Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Application of Remote Sensing-based Indices in Determining "Snow Gone"Stage over Forest-dominant Regions
Quazi K. Hassan&
Navdeep S. Sekhon
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Outline
1. Introduction 2. Objectives
3. Study area and data used
4. Methodology
5. Results & discussion
6. Concluding remarks
7. Acknowledgements
8. Q & A
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
http://atlas.nrcan.gc.ca
Introduction (1)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Boreal Phenology• The phenological events of boreal forest are
broadly classified into the four categories namely:
i.snow stages;ii.understory stages;iii.deciduous stages; and iv.conifer stages.
Introduction (2)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
“Snow gone” (SGN) can be defined as the date when 25% or less of the area surrounding a site is covered by snow.
In practice, the SGN stage is critical for determining the onset of the forest fire season.
In Alberta, the Provincial Government acquires SGN stage at ~120 lookout tower sites across the landscape. It has two limitations:
(i) As it is based on visual observation, and thus the results potentially may vary from person to person; and
(ii) it fails to address spatial variability as the lookout tower network provides only point type information over spatial areas of a few hundreds of hectares.
Introduction (3)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
One option to address these concerns is to employ remote sensing-based techniques, which have already been proven as an effective method for delineating forestry-related variables at the landscape level.
Thus, we intend to explore the potential of remote sensing-based techniques for determining the SGN stage over the forest-dominant regions in Alberta.
Remote sensing is the acquisition of information about the earth surface in the form of images using electromagnetic radiation without having any physical contact.
In general, the reflective portion of the electromagnetic radiation (i.e., in between 0.4-3.0 µm) is used to understand SGN and other periodic changes.
Introduction (4)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
The physical characteristics of the vegetation cover (e.g., greenness, water content, reflectance etc.) are different during various phenological stages
Temporal trends of vegetation indices capturing a physical characteristic(s) of vegetation cover reflect the changes caused by phenological transformations
The most commonly used remote sensing-based indices in phenological studies are: Normalized difference vegetation index (NDVI) and enhanced
vegetation index (EVI); both are a measure of greenness Normalized difference water index (NDWI: a measure of water
content in the vegetation) Normalized difference snow index (NDSI: a measure of snow on
the ground)
Introduction (5)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
REDNIR
REDNIRNDVI
15.765.2
Re
Re
BluedNIR
dNIREVI
13.2
13.2 13.2
matSWIRNIR
matSWIRNIRmNDWI
m1.64at
m1.64at
SWIRGreen
SWIRGreenNDSI
is the reflectance from the respective spectral band
Modified from Zarco-Tejada et al. 2002
0.4 0.8 1.2 1.6 2.0 2.4Wavelength, µm
50
40
30
20
10
0
%
Ref
lect
ance
Blue Green Red NIR SWIR 0.469 0.555 0.645 0.858 1.24 1.64 2.13
Introduction (6)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
i. Perform a qualitative evaluation of four MODIS-based indices (i.e., EVI, NDWI1.64μm, NDWI2.13μm, and NDSI) in predicting the SGN stages;
ii. Compare the SGN values predicted by the efficient indices as determined in objective (i) with the observations available at lookout tower sites across the landscape; and
iii. Generate a SGN map using the best predictor as determined in objective (ii) to discuss the spatial variability over the entire Province of Alberta.
Objectives
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Central Mixedwood (25.4%)*Dry Mixedwood (13.43%)*Northern Mixedwood (3.5%)*Boreal Sub-Arctic (1.76%)*Peace-Athabasca Delta (0.83%)Lower Boreal Highlands (8.5%)*Alpine (2.03%)*Sub-Alpine (3.9%)*Montane (1.4%)*Upper Foothills (3.44%)*Lower Foothills (6.94%)*Athabasca Plain (2.04%)*Kazan Uplands (1.43%)*Foothills Parkland (0.57%)Peace River Parkland (0.48%)Central Parkland (8.1%)Dry Mixedgrass (6.99%)Foothills Fescue (2.12%)Northern Fescue (2.26%)Mixedgrass (3%)Upper Boreal Highlands (1.84%)*
Lookout tower sites in year 2006 (115)N
E
Study Area and Data Used (1)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Five hundred fifty two scenes of the MODIS-based 8-day composites of surface reflectance data (i.e., MOD09A1 v.005) at a 500 m resolution for the years 2006–2008 were obtained from NASA.
