PREDICTING LONG-TERM EFFECTS OF SILVICULTURAL PRACTICES

ON FOREST SITE PRODUCTIVITY

Phillip Sollins, Joseph E. Means, and Russell Ballard

ABSTRACT: We describe a system for predicting long-term consequences ofsilvicultural practices, especially those that may decrease long-termforest productivity. The system requires: (1) conceptual models thatincorporate current understanding of interactions among ecosystemprocesses; (2) process studies that, guided by the conceptual models,allow us to establish equations for the transfer of material and energyamong ecosystem components and to refine the conceptual models; (3) amanagement-oriented simulation model, developed from the conceptual model,used to predict long-term consequences of silvicultural practices; and(4) validation studies that test those predictions. Conceptual modelsmust account for interactions among processes as well as for all materialflow. Process studies should clarify the relations between processes andtheir controlling factors; operational trials should duplicate silvicul-tural practices to determine their effectiveness. In general, processstudies should be replicated at each site, operational trials across manysites. Experimental treatments selected for process studies need notadhere to standard silvicultural practice. Development of a management-oriented simulation model must be a high priority. FORCYTE, developed byJ. P. Kimmins and K. A. Scoullar, may offer the best starting point forforesters and researchers in the Pacific Northwest. Operational trialsshould validate the simulation model rather than merely provide informa-tion for specific sites, species, and treatments.

INTRODUCTION

Our understanding of how forest ecosystemsfunction has increased greatly in the last 15years. Nonetheless, although new silviculturalpractices have been proposed and implemented, our

PHILLIP SOLLINS is a research associate professor,Department of Forest Science, Oregon StateUniversity, Corvallis, Oreg. 97331. JOSEPH E.MEANS is a research forester at the PacificN orthwest Forest and Range Experiment Station,3200 Jefferson Way, Corvallis, Oreg. 97331.RUSSELL BALLARD is department manager, WesternF orestry Research, Weyerhaeuser Company, Tacoma,Wash. 98477.

ability to predict the long-term effects of thesepractices has increased little.

For example, there is increasing concern thatmore complete removal of forest biomass duringharvest will lead to long-term declines in produc-tivity (Boyle 1976, Boyle et al. 1973, Harvey etal. 1980, Kimmins 1977, Leaf 1979). A decline insecond-rotation yield has been demonstrated atsites that were windrowed (Ballard 1978) andslash-burned (Keeves 1966), but such long-termstudies are few. An even greater problem is thatconceptual models are not adequate for addressingthe problem (Morrison and Foster 1979, Stone 1979,Tamm 1979).

We urgently need an efficient way to use existingknowledge of forest ecosystems to predict thelong-term impacts of silvicultural practices. For

example, the National Forest Management Act re-quires that the National Forests be managed with-out - impairment of the productivity of the land"(U.S. Government 1982). Obviously, decisions mustbe made today based on existing knowledge, and newstudies must be initiated as well.

In this paper we describe a system which can helpus synthesize information, coordinate research,and utilize existing data to make long-term pro-jections. This system requires four components:conceptual models, process studies, simulation,and validation studies. We define and provideexamples of each of these, and show how they canfunction together as a coordinated system.

CONCEPTUAL MODEL

A conceptual model helps us integrate our under-standing of how the forest ecosystem functions.It identifies the important compartments of anecosystem, the processes by which material andenergy are transferred among them, and the rela-tionships among these processes. For example, aconceptual model of forest growth must include theeffects of nutrient availability and plant com-petition (e.g., Linder 1981) as well as the rela-tions between organic matter in the soil and theability of the soil to supply nutrients to plants(e.g., McGill et al. 1981).

Separate conceptual models may be needed for dif-ferent subsystems within an ecosystem, such as theforest floor or the tree canopy. But it may bedifficult to integrate information about the sub-systems without a conceptual model of the entireecosystem that explicitly shows how such submodelsinteract.

Conceptual models are often presented as box andarrow diagrams illustrating the transfer of •material between compartments (fig. 1). In suchmaterial-flow models, an arrow or similar symboldenotes a process and the boxes represent pools ofmaterial. These diagrams can easily be convertedinto quantitative budgets showing the amounts ofmaterial transferred along each pathway andchanges in pool sizes over a convenient timeperiod such as 1 year (Bormann et al. 1977, Coleet al. 1967, Sollins et al. 1980).

Because the budget accounts for all materialflowing through the forest ecosystem, inconsisten-cies in the data can often be discovered bychecking for conservation of mass (Sollins 1982).Ranking the transfers according to the amount ofmaterial flowing each year provides an initialestimate of the importance of each process withinthe ecosystem. Although quantity does notnecessarily equate with importance, this procedureoffers a useful starting point.

A disadvantage of conceptual models based onmaterial flow is that they cannot easily be usedto illustrate interactions among processes.

IrJ ›-> a.0a) Z4 41

CO2 atmosphericinput

14 1-4,

o #i-,a .1.- .C7 :La .,> 1

•0Z ,.,.-o rg i

olaiZi,O 203 4„4 ._

illZ0N

.r' l1.-oo

1.

-4- - 1 2—— 41--'',. ...-- I :,

_ -01oc

,..

tal •, . •,-- foliage-1 1_.,-._-,

vegetation mineralnutrient 001

reproduCtirestructures t....• — eccoo nsolop my ; 1:

1 o •.. wood...,—Ig,i , •••"'' '1-

- 07"-—

.4.vi

r2 -

.c 02,

"P"•i

= —o• ..,..

