1
Modeling Soil Organic Matter Distribution in a Modeling Soil Organic Matter Distribution in a Northern Hardwood Forest and Tropical Northern Hardwood Forest and Tropical Watershed. Watershed. Kris Johnson* Fred Scatena* and Yude Kris Johnson* Fred Scatena* and Yude Pan** Pan** *University of Pennsylvania **USDA Forest Service *University of Pennsylvania **USDA Forest Service The Green Mountains The Green Mountains Bisley Bisley Ponce M ayaguez San Juan Puerto Rico Caribbean National Forest Bisley E xperimental Watershed 0 5000 10000 15000 20000 Mean Organic Matter (g/m2) Valley Slope Ridge Field Observed Organic Matter 0 5000 10000 15000 20000 Mean(SOM_Top60) Valley Slope Ridge Organic Matter Distribution By 0 2500 5000 7500 10000 12500 15000 Soil Organic Carbon (g/m2) 5807 5901 5914 5903 5913 5704 5917 5804 5803 5821 5909 5826 5911 5802 5820 5806 5819 5902 5915 5706 5703 site Y Century Simulated 1958-60 Site Measuremen 1990-91 Site Measuremen Quantitative Pit Measur Soil Organic Carbon (g/m 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000 5703 5704 5706 5708 5802 5803 5804 5806 5807 5819 5820 5821 5826 5836 5901 5902 5903 5909 5911 5913 5914 5915 5917 site Std Dev 0 1000 2000 3000 4000 5703 5704 5706 5708 5802 5803 5804 5806 5807 5819 5820 5821 5826 5836 5901 5902 5903 5909 5911 5913 5914 5915 5917 site Estimating Soil Organic Matter Content - The Post and Curtis and University of Pennsylvania Studies During the years 1957-1960 B. W. Post and R. O. Curtis set out to establish forest plots for the purpose of modeling a site index useful for timber production. Plots were selected to minimize variations in vegetation type and parent material. All the plots are northern hardwood forest over well-drained acid-till soils. A total of 78 plots were measured for biomass and soil nutrients – including Soil Organic Matter (SOM) (Figure 2). To estimate SOM for the entire site, three pits were excavated and described. Each horizon was sampled and combined with the other horizons to make a composite sample. Next, soil depth was measured in the pits and also by using a soil auger at 17 points along a transect in each plot. SOC content was then calculated for the whole plot by estimating bulk density by regression on percent loss on ignition (%LOI). These plots were revisited during the years 1990- 1992 by the University of Pennsylvania team led by Arthur Johnson. The soils were sampled in a similar way by digging three pits, however this time one of the pits was also measured quantitatively. The most direct measure of SOM is the quantitative pit method that divides the soil profile into forest floor, 0-10 cm, 10-20 cm, and 20+ layers. This method accurately measures bulk density by weighing the rocks from the pit as they are excavated (Huntington et al. 1988). A site average for SOM content was also calculated in the same way as the Post and Curtis study (by regression on %LOI). Therefore, three separate measurements of SOC are possible for each site, although the quantitative pit estimate is only representative of the soils in the precise location where it was excavated (Figure 1). SOM was multiplied by 0.45 to convert to SOC. Objectives Use the Green Mountain dataset to: 1)Estimate SOC content for a Northern Hardwood forest (Table 1). 2)Parameterize the Century model for a Northern Hardwood forest. Results - Century Model Performance The SOC content for the top 20 cm of soil was simulated by the Century model (Parton et al. 1988) and compared to the three measurements of SOC for 20 sites in the Green Mountains. The model was allowed to spin up for 2000 years. Precipitation, temperature, soil texture, bulk density, soil depth and drainage were varied from site to site. With a few exceptions, the Century model simulated SOC in the top 20 cm of mineral soil within the range of measurements made for each site (Figure 3). However, the measured SOC and simulated SOC were poorly correlated (adjusted R2 < 0.05; p-value > 0.20) no matter which measurement was used as a benchmark. This reflects the challenge of simulating forest soils, especially in this region which is dominated by glacial till soils, tree- throws (i.e. “pit and mound topography”) and the formation of spodic horizons. Many sites that were over-estimated were more developed in their soil formation, having well-defined spodic horizons. In addition, a few very rocky sites (>50% rocks) were also overestimated by the model. Therefore, it seems that the model could not account for leaching of organic materials from the surface horizons as well as the fraction of rocks which makes the carbon content smaller. It is not as clear why some sites were underestimated by the model. One source of error is that soils with true A horizons and thick forest floors (10-20cm) are sometimes difficult to separate into mineral and organic horizons, which make the measurements uncertain. Also, the climate data, which is also modeled, may be inaccurate for some locations. Model errors for 20 