| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Dep. of Forest Resources, Univ. of Idaho, Moscow, ID 83844-1133 USA
b Dep. of Geography, Univ. of California, Santa Barbara, CA 93106 USA
c Dep. of Ecology, Evolution & Marine Biology, Univ. of California, Santa Barbara, CA, 93106 USA
paulg{at}uidaho.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterial and methodsResultsDiscussionREFERENCES |
|---|
Abbreviations: CEC, cation-exchange capacity • CTI, compound topographic index • Cmass, profile total C mass • DEM, digital elevation model • GIS, geographic information system • GPS, global positioning system • NPP, net primary productivity
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterial and methodsResultsDiscussionREFERENCES |
|---|
The advent of geographic information system (GIS)-based digital elevation modeling has facilitated quantitative modeling of catenas at local hillslope scales (McSweeney et al., 1994). Moore et al. (1993), building on the work of Speight (1968) and others, developed quantitative soil–landscape models using an integration of digital terrain analysis and statistical modeling methods to map predicted soil properties for a small hillslope in Colorado. In Australia, this work was continued and expanded to a broader suite of parent materials and geomorphic settings to test an integration of tools for digital terrain analysis, field sampling, remote sensing, statistical modeling, and GIS (Gessler et al., 1995, 1996; Gessler, 1996; McKenzie et al., 2000). These workers demonstrated the successful implementation of quantitative soil–landscape modeling on a broad landscape scale useful for soil resource inventory and as a framework for understanding soil–landscape function. Moore et al. (1993) and Gessler et al. (1995) showed strong correlation and predictive utility between a digital terrain index, the compound topographic index (CTI), and several soil properties. The CTI, often referred to as the steady-state wetness index, is a quantification of catenary landscape position. It integrates both landform position and context through an index defined as
 | (1) |
The purpose here was to continue development of the quantitative catena concept at the hillslope scale for a detailed study site in southern California. Whereas the earlier catena study by Moore et al. (1993) focused entirely on patterns of surficial soil properties (in the top 15 cm), we began the process of integrating catena patterns with pedologic processes. Specifically, we combined developed methods of digital terrain analysis, soil and ecosystem field sampling, statistical modeling, and pedological and geomorphological process interpretation.
Detailed hillslope characterization and modeling of digital terrain and soil property relationships is key to understanding the biogeochemical cycling of nutrient elements in terrestrial ecosystems because they often point to the spatial distribution of significant processes. Thus, one of our results is development of explicit soil–landscape models describing the distribution of ecosystem net primary production of C and soil C along the catena. The models provide preliminary context for developing hypotheses about soil–landscape function with respect to C cycling and water balance. We believe that understanding soil–landscape dynamics at the local hillslope catena level is key to understanding how to scale C dynamics to broader regional and global levels. In this paper, we provide a methodological blueprint for quantitative catena studies and results that have regional significance for soil survey and the biogeochemical interpretation of landscapes.
|
Material and methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterial and methodsResultsDiscussionREFERENCES |
|---|
|
Zero-Order Hillslope Geomorphometry and Sample Site Selection
The rectangular hillslope is 
100 m wide by 200 m long encompassing both convex and concave landform components spanning from ridge top to valley bottom (Fig. 1). A Trimble GPS 4400 Total Station Survey system (Trimble, 1996) was used to collect an 5-m spacing regular grid of easting, northing, and elevation points by taking measurements along transects over the study site. The reported accuracy of the equipment and survey type is ±1 cm in the horizontal plane and ±2 cm vertically, relative to the accuracy of a local basestation. The basestation location was surveyed with three-dimensional accuracy of 2 cm (Holmes, 1999), indicating that the GPS points may have as much as ±5 cm of absolute error in three dimensions. Relative accuracies are better.
