David R. Butler. Department of Geography. University of North Carolina ..... McGuinness, 1973). Meterologic data from three nearby first order stations.
ABSTRACT
Variation in Grassland Biomass: A Spatial and Temporal Perspective Stephen J. Walsh Therese M. DeGuire and David R. Butler
During a sampling period extending from May 07 through August 13, 1985, plant biomass, soil temperature, soil moisture, and other selected biophysical variables were collected on a weekly basis at 12 sample sites distributed along a 200 km east-west trending transect in west-central Oklahoma. Correlation coefficients showed temporal and spatial relationships between dry mean biomass and the relative location of sample sites along the transect, Universal Transverse Mercator coordinate location of each sample site, sample site elevation, and sample week (- 0.43, + 0.43, - 0.54, and + 0.42, respectively) . A multiple regression analysis showed that 84.5 percent of the variation in plant biomass could be explained by a combination of collected biophysical variables: soil texture, week of data collection, location of the sample site, soil moisture, soil temperature , heat units, and potent i al evapotranspi ration. INTRODUCTION
Department of Geograph y University of North Carolina Chapel Hill, North Carolina 27514 Department of Geography Oklahoma State University Stillwater, Oklahoma 74078 Department of Geography Un iversity of Georgia Athens, Georgia 30602
10
Plant communities occupy a definite form, pattern, and location in space at any given time period. Assessment of vegetation communities should encompass the dynamic nature of vegetation and the interactive forces associated with their development. Climatic and edaphic factors act upon plant communities and ultimately determine their existence and survival (McNaughton, 1983). Examination of relationships between vegetation and environmental factors can be accommodated through a study of the spatial pattern in plant communities over time. The edaphic influence of soil physics relative to texture, holding capacity, and ease of water movement through the soil, in combination with the atmospheric influence of meteorologic variables on evaporation and transpiration affect the energy balance between the plant and its surrounding environment. The interplay between abiotic driving variables, temperature, terrain orientation, evapotranspiration, and the ability of the soils to store moisture for plant utilization, will
directly influence the moisture supply and demand of the ecosystem (Griffiths and Driscoll, 1982). The intent of this study is to evaluate the geopedoclimatic influences that affect biomass fluctuations associated with field sites positioned along an environmental gradient and to examine the impact of spatial and temporal variables combined with selected biophysical variables in explaining biomass variability. STUDY AREA The study area transect extends 200 km from Stillwater in the east to Woodward in the west (Figure 1). The study region extends through a zone of climatic transition and within a relatively pronounced moisture, temperature, and vegetation gradient. The climate along the transect is relatively dynamic given the short distance between sites 1 and 12, approximately 200 km . The average annual precipitation and the average annual air temperature decreases from east to west along the transect. Payne county, in the eastern part of the transect, receives an annual average precipitation of 86 cm; Blaine county, in the central portion of the transect, has 65 cm; and Woodward county, in the western part of the transect, has 54 cm of precipitation per year. Average summer precipitation (months of June, July, and August) for Payne, Blaine, and Woodward counties is 7.8, 7.4, and 6.3 cm, respectively. The average annual temperatures for Payne, Blaine, and Woodward counties are 17.8° C, 16°'C, and 15° C, respectively. Summer temperature averages are 26° C for Payne, 28S C for Blaine, and 27° C for Woodward county. A transition between alternating forest-shrub and prairie cover to almost continuous grassland occurs across the central portion of the transect because of changes in water content and water holding capacity of different soil types. All sites contain native grasses, which have not been subjected to fire in recent years. A change from tall grass to short grass prairie occurs in an east-west direction associated with the east-west