I C O2

•• • .. ' •

--- —- 1 c,

ii. g12A

, •• •'•-

''—:" —=-212=1:11=—"zrcoa r se ... .. . roots .....— .1—, ..amm .. . - ... FOIO• NO -WIN.dm all ;:, a; 5.7-aiitaNi..rilifis.1.■....ma sis• . ..rm. ligmonouswatillimMilvi.mefeewee.tsmares■mmi veer mom .... .... VMS IND,11114111MMINNI.1111MMAWN II dillriall IIIIMINNIAIMI/bral INI11s 01 I II IM InglffelfaMN • 1111 fine roots, "rmil ‘00.11.1. -v•-■ f owl°MP 111111

s m orrhlzoo laill NW iiniNMI- soul mineralAlta —Aii, Nuamilutrient .00! .

,,0,..u,_,,-zw00COON

M

I— c,,, c___, IP minerals pminerals

iiii ., 411116 41101" leadN ing loss--

Figure 1.--A conceptual model of material transferin a coniferous forest (from Johnson et al. 1982).

Although Forrester (1961) proposed a schematicnotation for superimposing interactions upon amaterial-flow diagram, our experience has beenthat such notation inevitably creates an unin-telligible diagram.

Interactions among processes, however, can be rep-resented clearly with diagrams that emphasize theprocesses rather than the compartments (fig. 2).Word models, written descriptions of interactionsamong processes (table 1), are also effective. Ineither, the interactions among processes must bedescribed explicitly. Material-flow diagrams andbudgets alone cannot tell the whole story.

Descriptions of interactions, however, can bequite misleading unless they are based on anddeveloped in conjunction with a material-flowbudget. Without a material-flow budget, it is tooeasy to omit an important interaction or processby assuming that we know what is important.Omissions and erroneous assumptions can becorrected, of course, but the trial-and-errorapproach may prove expensive. First constructinga material-flow budget that accounts for all flowmay save considerable effort in the long run.

202

Comparison

PredictedBiomoss and

NutrientIncrement

GrowthEquotionsby SiteOuolity

Overstory and UnderstoryBiomoss and Nutrient

Content

AddNutrient and

BiomossIncrement

ReduceGrowth

IncreaseGrowth Reduce

SiteQuality

IncreaseSite

Quo lit

If D> S or D<S for3 Consecutive rears

inte notCyc

Nutrient Nee ded to ProducePredicted Biomass Increment

BiomassNutrient

Concentration

ActuolIncrement

Figure 2.--Diagram showing interactions between nutrient availability and biomass in FORCYTE (from Kimminsand Scoullar 1979).

Table 1--Interactions among modules of CONIFER (from Sollins et al. 1979)

4

•Effect of carbon variables on water and energy flows

A. Foliage biomass affects:TranspirationFraction of rain incident to canopy thatstrikes foliage, and therefore alsofraction striking nonfoliage. (This andfollowing two affect drip, litter, andsoil moisture dynamics. There are alsoindirect effects through percent cover.)Water retention capacity of canopyDistribution of retention capacity betweenfoliage and nonfoliageFraction of rainfall passing directly toforest floor (through percent cover)Net longwave radiation input to canopy(through percent cover, which affectsinput and loss)

B. Stem biomass affects:1. Percent cover (and therefore numbers 2-6

above)C. Fine leaf and woody litter mass affects:

1. Water retention capacity of litter

Effect of energy variables on carbon and waterflows

A. Heat input to canopy affects:Potential evaporation from canopyTranspiration

B. Litter temperature affects:Litter decomposition processesPotential evaporation from litter

C. Soil temperature affects:1. Large and fine root respiration and growth

D. Net heat input to snowpack and heat deficit ofsnowpack affect:1. Net transfer between free water and ice in

snowpack

Effect of water variables on carbon and energyflows

A. Soil moisture affects:New and old foliage photosynthesis (viastomatal resistance)Fine root death (via plant moisture stress)

3. Dead root + soil organic matter decomposi-tion processes

B. Litter moisture affects:1. Litter decomposition processesSnowpack ice affects:1. Litter temperatureSnowfall affects:

Heat loss from snowpack due to snowfallAlbedo of snowpack

E. Drip plus direct rainfall affect:1. Litter and soil temperature

Table 2--Experimental treatments suitable for process studies

Treatment Examples of processes affected directly

Trenching (root exclusion)

Fertilization

Defoliation

Acidification

Soil compaction

Thinning

Devegetation

Uptake, decomposition, respiration

Leaching, decomposition, uptake, nitrification, denitrification

Transpiration, litterfall, light and water interception, photosynthesis

Cation exchange, weathering, Fe and Al eluviation

Water infiltration, soil aeration, surface erosion

Photosynthesis, respiration, light and water interception

All heat and water transfers, uptake, decomposition, etc.

PROCESS STUDIES

Process studies, the second component of a syste-matic approach, attempt to define the relationsbetween each process within the conceptual modeland the factors regulating each process. Examplesof ecosystem processes include decomposition offoliage litter, N fixation, and photosynthesis.If a conceptual model is built around transfersand accumulation of material and energy, the pro-cesses are simply all the flows of material andenergy among compartments. Jorgenson et al. (1975)review many of the material-flow processes impor-tant in forest ecosystems; Gorham et al. (1979)and McColl and Grigal (1979) describe others.

