MONITORING FOREST SOIL PROPERTIES TO MAINTAIN PRODUCTIVITY

J. M. Geist
R. T. Meurisse
T. A. Max

ABSTRACT

This paper addresses considerations necessary to construct a scientifically based sampling system for monitoring soils. A statistically sound method, which uses line transects, is described, and several applications for various sampling objectives are discussed. Results from sampling forest harvest units illustrate how data variability and choice of statistical precision levels affect sample size requirements. The transect system has been found to be an objective and easy-to-apply approach to monitoring forest soils in a variety of situations.

INTRODUCTION

Various Federal legislation directs natural resource management agencies to steward the public land in a manner that maintains or enhances productivity. The Multiple-Use Sustained Yield Act of 1960 called for management of renewable resources of the National Forests without impairment of land productivity, and more recently, the National Forest Management Act of 1976 emphasized protection and improvement of soil resources. The latter act also included instructions to monitor the effectiveness of management in meeting planned goals. Such monitoring includes soil resource monitoring.

Monitoring needs may vary relative to management objectives and the soil characteristics measured. To date, however, most forest soil monitoring has focused on alteration of physical properties. Ultimately, there is a need to translate soil status to site productivity commonly measured as tree growth. To do this, research is needed to establish relations between specific changes in soil condition and plant growth. Managers, then, have two major jobs: detection of changes in soil characteristics through monitoring, and interpretation of the effects of these changes on site productivity based on growth or other measures provided by research. It is imperative that research scientists and land managers work together to improve detection of changes in soil characteristics and to translate these changes into effects on forest soil productivity. Improved assessment methodology could be developed as a part of monitoring, as an independent research effort, or a combination of both. In this paper, we will only discuss monitoring to detect changes in soil characteristics.

A sound monitoring system should contain several key elements to ensure the sampling system will accurately reflect the properties of the area(s) assessed. These key elements include:

  1. Ability to objectively assess soil characteristics regardless of their spatial distribution.
  2. Unbiased choices of sampling locations.
  3. Well-distributed coverage of the monitored area.
  4. Flexibility in the kind of information gathered and the tools that might be used.
  5. Statistically valid sampling design to provide objective estimates of sampling error; statistical tests, if needed; and methods for controlling precision of estimates.

STANDARDS AND EXAMPLE SYSTEM

A transect system that includes these key elements and utilizes existing statistical theory for monitoring forest soils in the northwestern United States was developed and described by Hazard and Geist (1984). The rationale and steps used were later related to National Forest planning by Miller and Hazard (1988). The field and office procedures for conducting the monitoring were published as a "how-to" guide for soils specialists and other monitoring personnel of the Forest Service (Howes and others 1983). The transect system was subsequently adopted as the "standard" for monitoring soil physical conditions of National Forests in the Pacific Northwest Region, and formed the basis for nationwide soil monitoring methods adopted by the Forest Service in the National Soil Management Handbook (FSH 2509.18).

Various standards or guideposts are a necessary part of monitoring so the effectiveness of management prescriptions can be judged relative to provisions of forest plans. These standards relate to limits of overall operation or individual operational effects, which on the basis of biological or productivity knowledge reflect the tolerance of soil-vegetation systems to manipulations of various kinds.

The Pacific Northwest Region, and the Forest Service nationally, have adopted several standards for maintaining both long- and short-term soil productivity. The Pacific Northwest regional standards are based on the premise that productivity is a function of several state variables including available soil moisture, available soil nutrients, and soil aeration as discussed by Meurisse (1988). Further discussion about the soil capacity factors is included in these proceedings (Meurisse and others). The national standard for determining significant changes in productivity of the land is from a consensus of experts based on available research and current technology. National Forests are encouraged to establish standards of their own, if justifiable. The general standard for National Forests in the Pacific Northwest Region is: "a minimum of 80 percent of an activity area, including transportation system, should be left in a condition of acceptable (non detrimental) productivity potential for trees and other managed vegetation following land management activities." Defined detrimental conditions are:

