Variable Area Local Maxima
The algorithm has been updated to include an inverse watershed segmentation (right panel) which reports the minor and major axis of canopy diameter and the total canopy area. The equivalent diameter of the canopy (min+max/2) is shown in the left panel (yellow circles). Some watersheds are empty if there was no tree identified within it (this can be caused by deformed canopy shapes that result in the watershed segmentation splitting the canopy up in to multiple areas. Conversely, some watersheds may have more than one tree identified within them, e.g. a very large tree and a smaller seedling tree off the edge of its canopy.
The code for the VLM:
After I generated the stem maps for each tile I merged the tiles, reordered the tree list by UTM easting, and assigned new identifier numbers:
After I created the tiled data, I clipped the stems to each of the three catchments based on the GIS shape file polygon.
Estimating Individual tree aboveground biomass and carbon
To calculate the biomass of each individual tree I searched the literature for allometric models based on tree height and canopy diameter. There is not much research to date on tree height-to-biomass calculation alone, so I am using my published diameter at breast height (DBH) prediction model from Swetnam and Falk (2014):
DBH (cm) = Beta * HT (m) * √CanopyDiameter (m); RMSE = 7.66 cm, r2 = 0.811
I can estimate the above ground biomass/Carbon (AGB/AGC) using published wood specific gravity (SG, 1Chojnacky et al. 2014) and C content (C%, 2Lamlom and Savidge 2003). Biomass equations are from Chojnacky et al. (2014) where the form of the equation is: ln(AGB) = β0 + β1ln(DBH).
|Species||Specific Gravity (SG)1||Carbon Density (%)2||β0||β1||R2-statistic1|
|Abies spp.||48.55 - 50.08|
|Picea spp.||0.37||49.95 - 50.39|
|Pinus ponderosa||0.38||52.47 ± 0.38||-2.6177||2.4638|
|Pinus contorta||0.38||50.32 ± 0.43||-2.6177||2.4638||0.83|
|Populus tremuloides||0.35||47.09 ± 0.75|
|Pseudotsuga menziesii||0.45||50.50 ± 0.36||-2.4623||2.4852||0.86|
Lodgepole pine (Pinus contorta) allometry is reported by Litton et al. (2004) as: (height = 2.89 * DBH^0.59, r2 = 0.82, P= 0.01, n = 200). The inverse model for height based DBH is: 0.1655 * Ht ^ 1.695. The derived model for AGB based on height assuming the pipe-model, and a wood specific gravity SG = 0.38 for lodgepole pine from Chojnacky et al. (2014): AGB (kg) = 0.0008175 * Ht (m) ^ 4.39
As an example of the estimated above ground carbon (AGC) derived from the Chojnacky et al. (2014) Pinus spp. DBH based AGB model for P. ponderosa and P. contorta (same model) versus the pipe model which uses height and DBH (as radius), I graph the Betasso catchment trees below, note that both the X and Y axis use the same estimated DBH values.
I used the LANDFIRE Existing Vegetation Type (EVT) to characterize individual tree forest types. By catchment the dominant EVTs are:
- Southern Rocky Mountain Ponderosa Pine Woodland (98.166%)
- Developed or Ruderal Shrubland (0.765%)
- All Other Classes (1.1%)
- Southern Rocky Mountain Ponderosa Pine Woodland (41.923%)
- Rocky Mountain Lodgepole Pine Forest (41.669%)
- Inter-Mountain Basins Aspen-Mixed Conifer Forest and Woodland (4.338%)
- Southern Rocky Mountain Dry-Mesic Montane Mixed Conifer Forest and Woodland (4.064)
- Rocky Mountain Montane Riparian Forest and Woodland (2.933%)
- Southern Rocky Mountain Mesic Montane Mixed Conifer Forest and Woodland (2.660%)
- Rocky Mountain Aspen Forest and Woodland (2.042%)
- All Other Classes (0.371%)
- Rocky Mountain Subalpine Dry-Mesic Spruce-Fir Forest and Woodland (57.384%)
- Rocky Mountain Lodgepole Pine Forest (17.384%)
- Rocky Mountain Subalpine Mesic-Wet Spruce-Fir Forest and Woodland (16.965%)
- Inter-Mountain Basins Aspen-Mixed Conifer Forest and Woodland (2.429%)
- Snow-Ice (4.084%)
- All Other Classes (1.754%)
Table: The EVT's by Catchment and above ground biomass/carbon models available for each.