sites ranged from about 20% to 30% (Table 2). Figure 1 (above). Variability chart for 20 selected sites used for Century Model parameterization. Figure 2 (right). The Green Mountain Physiographic Region. Green areas are Northern Hardwood (i.e. maple-beech-birch) stands. White dots are plot locations and bigger white dots correspond to higher SOC. Table 2 Sim ulation Error 1958-60 Site Measurem ent 25% 1990-91 Site Measurem ent 31% Q uantitative Pit Measurem ent 21% SOM_Top60(g/m2) 0 5000 10000 15000 20000 25000 30000 35000 40000 45000 000 560 000 600 040 520 040 560 040 600 080 440 080 480 080 520 080 560 080 600 120 200 120 240 120 280 120 320 120 360 120 400 120 440 120 480 120 520 120 560 120 600 120 680 120 720 120 760 160 240 160 320 160 360 160 400 160 440 160 480 160 520 160 560 160 600 160 640 160 680 160 720 160 760 200 240 200 280 200 320 200 360 200 400 200 440 200 480 200 520 200 560 200 600 200 640 200 680 200 720 200 760 240 360 240 400 240 440 240 480 GRID Std Dev 0 5000 10000 15000 20000 000 560 000 600 040 520 040 560 040 600 080 440 080 480 080 520 080 560 080 600 120 200 120 240 120 280 120 320 120 360 120 400 120 440 120 480 120 520 120 560 120 600 120 680 120 720 120 760 160 240 160 320 160 360 160 400 160 440 160 480 160 520 160 560 160 600 160 640 160 680 160 720 160 760 200 240 200 280 200 320 200 360 200 400 200 440 200 480 200 520 200 560 200 600 200 640 200 680 200 720 200 760 240 360 240 400 240 440 240 480 GRID The Bisley Experimental Watershed is a small tropical watershed covering 13 hectares (about 3 city blocks) located in the Caribbean National Forest, Puerto Rico (Figures 4). Three main vegetation types occur throughout the National Forest at different elevations. Bisley is located in the lower elevation “tabonuco forest” (Dacryodes excelsa). The tabonuco is a hurricane-resilient tree that has smaller leaves than the surrounding vegetation. During high winds, the leaves are quickly dropped in order to reduce wind resistance and minimize stress on the tree. Additionally, individual trees are often linked to each other in a network of root grafts so as to avoid being uprooted. The tree and its roots provide physical stability for the soil so that much of the SOM is saved from being transported during heavy rainfall. Further, the landscape is dissected with many small ephemeral stream channels sometimes separated by only a few meters. Therefore, more SOM preferentially accumulates in stable ridge areas where tabonuco vegetation dominates (Scatena and Lugo 1995). SOM was measured at gridpoints spaced every 40 m (85 total observations) and at three depths (0-10cm, 10-35cm, 35-60cm) with a 1-inch coring device (Figure 6). Objective To model ridge, slope and valley landforms for SOM estimation. Topographic variability, even within individual plots, is very high in this dissected landscape. To improve results, only those plots which were believed to have reliable SOM measurements were used. To assess this, two separate measurements of the same plots in the years 1988 and 1990 were compared. Only those plots with relatively low variability between measurements were used for analysis (about 60 of 85 total plots) (Figure 5). Topographic position index (TPI) was used to identify ridge areas from valley areas (Figure 6). This was the best way to account for spatial variation of SOC in the watershed, even after considering spatial autocorrelation models. TPI classes adequately describe the differences in SOC variability in the watershed (Figure 7). Figure 4 (above). The location of the Bisley Experimental Wat Figure 5 (below). Variability chart of SOC measurements at Bi Figure 7. SOM distribution among landform types as observed in the field (above left) and SOM distribution among landforms modeled by TPI (above right). Figure 6 (left). Ridge, slope and valleys identified by TPI. Darkest colors are ridges, medium are slopes and lightest colors are valleys. White dots represent SOM in the top 60 cm where bigger dots are higher in SOM. Acknowledgements Many thanks to Arthur Johnson and David Vann for providing the data and helping with interpretations. William Parton and Cindy Keogh at the Natural Resource Ecology Lab, Fort Collins, CO for training and helping with calibration of the model. Also, Richard Birdseye and the U.S. Forest Service for support and collaboration. Table 1 M gC/ha ForestFloor 0-10cm 10-20cm Top 20cm 20plus cm Total M ineral **Total Solum G reen M ountains (this study) 83 (17.6) 38 (2.0) 28 (2.0) 65 (3.7) 62 (7.5) 124 (10.3) 180 (13.5) *H ubbard Brook (H untington etal 1998) 30 (3.0) 32 (1.2) 27 (1.6) 73 (6.8) 130 (7.7) 160 (8.2) Total SO C in the G reen M ountain Physiographic Region 2.6E07 (5.5E06) 1.2E07 (6.4E05) 8.8E06 (6.2E05) 2.0E07 (1.2E06) 2.0E07 (2.4E06) 3.9E07 (3.2E06) 5.7E07 (4.3E06) *H ubbard Brook data from W atershed 5 and includes som e spruce/firforest **Total Solum refers to the forestfloorplus m ineral horizons ***Extrapolated from the this study's dataset Figure 3

