sentinel 2 science for rangeland applications: a case...
TRANSCRIPT
Sentinel 2 Science for Rangeland Applications: A Case
Study from Alberta Canada
Michael J. Hill1 and Anne M. Smith2
2Earth System Science and Policy, University of North Dakota, Grand Forks, ND, USA
2Agriculture and Agri-Food Canada, Lethbridge, Alberta, Canada
Sentinel 2 for Science Workshop
ESA/ESRIN, Frascati. Italy, May 19-21 2014
This research is supported by a grant
from the Canadian Space Agency
Rangelands
Southern
Alberta
http://www.webpages.uidaho.edu/what-is-range/rangelands_map.htm
“Rangelands produce a wide variety of goods and services desired by society, including livestock forage,
wildlife habitat, water, mineral resources, wood products, wildland recreation, open space and natural beauty.”
Grazed Lands
http://www.uni-klu.ac.at/socec/inhalt/1189.htm Erb, K.-H., Gaube, V., Krausmann, F., Plutzar, C., Bondeau, A., H. Haberl. 2007. A comprehensive
global 5min resolution land-use dataset for the year 2000 consistent with national census data. Journal
of Land Use Science 2(3), 191-224.
Rangeland Concepts – State and Transition
Sentinel2
Spectral
Spatial
Temporal
Vegetation
Indices
Rangeland Concepts – Landscape Health
Landscape
Health
Better and stable
Stress
Poorer and irreversible
Vigor
Organization
Resilience
Ecosystem
service delivery
Sources: http://nmnh.typepad.com/rogers_archaeology_lab/weblogs/
http://oregonstate.edu/instruct/anth481/ectop/ecservices.html; http://www.ranchadvisory.com/about_us/news/
http://www.intechopen.com/books/advances-in-landscape-architecture/understanding-landscape-structure-using-landscape-metrics
How well does
the landscape
function?
State
change
Landscape Health from Remote Sensing
Vigor
Organization
Resilience
Measures Criteria
Biomass, fractional cover, GPP
Composition, 3-dimensional structure
Time series, temporal comparisons
Indicators
Pigment indices, VNIR/SWIR
index relationships
LiDAR, patch delineation,
plant functional type
classification, fractional
cover unmixing, clumping
Temporal stability of signal,
pixel distribution shapes and
metrics, time series trends and
decompositions
Description of Landscape States
Indi
cato
rs o
f lan
dsca
pe h
ealth
State Change in Managed Rangelands
- changes in metrics of suites of VIs linked to biophysical properties
Baseline
Time-of-season specific
Climatic oscillation
Management issue – fire/grazing
State change
Minimum detection threshold
Based on multiple metrics
Includes persistence of trend
Hill (2013). Remote Sensing of Environment, 137, 94 –111
Sentinel 2
Specifically, Sentinel 2 provides:
– a temporal frequency that captures serially dependent
processes of degradation and regeneration;
– high spatial resolution to resolve finer patch structures
vital to the process of water harvest;
– high spatial resolution to improve retrieval of bare soil as
an indicator of degradation;
– spectral resolution in pigment wavelengths to improve
discrimination of community structure and detection of
invasive species
– Landsat-match bands for synergies from enhanced
temporal coverage at fine spatial resolution.
350 M ha North America
25 M ha Western Canada
Canadian Grasslands – Southern Alberta
Canadian Grasslands – Southern Alberta
Grassland Cultivated crop
C emission 26 Mg ha-1
C sequestration
0.5 Mg ha-1 year-1
Photographs courtesy of
G. Bourgeois, M. Didkowsky,
R. Bourchier, J. Nicholsen, G. Larson, C. Kloppenburg.
Image courtesy of Barry Adams, Alberta Environment and Sustainable Resource Development
Native Prairie Vegetation Inventory
1990-1993
1:30,000 aerial photography
¼ section based
% ground cover of tree, shrub,
graminoid, riparian, lake or wetland
Grassland Vegetation Inventory
colour infrared digital airborne imagery 0.5 m
Comprehensive biophysical, anthropogenic and
land use inventory
polygon-based, 5.0 ha for upland site types and
1.0 ha for wetland site types.
Accuracy estimated at >90%
Grassland
Vegetation
Inventory
Site Types – Complex units for assessment
Predominantly loamy
Cropland
No site type > 65%
of polygon
Dryland pasture
or hayfields
Updated GVI
2007 SPOT Image 2006 GVI
GVI updates through satellite imagery?
Database
Record
Add/edit/delete
Slides courtesy of Barry
Adams AESRD
Native Grassland, Hyperion Sites, Vegetation Indices
Sensitivity Hyperspectral Broad and Narrow-Band
Vegetation Indices Photosynthetic
activity (PS)
NDVI (Normalized Difference Vegetation Index)
PS EVI (Enhanced Vegetation Index)
PS RENDVI (Red Edge Normalized Difference Vegetation Index)
PS CRI1 (Carotenoid Reflectance Index I)
PS/stress ARI1 (Anthocyanin Reflectance Index I)
PS/stress RGR (Red-Green Ratio)
Water NDII (Normalized Difference Infrared Index)
Senescence/Soil PSRI (Plant Senescence Reflectance Index)
Senescence/Soil SWIR32 (Short Wave Infrared Reflectance Ratio
Biomass SATVI (Soil Adjusted Total Vegetation Index)
Late season Hyperion used to simulate Sentinel 2
Site Hyperion
Cressday, Alberta September 5, 2009
Aden, Alberta September 8, 2009
Twin River, Alberta August 8, 2009
Taber, Alberta August 21, 2009
Some indices are
approximate due to
mix of broad and
narrow bands
Characterizing Site Type – Photosynthetic Activity
ADEN TABER NDVI
EVI
RENDVI
RGR
Dry site – Aden (Sentinel 2-specific VIs)
NDVI
ARI1
CRI1
RENDVI
Mesic Site – Taber (VIs sensitive to water, soil,
senescence, biomass)
Converted polygon
Partially converted
NDII
PSRI
SWIR32
SATVI
Indicator Maps PSRI
RGR SWIR32 CRI1
RENDVI NDII SATVI
Relative importance of
indicators will vary between
- dry and wet sites,
- over seasons, and
- with different land cover
types and phenologies.
19
Conclusions
Follow-up work is expanding analysis for CSA project
– using Landsat archive and Landsat 8 data, as well as additional Hyperion for
Sentinel 2 simulation
– developing routine methods for GVI change detection and range monitoring
The Group on Earth Observations (GEO) and its GLobal
Agricultural Monitoring (GEOGLAM) initiative
– collaborative effort across international and institutional communities to
develop global systems for monitoring condition and measuring productivity
in pastures and rangelands (RAPP: Rangelands and Pasture Productivity).
Sentinel 2 provides:
– important enhancements in frequency of coverage, spectral bands and
synergies with Landsat
– enable more detailed and intensive monitoring and implementation of recent
conceptual and methodological developments across the global rangelands