The point coordinates were processed using the ANUDEM interpolation software (Hutchinson, 1989, 1996) to develop five digital elevation models (DEMs) at grid point spacings of 2, 4, 6, 8, and 10 m. These were developed to compare the utility of digital terrain attributes for predicting soil patterns at different grid-point spacing resolutions. Primary and secondary terrain attributes (Moore et al., 1991) that quantify landform (e.g., slope gradient, profile and plan curvatures, flow direction, flow accumulation, CTI, and upslope mean statistics) were computed for each DEM using the TAPESG and UPSUMG software programs (Moore et al., 1993; Gessler et al., 1995; Gallant and Wilson, 1996). Terrain attributes provide a quantification of landform shape, connectivities and adjacencies that define external landscape geomorphometry and water flow patterns.
Histograms and summary statistics that characterize the distribution of the quantitative topographic attributes along with spatial displays were used to guide the distribution of 20 sample locations across the range of elevation, slope, and landform curvatures. These preliminary soil pits indicated that the surface soil patterns strongly reflected landform curvatures; therefore, a subset of nine sample locations (Fig. 1; Pits 1, 2, 4, 8, 10, 12, 14, 18, 19) encompassing this variation was chosen for soil pit excavation to bedrock or 3 m, whichever occurred first. An additional soil pit, located on the planar surface of the terrace at the top of the catena was not used in the hillslope modeling, but was used as an index of soil C content on relatively level ground. Depth to bedrock was determined directly in the backhoe and hand-dug pits. In concave positions where the soil material was >3 m in depth, estimations of soil depth were made using a truck-mounted Geoprobe that measures electrical resistivity of lithologic layers (Geoprobe Systems, 1995).
Soil Morphological Descriptions, Sampling, and Laboratory Analyses
Soil profiles were described using standard procedures (Soil Survey Staff, 1993) and horizons were channel sampled at several locations along pit faces for chemical and particle-size analyses. Intact clods (2–4 per horizon) were excavated from each horizon for bulk density measurement. Samples were oven-dried, crushed, sieved, and the >2-mm portion was weighed. Particle-size and chemical analyses were completed using the <2-mm fraction. Organic C and N were measured using a Fisons Instruments NA 1500 Series 2 CN analyzer (Fisons Scientific Equipment, Loughborough, UK) at the University of California–Santa Barbara. Bulk density was measured at University California–Santa Barbara by the saran coated clod method (Soil Survey Staff, 1996). Texture, pH, cation-exchange capacity (CEC) and base saturation analyses were completed by the University of Missouri Soil Analysis Lab. Particle size was measured by the pipette method (Soil Survey Staff, 1996). Soil pH was determined both by CaCl2 and by water extract. Base concentrations and CEC were determined by the NH4OAc method and CEC was also calculated as the sum of bases (Soil Survey Staff, 1996).
The mass of C in each horizon was calculated by multiplying the measured organic C percentage, corrected for gravel content, by the measured bulk density. Bulk density was determined as the mean of at least two clod samples per horizon. The mass of C was then multiplied by soil horizon thickness and added to provide an estimate of the mass of C stored (g cm-2) in the entire soil profile. The profile C mass (Cmass) is a measure of the C stored in a 1-cm2 area to the depth of the entire soil profile (A + B horizons).
Site Vegetation Sampling—Net Primary Productivity (NPP)
Aboveground annual grassland productivity was determined at the height of the annual growing season in April 1997. Tree biomass was not sampled because of its sparse areal extent on the study site. Clip-plots (20 by 50 cm; 0.1 m2) were collected to bare soil and sorted according to live species composition (biomass) and litter. Plant material was dried at 65°C for 48 h and weighed. The NPP for 1997 was determined as grams of biomass per square meter at each sample location. Duplicate plots were placed 1 m above each of 20 pit faces to correlate NPP with soil properties along the hillslope. All sample sites were geo-referenced with the Trimble GPS described above. Data from the nine soil pit locations are used here.