We have said that process studies attempt toclarify the relations between processes andcontrolling factors. It is important to separatethem from operational trials, which attempt to .duplicate a silvicultural management practice todetermine its effectiveness. The two have dif-ferent objectives and therefore different designcriteria. An experimental treatment intended tobe part of a process study must be designed foreasy interpretation of results (table 2). If, forexample, litter is to be raked off a plot, thenthe same amount of litter should be removedeverywhere from each plot. To assure uniformtreatment, such an experiment is usually conductedby researchers or under their direct supervision.In an operational trial, as in actual management,the amount of vegetation removed, clipped, orsprayed varies greatly within each treatment area.For process studies, however, the treatment needsto be more carefully controlled.

Because most management treatments affect manycomponents of the forest ecosystem simultaneously,such treatments are of limited value in processstudies. For example, any site-preparation treat-ment or harvest method affects the litter layer,the soil, and the vegetation. While it may beeasy to identify components that are affected by amanagement practice, it can be almost impossibleto separate cause and effect.

For process studies, then, it is preferable todevise a treatment that affects only one componentof a system. Then, when an effect is detected,

its cause will be clear. We do not mean, however,to downplay the importance of operational trials.They are critical to the overall research/management process and will be discussed later.

An important feature of any process study isreplication. This procedure, however, must betailored to the researcher's objectives. Forinstance, by repeating an experiment at manysites, mean and variance can be estimated across abroad range of sites. Such replication, however,gives no estimate of mean or variance at eachsite, a constraint that may severely limit ourability to extract information about processes.The problem here is that many relations betweenprocesses and their controlling factors arenonlinear; for example, a two-fold change in tem-perature will usually not cause a two-fold changein decomposition rate. Thus, when we try torelate average values for decomposition rates overbroad regions to average temperatures over broadregions, we find poor correlations. A singleequation can describe the relation (Bunnell et al.1977a, 1977b), but only if we have a good estimateof the mean temperature and decomposition rate ateach site. To obtain site-specific data, measure-ments and treatments must be replicated at eachsite. To obtain information applicable to a broadregion, studies must be replicated across a widerange of sites. As a rule of thumb, process stu-dies require site-specific information; opera-tional trials require replication across a broadrange of sites.

Replication, however, must be approachedcautiously. Given deadlines and finite budgets,it will never be possible to measure all processeswith the degree of statistical elegance that wemight like. To obtain accurate data for one pro-cess and ignore other processes entirely may nothelp, particularly if the process closely studiedturns out to be insignificant. An alternative isto first obtain at least some information aboutall major processes at a site, even if thisprecludes adequate replication. In poorly definedsystems, such as are dealt with in forestry, someinitial study of all the processes may be essen-tial. This done, rigorous investigations canbegin on those processes that seem to he mostimportant.

204

In the system we describe here, the object of pro-

cess studies is to define the factors regulatingeach process at each site, not to test managementstrategies. Research objectives should thereforedictate the choice of sites. Experimental forestsand Research Natural Areas, such as those operatedby the USDA Forest Service and many universities,offer advantages for process research. Whenresearch is concentrated at a specific site, theinformation from one study can be used directly inanother without having to account for differencesbetween sites. Baseline meteorological and hydro-logical data are also frequently available forsuch sites, increasing the efficiency of processresearch.

Treatments selected for process studies need notfollow standard silvicultural practice or be eco-nomically feasible. An excellent example of thisis an experiment by Turner (1977) in Douglas-firstands in western Washington in which N availabi-lity was decreased temporarily by adding car-bohydrate (sugar) to the forest floor, anexpensive and operationally impractical treatment.The C:N ratio of the forest floor was increasedmarkedly, affecting decomposition, leaching of Nfrom the forest floor, and internal redistributionof N within the trees. Occasionally, treatmentsthat adhere to standard practice will provevaluable as part of a process study. If resultsare monitored for a sufficient time, the treatmentcan then serve as both an operational trial and aprocess study.

Simulation models can be important tools for pro-cess research. A research-oriented simulationmodel can be constructed for a selected process orset of processes such as decomposition (Bunnell etal. 1977a, 1977b) or water and energy exchange inforest canopies (Halldin 1979). The output can becompared to the results of process studies inorder to verify our understanding of that portionof the system. A sensitivity analysis can be per-formed on such a model, a procedure in which eachparameter value is varied by a fixed percentage inorder to see which parameter values influence theprocess most. Such factors are then obvious can-didates for further study.

MANAGEMENT-ORIENTED SIMULATION MODELS

The third component of a systematic approach topredicting forest productivity is simulation. A" management-oriented simulation model," as definedhere, is one designed specifically for land mana-gers interested in the long-term effects of silvi-cultural practices. To be useful, a managementmodel must predict ecosystem behavior over a widerange of environments and time intervals. Themodel should predict yield, and costs and benefitsin terms of both dollars and energy. The modelmust simulate processes if it is to be extrapo-lated to combinations of site, species, and treat-ment outside the experience of its authors.Effects of nutrition must be included. Inputrequirements and run cost must be kept modest andoutput format must be convenient or the model willnot be used.

