Eric B. Peterson, Ph.D.

1
Invasion of theAnnual Grasses
Mapping Annual Grasses

• Eric B. Peterson, Ph.D.
• Vegetation Ecologist
• Nevada Natural Heritage Program

Eric Peterson Nevada Natural Heritage Program
2
Vegetation Classification
3
Vegetation Condition
4
(No Transcript)
5
(No Transcript)
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
Training Data Percent Cover
29
11
Invasive Grass Mapping
Statewide 2006
Owyhee Uplands 2007
Pilot 2003
12
Use of the Maps
Landscape Condition or Ecological
Integrity Fire Planning Bad spots,
especially flammable How fast may fires
travel? Good spots, less flammable Conservati
on Planning Restore/Rehabilitate bad spots
greenstrips too? Protect the good spots!
ANGRIN
13
The basic howtouse seasonality (phenology)
Typical in Northern Nevada for Bromus tectorum
(cheatgrass)
Fall Winter Spring Summer
N
D
O
A
M
J
F
M
A
S
J
J
NDVI greenness, based on chlorophyll
reflectance in infrared relative to red
?NDVI a drop in greenness corresponds do the
abundance of annual grasses
So, 1. Sample training data points (field plots)
for annual grasses 2. Obtain imagery and extract
reflectance information for the sampled points 3.
Build statistical models using ?NDVI and other
parameters that improve accuracy 4. Use those
models to estimate annual grasses throughout the
satellite images 5. Assess accuracy.
14
Training Data 0.1 ha Plots
15
Training Data Percent Cover
29
16
Training Data Broad Gradients
412 Plots
17
Imagery challenges for large expanses

• Main imagery from Landsat 5
• 8 scenes to cover the Owyhee Uplands
• Two seasons makes 16 images
• Timing crucial!
• Cloud gaps
• covered by MODIS

18
Statistical Modeling

• Extract imagery and accessory data for the plots
  we have measured
• Use that for regression of annual grass cover
  with ?NDVI and accessory data

19
Accessory Data not just elevation
Elevation
Minimum Temperature
Average Annual PPT
others...
Heat Index
20
Data Model
LANDSAT (n 407) Parameter Coef.
Std. Error z p Intercept
54.2022 17.8060 3.044 2.33e-03 NDVI-E
22.2737 30.4038 0.733 4.64e-01 ?NDVI
111.3461 26.9516 4.131
3.61e-05 TMIN 0.0572 0.0147
3.905 9.42e-05 LL1 -0.7137
0.1820 -3.921 8.83e-05 ?NDVITMIN 0.3225
0.1554 2.076 3.79e-02 MODIS (n
412) Parameter Coef. Std. Error
z p Intercept -9.4831 16.1039
-0.589 5.56e-01 NDVI-E 39.9413
34.4409 1.160 2.46e-01 ?NDVI
58.8408 24.4824 2.403 1.62e-02 TMIN
0.0624 0.0140 4.457 8.29e-06 MLA1
-125.1069 141.9871 -0.881
3.78e-01 ?NDVITMIN 0.3416 0.1243
2.749 5.97e-03
21
Use Modelfor Map
22
Accuracy Assessment
SWReGAP n828 Shrubmap (Oregon) n662 Shrubmap
(Idaho) n276 Circular n412
23
Accuracy Assessment
Data n RMSE Correlation Correlation Correlation Percentile Percentile Percentile Percentile
Data n RMSE R Tau Rho 50 75 75 95
SWReGAP 828 9.7 0.33 0.36 0.39 0 2 2 20
Shrubmap OR 662 14.8 0.28 0.19 0.24 6 14 14 30
Shrubmap ID 276 22.3 0.60 0.40 0.50 11 25 52 52
Circular 412 16.2 0.55 0.45 0.56 3 13 35 35

24
Ultimately,accuracy is evaluated by biologists
with experience in the field
25
The map
Accuracy as estimatedfrom REGAP plots
RMSE 10.33 75 Conf. 5 95 Conf. 21

ANGRIN
26

• Thanks!
• Funding
• Nevada Biodiversity Initiative
• US Fish Wildlife
• BLM
• NV Dept Wildlife
• Carson Valley Chucker Assoc.
• Assistance
• Aaron Bruney
• David Hargrave
• NNHP staff !