  1. Compacted—an increase in soil bulk density of 20 percent or more in volcanic ash and pumice soils, or an increase in soil bulk density of 15 percent or more or macropore space reduction of 50 percent or more in other soils.
  2. Puddled—depth of rutting is 6 inches or more.
  3. Displaced—removal of 50 percent of the topsoil or humus-enriched A1 and/or AC horizons from an area 100 square feet or more which is at least 5 feet in width.
  4. Severely burned soils—those where the top layer of mineral soil has been changed in color, usually to red, and the next half-inch blackened from organic matter charring.

More detailed descriptions of these standards are included in Meurisse (1988).

This monitoring system and standards have been widely used in National Forests of the Pacific Northwest, primarily in monitoring soil compaction and displacement associated with timber harvesting. Sullivan (1988) monitored other effects as well. Briefly, the system uses a systematic grid of points that is randomly positioned over the area to be sampled. Randomly oriented transects radiate from each grid point (fig. 1). Surface and subsurface monitoring information is gathered by categorizing segments of each transect line according to soil condition and by collecting soil samples at intervals along the line. A 100-foot transect length is used. As a result, summed segment lengths, expressed in feet, are percentages of transect lines in a given soil condition. The average, by condition, over all transects represents the percentage of area in that condition for the monitored unit. Other line lengths can be used, so long as measurements are converted to percentages, but we have found that 100 feet is an adequate length. Soil condition percentages by line are used to calculate statistical attributes of the sampled area, such as means, standard deviations, and standard errors. Interval samples along the transects are used for assessments like soil compaction, for which intact cores or penetration readings may be taken. The system can also accommodate fertility samples. In the case of compaction, core samples are extracted from a specific depth or soil horizon at 10-foot intervals beginning at the 5-foot mark. Bulk density values are determined and then compared to the average bulk density of additional random transects (at least three) located on an untreated area (one which provides an appropriate predisturbance background value), probably adjacent to the sampled area. The average value of bulk density from transects in undisturbed areas plus 15 or 20 percent (that is, undisturbed average multiplied by 1.15 or 1.20) represents the lower limit of unacceptable or detrimental degree of compaction. Each core sample represents 10 percent of each transect length. Thus, if three core samples of a line transect exceed the bulk density limit, then 30 percent of the line, or area the transect represents, has been detrimentally compacted.

Figure 1—An example of the randomly positioned, systematic grid with randomly oriented transects. [view larger image - 24K] [Text description of this figure]

Diagram showing 15 transects within a study plot.

The soil condition categories monitored by practitioners in the Pacific Northwest vary somewhat, but they include slash piles, landings, skid trails, spur roads, compacted, displaced, puddled, eroded, miscellaneous, deposited, and undisturbed. Categories can readily be changed, so they apply to a wide variety of situations. The latter two categories are not considered to be detrimental.

Two computer programs were developed to save computation time. One calculated bulk densities from core measurements (Starr and Geist 1988), and the second summarized the data and computed statistical estimates (Hazard and others 1985). The calculated variance for overall detrimentally affected area provides users a statistical test for whether general limits of total damage were exceeded. A statistical test may not be needed or desired in some cases, but this option is available. The same program can compute estimates of numbers of samples (number of transects) needed to satisfy specified precision requirements. We present example data from using this option later in the paper.

APPLICATIONS

There are a number of ways this sampling scheme can be applied in meeting monitoring needs, and this flexibility increases its usability for land managers. We present some examples here.

Single-Activity Area Evaluations

To date, this system has primarily been used to estimate the final status of soil characteristics for an individual area, like one unit of a timber sale (fig. 1). Repeated application of the sampling design on the same unit can be used to assess the effects of a series of activities either individually or collectively (that is, cumulative effects). Repeated measures optimize sensitivity for detecting change by reusing the same grid points and transects at each sampling time. When performing repeated measurements, slight offsets are required where destructive samples are obtained; the procedural decisions for selecting the offset positions are part of the rules established before fieldwork begins.