Statistical Analysis and Soil–Landscape Modeling
Statistical summaries, exploratory data analysis, and statistical modeling for derivation of quantitative soil–landscape models were accomplished using the Splus statistical analysis package (Mathsoft, 1999). Regression based soil–landscape models of A horizon depth, soil depth, Cmass, and NPP were developed. We use models of A horizon depth and soil depth to provide initial approximations of the soil–landscape structure across the catena hillslope [Albic (E) horizons do not exist on this hillslope]. The A horizon is the most biologically active layer where most soil C is stored. Its thickness is indicative of biological activity, but it also provides clues to the location of erosion and deposition on the hillslope. Soil depth, defined here as the depth of the overall A horizon plus B horizon(s), provides an indication of water storage capacity, nutrient storage pools, overall productivity, and longer-term erosion, deposition, and soil building processes. Overall, these variables provide a basis for hypothesizing the spatial distribution of process zones and landscape function. Finally, we develop spatial models of Cmass and NPP to represent the distribution of C in the soil and in the aboveground biomass and compare them with the soil property models.
Empirical regression models were developed using correlations between the response variables (A horizon depth, soil depth, NPP, Cmass) and the corresponding quantitative terrain attributes (explanatory variables) computed for each sample site location. Scatterplot matrices (Cleveland, 1993; Moore et al., 1993) were used for exploratory data analysis to complement forward and backward stepwise regression searches for useful combinations of predictor terrain attributes. The regression searches employed the Akaike Information Criterion (Akaike, 1974), which develops a statistic using the residual deviance penalized by the number of parameters requiring estimation in a model fit. Model fits were then developed and intercept terms, regression model coefficients, and goodness of fit parameters were output. Due to the small sample number, only linear relationships were used for modeling. The intercept term and regression coefficients were used to create prediction surfaces for the response variables using grid-based map algebra tools.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterial and methodsResultsDiscussionREFERENCES |
|---|
|
Hillslope Data Synthesis
The soil texture classes from Table 1 show a generally uniform distribution of clay loam textures with slightly coarser textures toward the profile surface. Change in clay percentages down the soil profile suggests pedogenic eluviation–illuviation processes, particularly in the upper slope profiles. The concave profiles appear well mixed. Silt percentages are generally highest in the surface horizons and there is more silt in the lower landscape positions. Sand values are greatest in upper landscape positions. Gravel percentages are greater in the surface of the two mid-slope concave (Pits 10 and 14) positions than elsewhere, suggesting past high-energy erosional events or proximity to steep sideslopes. Gravel is also greater in the deeper horizons of soils in upper hillslope positions because of parent material weathering and mixing. Sand/clay ratios are greater in convex and shoulder hillslope pits because of a combination of down-profile and downslope removal of finer materials. Bulk densities are greater in the lower A horizons of upper slope positions than in concave positions. This suggests greater water storage capacities in concave positions because of increased pore space and soil depth.
Soil pH is generally neutral in the deeper horizons of concave landscape positions, which can be explained by the downslope and down-profile increase in base cations (Table 1). Cation-exchange capacities follow clay content and cluster around 30 cmolc kg-1. If we assume average clay content of 30%, the CEC of the clay fraction would be around 100 cmolc kg-1, which suggests the presence of 2:1 clay minerals in all landscape positions. Table 2 shows the correlation coefficients among chemical and physical properties presented in Table 1. As expected, clay content is highly correlated with CEC, whereas soil organic C and N are negatively related to clay and bulk density.
|
|
|
|
|
|
The best model for soil depth combined the CTI with upslope mean profile curvature. Profile curvature is the rate of change of slope in a downslope direction. It characterizes changes in flow acceleration that may differentiate erosion and deposition zones in landscapes. The upslope mean profile curvature variable therefore provides a quantification of relative erosional vs. depositional contributing areas. The correlation between the model coefficients of CTI and upslope mean profile curvature is only 0.37, suggesting that they provide complementary explanatory utility in the model (Table 3, Model 3). However, the simpler model using CTI alone results in an R2 of 0.84, a very strong and easily derived correlation useful for modeling soil depth across the hillslope catena.
