Transcript Estimating Foliar N in Human Dominated Landscapes

By: Paul A. Pellissier, Scott V. Ollinger, Lucie C. Lepine
Field Collection:
Components of Net Primary Production?
•In many terrestrial ecosystems Nitrogen (N) is central to, and
often limits, net primary production (NPP) of terrestrial plants.
•In cultivated grasslands N is often supplied, by way of fertilizer,
to achieve management goals. This may lead to other limiting
factors such as available water.
Photo 2: Collecting field measurements of
hayfield canopy reflectance with a portable
field spectrometer at the UNH Fairchild Dairy,
Durham, NH. (Photo by L. Lepine)
60%
Site Description-Lamprey River Watershed:
1.9
VIP Score
Predicted
3.00
2.50
2.00
Photosynthetic
Capacity
Prediction and importance
plots based on PLS regression
of 2 extracted factors. Root
mean PRESS=0.589011
PLS regression was preformed
using 8 modeled light response
curves incorporating 80
instantaneous photosynthetic
assimilation readings (plotted)
Importance plot shows
importance of short wave
infrared, red edge, and green
regions
24
y = 0.9423x + 0.1166
R² = 0.9426
1.3
0.7
RMSE=0.1
1.50
2.00
2.50
Actual
3.00
3.50
0.5
4.00
300 500 700 900 1100 1300 1500 1700 1900 2100 2300
Wavelength (nm)
1.8
26
1.6
1.4
VIP Score
22
20
18
1.2
1
16
0.8
14
y = 0.8981x + 2.0412
R² = 0.8981
12
RMSE=1.48
24
26
12
14
16
18
20
Actual
22
0.6
28
Dry Biomass
Prediction and importance
plots based on PLS regression
of 4 extracted factors . Sample
size, n=29; Root mean
PRESS=0.93348. Importance
plot shows the heavy
influence throughout the
visible wavelengths including
blue, green and red edge
regions.
1.5
0.9
1.00
1.00
28
1.7
1.1
1.50
300
500
700
900 1100 1300 1500 1700 1900 2100 2300
Wavelength (nm)
300
500
700
900 1100 1300 1500 1700 1900 2100 2300
Wavelength (nm)
300
500
700
900 1100 1300 1500 1700 1900 2100 2300
Wavelength (nm)
2
800
1.8
700
1.6
600
500
400
1.4
1.2
300
1
200
y = 0.6124x + 102.87
R² = 0.6124
RMSE=99.93
100
0.8
0
0
200
400
-100
600
800
1000
0.6
1200
Actual
1.6
Foliar Water
Content
Pasture
40%
Reflectance
2.1
Grass Canopy Reflectance by Management Practice
50%
Lamprey River Watershed
3.50
2.3
--Grass is More Than Just Green--
--Methods-• Fifth order river located in
southern New Hampshire
• Area: 479 square kilometers
• Encompasses nine towns
• Population density ranges from
0 to 630 people km-2,with an
average of 129.
• 17.5% of land area is nonforested
• Surface waters currently
impaired from excess N
Prediction and importance
plots based on Partial Least
Squares (PLS) regression of 11
extracted factors . Sample size,
n=27. Root mean PRESS=
0.73447. The Importance plot
shows the heavy influence of
chlorophyll absorption in the
blue region and reflectance in
the green region. The PLS
analysis also values the red
edge, and a possible water
absorption feature near
1300nm.
2.5
Hayfield
30%
Residental
Prediction and importance
plots based on PLS regression
of 5 extracted factors . Sample
size, n=29; Root mean
PRESS=1.02361. Regions of
influence include the blue
green transition zone,
chlorophyll absorption well,
red edge, and NIR plateau.
6
1.5
1.4
5
1.3
4
VIP Score
•Understanding how these systems function in terms of N
cycling and carbon storage at the landscape level is important in
quantifying their environmental impacts at the regional to global
scale.
Foliar %N
4.00
Important Regions
VIP Score
--Background--
Spectral Prediction
Predicted
Photo 1: UNH wants to cut my grass? Collecting biomass at
one of our sites located in Deerfield, NH. (Photo by L. Lepine)
Attribute
Predicted
• The current study aims to determine generalizable
relationships between aboveground NPP and
spectral reflectance in both turf and agricultural
grasslands.
• Hyperspectral measurements of canopy reflectance
collected over two growing seasons are correlated
with foliar and environmental attributes.
• Relationships derived from ground-based canopy
reflectance will be applied to watershed-scale
airborne imagery.
Two summers, two aims:
The goal of summer 2012 was
to collect data critical to relating
airborne and field-based spectra,
whereas during summer 2013
efforts were focused on
determining relationships
between plant /environmental
conditions and ground based
spectra.
Data 2012
•Airborne imagery: approx.
500km2 , spatial resolution 5m2
•Eighteen field sites established
•Ground-based canopy
reflectance
•Canopy height
•Foliar biomass and N content
Data 2013
•Intensive ground based
reflectance, sub-meter spatial
resolution
•Canopy height
•Foliar biomass and N content
•Leaf water content
•Soil moisture
•Leaf mass per area
•Photosynthetic capacity
Predicted
--Abstract--
Spectral Relationships:
3
1.1
1
0.9
2
0.8
1
y = 0.4893x + 1.1938
R² = 0.4893
RMSE=0.61
20%
Fallow Field
1.2
0
0
1
2
3
Actual
4
5
6
0.7
0.6
10%
Looking Forward:
0%
Visible Light
Near Infrared
Short-wave Infrared
Wavelength (nm)
Figure 1: Site location within New
Hampshire.
Figure 2: Averaged spectral reflectance curves of grass canopies by management regime. Curves represent over 3500
individual spectra taken at 18 sites located within the Lamprey River watershed. Average spectral reflectance differs
significantly (P<0.003) for several regions ( 750-920nm*, 920-1150nm, 1151-1350nm**)in the NIR plateau . *Hay fields
and pasture P=0.0165. **Hay field and pasture P=0.6933.
Future efforts will focus using the relationships shown here to
predict NPP, scaling these predictions to the watershed scale,
and interpreting any observed spatial patterns in NPP.
Prepared by Paul A. Pellissier for the NSF EPSCoR National Conference, 4th November 2013