Comparisons Among Areas

Sometimes the question arises: Are the average values for soil conditions significantly different among units? Statistics can help handle the variability that often confuses our ability to otherwise make objective comparisons. Monitoring is carried out on each of the units using the same sampling approach as for a single unit. The percentages by line are the observations used to statistically test differences among units. An unpaired t-test would be an appropriate statistical test to contrast two units. Comparisons among more than two units would require a different statistical approach (for example, a multiple comparison procedure) to properly account for multiple nonindependent statistical tests.

A word of caution is noteworthy here. Comparisons among single-unit means would not suffice as a comparison of different management activities or treatments that happened to be associated with each mean. This is because no measure of variability exists for the activities. We address activity comparisons next.

Sampling to Compare Differences Among Activities

Where a comparison of the effects of two (or more) kinds of activities on soil properties is desired (for example, random skidding vs. designated skid trails), sets of management units are chosen for each practice. Activities are assumed to be applied to the whole unit. Each set of units represents the range of conditions over which one of the activities occur (the sample of units represents a well-defined population of interest). The procedures for transect-sampling each area are the same as described for an individual unit. The unit average of transect percentages represents a single observation, a replicate, in the array of unit averages for a soil condition associated with the activity. The two arrays of averages are used for statistically comparing activities. A one-way analysis of variance would be an appropriate statistical test. The number of units used to assess the effects of each activity should be similar but not necessarily equal. Sensitivity of the test to detect differences is increased as the number of units (n) increases, but also depends on the magnitudes of the true differences among activities. Auxiliary information to identify differences among units, independent of the activity, may be used in covariance analysis or blocking to increase the sensitivity of the statistical comparison.

Stratified Sampling

Stratification can increase the efficiency of sampling. In general, improved efficiency occurs when a population can be divided into strata so that the mean values (of the variable of interest) differ among the strata, yet strata are relatively homogeneous. The transect system can be used here too, but the user may need to apply procedures for shifting transect positions because they cannot cross strata boundaries.

Sampling soil compaction is an example where stratification might work well. Much of the compaction occurring on a logged area may be concentrated in rather easily discernible skid trails. In such a case, two strata would be defined as skid trails and nonskid trails. The expectation is that the percentage of area compacted on skid trails is much higher than for the rest of the unit. Variation within each strata would be less than variation across the unit as a whole. Hence, we expect stratified sampling to be more efficient in this case, if it is not too expensive to determine strata boundaries.

Part of the increased efficiency of stratified sampling results from more information about the population being included in the estimators. In particular, strata sizes (or at least their relative sizes) must be known, which requires area determinations. Areas must be measured easily and cheaply for stratified sampling to be efficient.

A remedy for transects crossing boundaries would be to reorient them by choosing a new random direction for each. Other remedies might include a different grid size, a shorter transect length, or a different transect configuration for the skid trail stratum. Such solutions may require additional research to verify statistical adequacy.

Double Sampling

Core sampling is an accurate method of evaluating soil compaction, but it is too time consuming to meet the rapid survey needs that commonly arise. In such cases we would encourage exploratory tests of rapid surveys using double sampling. For double sampling, a large and a small sample are taken. The small sample is taken using a standard, more time-consuming method. The large sample is taken using a faster, less precise method. The small sample is a subset of the large sample, hence the term double sampling. For example, we will use core sampling and probing with a sharpshooter shovel as the two methods to assess compacted percentages of transects (still the basic sampling unit). The sampler must first obtain a "feel" for excessively compacted conditions in a controlled area where conditions are known. The sharpshooter measurements would be categorical, that is, classified above or below the specified detrimental compaction limit, rather than continuous values. Once an individual develops the needed feel for a given soil, rechecking calibration will require less core sampling. In this application, the large sample is taken by probing sample points along all transects on the area to determine the compacted percentage of each transect. The small sample, of transects selected systematically or randomly from all transects on the area, is taken by core sampling at the same sample points, offset to avoid disturbance from the shovel. Compaction percentages of the subsample transects are then computed from the core sampling data. The correlation between transect proportions, as determined by the two methods for individual lines in the subsample, as well as the relative costs of the two measurement techniques, determine the efficiency of double sampling and the optimal size of the small sample.