The two best models for NPP use CTI and flow accumulation as single linear predictors. Correlations are not as strong as for other variables. The scatterplots in Fig. 4 suggest that the relationship may be stronger at small CTI values but that the residual variance increases with larger CTI values when modeled with a linear fit. A much larger sample or alternative statistical methods that account for changes in residual variance would be required to test this hypothesis. The use of the flow-accumulation attribute implies that lateral redistribution of water in these landscapes results in higher plant productivity.
The best models for profile total C are derived by combinations of slope and flow accumulation. The combination of these individual variables provides only a slight improvement on a model using the CTI alone. This predictive relationship is very strong as suggested by the scatterplots in Fig. 4, accounting for 78% of the variation in C stored in the soil of this catena. The amount of soil C stored in the soil profile is probably the result of both plant productivity and lateral redistribution of materials over the hillslope. Again, this relationship emphasizes the potential for quantitative modeling of Cmass using easily derived digital terrain attributes.
Spatial Implementation
Model coefficients and intercept terms from Table 3 were used to implement models for the 2-m grid point DEM using map algebra to provide predicted surfaces for each of the four predicted soil–landscape variables (Fig. 5)
. The four predicted surfaces follow pedologically logical patterns as concavity increases in the downslope direction. The regression model patterns vary smoothly without abrupt changes except as an artifact along the footslope of the catena where sampling was minimal. Regression models inherently smooth information; implementation of different model types such as decision trees can produce much sharper boundaries (Gessler, 1996). Catena topography follows relatively smooth transitions that are well represented by regression models.
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterial and methodsResultsDiscussionREFERENCES |
|---|
A horizon depth varies from 8 to 80 cm and soil depth varies from 8 to >450 cm across the hillslope. This variation occurs within 10 to 15 m of horizontal distance representing changes in soil properties that could never be incorporated into a soil survey map at standard cartographic scales (i.e., 1:15 840; 1:20 000). A horizon depth and soil depth result from an integration of pedogeomorphic processes through time. A horizon depth may be more indicative of an integration of shorter-term processes relating to recent surficial dynamics and biotic activity. Soil depth represents a longer-term integration of processes that could include more than one event of A horizon development and/or the dynamics of A horizon gains and losses along with the deeper weathering of parent materials and hydrological dynamics. The A horizon and soil depth soil–landscape models are similar to other work in Australia (Gessler, 1996) where soil depth was more predictable than A horizon depth using a similar set of explanatory terrain variables. The results here suggest that external landform is an important modifier of long-term landscape processes and should be explicitly incorporated into biogeochemical cycling models for this region. The weaker relationship between A horizon depth and external landform suggests that shorter-term processes are more random and variable.
The distribution of soil C along the catena is well represented by our models; there is more C in the deeper soils in the downslope concave positions. We noted that the C/N ratios for the soil profiles split into an upslope, smooth to convex grouping, and a downslope concave group. The plots implied greater addition of relatively fresh C to the latter group. If C accumulation is favored on some parts of the catena compared with others, how does the Cmass in the diversity of hillslope positions compare with that on a flat surface that is less subject to mass and water movement? Is more C sequestered in the concave positions than is lost from the convex positions by erosion? We tried answering this question by measuring Cmass in soil on the planar terrace top whose morphology is very similar to those on the upper shoulder positions (argillic and mollic diagnostic horizons). Calculation of Cmass using the same methods outlined above gives an estimate of 121 g cm-2 for the ridgetop. Comparing this value with those listed in Table 1 indicates loss sites (Pits 1, 2, 4, 8, and 12), and gain sites (Pits 10, 14, 18, and 19), matching those differentiated by the two separate regression lines illustrated in Fig. 3. The average Cmass for the sample sites across the hillslope is 110 g cm-2. This suggests a relatively small loss of C on the hillslope compared with the ridgetop if we use simple spatial averaging. On the other hand, because the Cmass-landscape model incorporates CTI, it provides a more appropriate representation of the spatial variation of Cmass and produces a landscape-weighted average value of 99 g cm-2, suggesting a greater net loss of C on the hillslope compared with the ridgetop. We do not wish to overinterpret this back-of-the-envelope calculation, but these initial estimates provide comparative values helpful in development of more rigorous mass balance monitoring.