A management-oriented simulation model can bedeveloped from the conceptual model once process

studies have provided the necessary equationsdescribing each process. Without an adequateconceptual model, progress may be painfully slowand expensive.

Computer simulation is the only way land managerscan foresee the behavior of forest ecosystems overmany rotations, an interval longer than our lifespans. Although projects should be initiated thatwill span several rotations, we cannot afford towait for the results. Occasionally, we may findchronosequences of stands that allow us to gatherdata simultaneously on different stages of naturalstand development, but it is difficult to deter-mine how similar the stands were when established(Stone 1975). Moreover, intensive management hasonly been practiced for a few decades in mostforest regions, and few, if any, sites have beenmanaged intensively for several rotations. Thus,although simulation results may be tentative,there are few alternatives.

A good management model allows land managers toselect silvicultural alternatives on the basis ofthe long-term consequences. Most importantly, agood model can help us deal with the problem thateach component of the ecosystem is connected toevery other--that one cannot be altered withoutaffecting all. For example, fertilization affectsmany processes simultaneously. In addition tospeeding tree growth, it can promote nutrientimmobilization by the soil microbiota, increasethe rate of litter decomposition, burn roots, andinhibit mycorrhizal development. All of theseprocesses will interact to affect uptake by thetrees and their subsequent growth. An adequatedescription of the effect of fertilizer on treegrowth therefore requires a model of the entiresystem. A computer simulation model can be apractical way to organize such a description,which then becomes a hypothesis for the behaviorof the whole system and can be tested by comparingthe predicted and observed responses.

A realistic management-oriented simulation modelwill help research managers to assign researchpriorities. Through sensitivity analysis, criti-cal processes can be identified and researchdollars invested where the need and payoff islikely to be greatest. A model is not a substi-tute for creative thinking; but to the extent thatit reflects our current understanding of thesystem, it will be a powerful tool for guidingresearch.

We know of only two management-oriented simulationmodels for forests that include the effects ofnutrient availability. FORTNITE is described bythe authors (Aber et al. 1982) as a generalizedcomputer model for organic matter and N dynamicsin forest ecosystems. Developed by merging amodel of forest floor decomposition (Aber et al.1978, 1979; Aber and Melillo 1982) and a model offorest succession (Botkin et al. 1972), it treatsa 10- x 10-m plot and follows individual trees andage classes of litter through time. Because someprocesses are assumed to be random, the computerprogram reports averages over replicate plots.The model has been "parameterized" for a NewEnglar;d hardwood forest and used to predicteffects of rotation length, harvest intensity, andfertilization on fiber yield (Aber et al. 1982).

?ns

The authors show that it accurately predictstrends in basal area, forest floor biomass, anddead wood mass after clearcutting, although dataavailable for comparison are limited. Using themodel, they concluded that extremely short rota-tions would reduce yield by as much as two thirds,but that fertilization might offset some of thedecrease. Whole-tree harvesting on a 90-yearrotation yielded more fiber than conventionalclearcutting, while selective cutting yieldedslightly less.

Hemstrom and Adams (1982) have adapted JABOWA, theforest succession model used in FORTNITE, for usewith conifer forests in the Pacific Northwest buthave not incorporated the nutrient cycling portionof FORTNITE into their model (V. Dale, personalcommunication 1983).

The other management-oriented simulation modelwhich includes effects of nutrient availability isFORCYTE (Kimmins and Scoullar 1979). It isdesigned to predict, on a site-specific basis, thelong-term effects that various intensive forestmanagement and harvesting practices may have onnutrient budgets and productivity as well as onenergy and economic costs. It uses an input datafile to provide the necessary site- and species-specific information and to specify the regenera-tion, spacing, thinning, fertilization, andharvesting options for each rotation. Many basicecosystem processes are included in the model(fig. 2). Tree growth is predicted from infor-mation on yield in managed and unmanaged stands(see, for example, Hann and Riitters 1982). It isassumed that tree growth will decrease from thesemeasured rates if levels of available N are notsufficient to meet growth demands. The effects ofmoisture availability are included implicitly byrecognizing three site classes labeled good,medium, and poor. A site can change from one sit4quality class to another in response to proper (orimproper) management. Early versions of thismodel were designed for even-aged, single-speciesforests managed on rotations of less than 150years; but the model can be adapted to other standstructures and forest types. As of 1983, FORCYTEhad been used for 4 years as a teaching tool atthe University of British Columbia and at OregonState University. It was also being modified foruse with western hemlock in the Oregon CoastRange, subtropical Eucalyptus plantations inBrazil, black spruce in Alaska, radiata pine inSouth Australia, and several other forest types.

Both FORTNITE and FORCYTE incorporate most of theinformation on nutrition and productivityavailable in the regions for which they weredesigned. Both models, however, have limitations.Neither has yet been validated adequately againstindependent data. But with refinement and morevalidation, both could become valuable managementtools in their respective regions.

VALIDATION STUDIES

Under validation, the fourth component needed topredict long-term productivity, we include anyprocedure by which we increase our confidence inthe correctness of a model. This depends on the

similarities between observed and predicted re-sponses, the number of model variables checked,the treatments involved, and the length of time,range of sites and climatic conditions over whichcomparisons are made. Validation studies musteventually span two or three rotations if long-term predictions of the model are to be verified.