Modeling Soil Organic Matter Distribution in a Northern Hardwood Forest and Tropical Watershed. Kris Johnson* Fred Scatena* and Yude Pan** *University

Embed Size (px)

Citation preview

Page 1: Modeling Soil Organic Matter Distribution in a Northern Hardwood Forest and Tropical Watershed. Kris Johnson* Fred Scatena* and Yude Pan** *University

Modeling Soil Organic Matter Distribution in a Northern Modeling Soil Organic Matter Distribution in a Northern Hardwood Forest and Tropical Watershed.Hardwood Forest and Tropical Watershed.

Kris Johnson* Fred Scatena* and Yude Pan**Kris Johnson* Fred Scatena* and Yude Pan***University of Pennsylvania **USDA Forest Service*University of Pennsylvania **USDA Forest Service

The Green MountainsThe Green Mountains BisleyBisley

Ponce

Mayaguez

San Juan

Puerto Rico

Caribbean National Forest

BisleyExperimentalWatershed

0

5000

10000

15000

20000

Mea

n O

rgan

ic M

atte

r (g

/m2)

Valley Slope Ridge

Field Observed Organic Matter

Column 1 Valley Slope Ridge

Chart

0

5000

10000

15000

20000

Mea

n(S

OM

_Top

60)

Valley Slope Ridge

Organic Matter Distribution By TPI

TPIbroad3Classa Valley Slope Ridge

Chart

0

2500

5000

7500

10000

12500

15000

Soi

l Org

anic

Car

bon

(g/m

2)

5807

5901

5914

5903

5913

5704

5917

5804

5803

5821

5909

5826

5911

5802

5820

5806

5819

5902

5915

5706

5703

site

Overlay Chart

Y

Century Simulated

1958-60 Site Measurement

1990-91 Site Measurement

Quantitative Pit Measurement

Chart

Soi

l Org

anic

Car

bon

(g/m

2)

2000

3000

4000

5000

6000

7000

8000

9000

10000

11000

12000

5703

5704

5706

5708

5802

5803

5804

5806

5807

5819

5820

5821

5826

5836

5901

5902

5903

5909

5911

5913

5914

5915

5917

site

Std

Dev

0

1000

2000

3000

4000

5703

5704

5706

5708

5802

5803

5804

5806

5807

5819

5820

5821

5826

5836

5901

5902

5903

5909

5911

5913

5914

5915

5917

site

Variability Chart for Stack

Estimating Soil Organic Matter Content - The Post and Curtis and University of Pennsylvania Studies

During the years 1957-1960 B. W. Post and R. O. Curtis set out to establish forest plots for the purpose of modeling a site index useful for timber production. Plots were selected to minimize variations in vegetation type and parent material. All the plots are northern hardwood forest over well-drained acid-till soils. A total of 78 plots were measured for biomass and soil nutrients – including Soil Organic Matter (SOM) (Figure 2). To estimate SOM for the entire site, three pits were excavated and described. Each horizon was sampled and combined with the other horizons to make a composite sample. Next, soil depth was measured in the pits and also by using a soil auger at 17 points along a transect in each plot. SOC content was then calculated for the whole plot by estimating bulk density by regression on percent loss on ignition (%LOI).

