Red Spruce in Canaan Valley area WV.

Spruce and Hemlock Forests in West Virginia - Recovering from the great cut

A perspective on soil carbon resource impacts


Travis Nauman, Geology 659 Student Project

HomeSpruce, Hemlock, and CarbonEnvironmental ModelsMaps

spruce forest near Shavers Fork of Cheat River, WV

Importance of Spruce Systems

    Red spruce and hemlock stands are important for a multitude of reasons including habitat for Cheat Mountain Salamanders and Northern Flying Squirrels, timber production, recreational aesthetics, and carbon sequestration for global warming mitigation. Current and past documentation of Organic horizon (O-horizon) surface layers up to a meter thick in spruce stands shows that these systems can tie up a lot of carbon. These soils also collect dark-colored carbon in subsurface spodic horizons (Bh or Bhs horizons) where carbon is moved lower in the soil through a leached zone and then deposited usually anywhere from 20 cm all the way down to meters deep depending on the amount of horizon development (e.g. left). This potential for the combination of surface and subsurface carbon storage offers a lot of potential for sequestration efforts in areas that lost spruce historically (e.g. right). We suspect many of these sites may have lost the carbon rich spodic horizons from Spodosol soils as has been observed in similar areas in other parts of the U.S. (Barret and Schaetzl, 1998), and thus restoring spruce could potentially re-form these horizons and pull carbon back out of the atmosphere.
    Many studies have looked at the history of red spruce (Picea rubens) in the central Appalachians. Most agree that the spruce stands seen currently are a small fraction of what existed before 1800 due to vast logging and resulting fires (Adams et al., 2010; Hopkins, 1899; Pielke, 1981; Rollins et al., 2010).  In most areas where spruce was harvested, northern hardwood forest has recolonized
(see pictures on right) and we suspect that much carbon has been lost in these areas due to erosion after harvest, disturbance-related fire, and resulting changes in energy and carbon inputs. However, there are still more continuous spruce stands along the highest ridges and to a lesser degree in other parts of high elevation areas of the central Appalachians. Rockier and sandier high ridge areas that currently have spruce show a well documented relationship with Spodosol soils with high levels of organic matter and low nutrient contents (Flegel, 1998). However, other mountain  sideslope areas in siltier soils have little documentation of Spodosols despite the existence of isolated yet significant  spruce stands. Our recent investigations have documented the presence of Spodosol in  these areas and we are looking at  this data for guidance in future management efforts for these forests.  We hypothesize that much of these areas that are now in hardwood could be restored to spruce and that this shift could potentially resequester large amounts of carbon lost because of past harvest and burning.


hardwood forest in WV


hardwood soil

Study Objectives

1) Model any environmental or management patterns discernable in the distribution of Spodosol formation using surface O-horizon thickness as a proxy
  • Use spruce or hemlock presence and other environmental factors from spatial environmental data to create O-horizon models using regression trees (Breiman, 1984).
2) Build a predictive spatial model using random forest regression trees (Liaw and Wiener, 2002) with available spatial data to see if prediction between observations is possible.

  • 303 soil observation holes in southern Randolph and northern Pocahontas counties in West Virginia to use in analysis ( Maps page).
    • Observations sampled:
      • randomly by landform curvature strata
      • uplands soils that formed in bedrock parent material (residuum) 
      • in Hampshire and Chemung geologies which produce siltier soils
  • This design was used to help isolate geology and sediment transport from environmental models.

Results Summary

Results Diagram

Conceptual diagram showing how the analytical approaches are meant to isolate areas with unique environments in a way that is interpretable relative to O-horizon depths

Results showed patterns accross models with deeper O-horizons in Spodosol soils associated with spruce-hemlock influenced forest stands in west to northwest facing aspects and more concave environments. Simple regression tree models could only reasonably represent 50-60% of the O-horizon variability of the full data set and random forest modeling showed that these relationships are really only robust in predicting about 20% of that variation when implemented for interpolative mapping. This was determined using independent validation by randomly withholding observations in model-building. We expect that the complexity of the timber industry impact on these areas due to intensity of harvest(s) and fires that might have frequently burned off O-horizons following harvest historically may be contributing to the unexplainable variation. Landsat seems to account for some variation in current management and vegetation in models, but it is hard to make specific conclusions by interpretation of tree splits of these bands.

It is likely that most of the the areas with spruce or hemlock observed are regenerated from pockets that survived intense cutting and fires in these areas between 1860 and 1940. The west-northwest aspects and concave landforms might be favoring their recovery in these areas, and would also favor organic material build-up by providing a more moist and cool microclimate. Futher characterization of spruce and hemlock age and dominance in these areas will be necessary to see how these vegetation-topograph-climate dynamics are weighted in their influence. Our results do indicate that focusing spruce or hemlock restoration in these environments might deliver the quickest benefits in carbon sequestration efforts. We can associate just the presence of spruce and/or hemlock with a increased O-horizon depth of approximately 5 to 10 centimeters. This figure could also be interpretted to mean that we may have lost at least that much carbon in areas where we lost spruce-hemlock forests historically. There were probably areas that lost much more, but it's hard to document that without having a reference site for comparison.

It must also be noted that there were some limited areas with deep O-horizons and spodic like soil properties that were not under current spruce or hemlock canopies and environments that don't match the recipe described above. I would suspect that these areas are most likely remnant sites that used to have spruce or hemlock cover before the 1860's, but do not have regeneration. Future analysis should look at the distribution of these areas to see if there is any environmental patterns in these potential relic soils. Since O-horizons can be quite dynamic, other proxies of Spodosol development like sesquioxide horizon observations should also be brought in to help identify and analyze remnant sites.   

Collaborating Scientists in Field Work and Data Collection:
Special Thanks to Skip Bell and Tim Dilliplane of the Morgantown WV USDA-NRCS Field Office


Adams, H.S., Stephenson, S., Adams, A.W. and Adams, M.B., 2010. The isolated red spruce communities of Virginia and West Virginia, The conference on the ecology and management of high-elevation forests in the central and southern Appalachian Mountains. Department of Agriculture, Forest Service, Northern Research Station, Slatyfork, WV, pp. 1-12.

Barrett, L.R. and Schaetzl, R.J., 1998. Regressive Pedogenesis Following a Century of Deforestation: Evidence for Depodzolization. Soil Science, 163(6): 482-497.

Breiman, L., 1984. Classification and regression trees. Wadsworth International Group, 358 pp.

Flegel, D.G., 1998. Soil Survey of Pocahontas County, West Virginia (SSURGO). In: N.R.C.S. U.S. Department of Agriculture, W.V.D.O.A.  and U.S. Forest Service  (Editors).

Liaw, A and Wiener, M., 2002. Classification and Regression by randomForest. R news.

Pielke, R.A., 1981. The Distribution of Spruce in West-Central Virginia before Lumbering. Castanea, 46(3): 201-216.

Rollins, A.W., Adams, H.S. and Stephenson, S.L., 2010. Changes in Forest Composition and Structure across the Red Spruce-Hardwood Ecotone in the Central Appalachians. Castanea, 75(3): 303-314.