Evaluating Potential of MODIS-based Indices in Determining “Snow Gone” Stage over Forest-dominant Regions

Abstract

"Snow gone" (SGN) stage is one of the critical variables that describe the start of the official forest fire season in the Canadian Province of Alberta. In this paper, our objective is to evaluate the potential of MODIS-based indices for determining the SGN stage. Those included: (i) enhanced vegetation index (EVI), (ii) normalized difference water index (NDWI) using the shortwave infrared (SWIR) spectral bands centered at 1.64 µm (NDWI 1.64µm ) and at 2.13 µm (NDWI 2.13µm ), and (iii) normalized difference snow index (NDSI). These were calculated using the 500 m 8-day gridded MODIS-based composites of surface reflectance data (i.e., MOD09A1 v.005) for the period 2006-08. We performed a qualitative evaluation of these indices over two forest fire prone natural subregions in Alberta (i.e., central mixedwood and lower boreal highlands). In the process, we generated and compared the natural subregion-specific lookout tower sites average: (i) temporal trends for each of the indices, and (ii) SGN stage using the ground-based observations available from Alberta Sustainable Resource Development. The EVI-values were found to have large uncertainty at the onset of the spring and unable to predict the SGN stages precisely. In terms of NDSI, it showed earlier prediction capabilities. On the contrary, both of the NDWI's showed distinct pattern (i.e., reached a minimum value before started to increase again during the spring) in relation to observed SGN stages. Thus further analysis was carried out to determine the best predictor by comparing the NDWI's predicted SGN stages with the ground-based observations at all of the individual lookout tower sites OPEN ACCESSRemote Sens. 2010, 2 1349(approximately 120 in total) across the study area. It revealed that NDWI 2.13µm demonstrated better prediction capabilities (i.e., on an average approximately 90% of the observations fell within ±2 periods or ±16 days of deviation) in comparison to NDWI 1.64µm (i.e., on an average approximately 73% of the observations fell within ±2 periods or ±16 days of deviation).

Evaluation of the Plant Phenology Index (PPI), NDVI and EVI for Start-of-Season Trend Analysis of the Northern Hemisphere Boreal Zone

“…The low r-values reported here indicate very limited to no sensitivity of EVI and NDVI to variability in ground observed SOS for areas characterized by winter snow. This result severely challenges the utility of NDVI/EVI for phenology assessment and change studies in high-latitude regions as such VI SOS estimates are likely to coincide with the disappearance of snow as also discussed in [11,35,42,50] unless careful data screening and preprocessing are done to avoid the impact from snow [28].…”

“…The sources of uncertainty are mainly attributed to insufficient methods for overcoming the presence of snow that significantly affects VI signal [16,[40][41][42][43], as well as difficulties in capturing subtle seasonal variation in canopy greenness of evergreen forests [15,[44][45][46][47].…”

Remote sensing-based determination of understory grass greening stage over boreal forest

“…During the validation phase, we observed relatively high deviations (i.e., > AE 2 periods) between the MODIS-predicted and ground-based observations for ∼6% and ∼35% of the incidents using AGDD and NDWI thresholds, respectively. These might be related to one or a combination of the following causes: (1) due to the use of visual observations in collecting the understory GGS records, it would be possible to have variation from one to another operator; 15 (2) a single threshold for the predictor of interest would not be suitable in capturing the spatial dynamics over the study area 18,24 ; and (3) in some instances, the spatial resolution of the MODIS-based predictions might not commensurate with that of the ground-based observations.…”

“…The lookout tower operators recorded these data on a daily basis and reported in the form of day of year (DOY: 1 to 365 or 366 depending on the leap year); thus, we transformed them into 8-day period to align with 8-day composites of MODIS-based data using the following equation described in Ref. 15: where P (¼1 to 46) is the period of MODIS-based 8-day composites throughout the year. Note that the value of P should always be an integer, e.g., P ¼ 20 if Eq.…”

“…In order to accomplish this, we chose to employ the normalized difference water index (NDWI: a measure of water content in the vegetation), because it has demonstrated better prediction capability in determining the vegetation green up stages independent of the snow conditions (i.e., common over boreal forested regions). 14,15 As climatic variables (in particular to temperature) largely control vegetation growth, remote sensing-based surface temperature regimes have also been used in vegetation phenological studies. For example: (1) Zhang et al 16 determined the vegetation green-up onset using MODIS-based surface temperature (T S ) and vegetation indices (i.e., NDVI and EVI) over northern mid-to-high latitudes (i.e., between 35°N and 70°N); (2) Hanes and Schwartz 17 modeled deciduous phenology using MODIS-based T S and leaf area index in the Chequamegon National Forest in northern Wisconsin, USA; and (3) Hassan and Rahman 18 determined deciduous phenology using MODIS-based T S and EVI-images over boreal-dominant regions in Alberta, Canada.…”