Details about the theory of double sampling are covered by Cochran (1977). Some helpful discussions on the use of double sampling are given by Geist and Hazard (1975), even though the application is different. Another potential application of double sampling was reported by Clayton and others (1987), who used a penetrometer and sharpshooter shovel to classify compaction.

RESULTS FROM SINGLE-AREA SAMPLING

The transect sampling system was developed in cooperation with the Pacific Northwest Region of the Forest Service, and the Umatilla, Malheur, and Wallowa-Whitman National Forests. One objective was to assess the amount of detrimental conditions (excluding transportation system in this case) for an array of 11 harvest units (Geist and others 1989). These efforts tested the system, gave useful information on harvest effects, and initiated a data base for the variability of soil characteristics. This information provides a basis for estimating numbers of samples at prescribed levels of precision at specified statistical probabilities. For example, the number of transects (grid points) needed to determine the average detrimentally impacted area within 20 percent error at two levels of probability are shown in table 1. More transects are necessary when variability is higher, when a higher probability is desired (90 percent vs. 80 percent) or when a smaller percentage of error is desired. An 80 percent probability translates to a 1 in 5 chance of error; a 90 percent probability means a 1 in 10 chance.

TABLE 1 
Number of transects required to estimate average detrimental soil conditions within 20 percent of the true mean percentage, at two probability levels for 11 harvest units (adapted from Geist and others 1989)
Harvest unit Summary statistics1 Probability level
X
Percent of area
s
Percent of area
0.80
No. of transects
0.90
No. of transects

1Mean and standard deviation, respectively, of percent area in detrimental condition.

1 25 20 27 45
2 21 25 58 95
3 44 25 13 21
4 37 17 8 13
5 27 14 12 20
6 23 18 27 43
7 21 15 21 35
8 19 15 24 40
9 34 22 17 28
10 41 15 5 9
11 24 16 20 33

Sullivan (1988) used this monitoring system to assess soil physical conditions following tractor logging and machine piling in 24 timber sale units in the Malheur National Forest in Oregon. He determined the "Average Percent Detrimental Impact" (APDI) and found that 15 of the units exceeded the general regional standard of 20 percent area. He summarized results in various ways: by general soil type, by timber type, by silvicultural prescription, and by previous activity. The latter involved repeated measures to determine cumulative effect. Some 95 percent of the APDI detected was from compaction. The results of his sampling have been used to take preventive or remedial measures including reducing the degree of slash disposal, using designated skid trails approved prior to tree felling, and using a winged subsoiler to treat portions of units that exceeded the general standard.

In a spinoff application of the sampling system, Cochran and Brock (1985) used the point grid and transect approach to study soil compaction and growth of young trees planted in two clearcuts. Growth measurements of trees within 1.2 m of the transects were compared to soil bulk density near the tree. Tree growth and bulk density data were summarized on a transect basis, and regression analysis was used to test relations between variables.

CONCLUSIONS

The sampling system discussed here should not be considered the last word in strategies for soil monitoring; however, it is flexible, objective, and easy to apply. It is hoped these pioneering efforts will evolve into progressively better sampling methods. If other applications of this system are explored, statistical consultation should be sought to help meet needs and retain statistical validity.

Those specialists who become involved in monitoring natural resources often find knowledge gaps that limit their ability to measure and interpret changes in productivity or other environmental attributes. Some knowledge gaps will not be filled soon, and those who seek to fill them will find a challenging arena for investigation because of the complex interactions and frequently confounded effects of management on ecosystems. Studies relating adverse soil conditions to plant growth have commonly been after-the-fact observational studies rather than designed experiments. Some future research should be designed to quantitatively define cause-effect relationships. The results would help guide monitoring to maintain forest soil productivity.