For a particular 8-day period, there were 4 scenes required to produce the entire extent of Alberta. Thus, for each of the years, there were forty six 8-day periodical images spanning from January 1 to December 31.
In addition, we also acquired ground-based observations of SGN at ~120 lookout tower sites during 2006-2008. There data were then converted into period to match with the satellite data as follows:
Study Area and Data Used (2)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
MODIS–based 8-day composites of surface reflectance at 500 m resolution
(i.e. MOD09A1); Four scenes per period
Natural subregion map of Alberta
Mosiacking of the four scenes
Calculating remote sensing-based indices of:
EVI NDSINDWI1.64µm NDWI2.13µm
Qualitative evaluation of the indices over the natural subregions of interest• Extracting the temporal dynamics for each of the indices at all of the lookout tower sites; • Generating natural subregion-specific average temporal dynamics; • Comparing with the average ground-based observations of SGN; and• Determining the efficient indices in predicting SGN
Averaging the ground observed SGN day for natural subregions of:(i) central mixwood, and(ii) lower boreal highlands.
Determining the best index in predicting SGN periods
Converting the DOY of SGN into no. of periods of MODIS-based indices
Generating the SGN map for the entire study area
Ground-based observations of “snow gone” (SGN) day at
lookout tower sites (in DOY)
Methodology
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Temporal dynamics of averaged values forEVI, NDWI1.64μm, NDWI2.13μm, and NDSI for central mixedwood(i.e., occupies ~25.5% of the province) during 2006–08. The average snow gone day from ground-based observations for the same area is shown bythe dotted vertical line.
Results & Discussion (1)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Temporal dynamics of averaged values forEVI, NDWI1.64μm, NDWI2.13μm, and NDSI for lower boreal highlands(i.e., occupies ~8.5% of the province) during 2006–08. The average snow gone day from ground-based observations for the same area is shown bythe dotted vertical line.
Results & Discussion (2)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Results & Discussion (3)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Results & Discussion (4)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
An example SGN map generated using NDWI2.13μm (the best predictor) for the year 2008.
It revealed that approximately 56% of the times the SGN stages fell in the rangeof 121–136 DOY.
Results & Discussion (5)
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Here we evaluated the potential of four MODIS-based indices (i.e., EVI, NDWI1.64μm, NDWI2.13μm, and NDSI) for determining SGN stages in Alberta.
A qualitative evaluation over two forest fire prone natural subregions demonstrated that both of the NDWI’s had better capabilities with
compare to EVI and NDSI.
Our quantitative analysis revealed that the NDWI2.13μm could better predict the SGN stages in comparison with NDWI1.64μm.
Thus, it will potentially be incorporated in the framework of forest fire management in the Province of Alberta.
Concluding Remarks
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Recommendation & Further Research
We have generated SGN maps at 500 m spatial resolution using 8-day composites of MODIS data. Thus, it would be important to enhance:
temporal resolution to 2-4 days; and spatial resolution to 250 m.
Despite good results, we recommend to quantify the applicability of the described approach before implementing over other biomes/regions in Canada or elsewhere in the world.
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Acknowledgements
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
References
Sekhon, N.S., Hassan, Q.K., and Sleep, R.W. 2010. Evaluating potential of MODIS-based indices in determining "snow gone" stage over forest-dominant regions. Remote Sensing, 2, 1348-1363.
Zarco –Tejada, P.J., Rueda, C.A., and Ustin, S.L. 2003. Water content estimation in vegetation with MODIS reflectance data and model inversion methods. Remote Sensing of Environment, 85, 109-124.
Earth Observation for Environment Laboratory, Dept. of Geomatics Engineering
Thank You.
Questions?