To be meaningful, validation studies must comparemodel predictions with experimental data that werenot used to construct the model. Once the currentversion of the model has been validated, theresults of the study can be used to construct arefined version. A new, independent set of datais then needed to validate this refined model.

Operational trials can play a critical role in thevalidation process. Such trials allow us to testmany parts of the model at once because theyaffect various ecosystem processes simultaneously.Note that this is precisely the reason why wesuggested that most operational trials are notsuited for process studies.

Extensive operational trials, conducted byindustry and agencies, already cover a wide rangeof sites and treatments (table 3). Because thedata have already been collected, such studies canbe used to validate the model at an extremely lowcost. All too often, the data from operationaltrials are used only to provide information on theeffects of a particular silvicultural practice ata particular site or set of sites. Their valueincreases if they are used to validate a modelthat can then be extrapolated to other sites andsilvicultural treatments and through time.

Many studies not generally considered operationaltrials may also serve to validate management-oriented models. Several long-term studies ofeffects of insects, animal activity, and diseaseon growth and yield (table 3) could be par-ticularly valuable because they include processesand interactions seldom addressed in silviculturaltrials. For example, if disease or severe damageby insects or other animals is noted during a fer-tilizer trial, measurements are sometimes discon-tinued on the affected plots. Such action may beunderstandable because the object of the study wasto gauge fertilizer response. But insects anddisease are important components of the forestecosystem and have considerable impact on yield;including such interactions in a model and vali-dating it accordingly will increase its usefulnessas a management tool.

Operational trials, too, will be more useful formodel validation if key ecosystem processes andpools are monitored, not just the amount of bio-mass removed. The measured processes should bethose that sensitivity analyses have shown willhave a critical effect on the accuracy of the pre-dictions.

Such monitoring can provide two important bene-fits. First, weak parts of the conceptual modelcan be identified and then improved with processstudies. Second, the process data can provide anadditional check on the validity of the underlyingconceptual model. For example, it is conceivablethat a management model could predict yield

206

Table 3--Operational trials and other experiments suitable for validating models of long-term productivity

Treatment

Example

Reference!/

Fertilization Northwest Regional Forest Fertilization Project

(1)

Levels of utilization Department of Energy study at Pack Forest and elsewhere

(2)(harvesting intensity)

Multiple rotations Long-term CFI plots in southern Australia already spanning several

(3)rotations

Thinning Levels of Growing Stock Study in Pacific Northwest

(4)

Sludge application University of Washington/Metro Study

(5)

Vegetation control CRAFTS Project

(6)

Mixed species plantings Alder/Douglas-fir mixes at Wind River and Cascade Head; Oregon State

(7)University LTER study on Douglas-fir/Ceanothus mixes

Insect control

Effects of levels of tussock moth defoliation on Douglas-fir growth

(8)

Animal damage

Animal damage impacts on conifer plantations

(9)

Disease

Levels of Dothistroma pini control

(10)

Irrigation

SWECON (in combination with fertilization and insect control)

(11)

Site/residue treatment

Pacific Northwest residue treatment studies

(12)

Tire

Prescribed burning in southeastern U.S.A. (13)

Soil removal

Soil removal during site preparation

(14)

Drainage

Drainage intensity, southeastern U.S.A., coastal plain

(15)

I! (1) Peterson and Gessel 1983Cole and Bigger 1983Keeves 1966Williamson 1976Edmonds and Cole 1980Preest 1975Miller and Murray 1978

(8) Wickman 1963

Black et al. 1979Gilmour and Noorderhaven 1971Aronsson and Tamm 1982Cramer 1974Stone 1971Glass 1976

(15) Terry and Hughes 1978

correctly on occasion yet still he substantiallyincorrect in its representation of internal pro-cesses. A model that accurately predicts rates ofsuch processes and crop yield is much more likelyto be correct in a wide range of circumstances.

Long-term growth and yield plots deserve specialmention. The Pacific Northwest has an unusuallylarge number of these with records spanning up to48 years and areas as large as 42 ha (Sollins1982, Williamson 1976). Additional plots havebeen established by the USDA Forest Service that,if maintained over the next few decades, will pro-vide comparable information (Hawk et al. 1978).These plots offer our only opportunity to checkmodel predictions over long periods of timethrough the use of records of tree growth and mor-tality. They also give us the only opportunity toconduct process research at sites for which suchrecords are available. Consequently, these plotsmust be protected. Buffers must be maintained

because clearcutting to the edge of the plots willinevitably increase mortality within them.Salvage operations within the plots also must beprevented if forest floor and soil processes areto be studied.

INTEGRATING THE FOUR COMPONENTS

The overall object of the system described here isto increase our understanding of forest ecosystems(the conceptual model) and our ability to predictthe long-term consequences of silvicultural prac-tices (the management model). We illustrate thiswith a diagram containing two feedback loops (fig.3).

The left-hand loop through the conceptual modeland process studies is the primary way to improvethe conceptual model. The loop begins whenavailable knowledge is synthesized into an initial

refineconceptual

modelrefineconceptual

model

comparepredicted

and observedresponse

identifyimportant

..cluestionscollate

availableinformation predict

response tomanagement

practices

CONCEPTUALMODEL

PROCESSSTUDIES

VALIDATIONSTUDIES

(operationaltrials)

MANAGEMENTORIENTED

SIMULATIONMODEL

Figure 3.--Feedback loops between research and management operations.

conceptual model describing ecosystem structureand function. The conceptual model helps deter-mine the priorities for process studies. In turn,the results of the process studies serve to refinethe conceptual model.

