV, SGS).
•
•
Compare production & species richness,
using g/m2 per species per
quadrat & number of species per quadrat.
Compare biomass over areas of ecological interest using measures of
central tendency (mean, median, mode, and standard deviation) of g/m2 over
biomes at each site.
Original Goals (Eco-informatics/CS)
•
•
•
•
•
Published ecology data integration case study
Proof of concept for DataBank integration
Use of CLIO for ecology data integration
Example of data integration and use of site databases at LTER
Sample ontology for data integration
Adjusted Goals (Ecology): to know we have done it ‘right’
…something of value to the ecologists….
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
2
Grasslands Biomass
Data Integration Schema
unit
LTER
site
veg
zone
m
LTER
subsite
m
m
AANPP m
(weight)
species
m
m
1
1
Site1LocationMap
Site1LocationMap
Site1LocationMap
m
m
date
location
location
season
Year?
m
m
TRT?
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
3
Scientific Background
Modeling Annual Aboveground Productivity
in Grasslands
Soil &
veg type
Data Inputs
Measurements of
Biomass
Precipitation
Wind Speed
Radiation
Satellite imagery
Computational
Flux Tower array
Model
Plot level harvest
Soil Type
Parameter: Biome Type
Productivity or
Carbon Flow
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
Productivity
4
Methods for Above Ground NPP
(Collection of Productivity Data)
Satellite Imagery
10 – 100 km
Plot Level
Harvest
Flux Tower
100 ha
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
.25 m2
5
Collection Methods for Above Ground Net
Primary Productivity at SGS
Site: SGS
Sampling Design:
6 Sub-Sites: esa, swale, mid-slope, ridge, section 25, and owl creek
3 plots: (called transects) at each sub-site
5 sub-plots: (called plots) at each plot
Total of 90 ¼ m2 sub-plots harvested
Plot Level
Harvest
Harvest Methods:
Clip at crown-level, except for shrubs.
Plots are clipped by species.
.25 m2
Drying oven at a temperature of 55 C and weighed in the lab
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
6
Our Test-Case Integration
What’s in the integrated database?
• Aboveground net
primary productivity,
measured or calculated
in autumn.
• Three LTERs:
Sevilleta, Jornada, SGS
NPP observation
year
species
weight
plot
plot
area?
study site
• NPP by species by plot
species
• Data from grasslands
family
c. path
form
com. name
sci. name
only: nothing from
Sevilleta’s PinonStudy site
Juniper woodland
LTER
easting
northing
elevation
Vegetation type
• Contextual information
on species and plots.
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
7
Size of the integrated database
• 1093 species
Number of plots
• 44080 NPP
measurements
• 1065 plots
Se ville ta
• Covers 1989 - 2004 Jornada
90
240
735
SGS
1989
1990
1991
1992
1993
1994
1995
1996
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
1997
1998
1999
2000
2001
2002
2003
2004
8
Database Creation
NPP observation
plot
year
species
weight
plot
Area
study site
Study site
Published LTER NPP data
LTER
easting
northing
elevation
Vegetation type
Conversation with IMs
species
family
c. path
form
com. name
sci. name
Published LTER site metadata
(species list, study protocol)
Conversation with ecologists
USDA PLANTS
database
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
9
Integrating Dominant Vegetation Type
The three LTERs have overlapping vegetation types
Try: cross-site comparison of
productivity by equivalent vegetation type.
Jornada
Sevilleta
Blue grama grassland
Creosote bush scrub
SGS
Grassland
Larrea core
Grassland Black grama grassland
Tarbush flats
Mesquite dunes
Playa
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
10
Integrating growth forms
A lowest-common-denominator classification
for growth forms across 3 LTERs.
Sevilleta
Jornada
SGS
Tree
Succulent
Integrated
Tree
Leaf succulent
Succulent
Succulent
Shrub
Shrub
Stem succulent
Shrub
shrub
Sub-shrub
Sub-shrub
Herb
Fern
Sub-shrub
Forb
Forb
Forb
Sedge
Herbaceous vine
Grass
Herb
Grass
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
Grass
Grass
11
JRN: 203
Integrating species
SEV: 660
JRN + SEV: 126
JRN + SEV + SGS: 11
SGS + SEV: 14
JRN + SGS: 5
•
SGS: 41
1997
1998
1999
Gilia mexicana
Gilia
flavocincta
•
•
•
G. flavocincta
Integrating species very difficult
156 species found in > 1 LTER
Species are constantly reclassified,
so a timeline was constructed using
author and reference.