|LANDFIRE EVT Code||Vegetation Class Descriptions|
(n, 30 m2)
(% of Area)
(n, 30 m2)
(% of Area)
(n, 30 m2)
(% of Area)
|Height-to-Biomass Allometry Model Parameters||RMSE (kg)||R2||Reference|
|3011||Rocky Mountain Aspen Forest and Woodland||10||0.058||3646||2.042||1349||0.425||AGB (kg) = 0.006 * ht (m) ^ 3.947||643.5||0.492||Swetnam 2013|
|3049||Rocky Mountain Foothill Limber Pine-Juniper Woodland||18||0.010|
|3050||Rocky Mountain Lodgepole Pine Forest||74365||41.669||55154||17.384||Lefsky et al. 1999|
|3051||Southern Rocky Mountain Dry-Mesic Montane Mixed Conifer Forest and Woodland||7253||4.064||AGB (kg) = 0.027 * ht (m) ^ 3.605||668.2||0.717||Swetnam 2013|
|3052||Southern Rocky Mountain Mesic Montane Mixed Conifer Forest and Woodland||4748||2.660||AGB (kg) = 0.039 * ht (m) ^ 3.455||779.8||0.755||Swetnam 2013|
|3054||Southern Rocky Mountain Ponderosa Pine Woodland||16968||98.166||74819||41.923||AGB (kg) = 0.010 * ht (m) ^ 3.935||526.7||0.760||Swetnam 2013|
|3055||Rocky Mountain Subalpine Dry-Mesic Spruce-Fir Forest and Woodland||79||0.0442||181827||57.384|
|3056||Rocky Mountain Subalpine Mesic-Wet Spruce-Fir Forest and Woodland||53825||16.965|
|3057||Rocky Mountain Subalpine-Montane Limber-Bristlecone Pine Woodland||108||0.061||171||0.054|
|3061||Inter-Mountain Basins Aspen-Mixed Conifer Forest and Woodland||10||0.058||7742||4.338||7706||2.429||AGB (kg) = 0.027 * ht (m) ^ 3.605||668.2||0.717||Swetnam 2013|
|3070||Rocky Mountain Alpine Dwarf-Shrubland||83||0.026|
|3080||Inter-Mountain Basins Big Sagebrush Shrubland||123||0.069|
|3086||Rocky Mountain Lower Montane-Foothill Shrubland||220||0.123||40||0.013|
|3092||Southern California Coastal Scrub||40||0.013|
|3144||Rocky Mountain Alpine Turf||916||0.289||N/A|
|3145||Rocky Mountain Subalpine-Montane Mesic Meadow||27||0.009||N/A|
|3159||Rocky Mountain Montane Riparian Forest and Woodland||5235||2.933|
|3182||Introduced Upland Vegetation-Perennial Grassland and Forbland||61||0.034||N/A|
|3219||Inter-Mountain Basins Sparsely Vegetated Systems II||108||0.034|
|3220||Artemisia tridentata ssp. vaseyana Shrubland Alliance||18||0.104||19||0.006|
|3222||Rocky Mountain Alpine/Montane Sparsely Vegetated Systems II||24||0.008|
|3252||Rocky Mountain Subalpine/Upper Montane Riparian Shrubland||51||0.029||3025||0.953|
|3900||Western Cool Temperate Urban Deciduous Forest||13||0.075|
|3923||Western Cool Temperate Developed Ruderal Shrubland||34||0.197|
|3924||Western Cool Temperate Developed Ruderal Grassland||2||0.0006||N/A|
Table: Muldavin et al. (2006) Vegetation Map Types
|VALUE||COUNT||CLASS_NAME||Hectares (ha)||SG||C||La Jara||La Jara||Jaramillo||Jaramillo||History Grove||History Grove|
|1||4353949||Spruce-Fir Forest and Woodland (Dry Mesic)||10633.97||0.33||0.5|
|2||2732982||Spruce-Fir Forest and Woodland (Moist Mesic)||6674.971||0.33||0.5|
|4||22084678||Mixed Conifer Forest and Woodland (Dry Mesic)||53939.1||0.38||0.5|
|5||14126495||Mixed Conifer Forest and Woodland (Moist Mesic)||34502.14||0.38||0.5|
|7||783518||Blue Spruce Fringe Forest||1913.648||0.33||0.5|
|10||3241666||Aspen Forest and Woodland (Dry Mesic)||7917.36||0.35||0.47|
|11||1921226||Aspen Forest and Woodland (Moist Mesic)||4692.357||0.35||0.47|
|13||9348740||Ponderosa Pine Forest||22833.13||0.38||0.52|
|14||1459721||Gambel Oak-Mixed Montane Shrubland||3565.192||0.61||0.5|
|16||4990766||Upper Montane Grassland||12189.32||0.33||0.5|
|17||12778582||Lower Montane Grassland||31209.98||0.33||0.5|
|21||14647||Montane Riparian Shrubland||35.78066||0.33||0.5|
|24||161175||Sparsely Vegetated Rock Outcrop||393.6614||0.33||0.5|
|25||925811||Felsenmeer Rock Field||2261.175||0.33||0.5|
|28||16769||Post-Fire Bare Ground||40.94513||0.33||0.5|
Matlab Script for EVT classification of Jemez_stems_evt.csv imported into Matlab
Next I estimate the DBH of each tree based on the tree height multiplied by the square root of its estimated canopy diameter (as measured by the VLM's inverse watershed segmentation). If the tree had an estimated canopy diameter that was >150% of its predicted canopy diameter the predicted canopy diameter is used instead. This is because the watershed segmentation does not always represent a canopy properly, in particular for smaller trees next to very large trees. The normalization constant is estimated from data to be ~0.82 for all trees (data are from a mix of Pseudotsuga, Picea, and Pinus trees measured in Arizona and New Mexico, Swetnam and Falk 2014).