The right-hand loop through the conceptual model,the management model, and the operational trials(fig. 3) enables us to make increasingly reliablepredictions for forest management. Onceconstructed, the management-oriented simulationmodel can be updated to incorporate refinements in,the conceptual model that have resulted from pro-cess studies. Furthermore, validation of themanagement model will improve both it and the con-ceptual model. Note that discrepancy between pre-dicted and observed responses forces modificationof both models; agreement reinforces confidenceand discourages change. Neither discrepancy noragreement is necessarily good or bad. Discrepancyopens up exciting possibilities for research;agreement means that land managers have a usefultool on which to base their decisions.

With diligent and creative effort by researchersand land managers, steady progress is inevitable.But the rate of progress cannot be measuredwithout a clear goal. There is no reason torefine the management model unless it is inade-quate for predicting responses within prescribedlimits. Selection of such responses and limits isthe responsibility of researchers and land mana-gers together.

Efficient progress toward an understanding of thelong-term effects of silvicultural practicesrequires cooperation among forest managers andmany research disciplines: silviculture, geology,soils, nutrition, mensuration, microbiology, plantphysiology, entomology, pathology, economics, andperhaps others. Information must flow freelyacross disciplines, as well as between researchers

and land managers. Some validation studies mustbe long-term, perhaps as long as two or threerotations. These require thorough documentationand conscientious protection of study sites. Inaddition, existing studies and data sets should beused by many researchers if the studies are torealize their full potential. Cooperative studiesare essential for all these reasons.

Existing cooperative research projects can helpwith the coordination of new studies. Such pro-jects include the Regional Forest NutritionResearch Project (Peterson and Gessel 1983), theCRAFTS (Coordinated Research on AlternativeForestry Treatments and Systems for ForestVegetation Management) program (Walstad et al.1982) and the North Carolina State ForestFertilization Cooperative in the southeasternUnited States (Allen and Duzan 1983), as well asthe entire IUFRO program. With modest funding,such cooperatives can help (1) coordinate researchto avoid duplication, (2) coordinate large-scaletesting of models already under development, and(3) provide forums where researchers and landmanagers can discuss results and needs.Assignment of research priorities and standar-dization of methods are other possibilities, butwe must keep in mind that the objective is topromote progress, not stifle creative thinking.In general, our goal is to use data efficiently tosolve land-management problems. If we can assurethat this will happen, we have a logical basis forseeking expanded funding for research into thelong-term behavior of forest ecosystems.

ACKNOWLEDGMENTS

We thank J. P. Kimmins, D. Perry, F. Swanson, andJ. S. Trappe for helpful reviews of this paper.This is contribution number 1785 from the ForestResearch Laboratory, Oregon State University.

208

Bunnell, F.L.; Tait, D.E.N.; Flanagan, P.W.; VanCleve, K. Microbial respiration and substrateweight loss -- II. A model of the influences ofchemical composition. Soil Biol. Biochem.9:41-47; 1977b.

Aber, J.D.; Botkin, D.B.; Melillo, J.M.Predicting the effects of different harvestingregimes on productivity and yield in northern

hardwoods. Can. J. For. Res. 9:10-14; 1979.

Aber, J.D.; Melillo, J.M. FORTNITE: a computermodel of organic matter and nitrogen dynamicsin forest ecosystems. Madison, WI: Univ.Wisconsin Res. Bull. R3130; 1982. 49 p.

Aber, J.D.; Melillo, J.M.; Federer, C.A.Predicting the effects of rotation length,harvest intensity, and fertilization on fiberyield from northern hardwood forests in NewEngland. For. Sci. 28:31-45; 1982.

Allen, H.L.; Duzan, H.W., Jr. Nutritionalmanagement of loblolly pine stands--a statusreport of the North Carolina State ForestFertilization Cooperative (this volume); 1983.

Aronsson, A.; Tamm, C.O. Effects of irrigationon young stands of pine (Pinus sylvestis L.)

and spruce (Picea abies Karst.). In: B.Ranneby, ed. Statistics in Theory andPractice. Umea: Seed. Univ. Agric. Sec.For. Biometry; 1982: 53-69.

Ballard, R. Effect of slash and soil removal onthe productivity of second-rotation radiatapine on a pumice soil. N.Z.J. For. Sci.8:248-258; 1978.

Black, H.C.; Dimock, E.J., II; Evans, J.;Rochelle, J.A. Animal damage to coniferousplantations in Oregon and Washington.Corvallis, OR: Oregon State Univ., For. Res.Lab. Res. Bull. 25; 1979. 45 p.

Bormann, F.H.; Likens, G.E.; Melillo, J.M.Nitrogen budget for an aggrading northernhardwood forest ecosystem. Science 196:981-983;1977.

Botkin, D.B.; Janak, J.F.; Wallis, J.R. Someecological consequences of a computer model offorest growth. J. Ecol. 60:849-872; 1972.

Boyle, J.R. A system for evaluating potentialimpacts of whole-tree harvesting on sitequality. TAPPI 59(7):79-81; 1976.