USDA Plants used to fill in missing
species information.
1999
2001
2002
G.
mexican
a
G. sinuata
Gilia
opthamoides
Gilia sinuata
Gilia sinuata
Gilia flavocincta
GiliaLTER
mexicana
Cushing;
ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
12
Sample analysis – by family
160
140
120
100
80
60
40
20
0
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
ER
A
ST
FA
BA
C
ZY
G
O
P
M
A
LV
A
PO
R
TU
PO
A
C
E
SGS
Sevilleta
Jornada
13
Issues 1: Species Codes
•
•
•
•
Codes are site specific … ACNE Acacia Neovernicosa or
Acalypha Neomexicana”
Over time, species differentiate : Bothriochloa saccaroides 
Bothriochloa laguroides,
LTER sites update at different times.
Some LTERs use subspecies & varieties; some do not
distinguish below species level.
We integrated species across two dimensions.
1. updated older data with newer species codes using “authority” (author,
publication). Jornada’s species database change log listed date of switch
to a new “authority” – it was a BIG help!
2.
The three separate updated species lists were merged with the official
USDA species list.
For species diversity queries, we’re treating all subspecies as a single species.
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
14
Issues (cont)
2. Form. Different LTERs use different categories as
forms, e.g., all non-woody leafy herbs might be
classed as forb, or separated into herbacious vine
and herb….
3. Timing. Biomass was measured at different times
of year. We took only fall measurements, but….
4. Plot Organization, Size. We did not combine
hierarchies, but just used data at plot level.
5. Site types. Each research area is classified as a site
type, but different terms are used, e.g.,
JRN Grassland = SEV black gramma
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
15
Analysis Questions
Which analyses OK if missing data for one site for one year?
Surprizing result:
JRN ANPP is higher
Not surprizing result:
SEV has highest species diversity.
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
16
Analysis Summary
JRN, SEV, SGS Plant Communities
•Grassland LTER Synthesis 1999, Knapp&Smith 2001
•Average differences by LTER site and dominant vegetation type
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
17
All photos shamelessly taken from various websites
Analysis
• NPP – Total net primary productivity in a
1m2 plot
• Species richness: Number of different
species present in a 1m2 plot
• Community analyses weighted by NPP or
by species Presence/Absence
• Indicator Species Analysis
• Correlations with Environmental Variables
– still organizing data
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
18
Grassland Community Analysis: Ordination
Significant differences among LTER sites
site
LTER Site
1
2Jornada
3 SGS
Axis 3
 Sevilleta
Based on
Presence/
Absence
A = 0.0945
P < 0.0001
Axis 2
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
19
Indicator Species Analysis
23 Indicator Species for Jornada LTER
29 Indicator Species for Shortgrass Steppe LTER
32 Indicator Species for Sevilleta LTER
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
20
Extensions
May and June Workshops
•
•
•
•
•
•
•
Add more years, look at trends through time
Add more sites: KNZ, two S. African sites, ….
Move to a “big-iron” database….
Compute biomass by species (are these data available?)
Compute Presence/Absence by species (except SGS?)
Do cover-based ordinations (except SGS?)
Correlate ANPP with env. variables: precip, temp, soil texture,
soil type, elevation, AET, soil moisture, PAR, soil temp
Identify standard analyses (derived data), ala Trends?
INVESTIGATE & DOCUMENT THE “CAN’T DOs”:
• Relative frequency, diversity, species abundances, species
richness (based on SGS methods)
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
21
Revisit Simplifying Assumptions
Differences in data collection or methodologies ….
Differences in ANPP calculation ….
data result from regressions particular to each site.
Differences in Plot Size
ANPP probably scales up….
Species Richness (#species per plot) probably doesn’t….
Differences in plot designation ….
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
22
QA/QC* Wish List
How do we automate integration and mark-up (even a little) ?
• Our integration done by hand… not feasible….
• Tracking was ad hoc….
How do we track and distribute changes to data?
• Species and species family changes (SEEK?)
• Assignments to Form
How do we document differences among data:
• Methodology and plot differences, e.g., Sub-plot based analysis is below
the scale of interest. Statistical n becomes 3 or 5, typical of ecological
data, but ok once all years of data are analyzed
How do we determine (and fix) critical ecology issues:
• Under- or over-estimates, e.g., “SGS 14-17% under-estimate of cool
seasons based on C14 data.”
* Las Cruces, Jan 31-Feb 1, 2007.
Cushing; LTER ASM QA-QC Las Cruces, Jan 31-Feb 1, 2007
23