Technical Thematic Report No. 17. - Monitoring ecosystems remotely: a selection of trends measured from satellite observations of Canada
Average Dynamic Habitat Index, 2000 to 2006
The Canadian Dynamic Habitat Index (DHI) is a composite image of three indicators of vegetation dynamics. It is a relatively new remote sensing index based on a data set that begins in 2000, with potential uses for this index still being tested and refined. The DHI as presented here is currently a point in time measure; trends are not yet shown.
The three indicators that make up the Canadian DHI are: (1) cumulative annual greenness; (2) minimum annual fPAR; and (3) seasonal variation in greenness (Coops et al., 2008). All three indicators are derived from estimates of the fraction of photosynthetically-active radiation that is absorbed by the earth’s surface (fPAR), obtained from the MODIS sensors launched in 1999 and 2001 (Heinsch et al., 2006). fPAR ranges from 0 to 1 with higher values of fPAR averaged over the growing season corresponding to more densely vegetated productive landscapes and lower values of fPAR averaged over the growing season corresponding to less productive landscapes (Coops et al., 2008). The fPAR of areas covered in snow approaches zero, though areas of shadow and chlorophyll absorption by conifers will still contribute to a positive fPAR value.
Although fPAR is similar to NDVI, the estimation of fPAR is more directly related to plant physiology than NDVI as it is calculated from a physically based model of the propagation of light in plant canopies (Coops et al., 2008). fPAR also does not have the same issues with saturation at higher values as the NDVI (Coops et al., 2009b) (see Quality checks and limitations on page 19). Estimates of fPAR utilize a number of spectral bands from the MODIS sensor (up to 7) whereas the NDVI is based on two spectral bands, the red and near infrared (see NDVI on page 17).
The DHI was first developed in Australia (Mackey et al., 2004) and has been adjusted for use in Canada by Coops et al. (2008). Their results form the basis of this section and have been summarized by ecozone+ for the purposes of the ESTR.
Coops et al. (2008) obtained estimates of 1 km resolution monthly maximum fPAR values derived from data collected by MODIS sensors on the Terra and Aqua satellites from the NASA Earth Observing System from 2000 to 2005. This data is calibrated by NASA to take sun angle, background reflectance, and view angle into account, and is made available publicly. Monthly maximum fPAR values are used in order to minimize the influence of cloud and snow cover, atmospheric variation, and other confounding environmental conditions. From these values, Coops et al. (2008) calculated the three components of the DHI, for each year:
- Cumulative annual fPAR: the integrated annual fPAR, based on the monthly maximum fPAR values for the year. This has been interpreted as the cumulative annual greenness;
- Annual minimum fPAR: the minimum value out of the monthly maximum fPAR values in a year. This has been interpreted as the minimum annual green vegetation cover. However, this interpretation is not accurate in winter because of the effects of shadows, so we have avoided this interpretation;
- Annual coefficient of variation of fPAR: the standard deviation of the monthly maximum fPAR values in a year, divided by the mean of the monthly maximum fPAR values in a year. Interpreted as the annual degree of vegetation seasonality.
They then calculated an average over the six-year dataset for each of the three components and produced a map for each component and one composite map (Coops et al., 2008). Their results are further analysed by ecozone+ here. As the data represent a six-year average, it currently only provides an estimate of the status in each ecozone+ without providing any information on trends.
Quality checks and limitations
Significant algorithm refinement of the MODIS sensor has been conducted since its inception (Yang et al., 2006a), however, limited validation has been conducted on the fPAR estimates specifically (Yang et al., 2006b).
We noted a minor but distinct seam along the 60th parallel in the annual minimum cover component that may have resulted from some change in processing algorithm at that latitude. This does not affect our interpretation, since that component is so low at that latitude.
The three components of the DHI are shown separately in Figure 11a to c. Vegetation dynamics by component are summarized by ecozone+ in Table 2.
From the national perspective of average cumulative annual greenness (Figure 11a) it is apparent that the most productive regions are Canada’s forests, with the greatest greenness in the southern maritime climates (both Atlantic and Pacific with greenness decreasing toward the tree line). The altitude effect of Canada’s high mountains is apparent, as is the influence of aridity of Canada’s prairie region.
The national pattern of average annual minimum fPAR (Figure 11b) is similar to the average cumulative greenness, however it is greatly compressed toward the south. In other words, the minimum cover is quite low throughout the country except in the south and at low elevations. The highest values of average minimum fPARcover occur in a swath of forest across eastern Canada. This corresponds to a band of dense mixed forest. In this area, fPAR is relatively high, even in winter, because of the shadowing effects of all trees, as well as the chlorophyll absorption by conifers.
The average seasonal variation in greenness (Figure 11c) shows different patterns than the other two components of the DHI. Most dramatically, there is no difference between the prairie region and the forests. The seasonal variation increases towards the north and at higher altitudes in the south. This increase is the result of lower mean fPAR in these areas, rather than an increase in the standard deviation. This component shows some subtle and unexpected variations within ecozones+, described briefly in Table 2. When this component is inspected closely, it is apparent that it is very sensitive to relatively small variations in altitude.
Coops et al. (2008) also found that the Boreal Plains, Mixedwood Plains, and Hudson Plains ecozones (using the National Ecological Framework (Ecological Stratification Working Group, 1995)) had the highest variation in the DHI during looking at each year from 2000 to 2005.