trending climatic gradient. Typical perennial grasses along the transect are bluestem, grama grasses, brome grass, switchgrass, and buffalograss. Specific grass species for each site are given in Table 1. Available soil water is considered to be one of the principle determinants in shortgrass prairie productivity (Detling, 1979). As part of this study, soil samples for particle size analysis were collected at each depth (15 cm, 61 cm, and 91 cm) for the sample sites. Soil texture is a critical variable in controlling moisture availability for plant growth relative to water holding capacity, permeability, infiltration rates, runoff levels, and internal redistribution of available moisture (Hillel, 1982). Clay loam and sandy loam soils predominate at the 12 sample sites (Table 2). The physiography of the study area on a broad scale is diverse; however, physiographic characteristics at each site location were deliberately chosen to be relatively uniform. The upland locations of the loam soils were chosen to minimize site variations due to slope runoff, infiltration rates, and aspect. The 12 sample sites distributed along the transect are characterized by slopes of 0-9 percent. BIOMASS RESPONSE TO BIOPHYSICAL VARIABLES The transect sampling scheme utilized in this study combines not only the biomass evaluation at a site over time, but also the spatial assessment of biomass over a broad area across a zone of climatic gradient. Marotz (1983) found that vegetation gradients appear to coincide with climatic gradients. White and Glenn-Lewin (1984) examined vegetation associations relative to both species composition and vegetation/environment relationships along a topographicmoisture gradient within the tallgrass prairie vegetation of Iowa and eastern Nebraska. They report that topographic position and soil moisture are fundamental factors influencing vegetation composition and structure, and that variations within tallgrass prairie reflect vegetational responses at various scales 11
• sample sites o meteorologic station
OWoodword
-12
Blaine
o 20 40miles ~~~i !""'**= ;;;;;gj .. o 20 40 60 km Figure 1. Transect and sample site locations in west-central Oklahoma. to topographic position, local edaphic characteristics, and geography. Soil moisture reserves play an im portant role in determining plant activity (Dejong and MacDonald, 1975). Water is one of the principle factors in limiting growth in the shortgrass prairie (Detling, 1979). Seasonal precipitation amounts
12
strongly influence the growth and development of the native perennial grasses (Smoliak, 1956; Cable, 1975). Rains of short duration and magnitude, however, may evaporate too quickly to be effective and large rain amounts of intense storms may rapidly run off because the soil reaches field capacity and lacks the
TABLE 1 Sample Site Vegetation Composition
Type
Vegetation Density
Grass %
1
40 30 20 10
switchgrass Little Bluestem Big Bluestem Serial
(Panicium viragatum) (Andropogon scoparious) (Andropogon gerardi)
Tall Mid Tall
Dense
2
70 30
Little Bluestem Serial: Western Ragweed
(Andropogon scoparius)
Mid
Medium
Mid
Medium
3
70 10 10 10
Native Grass
Species Name
site
(Ambrosia psilostachya) (Artimisia)
Little Bluestem Western Ragweed Switch Big Bluestem Buffalograss Hairy Grama
(Andropogon scoparius) (Ambrosia psilostachya) (Panicium viragatum) (Andropogon gerardi) (Buchloe dactyloides) (Bouteloua hirsuta)
Tall Tall Short Short
4
33 33 33
Silver Bluestem Japanese Brome Hestern Ragweed
(Andropogon saccharoides) Mid (Bromus Japoniaes) (Ambrosia psilostachya)
Dense
5
40 30 20 10
Silver Bluestem Japanese Brome Western Ragweed Sideoats Grama
(Andropogon saccharoides) Mid (Bromus Japoniaes) (Ambosia psilostachya) (Boutelova curtipendula) Mid
Medium
6
60 30 10
Japanese Brome Silver Bluestem Serial : Snowy Partridgepea Blue Wildindigo Doted Grayfeather
(Bromus Japoniaes) (Andropogon saccharoides) Mid
Medium
(Chamaecrista fasciculata) (Baptista australis) (Liastris puncata)
7
60 20 10 10
Buffalograss Japanese Brome Silver Bluestem \,estern Ragweed
Short (Buchloe dactyloides) ( Bromus Japoniaes) (Andropogon saccharoides) Mid (Ambrosia psilostachya)
MediumSparse
8
60 30
Japanese Brome Buffalograss Hairy Grama Serial: l'iestern Ragweed Snowy Partridgepea Sand Sage
(Bromus Japoniaes) (Buchloe dactyloides) (Bouteloua hirsuta)
MediumSparse
Sideoats Grama Hairy Grama Little Bluestem Japanese Brome Bare Soil
(Bouteloua curtipendula) (Bouteloua hirsuta) (Andropogon scoparius) (Bromus Japoniaes)
10
9
60 20 10 10
Short Short
(Ambrosia psilostachya) (Chamaecrista fasciculata) (Artimisia) Mid Short Mid
Medium
13
TABLE 1 (Continued) Sample Site Vegetation Composition
site
Grass
Species Name
Native Grass
Type
%
10
50 30 10 10
11
70 20 10
12
60 10 10 10 10
Japanese Brome Hairy Grama Buffa10grass Western Ragweed Little Bluestem Serial
(Bromus Japoniaes) (Boute1oua hirsuta) (Buchloe dactyloides) (Ambrosia psilostachya) (Andropogon scopar ius)
Little Bluestem Silver Bluestem Western Ragweed Japanese Brome
(Andropogon scoparius) Mid (Andropogron saccharodies) Mid (Ambrosia psilostachya) (Bromus Japoniaes)
Medium
Sideoats Grama Bare Soil Little Bluestem Japanese Brome Western Ragweed Snowy Partridgepea
(Bouteloua curtipendula)
Medium
Dense Short Short Mid
Mid
Mid (Andropogon scoparius) (Bromus Japoniaes) (Ambrosia psilostachya ) (Chamaecrista fasciculata)
ability to store additional water. They conclude that storm size, intensity, and spacing of the precipitation events influence the effectiveness of precipitation available for plant growth. Plant biomass production is often the most assessed response criterion indicating the effectiveness of rainfall for plant utilization. The analysis of the relationship between plant biomass production and evapotranspiration for grassland sites along a gradient of low to high evapotranspiration suggests that the rate of increasing biomass production is a function of decreasing water use efficiency (Webb et aI., 1978). Precipitation is less effective in promoting an increase in biomass production during the middle to late growing season . The rate of early spring growth is controlled primarily by water availability (Detling, 1979), which in turn is at least partially dependent on antecedent moisture conditions from the previous fall and winter. Lauenroth (1979) states that precipi-
14
Vegetation Density
tation and temperature are important determinants of the average annual biomass production of grasslands. Factors including soil properties, terrain orientation, elevation, site location, and successional status also influence biomass levels in grasslands. Temporal variations in temperature responses reveal two distinct grassland types. Warm season plants exhibit higher photosynthetic rates, more efficient water use, and a short, active growing season in the early summer; cool season plants exhibit a longer growing season, beginning in early spring, and a later attainment of peak biomass levels (Doliner and Jolliffe, 1979). Temperature and moisture act as composite factors interacting in complex ways with other environmental factors to affect plant growth. Single factors gradient analysis is not as useful in accounting for variations in vegetation as multivariate gradient analysis. Independent moisture and temperature relationships are often not straightforward
alone, and it is evident that other combined factors are interacting with precipitation and temperature to affect biomass growth rates (French, 1979; Alaback, 1986). DATA COLLECTION During a sampling period extending from 07 May through 13 August, 1985, measurements of plant biomass, soil temperature, and soil moisture were taken on a weekly basis. Sampling started shortly after sunrise in Stillwater (site 1) and ended at dusk approximately 13 km south of Woodward (site 12) in order to complete all measurements in a single day. Each site was sampled at approximately the same time each sample day. The 12 sample sites were initially selected from topographic maps and county soil surveys of the U.S. Department of Agriculture's Soil Conservation Service (SCS). The SCS maps were used to identify easily accessible sites, and to locate areas with uniform vegetation , and flat, upland slopes. Sites with soils having relatively uniform textural and structural characteristics were chosen to minimize the effects that site variations in these variables could produce . Initial slope, elevation, and locational measures were collected from the topo graphic maps. The final selection of a site was determined by field observation. Field measures of topography were completed with a Brunton compass. Terrain conditions and locational elements are important to this research because elevation and latitude contribute to a decrease in mean annual soil temperature. Wire fences were built to protect the vegetative cover at each 8 x 10m site and to shield the instrumentation from disturbance. A quadrat of 24 x 42 cm (0.1 m 2 ) area was constructed of a durable alloy to retain a uniform shape and size throughout the study period. At each study site, 10 quadrat samples were collected on a weekly basis. A systematic sampling approach was utilized to distribute the 10 quadrat samples throughout each study site without duplication of quadrat locations for any of the 15 study weeks. The above-ground live biomass was clipped, bagged, dried, and
weighed . A dry mean biomass value for each site per week was derived for further analysis. Within each site, two sets of thermocouple/psychrometers were implanted to measure soil temperature. The sensors were planted into the ground by augering two holes in the soil to a depth of 122 cm: the augered holes were approximately 3.0 m apart. Sensors were implanted in each hole at depths of 15, 61, and 91 cm. These depths were chosen to observe the changes in soil temperature and soil moisture with depth along a vertical gradient which responds to diurnal and seasonal fluctuations (Panciera et. aI., 1986). Soil moisture also was measured at the 12 sample sites along the transect. Two sets of galvanized steel access tubes were installed at each site. A neutron probe was used to measure soil moisture because of its rapidity and accuracy of measurement. The probe allows the characteristics of the soil to be measured in any physical state covering a large volume of space (Panciera et ai , 1986). In addition to the collection of soil moisture and soil temperature variables at each site, a group of biophysical variables, which were hypothesized to affect plant biomass variability along the transect during the sample period , were measured and/or calculated. Table 3 lists the 23 biophysical variables identified as important elements affecting soil conditions along the transect. Week (WK) of observation relates to the time of data collection extending from May 7 through August 13, 1985. Weekly total precipitation (WPREC) is the total daily precipitation amount recorded at a designated Oklahoma cooperative weather station and assigned to represent a specific sample site along the transect. Figure 1 shows the location of the local cooperative weather stations which collect daily precipitation and air temperature data adjacent to the study transect. The values recorded at each station were assigned to specific field sites through use of the Thiessen polygon approach. The weekly total cumulative precipitation (TPREC) was derived in an attempt to 15
TABLE 2 Soil Characteristics for Transect Sample Sites
(7)
Site
Depth
Percent Sand
Percent Silt
Percent Clay
(not collected for this study)
Soil Texture
loam/clay loam
Soil Structure A Horizons B Horizons AI: moderate, fine, granular
B2t: B3:
1 2 3
39.5
35
25.5
37
35
28
1
39.5 42.0 22
38.75 37.5 42.5
21. 75 20.5 35.5
2 3
17 14.5
42.5 37.5
40.5 48.0
5
1 2 3
37 38.25 37
45 31. 25 30
6
I 2 3
29.5 49.5 44.5
36.25 22.5 22.5
2
3
2 3 4
loam/clay loam
AI: moderate, fine, granular
B2t: B3:
loam
AI: moderate, fine, granular
B2t: B3:
weak, medium, blocky massive weak, medium, blocky massive weak, medium, blocky massive
silty clay
AI: moderate, fine, granular
B:
18 30.5 33
clay loam
Al: moderate, fine, granular
BI: moderate, fine, subangular blocky B2t: weak to moderate, fine to medium, subangular blocky B3: weak, fine, subangular blocky
34.25 28 33
clay loam
AI: moderate, fine, granular
BI: moderate, fine, subangular blocky B2t: weak to moderate, fine to medium, subangular blocky B3: weak, fine, subangular blocky
1
moderate, medium, granular
7
1 2 3
32 29.5 33.25
40 32.5 46.25
28 38 20.5
clay loam
AI:
Bl: moderate, fine, subangular blocky B2t: weak to moderate, fine to medium, subangular blocky B3: weak, fine, subangular blocky
8
1 2 3
62.0 49.5 53.25
27.5 35 36.25
10.5 15.5 10.5
sandy loam
AI:
moderate, medium, granular
B2t and B3: weak, medium, prismatic structure breaking to subangular blocky and granular
9
1 2 3
37 34.5 32
48.75 43.75 40
14.25 21. 75 28
loam
AI:
Bl: moderate, fine, subangular blocky B2t: moderate, medium, subangular blocky B3: weak, fine, subangular blocky
10
1 2 3
30 26.5
20.5 21. 75
sandy clay loami loam
AI:
49.5 52
moderate, medium, granular
B2t and B3: weak, medium, prismatic structure breaking to subangular blocky and granular
1 2 3
52 54.5 39.5
28.75 26.25 35
19.25 19.25 25.5
loam/sandy loam
AI:
moderate, medium, granular
B2t and B3: weak, medium, prismatic structure breaking to subangular blocky and granular
1
40.75 38.25 39.5
41.25 38.75 37.5
18 23 23
loam
AI:
moderate, medium, granular
B2t and B3: weak, medium, prismatic structure breaking to subangular blocky and granular
11
12
2 3 -..J
moderate, fine, granular
moderate, fine, granular
TABLE 3
Variables Evaluated for Multiple Regression Model: Variable CrossProducts and the 1st, 2nd, 3rd, and 4th Powers of Each Variable are Included. Only One Variable is Shown to Demonstrate the Cross-Product and Powers Method
variable Code
18
Variable Code Description
WK
Week of observation
WPREC
Weekly total precipitation
TPREC
Weekly total cumulative precipitation
W2PRE
Two week total cumulative precipitation
W3PRE
Three week total cumulative precipitation
PET
Potential evapotranspiration
MAl
Moisture availability index
SN
Percent sand
CL
Percent clay
SL
Percent silt
ELE
Elevation
MTEMP
Weekly mean air temperature
HUN
Weekly total heat units
THU
Weekly total cumulative heat units
RG
Climatic region/division
STA
Climatic station
UTM
Universal Transverse Mercator coordinate location
M15
Soil moisture at 15 cm depth
M61
Soil moisture at 61 cm depth
M91
Soil moisture at 91 cm depth
T15
Soil temperature at 15 cm depth
T61
Soil temperature at 61 cm depth
T91
Soil temperature at 91 cm depth
TABLE 3 (Continued)
M15T15
Cross product of soil moisture and soil temperature at 15 ern depth
M152
Soil moisture at 15 ern depth 2
M153
Soil moisture at 61 ern depth 3
M154
Soil moisture at 91 ern depth 4
evaluate the overall precipitation amount occurring at each site and the timing of the precipitation during the study period . A two-week (W2PRE) and a threeweek (W3PRE) total cumulative precipi tation were calculated for the 15 study weeks in order to investigate the impact of precipitation amounts received during prior weeks on the soil moisture and soil temperature conditions measured at the field sites. Potential evapotranspiration (PET) was calculated through use of the Christiansen method (Bordne and McGuinness, 1973). Meterologic data from three nearby first order stations were merged with data from the cooperative stations for the calculation of PET. The moisture availability index (MAl) was calculated by dividing total weekly precipitation by potential evapotranspira tion for the same time period (Wang, 1960). Soil samples collected at 15, 61, and 91 cm depths for each of the 12 sample sites were analyzed through particle size analysis. Percent sand (SN). silt (Sl), and clay (Cl) were calculated for the three depths at each site. Elevation (ElE) was measured for each site through use of topographic maps. Weekly mean air temperatures (MTEMP) were calculated from daily measures at specified cooperative weather stations. Heat units (HUN) were calculated by subtracting 10° C from daily mean temperature values to obtain daily heat units occurring at each site. The weekly total cumulative heat units (THU) were determined for each site for the 15 study weeks. The concept of heat units was utilized to provide a measure of the impact of prior
weeks on the soil conditions measured at a specific site. Each field site was described as to its climatic region (RG) and ordered from 1 through 12 as to location along the transect. The numbering of the sites was from an east to west direction. Finally, the Universal Transverse Mercator (UTM) easting coordinate was determined for each field site to further describe the longitudinal variation in the position of the field sites along the transect. M15, M61, and M91 represents soil moisture measures taken through use of the neutron probe at the depths of 15, 61, and 91 cm. T15, T61, T91 represent soil temperature values acquired at 15, 61, and 91 cm depths through use of thermocouple/psychrometers. ANALYSIS The overall objective of this study is to analyze the spatial and temporal characteristics of plant biomass through a transect located within a climatic transition zone. The analysis seeks to explain the relationships between plant biomass and selected temporal and spatial biophysical variables along an environmental gradient collected for the 12 sample sites distributed along a 200 km transect. Figure 2 presents a three-space plot of dry mean biomass amounts, sample week, and site location along the transect for site 1 (eastern site). site 6 (central site). and site 12 (western site). The three-space plot shows a decrease in dry mean biomass by weight in an east to west direction along the transect. This decrease in dry mean biomass is par19
DRY BIOMASS (g/O.lm2)
g/Olm'
60
40
5
20
6
o Figure 2 Plant biomass measured over the 15 week study period for sites 1, 6, and 12.
tially a result of observed density differences associated with regional differences in the transition from tall grass to mixed grass to short grass prairie in an east to west direction along the transect. Figure 3 presents graphics of mean air temperature, total precipitation, sample week, and dry mean biomass levels for sample sites 1, 6, and 12. Figure 3 shows that plant biomass increases with time at all three sites. Peak levels of plant biomass occur during week 10 (09 July 1985) for sites 1 an_d 6 and somewhat earlier at site 12 (02 July 1985). A second peak in plant biomass is suggested at sites 6 and 12 at the end of the sample period, 13 August 1985. Precipitation levels are greatest at site 1 and least at site 6. Normally, precipitation levels decrease in an east to west direction. The summer of 1985 produced atypical precipitation conditions both in magnitude and interval of the precipitation events. Figure 3 also shows that air temperature increases with time at
20
all three sites, and that the warmest temperatures are found at site 6 and the coolest at site 12. Correlation coefficients of plant biomass versus air temperature, precipitation, sample week, and site, are + 0.42, + 0.11, + 0.42, and - 0.43, respectively. The correlations are significant at the 0.05 level. The weekly precipitation totals across the transect illustrate an erratic distribution independent of biomass trends; however, monthly precipitation totals do reveal decr"easing spatial and temporal precipitation trends. McNaughton (1983) recognized that monthly or annual effects have a greater influence on biomass than individual weekly effects. Multiple regression analysis was used to explain the variation in plant biomass through the collection and measurement of selected biophysical variables. A stepwise "MAXR" technique was employed to derive the multiple regression model because numerous independent variables were to be considered during
II .....
All UMl'( .... n,,.(
A'I1fMl'{lAfUfll
I'tIECI'!TAIION
~IA'ION
,,",
35°C
o.7S'.
US,·
, o ).
.
60
s"c
."c
2.50, ,,,
1.50,.
'0
.0
30
30
-s't
-.'1:
U S c...
20
l,25,,",
20
10
10
..,0'" I
'2
3
A
5
6
7
8
9
10
II
"
13
w.
IS
1
WfU
,?,"o
",n.' ",tf'
,
3
-4
#'
~'\
'0,0.
5
6
1
'0...'"
." •
,,0'" ",rI' 9
10
,\''0
,\,\"J
11
l'
"It§> 'O~ "b,"J
13
••
15
WI"
_ A .. r~o''''.
..o.c.. ".'p'IO'_
_
Or,a-..
_
A " Ie",,,o ,o ,,.. .
.D..c.. ' ,u.p,Ial._
_
0.,"0"'0 ..
A'I'fl'O\1f1!"TUllf ~.E(I 'IT"ttON
35°, 6.75< ..
-, 15~
175,,",
.. .. 30
- . '\:
l25 c...
10
I
'2
3
..
5
6
7
8
9
10
11
I'
13
,.
15
WUI(
Site 12
_ A " I."'p-.ot ....
.D.c..
"'."""0'_
Figure 3_ Plant biomass, air temperature, precipitation, site, and week for sites 1, 6, and 12_
model building_ Table 3 shows the selected biophysical variables (independent variables) and codes used to iden-
tify each independent variable during computer analysis_ In addition to utilizing the 23 biophysical variables in the 21
regression analysis, the 2nd, 3rd, and 4th powers of each variable, and the crossproduct of all variable combinations were calculated and entered into the pool of variables for regression analysis. Soil moisture, soil temperature, and soil texture values collected at 15, 61 , and 91 cm depths were all entered into the pool of variables for regression analysis. Soil structure, while undoubtedly significant in its effect on soil porosity and permeability, and therefore on water avail ability for plants, was not quantified fo r use in the multiple regression analysis because of the relatively uniform characteristics of structure. The "MAXR " regression strategy attempts to identify the best one-variable model from the pool of derived variables which produces the highest Ff value. Another variable, producing the greatest increase in Ff, is then added. Once the two-variable model is ob tained, each of the variables in the model are compared to each variable in the pool not included in the model. For each comparison , " MAXR" determines if the Ff would increase if one variable was replaced by another selection . The appropriate substitution is made, if deemed necessary, to produce the largest in crease in R2. The comparison process continues until " MAXR" finds that no remaining substitution would increase the ~ . The user decides on the number of steps to be included in the regression , usually based on a minimal increase in Ff with additional steps or a minimal decrease in the sum of squares error (SAS, 1985). Table 4 presents the 12 step analysis of dry mean biomass (dependent vari able) regressed against selected environmental variables. The model, significant at the 0.05 level, explains 84.5 percent of the variation in dry mean biomass collected during the 15 study weeks (May 07-August 13, 1985) and for each of the 12 sample sites distributed along the 200 km transect. The independent variables selected for the model during the regression procedure show that soil texture at all three depths (15, 61, and 91 cm) , week of data collection, eastwest location of the sample site along 22
the transect, elevation of the sample site, soil moisture and so il temperature at 15 and 61 cm depths, heat units, and potential evapotranspiration are important elements in explaining variation in plant biomass in this analysis. The edaphic textural influence is related to the physics of water movement and water holding capacity of the soil for plant utilization. The emphasis on the 15 cm and 61 cm depths is associated with both texture and upper reaches of the horizons influenced by atmospheric fluctuations in moisture and temperature (i.e. heat unit factor) over time ; the 91 cm depth being significant only with respect to textural associations. The lower depths are not easily influenced by climatic fluctuations, but texture is still important to moisture availability for plant roots at depth. The availability of soil water is determined by the potential of soil water in the boundary layer closely surrounding the roots ; the finer the soil texture (i.e. clay at 91 cm), the higher the field capacity to retain water (Hillel, 1982). Texture is also influential in infiltration rates and evapotranspiration rates associated with increased ease of water movement with increased grain size . Structure undoubtedly interacted with soil texture in the field to produce variations in water retention, infiltration, and evapotranspiration rates, but as stated earlier was not quantified or entered into the model. The term " multicollinearity" refers to the situation in which there are strong inte rcorrelations among the independent variables. Although multicollinearity makes it difficult to assess partial effects of independent variables, it does not hinder the assessment of their joint effects. If newly added variables are highly correlated with those variables in the model, then the Sum of Squares Errors will not decrease very much, but the fit will not be poorer. So, the presence of multicollinearity does not diminish the goodness of the fit of the equation to the observed points. However, it may inhibit the ability to make predictions about the dependent variable, since the standard error of those predictions tends to become inflated (Agresti and Agresti, 1979).
TABLE 4 Multiple Regression Analysis of Dry Mean Biomass Variability Along the Sample Transect Dependent Variable
Beta Value
Independent Variable
*Dry Mean Biomass All Depths (15, 61, 91 cm) R2
=
.845
+49.22830
Intercept
+ 0.00001
SN614
- 0.00023
CL913
+ 0.00719
SN15SL15
- 0.02807
SN15WK
- 0.00001
SN61UTM
- 0.00037
SL91ELE
+ 6.79695
M15WK
- 0.03694
M15ELE
+ 0.19021
M61HUN
+ 0.05778
T61WK
- 0.35677
WKPET
+ 0.00001
WKUTM
*Significant at 0.05 level
The multiple regression models developed in this research indicate marginal effects of collinearity. To reduce or eliminate undesireable collinearity, the following strategy was employed. When polynominal expressions are introduced into the analysis, the mean of the variable can be subtracted from the variable in order to provide "standardization" of that variable. In this way, the entrance of )f into the model is not predicated on X entering the model first. Collinearity further can be appraised by looking for large changes in the parameter estimates when an additional variable is added to the regression model. No sig-
nificant fluctuations in the beta values were observed during model building. Correlation coefficients for the various combinations of independent variables were squared, thereby, indicating the percent of variability in one variable explained by another variable in a linear model. Large intercorrelations between variables could be detected and inclusion of such variables into the regression models could be prevented. Changes in F and R2 values produced by the inclusion of an additional variable into the model were evaluated since the relative "usefulness" of the variable in contributing to the accountability of the model
23
can be assessed . Model bu ilding was term inated when F and R2 fluctuations were insignificant. Fi nally, t he multiple regression algo rit hm uti l ized i n th i s analys is is not t he standard stepwise regress.ion procedure. During each iteration in the model bu ilding process, the enti re model is reevaluated and substitutions in the independent va riables can be made if the effect on t he ~ value is replicated by a re lated variabl e. The possibility of add ing a v ~ri able to the model wh ich is highly intercor related with a variable al ready in th e regression is unl ikel y. CONCLUS IONS The overall objective of this study was to ana lyze the tem poral and spatia l variability of plant biom ass sam pled for 15 study weeks alon g a tra nsect located with in a zone of cl imatic transition . Correlation co effi ci ent s indicate that dry mean biomass was sign if icant ly related to space and ti me va ri ab les, along the transect : site location (- 0.43), and sam ple week (+ 0.42). The three-space plot of biomass illustrated