Aboveground net primary productivity provides a measure of robustness for short-term biotic processes that integrate water balance and nutrient availability, in contrast to Cmass, which is a result of longer-term mass balance in landscapes. The NPP landscape model was less reliable than the others for various possible reasons. First, plants have an inherent plasticity in their response to environmental variation. Second, subtle changes in species composition with microtopography may alter NPP in poorly understood ways. Finally, while trees were relatively sparse on the catena, their exclusion from our measurement protocol limits our interpretive ability.
Although the modeling focused on use of terrain attributes computed at 2-m grid spacing, exploratory data analysis suggested that the functional relationships hold at broader gridpoint spacings. The range of each potential explanatory terrain attribute constricts with increases in grid spacing. Implementation of the model coefficients and intercept values for spatial prediction in some instances produced negative and unrealistic values. However, extrapolation beyond the range of the explanatory variables generated nonrealistic values. In scaling models to other areas, it is important to sample the entire range of explanatory variables during model development. Future research needs to evaluate these issues carefully as they relate to extrapolation of developed models.
Implications for Modeling Soil–Landscape Function and Process
Our results show a tremendous amount of variation in basic soil properties that strongly influence C distribution and dynamics in this 2-ha catena. This variation relates strongly to landform as shown by comparative evaluation of soil classes, soil physical and chemical data, and measurements of A horizon depth, soil depth, and net primary production and mass of C stored in the soil profile. Our sample size was small—a constant problem in pedology. Yet, correlation of soil properties with digital terrain attributes (landform) can be quantified statistically and applied with a GIS to predict values across landscapes. Our procedures provide explicit and quantitative models that can be tested, updated, and used for testing hypotheses about the spatial distribution of landscape and ecosystem processes. They can also be used for parameterizing spatially distributed hydrological and ecosystem simulation models to help understand landscape dynamics that drive biogeochemical cycling.
Soil processes controlling C content and flux are complex. They are usually studied in detail locally and extrapolated using simulation approaches. Regional to global distribution of soil C content and turnover are often estimated using simulation models that are based on a few easily measured soil and ecosystem parameters (e.g., the CENTURY model, Parton et al., 1987). Simulation models provide greater flexibility than earlier approaches where properties and processes measured in broadly defined ecosystems are scaled up using global maps of ecosystem distribution (e.g., Ajtay et al., 1979; Post et al., 1982). Inherent in the simulation model approach as implemented by CENTURY is the assumption that climate controls vegetation distribution which in turn controls soil C properties. Macroclimate is assumed the dominant control on soil C dynamics. In flat to gently sloping terrain, this is probably valid, but in most hillslope areas there is considerable redistribution of water by overland flow or subsurface flow. On hillslopes, local differences in soil processes related to differences in topographic position are underestimated. The effect of landscape topography is an important missing component in simulation models.
Ignoring the effect of topography on ecosystem processes introduces a serious weakness in present approaches to regional and global C modeling and probably exacerbates our inability to account for all the sinks of anthropogenically produced C. Water distribution in landscapes tightly controls soil C dynamics. In fact, water redistribution between convex and concave landscape positions, as demonstrated here, may account for differences in soil C similar to dramatically different climatic zones. Topography and its influence on hydrology and soil C dynamics has been difficult to quantify because of the continuously varying nature of hillslopes and our inability to sample and model the variation in a systematic fashion.