Figure 11. Vegetation dynamics in Canada’s ecozones+ by Dynamic Habitat Index components, averaged over 2000 to 2005. DHI components are (a) cumulative annual greenness, (b) minimum annual fPAR and (c) degree of vegetation seasonality.
a. Cumulative annual greenness (2000 to 2005 average)
b. Minimum annual fPAR (2000 to 2005 average)
c. Seasonal variation in greenness (2000 to 2005 average)
Long Description for Figure 11
These three maps show the values of the three indicators that compose the Canadian Dynamic Habitat Index (DHI) averaged from 2000 to 2005: (a) cumulative annual greenness; (b) minimum annual fraction of photosynthetically-active radiation that is absorbed by the earth’s surface (fPAR); and (c) seasonal variation in greenness. Cumulative annual greenness is highest for Canada’s forests, with the greatest greenness in the southern maritime climates (both Atlantic and Pacific with greenness decreasing toward the tree line). Higher altitude areas have lower values as do arid regions, such as Canada’s Prairie region. The national pattern of average annual minimum fPAR is similar to the average cumulative greenness, however it is greatly compressed toward the south. The average seasonal variation in greenness shows different patterns than the other two components of the DHI. Most dramatically, there is no difference between the prairie region and the forests. The seasonal variation increases towards the north and at higher altitudes in the south.
Source: adapted from Coops et al. (2008)
|Ecozone+||Cumulative annual greenness||Minimum annual fPAR||Annual degree of vegetation seasonality|
|Arctic||Variable: very low in the Arctic Cordillera, higher on western Banks Island, the Great Plain of the Koukdjuak on Baffin Island and Ungava||Very low||High except slightly lower in some locations, such as Ungava.|
|Taiga Plains||Low to medium: decreasing from south to north||Very low to low||Decreasing toward south: some interesting patterns in Quebec and Labrador, especially around Hamilton inlet.|
|Taiga Shield||Variable: high in south, low in north; effects of elevation are evident.||Low but increasing toward south||High in north becoming medium in south; effects of elevation are particularly evident.|
|Hudson Plains||High in south becoming medium along the shore of Hudson Bay.||Low throughout||Grading from high in north to low in south.|
|Boreal Shield||High in south becoming lower toward north, particularly in northwest where fire is frequent||Variable: very low in north to very high in south||Grading from high in north to low in south. Lowest in southernmost Canadian Shield in eastern Ontario.|
|Atlantic Maritime||High to very high in southern New Brunswick and Nova Scotia||Variable: low in Gaspé and northern Cape Breton to very high in southern New Brunswick and Nova Scotia||High except lower in northern Cape Breton and parts of Gaspé.|
|Mixedwood Plains||Low in urban areas, medium in agricultural areas, and high in forested areas||Variable: highest in SW Ontario and Manitoulin Island; lowest south of Georgian Bay and north of Lake Simcoe.||Low throughout|
|Boreal Plains||Variable; lowest in agricultural areas||Variable; lower in agricultural areas; lowest in patches that may be burn scars||Variable: higher in agricultural areas and in patches that may be burn scars|
|Prairies||Medium to low: highest in forested Cypress Hills||Uniformly low||Uniformly medium|
|Taiga Cordillera||Variable: very low in mountains; higher in northern Yukon||Low||Varies with elevation|
|Boreal Cordillera||Variable: very low near mountain tops; highest in northern BC||Low; higher in valleys||The elevation effect is very pronounced: highest at highest elevation and lower in valleys|
|Pacific Maritime||Variable: very low on mountain tops to very high in coastal regions of Vancouver and Queen Charlotte Islands||Very low on mountain tops to high in coastal regions of Vancouver and Queen Charlotte Islands||The elevation effect is very pronounced: highest at highest elevation and lower in valleys|
|Montane Cordillera||High except at highest elevations and dry valleys.||Variable||Low except high at high elevations|
|Western Interior Basin||High except in dry interior valleys.||Variable||Low except high at high elevations|
|Newfoundland Boreal||High in east to medium to high in west.||Low except coastal fringe||Pronounced elevation effects|
Each of the three components of DHI provides information about vegetation production. The values of fPAR integrate the effects of various factors on productivity. Forests that consist primarily of evergreen trees, for example, will have a higher cumulative annual fPAR, a higher minimum annual fPAR, and a lower degree of seasonality than a forest consisting of primarily deciduous trees. Productivity (and therefore fPAR) tends to increase with increasing temperature, though will be reduced by periods of drought.
The DHI provides a potential approach for using standard earth observation products to monitor long-term vegetation productivity trends (from 2000). Specific uses for this relatively new dataset are still being explored and tested. One key area of research is in using the DHI components to predict patterns of biodiversity. This research is based on the assumption that the distribution and abundance of species across landscapes is driven by key environmental parameters (Turner et al., 2003), with vegetation productivity being one of these key parameters (MacArthur, 1972). Coops et al. (2008) argue that direct measurements of fPAR, which is included in the three DHI components, provide better estimates of vegetation productivity than the more traditionally used NDVI. It would be interesting to compare similar indicators of productivity based on NDVI data in order to better compare these methods.
Andrew et al. (2011) examined the ability of the DHI components to explain patterns of butterfly community composition and species affinities in Canada. They found that these components on their own were not a good predictor, but suggested they could be used to improve predictions of community composition within the qualitatively defined ecozones and ecoregions of Canada (Ecological Stratification Working Group, 1995). Coops et al (2009a)(2009c) used the DHI components to predict the species richness of breeding birds in the United States, and, in conjunction with land cover and topography, in Ontario with more promising results. The ability to predict species distribution and the relevant parameters for predicting species distribution are species dependent. In many cases, this type of analysis more specifically predicts potential species distribution, as opposed to actual species distribution (Kerr and Ostrovsky, 2003).
Using remotely sensed data to indirectly describe broad patterns of biodiversity could be used to develop an “early warning system” of large scale changes in biodiversity (Duro et al., 2007). The utility of the DHI components in this type of work remains to be seen. Other potential uses may also be developed.
- Date Modified: