Kamel Didan1*, Tomoaki Miura4 , Armando Barreto1
*1 The University of Arizona, 2 University of Hawaii
* [email protected] - http://measures.arizona.edu
Vegetation Phenology and Vegetation Index Products from Multiple Long Term Satellite Data RecordsPhase I : Towards Continuity
Introduction
Changes in vegetation phenology depict an integrated
response to change in environmental factors and provide
valuable information to global change research. Typically,
remote sensing of vegetation phenology is based on the
analysis of vegetation index temporal profiles, because of
their simplicity, stability, and inherent resistant to noise. Most
phenology estimates are, however, limited to using one
sensor owing to the inter-sensor continuity challenges.
Although, phenology is used for a variety of research and
application topics, the central objective remains the study of
vegetation dynamics change in response to change in climate
and other factors. Consequently, the consistency and length
of data records are key requirements.
With satellite missions lasting few years only, long term
phenology measures will have to be based on a mixture of
satellite data records.
Objectives
In this MEASURES' project, our multi-institution team of
investigators will generate a seamless and consistent sensor
independent long term (30+ years) Earth Science Data
Records (ESDRs) of land surface phenology parameters and
vegetation index, by using measurements from different
satellite missions and sensors. We’ll validate, characterize
and establish the error and uncertainty of these products .
Project Overview
We’re using AVHRR, MODIS, VIIRS and VGT daily land
surface reflectance and we’re designing a series of
sensor independent algorithms to apply to these data.
Validation
In collaboration with the National Phenology Network
(USA-NPN) we plan to correlate these remote sensing based
ESDR estimates of VI and phenology with ground
observations. We seek to evaluate their consistency and
accuracy by comparing them with in situ growing season
phenophases observations over different biomes, latitudinal
and elevation gradients.
Project flow Applications & Distribution
Changes in phenology provide valuable information for global
change research, land degradation studies, integrated pest and
invasive species management, drought monitoring, wildfire
risk assessment, and agricultural production.
This project is expected to generate, document, and deliver
30+ years of consistent and well characterized ESDR/CDR
quality daily measurements of Vegetation Index and land
surface phenology metrics. These ESDR products will be
distributed through the LP-DAAC, and separately via a PI
based interactive visualization and analysis system.
We plan to monitor the growing season along elevational and biotic
gradients. We will also leverage from ongoing national and international
efforts, particularly the ones lead by the US-NPN
Conclusions
We ‘re currently building more complex cluster maps and
we plan to generate separate continuity equations for each
of these clusters. The goal is then to reconstruct a new VI
record using these continuity equations. The phenology
algorithm will be driven by this VI ESDR.
Acknowledgements: This work is supported by a
NASA CA# NNX08AT05A, PI Kamel Didan
Others
VIIRS ?
One time data acquisitionOn demand data acquisition Automatic data acquisition
VITO
Global at 0.05o
Phenology and VI
ConterminousUS at 1km
Global daily CMG LSR (and
possibly continental US @ 1km)
from different sensors and
time periods
Discard
Data preprocessing and Quality Analysis
Surface Reflectance
Data QA& Reliability
VI ESDR
Annual PhenologyOutput QA
& Reliability
Phenology ESDR
Daily sensor independent and gap
filled VIOutput QA & Reliability
Product Evaluation Error & Accuracy
Fe
ed
ba
ck
Data disseminationWorkshops and Users feedback
Use
r co
mm
un
ity f
ee
db
ack
into
th
e p
roce
ssin
g a
nd
d
istr
ibu
tio
n s
yste
m
Discard/Re-order
VI and Phenology Processing Facility
(MEaSUREs Lab., The Univ of Arizona)
AVHRR
MODAPS
DAAC
Task-1
Task-2
Task-3/4/5/6
Task-7
Task-8
Task-9
Fe
ed
ba
ck
Fe
ed
ba
ck
Phenology QA& Reliability
USGS for EROS
DAACPI Web
Initial Data checks
1981 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2020
NOAA-7, -9, -11 AVHRR
(~1:30pm)
NOAA-14 AVHRR
(1:30pm at launch, 5:00pm at the end)
NOAA-16 AVHRR
(2:00pm)
NOAA-17 AVHRR
(10:00am)
SPOT-4 VEGETATION
(10:30am)
SPOT-5 VEGETATION
(10:30am)
Terra MODIS
(10:30am)
Aqua MODIS
(1:30pm)
NPP VIIRS
(1:30pm)
NPOESS VIIRS
(9:30am and 1:30pm)
Jan 1995 Sep 2001
Mar 2001
Aug 2002
Feb 2000
Jun 2002
Apr 1998
May 2002
Apr. 2008 (Launch to be pushed back)
Sep 2009*
Feb 2010
Sep 2003**
Sep 2008*
Jul 2007*
Dec 2002
Apr 2007*
Jul 1981
Sep 1994
Not RecommendedData Gap
Prior to 2000, only AVHRR Post 2000 many newer sensors
with varying characteristics.
Phase I : Continuity – The Cluster method
The cluster method, assumes that phenology (i.e. vegetation
dynamic) is controlled by factors that capture climate, soil,
elevation gradient, sun- shade exposure, and biophysical
limitations. The concept of phenology cluster is similar to
the biotic life-zones used to classify ecosystems. In that
regard, a phenology cluster is also a biotic zone, with similar
plant species, at around the same elevation gradient, and
governed by similar temperature, precipitation and radiation
regimes.
For this initial continuity phase we constructed a global
CMG resolution cluster map to drive the continuity work:
• Global DEM GTOPO30 data
• FAO Global Average annual Temperature
• FAO Global Precipitation map
• USDA-NRCS Global soil
• MODIS CMG Land Cover
Issues with:
Sensor/Platforms
• Bandpass
• Spatial resolution
• Orbital characteristics
• Sensor degradation
Algorithms/Re/Processing
• Atmosphere correction
• Temporal compositing
• Spatial aggregation
• VI formula
San Juan Mountains, CO
Wavelength (nm)
400 500 600 700 800 900 1000 1100
Norm
aliz
ed
Spectr
al R
espo
nse
Terra MODIS
NOAA-14AVHRR
NOAA-16AVHRR
ASTER
ETM+
M2 M3 I1 I2
SPOT-4VEGETATION
Global Temperature Map
Global Land Cover Map
Global DEM Map
Global Homogeneous Cluster Map (2000+units). VI continuity will be
established separately for each cluster.
Application of the cluster method to VI continuity
Global Soils Map
Global Precipitation Map
AVHRR 80ies AVHRR 90ies AVHRR 00ies AVHRR full record MODIS 00
Across sensor continuity: Clustered NDVI Decadal comparisons.
Data from most clusters exhibited robust continuity. Clusters from the
Rainforest and Boreal forest exhibited differences. These differences are due
to band pass, MODIS saturation of the Red/NIR (narrower bands) and
processing differences.
Using the clusters from above, the VI record is separated into data classes
and simple polynomial continuity continuity could be applied to the records.