Forest Ecology and Management 107 Ž1998. 275–289
Predicting response of coastal Douglas-fir to fertilizer treatments R.E. Carter
a,)
, E.R.G. McWilliams b, K. Klinka
c
a
First Marathon Securities, Suite 2000, 400 Burrard St., VancouÕer, BC, Canada, V6C 3A6 Forest Biometrics Consultant, 409 Tempe Crescent, North VancouÕer, BC, Canada, V7N 1E7 Department of Forest Sciences, Faculty of Forestry, UniÕersity of British Columbia, 270-2357 Main Mall, VancouÕer, BC, Canada, V6T 1W5 b
c
Received 3 June 1997; accepted 21 November 1997
Abstract A broadly-based, intensive Douglas-fir w Pseudotsuga menziesii ŽMirb.. Francox fertilization experiment throughout southern coastal British Columbia was used to examine 3 and 6 year crop tree growth responses to prescribed fertilizer applications. Absolute and relative basal area responses were evaluated in relation to site associations of the provincial ecosystem classification system, site index, and a large number of site and stand chemical and physical properties. Few of the site and stand variables examined as possible response prediction criteria appeared to have any real utility. The strongest relationships found were between relative basal area response and Ž1. site index Ž R 2 s 0.46 for both 3 and 6 year responses., Ž2. mineral soil mineralizable-N Ž R 2 s 0.50 and 0.46 for year 3 and 6 responses, respectively., and Ž3. total mineralizable-N Ž R 2 s 0.47 and 0.50 for year 3 and 6 responses, respectively.. In all cases average relative response declined with increasing site quality. However, there were highly productive sites ŽSI 50 G 35 m. characterized by an absence of growing-season water deficits and relatively low foliar N concentrations Ž12–13 grkg. showing significant fertilizer responses. These sites are where the greatest financial returns from fertilization may be realized. Relationships identified between site and stand variables and basal area responses were, in many cases, different from those found by other researchers for coastal Douglas-fir. This brings the portability of identified relationships into question. q 1998 Elsevier Science B.V. Keywords: Ecosystem classification; Site index; Mineralizable N
1. Introduction Forest fertilization research has provided foresters with few exacting tools or guidelines for diagnosing nutrient deficiencies and predicting responses to treatment. Diagnostic research has focused on foliar and, less often, soil analysis using critical limits
)
Corresponding author.
Že.g., van den Driessche, 1974; Morrison, 1974; Ballard and Carter, 1986., nutrient ratios ŽIngestad, 1971., and the concept of ‘nutrient balance’ in methods such as the Diagnosis and Recommendation Integrated System ŽDRIS. ŽBeaufils, 1973; Hockman and Allen, 1990.. These methods are generally applied in isolation from the many other genetic, environmental, and stand factors that may influence nutrient levels or requirements and consequently they tend to be imprecise.
0378-1127r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved. PII S 0 3 7 8 - 1 1 2 7 Ž 9 7 . 0 0 3 4 6 - 0
276
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
Prediction of fertilizer response, both in terms of predicting which sites will respond and predicting the magnitude of the response, has focused on empirical relationships. Examples of fertilizer response predictors include: site index ŽMiller and Fight, 1979; Miller et al., 1989., stand density ŽAllen, 1983., soil properties ŽShumway and Atkinson, 1978; Powers, 1980; Comerford and Fisher, 1982; Peterson et al., 1984; Lipas, 1985; Edmonds and Hsiang, 1987; Blake et al., 1988; Turner et al., 1988; Miller et al., 1989., foliar properties Žvan den Driessche, 1979; Blake, 1985; Turner et al., 1977, 1979, 1988; Timmer and Ray, 1988. and vegetation types ŽTimmer and Ray, 1988.. Most researchers have used correlation, regression or multivariate procedures to identify the ‘best’ response predictors for their particular sample. Predictive variables are normally subject to a wide variety of interactions with other stand structural and environmental factors limiting response predictions to rather broad generalizations. The overall goal of this study was to develop site-specific guidelines for operational fertilization of coastal Douglas-fir w Pseudotsuga menziesii ŽMirb.. Francox in British Columbia ŽCarter and Klinka, 1988.. The primary objective was to examine fertilizer responses across a wide range of sites in order to be better able to predict which sites will respond to fertilization. Fertilizer response is a function of the site, fertilizer uptake, stand history, and existing stand structure. To elucidate site effects, crop tree analysis of thinned immature stands was used to minimize the effects of stand history and structure on the observed response. The results of this study therefore provide average individual crop tree responses. These are not directly correlated to total stand response, as this is also a function of stand history, structure and site quality. However, the presumption is made that given a similar stand history and structure, the relative ranking of total stand response across sites would be the same as that observed for crop trees. Crop tree responses to fertilization were compared with site and stand factors in 48 stands of immature Douglas-fir to determine the utility of these factors as predictive variables for nutrient diagnosis and prescription. Response predictors tested included those shown to have utility in previous studies Žop. cit.. Že.g., site index, soil mineralizable-N, CrN
ratio, and extractable sulphate–sulphur, and foliar N and sulphate–sulphur..