These plots were revisited during the years 1990-1992 by the University of Pennsylvania team led by Arthur Johnson. The soils were sampled in a similar way by digging three pits, however this time one of the pits was also measured quantitatively. The most direct measure of SOM is the quantitative pit method that divides the soil profile into forest floor, 0-10 cm, 10-20 cm, and 20+ layers. This method accurately measures bulk density by weighing the rocks from the pit as they are excavated (Huntington et al. 1988). A site average for SOM content was also calculated in the same way as the Post and Curtis study (by regression on %LOI). Therefore, three separate measurements of SOC are possible for each site, although the quantitative pit estimate is only representative of the soils in the precise location where it was excavated (Figure 1). SOM was multiplied by 0.45 to convert to SOC.

ObjectivesUse the Green Mountain dataset to:1)Estimate SOC content for a Northern Hardwood forest (Table 1).2)Parameterize the Century model for a Northern Hardwood forest.

Results - Century Model PerformanceThe SOC content for the top 20 cm of soil was simulated by the Century model (Parton et al. 1988) and compared to the three measurements of SOC for 20 sites in the Green Mountains. The model was allowed to spin up for 2000 years. Precipitation, temperature, soil texture, bulk density, soil depth and drainage were varied from site to site.

With a few exceptions, the Century model simulated SOC in the top 20 cm of mineral soil within the range of measurements made for each site (Figure 3). However, the measured SOC and simulated SOC were poorly correlated (adjusted R2 < 0.05; p-value > 0.20) no matter which measurement was used as a benchmark. This reflects the challenge of simulating forest soils, especially in this region which is dominated by glacial till soils, tree-throws (i.e. “pit and mound topography”) and the formation of spodic horizons. Many sites that were over-estimated were more developed in their soil formation, having well-defined spodic horizons. In addition, a few very rocky sites (>50% rocks) were also overestimated by the model. Therefore, it seems that the model could not account for leaching of organic materials from the surface horizons as well as the fraction of rocks which makes the carbon content smaller. It is not as clear why some sites were underestimated by the model. One source of error is that soils with true A horizons and thick forest floors (10-20cm) are sometimes difficult to separate into mineral and organic horizons, which make the measurements uncertain. Also, the climate data, which is also modeled, may be inaccurate for some locations. Model errors for 20 sites ranged from about 20% to 30% (Table 2).

Figure 1 (above). Variability chart for 20 selected sites used for Century Model parameterization.

Figure 2 (right). The Green Mountain Physiographic Region. Green areas are Northern Hardwood (i.e. maple-beech-birch) stands. White dots are plot locations and bigger white dots correspond to higher SOC.

Table 2 Simulation Error1958-60 Site Measurement 25%1990-91 Site Measurement 31%Quantitative Pit Measurement 21%

SO

M_

Top

60

(g/m

2)