Boyle, J.R.; Phillips, J.J.; Ek, A.R."Whole-tree" harvesting: nutrient budgetevaluation. J. For. 71:760-762; 1973.

Bunnell, F.L.; Tait, D.E.N.; Flanagan, P.W.; VanCleve, K. Microbial respiration and substrateweight loss -- I. A general model of theinfluences of abiotic variables. Soil Biol.Biochem. 9:33-40; 1977a.

Cole, D.W.; Gessel, S.P.; Dice, S.F. Distributionand cycling of nitrogen, phosphorus, potassiumand calcium in a second-growth Douglas-fir

ecosystem. In: H. E. Young, ed. Symposium onprimary productivity and mineral cycling innaturay ecosystems; Orono, ME: College of LifeSciences, Univ. of Maine; 1967: 197-232.

Cole, D.W.; Bigger, C.M. Some initial effects ofharvesting second-growth Douglas-fir and redalder on productivity and nutrient losses(this volume); 1983.

Cramer, 0.P., ed. Environmental effects offorest residue management in the Pacific

Northwest: A state-of-knowledge compendium.Portland, OR: USDA For. Serv. Gen. Tech. Rep.PNW-24; Pac. NW For. and Range Exp. Stn.; 19-74.

Edmonds, R.L.; Cole, D.W., eds. Use ofdewatered sludge as an amendment for forestgrowth. Center for Ecosystem Studies Bull. No.3. Seattle, WA: College of Forest Resources,Univ. of Washington; 1980. 58 p.

Forrester, J. W. Industrial dynamics. Cambridge,MA: MIT Press; 1961.

Gilmour, J.W.; Noorderhaven, A. Influence of timeof application of cuprous oxide on control ofDothistroma needle blight. N.Z.J. For. Sci.1:160-166; 1971.

Glass, G.G. The effect of rootraking on an uplandPiedmont loblolly pine (Pinus taeda L.) site.Raleigh, NC: North Carolina State Univ. Schoolof Forest Resources Tech. Rep. 56; 1976.

Gorham, E.; Vitousek, P. M.; Reiners, W. A. Theregulation of chemical budgets over the courseto terrestrial ecosystem succession. Ann. Rev.

Ecol. Syst. 10:53-84; 1979.

Halldin, S, ed. Comparison of forest waterand energy exchange models. Proceedings, IUFROWorkshop, Uppsala, Sweden, Sept. 24-30, 1978;International Society of Modelling, Copenhagen,Denmark; 1979. 258 p.

Hann, D.W.; Riitters, K. A key to the literatureon forest growth and yield in the Pacific

Northwest: 1910-1981. Corvallis, OR: OregonState Univ., For. Res. Lab. Res. Bull. 39;1982. 77 p.

Harvey, A.E.; Jurgenson, M.F.; Larsen, M.J.Biological implication of increasing harvestintensity on the maintenance and productivity offorest soils. In: Environmental consequencesof timber harvesting in Rocky Mountainconiferous forests; Gen. Tech. Rep. INT-90.U.S. Department of Agriculture, Forest Service,

Intermount. For. and Range Exp. Stn.; 1980:211-220.

Aber, J.D.; Botkin, D.B.; Melillo, J.M.Predicting the effects of different harvestingregimes on forest floor dynamics in northernhardwoods. Can. J. For. Res. 8:306-315; 1978.

LITERATURE CITED

Hawk, G.M.; Franklin, J.F.; McKee, W.A.; Brown,R.B. H.J. Andrews Experimental Forest referencestand system: establishment and use history.Coniferous Forest Biome, Bull. No. 12.Seattle, WA: Univ. of Washington; 1978.79 p.

Hemstrom, M.; Adams, V.D. Modeling long-termforest succession in the Pacific Northwest.In: Means, J.E., ed. Proceedings, Forestsuccession and stand development research inthe Northwest; 1980 March 26; Corvallis, OR:Oregon State Univ., Forest ResearchLaboratory; 1982: 14-23.

Johnson, D.W.; Cole, D.W.; Bledsoe, C.S.; Cromack,K.; Edmonds, R.L.; Gessel, S.P.; Grier, C.C.;Richards, B.N.; Vogt, K.A. Nutrient cyclingin forests of the Pacific Northwest. In: R. L.Edmonds, ed. Analysis of coniferous forestecosystems in the western United States. US/IBPSynthesis Series, No. 14. Stroudsberg, PA:Hutchinson Ross Publ. Co.; 1982: 186-232.

Jorgensen, J. R.; Wells, C. G.; Metz, L. J. Thenutrient cycle: key to continuous forestproduction. J. For. 73:400-403; 1975.

Keeves, A. Some evidence of loss of productivitywith successive rotations of klvF;o ea0aTa inthe south-east of Australia. Aust. For. 30:51-63; 1966.

Kimmins, J.P.; Scoullar, K.A. FORCYTE: Acomputer simulation approach to evaluating theeffect of whole-tree harvesting on the nutrientbudget in northwest forests. In: Gessel, S.P.;Kenady, R.M.; Atkinson, W.A., eds. Proceedings,forest fertilization conference; ContributionNo. 40; Seattle, WA: College of ForestResources, Univ. of Washington; 1979: 266-273.

Kimmins, J. P. Evaluation of the consequences forfuture tree productivity of the loss ofnutrients in whole-tree harvesting. For. Ecol.Manage. 1:169-183; 1977.