growth over ti me. Successional vegetati on stages are of initial growth (May), completed growth (beginn ing of July), wit h ea rl y reproductive stages in Jul y, followed by August flower initiation . Th e stages represent a bimodal seasonal growt h pattern ; grassland dom inated by both cool-season and warm-season species exhibit a bimodal growth pattern (Sims and Singh, 1978). Lauenroth (1979) suggested that as vari ation among sites increased, influence of soil properties, terra in orientation and elevation, site location and successional status increased in significance . Continual biomass fluctuations over time can be expected, corresponding to fluct uations in the cl imatic precipitation and temperatu re reg ime. The results of the multiple regression analysis showed that 84.5 percent of the variation in plant biomass could be explained by the group of selected envi ronmental variables collected for this analysis. Variables were selected on the basis of providing a temporal and spatial description of each site (week, site location, and elevation ); moisture inputs 24
(soil moisture at three depths, Moisture Ava ilab ility Index, and precipitation) ; energy inputs (air temperature, soil temperature, and heat units) ; soil conditions (so i l texture) ; and system outputs (evapotranspiration ). The 15 percent of the va riation in plant biomass that was unexpla ined could be examined in a future study that assesses other characteristics such as clay mineralogy, percent organic matter, and cation exchange capacity. The geographic position of the sample transect and the placement and condition of the individual sample sites are critical concerns with regard to the effect of space, time, and environmental factors on plant biomass. The dynamic environmental conditions which affect central Oklahoma provide appropriate conditions in which to exam ine plant biomass variations in a relatively short east- west distance. SELECTED REFERENCES Agresti, A., and B. F. Agresti. 1979. Statistical methods for the social sci ences. Dellen Publishing Company. San Francisco. Californ ia, 322- 350. Alaback. P. B. 1986. Biomass regression equations for understory plants in coastal Alaska : effects of species and sampling design on estimates. Northwest Science. 60:90103. Bordne, E. F., and J. L. McGuinness. 1973. Some procedu res for calculating potential evapotranspiration. Professional Geographer. 25:22- 28. Cable. Dwight R. 1975. Influence of precipitation on perennial grass product ion in t he semi-desert Sout hwest. Ecology. 56:981 - 986. DeJong , E., and K. B. MacDonald. 1975. The soil moisture regime under native grassland. Geoderma. 14(3):207-221 . Detling, James K. 1979. Processes controlling blue grama production on th e shortgrass prairie. Perspectives in grassland ecology, Ed., Norman French. Ecological Studies. 32:25-42. Doliner, L. H.. and P. A. Jolliffe. 1979. Ecolog ical evidence concerning the adaptive significance of the C. dicarboxylic aci d pathway of photosynthesis. Oecologia. 38:2334. French, N. R. 1979. Principal subsystem interactions in grasslands. Perspectives in grassland ecology, Ed .• Norman French. Ecological Studies. 32:173-190. Griffiths, J. F., and D. M. Driscoll. 1982. Moisture in the atmosphere, clouds. and preci pitation. Survey of Climatology. Columbus. Ohio. Charles E. Merrill Publishing Company. 103-117. Hillel, O. 1982. Introduction to soil physics. New York, Academic Press, 235-249.
Lauenroth, W. K. 1979. Grassland primary production : North American grasslands in perspective. Perspectives in grassland ecology, Ed., Norman French. Ecological Studies. 32:3- 24.
Sims, P. l., and J. S. Singh. 1978. The structure and function of ten western North American grasslands. Journal of Ecology. 66 :547-572.
Marotz. G. 1983. Atypical precipitation areas : comparisons along an environmental gradient. central United States. Physical Geography. 4(1):1-24.
Smoliak. S. 1956. Influence of climatic conditions on forage production of shortgrass rangelands. Journal of Range Management. 9:89-91 .
McNaughton. S. J. 1983. Serengeti grassland ecology: the role of composite environmenta l factors and contingency in community organization. Ecolog ical Monographs. 53(3) :291-320.
Wang , J. Y. 1960. A critique of the heat unit approach to plant response studies. Ecology. 41 :785-790.
Panciera, S. E.. S. J. Walsh, and J. O. Vitek. 1986. Preliminary analysis of spatial and temporal variations of soil moisture in west -central Oklahoma . Physical Geography. 7(3) :258- 274. SAS Institute Inc. 1985. SAS user's guide : Statistics, Vers ion 5 Edition, Cary, North Carolina. 956.
Webb. W., S. Szarek, W. Lauenroth, R. Kinerson, and M. Smith. 1978. Primary productivity and water use in native forest. grassland , and desert ecosystems . Ecology . 59(6) :1239-1247. White, Jon A.. and O. C. Glenn-Lewin. 1984. Regional and local variation in taUgrass prairie remnants of Iowa and eastern Nebraska. Vegetatio. 57:65- 78.
25