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

00

0

56

0

00

0

60

0

04

0

52

0

04

0

56

0

04

0

60

0

08

0

44

0

08

0

48

0

08

0

52

0

08

0

56

0

08

0

60

0

12

0

20

0

12

0

24

0

12

0

28

0

12

0

32

0

12

0

36

0

12

0

40

0

12

0

44

0

12

0

48

0

12

0

52

0

12

0

56

0

12

0

60

0

12

0

68

0

12

0

72

0

12

0

76

0

16

0

24

0

16

0

32

0

16

0

36

0

16

0

40

0

16

0

44

0

16

0

48

0

16

0

52

0

16

0

56

0

16

0

60

0

16

0

64

0

16

0

68

0

16

0

72

0

16

0

76

0

20

0

24

0

20

0

28

0

20

0

32

0

20

0

36

0

20

0

40

0

20

0

44

0

20

0

48

0

20

0

52

0

20

0

56

0

20

0

60

0

20

0

64

0

20

0

68

0

20

0

72

0

20

0

76

0

24

0

36

0

24

0

40

0

24

0

44

0

24

0

48

0

GRID

Std

De

v

0

5000

10000

15000

20000

00

0

56

0

00

0

60

0

04

0

52

0

04

0

56

0

04

0

60

0

08

0

44

0

08

0

48

0

08

0

52

0

08

0

56

0

08

0

60

0

12

0

20

0

12

0

24

0

12

0

28

0

12

0

32

0

12

0

36

0

12

0

40

0

12

0

44

0

12

0

48

0

12

0

52

0

12

0

56

0

12

0

60

0

12

0

68

0

12

0

72

0

12

0

76

0

16

0

24

0

16

0

32

0

16

0

36

0

16

0

40

0

16

0

44

0

16

0

48

0

16

0

52

0

16

0

56

0

16

0

60

0

16

0

64

0

16

0

68

0

16

0

72

0

16

0

76

0

20

0

24

0

20

0

28

0

20

0

32

0

20

0

36

0

20

0

40

0

20

0

44

0

20

0

48

0

20

0

52

0

20

0

56

0

20

0

60

0

20

0

64

0

20

0

68

0

20

0

72

0

20

0

76

0

24

0

36

0

24

0

40

0

24

0

44

0

24

0

48

0

GRID

Variability Chart for SOM_Top60(g/m2)

Variability Gage

The Bisley Experimental Watershed is a small tropical watershed covering 13 hectares (about 3 city blocks) located in the Caribbean National Forest, Puerto Rico (Figures 4). Three main vegetation types occur throughout the National Forest at different elevations. Bisley is located in the lower elevation “tabonuco forest” (Dacryodes excelsa). The tabonuco is a hurricane-resilient tree that has smaller leaves than the surrounding vegetation. During high winds, the leaves are quickly dropped in order to reduce wind resistance and minimize stress on the tree. Additionally, individual trees are often linked to each other in a network of root grafts so as to avoid being uprooted. The tree and its roots provide physical stability for the soil so that much of the SOM is saved from being transported during heavy rainfall. Further, the landscape is dissected with many small ephemeral stream channels sometimes separated by only a few meters. Therefore, more SOM preferentially accumulates in stable ridge areas where tabonuco vegetation dominates (Scatena and Lugo 1995). SOM was measured at gridpoints spaced every 40 m (85 total observations) and at three depths (0-10cm, 10-35cm, 35-60cm) with a 1-inch coring device (Figure 6).

ObjectiveTo model ridge, slope and valley landforms for SOM estimation.

Topographic variability, even within individual plots, is very high in this dissected landscape. To improve results, only those plots which were believed to have reliable SOM measurements were used. To assess this, two separate measurements of the same plots in the years 1988 and 1990 were compared. Only those plots with relatively low variability between measurements were used for analysis (about 60 of 85 total plots) (Figure 5).

Topographic position index (TPI) was used to identify ridge areas from valley areas (Figure 6). This was the best way to account for spatial variation of SOC in the watershed, even after considering spatial autocorrelation models. TPI classes adequately describe the differences in SOC variability in the watershed (Figure 7).

Figure 4 (above). The location of the Bisley Experimental Watershed.

Figure 5 (below). Variability chart of SOC measurements at Bisley.

Figure 7. SOM distribution among landform types as observed in the field (above left) and SOM distribution among landforms modeled by TPI (above right).Figure 6 (left). Ridge, slope and valleys identified by TPI. Darkest colors are ridges, medium are slopes and lightest colors are valleys. White dots represent SOM in the top 60 cm where bigger dots are higher in SOM.

AcknowledgementsMany thanks to Arthur Johnson and David Vann for providing the data and helping with interpretations. William Parton and Cindy Keogh at the Natural Resource Ecology Lab, Fort Collins, CO for training and helping with calibration of the model. Also, Richard Birdseye and the U.S. Forest Service for support and collaboration.

Table 1 MgC/haForest Floor 0-10cm 10-20cm Top 20cm 20plus cm Total Mineral **Total Solum

Green Mountains (this study) 83 (17.6) 38 (2.0) 28 (2.0) 65 (3.7) 62 (7.5) 124 (10.3) 180 (13.5)*Hubbard Brook (Huntington et al 1998) 30 (3.0) 32 (1.2) 27 (1.6) 73 (6.8) 130 (7.7) 160 (8.2)Total SOC in the Green Mountain Physiographic Region

2.6E07 (5.5E06)

1.2E07 (6.4E05)

8.8E06 (6.2E05)

2.0E07 (1.2E06)

2.0E07 (2.4E06)

3.9E07 (3.2E06)

5.7E07 (4.3E06)

*Hubbard Brook data from Watershed 5 and includes some spruce/fir forest**Total Solum refers to the forest floor plus mineral horizons***Extrapolated from the this study's dataset

Figure 3