Leaf, A. L., ed. Impact of intensive harvestingon forest nutrient cycling: Syracuse, NY:College of Environmental Science and Forestry;State University of New York; 1979.

Linder, S., ed. Understanding and predictingtree growth. Stud. For. Suecica 160:1-87;1981.

McColl, J.G.; Grigal, D.F. Nutrient losses inleaching and erosion by intensive forestharvesting. In: Leaf, A. L., ed. Impact ofintensive harvesting on forest nutrientcycling; Syracuse, NY: College ofEnvironmental Science and Forestry, StateUniversity of New York; 1979: 231-248.

McGill, W.B.; Hunt, H.W.; Woodmansee, R.G.; Reuss,J.O. PHOENIX, a model of the dynamics of carbonand nitrogen in grassland soils. Ecol. Bull.(Stockholm) 33:49-115; 1981.

Miller, R.E.; Murray, M.D. The effects of redalder on growth of Douglas-fir. Portland, OR:USDA For. Serv. Gen. Tech. Rep. PNW-70; Pac.NW For. and Range Exp. Stn.; 1978: 283-305.

Morrison, I.K.; Foster, N.W. Biomass and elementremoval by complete-tree harvesting of mediumrotation forest stands. In: Leaf, A. L., ed.Impact of intensive harvesting on forestnutrient cycling; Syracuse, NY: College ofEnvironmental Science and Forestry, StateUniversity of New York; 1979: 111-129.

Peterson, C.; Gessel, S.P. Forest fertilizationin the U.S. Pacific Northwest: Results of theRegional Forest Nutrition Research Project(this volume); 1983.

Preest, D.S. Effects of herbaceous weed controlon young Douglas-fir moisture stress andgrowth. Corvallis, OR: School of Forestry,Oregon State Univ.; 1975. Thesis.

Sollins, P. Input and decay of coarse woodydebris in coniferous forests of western Oregonand Washington. Can. J. For. Res. 12:18-28;1982.

Sollins, P.; Brown, A.T.; Swartzman, G.L.CONIFER: A model of carbon and water flowthrough a coniferous forest. Reviseddocumentation. Coniferous Forest Biome, Bull.No. 15. Seattle, WA: Univ. of Washington;1979. 152 p.

Sollins, P.; Grier, C.C.; McCorison, F.M.;Cromack, K.; Fogel, R.; Fredriksen, R.L. Theinternal element cycles of an old-growthDouglas-fir ecosystem in western Oregon. Ecol.Monogr. 50:261-285; 1980.

Stone, E.L. Nutrient removals by intensiveharvest--some research gaps and opportunities.In: Leaf, A. L., ed. Impact of intensiveharvesting on forest nutrient cycling;Syracuse, NY: College of EnvironmentalScience and Forestry, State University of NewYork; 1979; 355-386.

Stone, E.L. Effects of prescribed burning on longterm productivity of coastal plain soils. In:Prescribed burning symposium. Asheville, NC:USDA For. Serv.; 1971: 115-126.

Stone, E.L. Effects of species on nutrient cyclesand soil change. Phil. Trans. Roy. Soc.London. B 271:149-162; 1975.

Tamm, C. O. Nutrient cycling and productivity offorest ecosystems. In: Leaf, A. L., ed.Impact of intensive harvesting on forestnutrient cycling; Syracuse, NY: College ofEnvironmental Science and Forestry, StateUniversity of New York; 1979: 2-21.

Terry, T.A.; Hughes, J.H. Drainage of excesswaters - how and why. In: W.E. Balmer, ed.Proceedings, Soil moisture - site productivitysymposium; Myrtle Beach, SC: USDA For. Serv.;1978: 148-166.

210

Wickman, B.E. Mortality and growth reduction of

white fir following defoliation by the Douglas-fir tussock moth. Berkeley, CA: USDA For.Serv. Res. Paper. PSW-7, Pac. SW For. and RangeExp. Stn.; 1963; p. 15.

Turner, J. Effect of nitrogen availability on

nitrogen cycling in a Douglas-fir stand. For.Sci. 23:307-316; 1977.

U.S. Government. Federal Register 47(190):43026-43052; 1982.

Walstad, J. D.; Gordon, J. C.; Newton, M.; Norris,L. A. CRAFTS: Comprehensive research programin forest vegetation management. Corvallis, OR:Oregon State Univ., Dep. For. Sci. (mimeo-graphed); 1982.

Williamson, R.L. Levels-of-growing-stockcooperative study in Douglas-fir. Report No. 4.Portland, OR: USDA For. Serv. Res. PaperPNW-210; Pac. NW For. and Range Exp. Stn.; 1976;p. 39.

211

I.U.F.R.O. Symposiumon Forest Site and Continuous ProductivitySeattle, WashingtonAugust 22-28, 1982

Russell Ballard andStanley P. GesselTechnical Editors

Sponsored by:USDA Forest Service, Pacific Northwest Forest

and Range Experiment StationNorthwest Forest Soils CouncilWeyerhaeuser CompanyUniversity of Washington, College of

Forest Resources

Published by:U.S. Department of Agriculture, Forest Service,Pacific Northwest Forest and Range Experiment Station,Portland, OregonGeneral Technical Report PNW-163December 1983