Through the power of digital elevation modeling implemented within a geographic information system, we can explicitly incorporate surficial processes into the hillslope context. To continue this effort we have instrumented the soil pits described here for monitoring soil water, gas, and temperatures fluxes over time. This will enable the testing of specific hypotheses about catenary soil–landscape function and biogeochemical cycling.Holmes Chadwick Kyriakidis 2000; Trimble Navigation Ltd 1996; Soil Conservation Service 1972
|
ACKNOWLEDGMENTS |
|---|
Received for publication September 28, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterial and methodsResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  P. Jiang, N. R. Kitchen, S. H. Anderson, E. J. Sadler, and K. A. Sudduth Estimating Plant-Available Water Using the Simple Inverse Yield Model for Claypan Landscapes Agron. J., May 7, 2008; 100(3): 830 - 836. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. P. D'Acqui, C. A. Santi, and F. Maselli Use of Ecosystem Information to Improve Soil Organic Carbon Mapping of a Mediterranean Island J. Environ. Qual., January 9, 2007; 36(1): 262 - 271. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. N. Kravchenko and G. P. Robertson Can Topographical and Yield Data Substantially Improve Total Soil Carbon Mapping by Regression Kriging? Agron. J., January 1, 2007; 99(1): 12 - 17. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. N. Williamson, P. E. Gessler, P. J. Shouse, and R. C. Graham Pedogenesis-Terrain Links in Zero-Order Watersheds after Chaparral to Grass Vegetation Conversion Soil Sci. Soc. Am. J., October 27, 2006; 70(6): 2065 - 2074. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. N. Kravchenko, G. P. Robertson, X. Hao, and D. G. Bullock Management Practice Effects on Surface Total Carbon: Differences in Spatial Variability Patterns Agron. J., October 3, 2006; 98(6): 1559 - 1568. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. D. Tomer, C. A. Cambardella, D. E. James, and T. B. Moorman Surface-Soil Properties and Water Contents across Two Watersheds with Contrasting Tillage Histories Soil Sci. Soc. Am. J., February 27, 2006; 70(2): 620 - 630. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. M. Ziadat Analyzing Digital Terrain Attributes to Predict Soil Attributes for a Relatively Large Area Soil Sci. Soc. Am. J., August 25, 2005; 69(5): 1590 - 1599. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. A. Thompson and R. K. Kolka Soil Carbon Storage Estimation in a Forested Watershed using Quantitative Soil-Landscape Modeling Soil Sci. Soc. Am. J., June 2, 2005; 69(4): 1086 - 1093. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. D. Tomer and D. E. James Do Soil Surveys and Terrain Analyses Identify Similar Priority Sites for Conservation? Soil Sci. Soc. Am. J., November 1, 2004; 68(6): 1905 - 1915. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. R. Schepers, J. F. Shanahan, M. A. Liebig, J. S. Schepers, S. H. Johnson, and A. Luchiari Jr. Appropriateness of Management Zones for Characterizing Spatial Variability of Soil Properties and Irrigated Corn Yields across Years Agron. J., January 1, 2004; 96(1): 195 - 203. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Fierer and J. P. Schimel A Proposed Mechanism for the Pulse in Carbon Dioxide Production Commonly Observed Following the Rapid Rewetting of a Dry Soil Soil Sci. Soc. Am. J., May 1, 2003; 67(3): 798 - 805. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Chamran, P. E. Gessler, and O. A. Chadwick Spatially Explicit Treatment of Soil-Water Dynamics along a Semiarid Catena Soil Sci. Soc. Am. J., September 1, 2002; 66(5): 1571 - 1583. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M. Schoorl, A. Veldkamp, and J. Bouma Modeling Water and Soil Redistribution in a Dynamic Landscape Context Soil Sci. Soc. Am. J., September 1, 2002; 66(5): 1610 - 1619. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. G. Fierer and E. J. Gabet Carbon and Nitrogen Losses by Surface Runoff following Changes in Vegetation J. Environ. Qual., July 1, 2002; 31(4): 1207 - 1213. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. N. Kravchenko and D. G. Bullock Spatial Variability of Soybean Quality Data as a Function of Field Topography: I. Spatial Data Analysis Crop Sci., May 1, 2002; 42(3): 804 - 815. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. N. Kravchenko and D. G. Bullock Spatial Variability of Soybean Quality Data as a Function of Field Topography: II. A Proposed Technique for Calculating the Size of the Area for Differential Soybean Harvest Crop Sci., May 1, 2002; 42(3): 816 - 821. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