For an EVT specific classification I tried:
At this point I'm ready to export the files back into a CSV or XLSX format
I suggest using the dlmwrite command in Matlab to write the export_utm.m outputs of the file to a *.CSV
Note: dlmwrite does not allow you to copy the text header directly.
Alternately, use the xlswrite command in Matlab to write to an XLSX file with header:
Note: You can only do this for ~1 million rows in an XLSX file.
Calculating the total biomass on a per area basis, derived from the individual trees:
After calculating the individual tree biomass I was also interested in calculating the total biomass on a per pixel basis. I kept this analysis at a larger pixel scale, e.g. 10m and 30m, because individual trees tend to take up a significant amount of space with their crowns and their root balls.
In QGIS I used the Vector > Research Tools > Vector Grid option to generate a polygon grid that aligns with the raster topographic surface models:
After I generated the gridded polygon I used the Vector > Analysis Tools > Points in Polygon module to sample the sum of the above ground carbon (calculated from the equations above) for each given EVT/species.
The final step is to set the viewing to a graduated color ramp:
Load the Catchment polygons (Upper Jaramillo, History Grove, La Jara, and ZOB) into QGIS and use the Select by Location feature.
Create new shape files for each set of stems.
Bright, B. C., Hicke, J. A., & Hudak, A. T. (2012). Estimating aboveground carbon stocks of a forest affected by mountain pine beetle in Idaho using lidar and multispectral imagery. Remote Sensing of Environment, 124, 270-281.
Falkowski, M. J., Smith, A. M., Gessler, P. E., Hudak, A. T., Vierling, L. A., & Evans, J. S. (2008). The influence of conifer forest canopy cover on the accuracy of two individual tree measurement algorithms using lidar data.Canadian Journal of Remote Sensing, 34(sup2), S338-S350.
Hudak, A. T., Lefsky, M. A., Cohen, W. B., & Berterretche, M. (2002). Integration of lidar and Landsat ETM+ data for estimating and mapping forest canopy height. Remote sensing of Environment, 82(2), 397-416.
Hudak, A. T., Crookston, N. L., Evans, J. S., Hall, D. E., & Falkowski, M. J. (2008). Nearest neighbor imputation of species-level, plot-scale forest structure attributes from LiDAR data. Remote Sensing of Environment, 112(5), 2232-2245.
Lefsky, M. A., Cohen, W. B., Hudak, A., Acker, S. A., & Ohmann, J. (1999). Integration of lidar, Landsat ETM+ and forest inventory data for regional forest mapping. International Archives of Photogrammetry and Remote Sensing, 32, 119-126.
Litton, C. M., Ryan, M. G., Tinker, D. B., & Knight, D. H. (2003). Belowground and aboveground biomass in young postfire lodgepole pine forests of contrasting tree density. Canadian Journal of Forest Research, 33(2), 351-363.
Swetnam, T. L., & Falk, D. A. (2014). Application of Metabolic Scaling Theory to reduce error in local maxima tree segmentation from aerial LiDAR. Forest Ecology and Management, 323, 158-167.
Turner, M. G., Tinker, D. B., Romme, W. H., Kashian, D. M., & Litton, C. M. (2004). Landscape patterns of sapling density, leaf area, and aboveground net primary production in postfire lodgepole pine forests, Yellowstone National Park (USA). Ecosystems, 7(7), 751-775.