REFERENCES

Clayton, J. L.; Kellogg, G.; Forrester, N. 1987. Soil disturbance-tree growth relations in central Idaho clearcuts. Res. Note INT-372. Ogden, UT: U.S. Department of Agriculture, Forest Service, Intermountain Research Station. 6 p.

Cochran, P. H.; Brock, Terry. 1985. Soil compaction and initial height growth of planted ponderosa pine. Res. Note PNW-434. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Forest and Range Experiment Station. 4 p.

Cochran, W. G. 1977. Sampling techniques. 3d ed. New York: John Wiley and Sons. 428 p.

Geist, J. Michael; Hazard, John W. 1975. Total nitrogen using a sodium hydroxide index and double sampling theory. Soil Science Society of America Proceedings. 39(2): 340-343.

Geist, J. Michael; Hazard, John W.; Seidel, Kenneth W. 1989. Assessing physical conditions of some Pacific Northwest volcanic ash soils after forest harvest. Soil Science Society of America Journal. 53(3): 946-950.

Hazard, John W.; Geist, Michael J. 1984. Sampling forest soil conditions to assess impacts of management activities. In: Stone, Earl L., ed. Forest soils and treatment impacts: Proceedings of the Sixth North American Forest Soils Conference; 1983 June. Knoxville, TN: The University of Tennessee: 421-430.

Hazard, John W.; Snellgrove, Jeralyn; Geist, J. Michael. 1985. Processing data from soil assessment surveys with the computer program SOILS. Gen. Tech. Rep. PNW-179. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Forest and Range Experiment Station. 26 p.

Howes, Steve; Hazard, John; Geist, J. Michael. 1983. Guidelines for sampling some physical conditions of surface soils. R6-RWM-146-1983. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Region. 34 p.

Meurisse, Robert T. 1988. Soil productivity protection and improvement: objectives, policy, and standards in the Pacific Northwest Region of the Forest Service. In: Slaughter, Charles W.; Gasbarro, Tony, eds. Proceedings of the Alaska Forest Soil Productivity Workshop; 1987 April 28-30; Anchorage, AK. Gen. Tech. Rep. PNW-219. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station: 63-68.

Miller, Richard E.; Hazard, John W. 1988. Strategy and tactics for monitoring long-term site productivity. In: Slaughter, Charles W.; Gasbarro, Tony, eds. Proceedings of the Alaska Forest Soil Productivity Workshop; 1987 April 28-30; Anchorage, AK. Gen. Tech. Rep. PNW-219. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station: 57-62.

Starr, G. Lynn; Geist, J. Michael. 1988. Soil bulk density and soil moisture calculated with a FORTRAN 77 program. Gen. Tech. Rep. PNW-211. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station. 6 p.

Sullivan, Timothy E. 1988. Monitoring soil physical conditions on a national forest in eastern Oregon: a case study. In: Slaughter, Charles W.; Gasbarro, Tony, eds. Proceedings of the Alaska Forest Soil Productivity Workshop; 1987 April 28-30; Anchorage, AK. Gen. Tech. Rep. PNW-219. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station: 69-76.

Paper presented at the Symposium on Management and Productivity of Western-Montane Forest Soils, Boise, ID, April 10-12, 1990.

J. M. Geist is Principal Research Soil Scientist, Pacific Northwest Research Station, Forestry and Range Sciences Laboratory, Forest Service, U.S. Department of Agriculture, 1401 Gekeler Lane, La Grande, OR 97850. R. T. Meurisse is Regional Soil Scientist, Pacific Northwest Region, Forest Service, U.S. Department of Agriculture, Portland, OR 97208. T. A. Max is Station Statistician, Pacific Northwest Research Station, Portland, OR 97208.