2. Methods and materials 2.1. Study sites Forty-eight study Žexperiment. sites were located in southwestern British Columbia between 488 and 518 latitude, 1218 and 1278 longitude, and 12 and 565-m elevation. The sites were within 3 climatically similar areas suited to the growth of Douglas-fir which were delineated by the Very Dry, Dry, and Moist Maritime subzones of the Coastal Western Hemlock zone ŽKlinka et al., 1991.. Soils across most of the sites were relatively young and were primarily derived from glacial tills or fluvial materials of mixed lithology Žprimarily andesites, basalts, or acid igneous materials such as quartz and granodiorites.. Within each subzone the sites were chosen to represent the widest possible range of site conditions and productivity of Douglas-fir. Chosen sites supported stands with immature Douglas-fir Žbreast height age 21–45 years. representing ) 80% of the stand basal area. All stands were thinned from below 3 to 7 years prior to treatment leaving crop trees with room for crown expansion. Site index ŽSI, in meters at a reference age of 50 years. for each stand was determined from site index tables for coastal Douglas-fir ŽMitchell and Cameron, 1985. and based upon the site curves developed by Bruce Ž1981.. Site and stand characteristics for the 48 experimental sites are given in Table 1. In British Columbia, stand management prescriptions are required by legislation to be based on the site units of biogeoclimatic ecosystem classification ŽKrajina, 1972; Pojar et al., 1987.. Basic site units—site associations—are groups of sites with similar or equivalent physical properties, vegetation and productivity potentials Že.g., Cajander, 1926; Daubenmire, 1968.. They are identified on the basis of plant indicator species and site characteristics such as slope position, aspect and gradient, and soil and forest floor physical properties ŽGreen and Klinka, 1994.. Thus, the 48 sites were stratified into one of the five site associations supporting productive growth
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
277
Table 1 Means, standard deviations, and minimum and maximum values of average site and stand characteristics for the 48 experiments Measure
Mean
Standard deviation
Minimum
Maximum
Initial stand characteristics Site index Žm at 50-yr bh. Age Žyr. Dbh Žcm. Height Žm. Live crown length Žm. % live crown Live crown width Žm. Stemsrhaa
29 28 23.8 19.0 9.9 55 4.4 636
5.5 8 6.9 5.3 2.3 16.1 1.0 218
16 21 11.0 9.2 5.6 30 2.3 290
39 45 39.4 29.5 15.7 96 7.5 1210
Foliar nutrient leÕels N Žgrkg. SO4 –S ŽgrMg. NrSO4 –S ratio P Žgrkg. NrP ratio
12 319 53 2.2 5.9
2 138 38 0.4 1.49
9 97 17 1.5 3.6
17 631 167 3.4 10.2
21 27.4
17 27.7
3 5.9
93 176.5
Soil characteristics Mineral soil mineralizable N ŽgrMg. Total mineralizable N Žkgrha. a
Calculated as Ž100rDNN. 2 , where DNN is average distance to nearest neighbour in m.
of Douglas-fir and according to soil moisture regime ŽSMR. and a soil nutrient regime ŽSNR. as follows: Ž1. Lichens Ž Cladina and Cladonia spp.., very dry and very poor to medium sites; Ž2. Salal Ž Gaultheria shallon., moderately dry and poor to medium sites; Ž3. Flat moss Ž Plagiothecium undulatum., slightly dry and poor to medium sites; Ž4. Flat moss, fresh and poor to medium sites; Ž5. Sword fern Ž Polystichum munitum., slightly dry and rich to very rich sites; and Ž6. Foamflower ŽTiarella trifoliata., moist and very moist, rich to very rich sites ŽGreen and Klinka, 1994.. Dry to very moist soil moisture regimes are defined by actual evapotranspiration Ž Et .:potential evapotranspiration Ž Emax . ratio plus depth to water table ŽPojar et al., 1987.; soil nutrient regimes are defined by mean mineral soil Ž0–30 cm. mineralizable-N concentrations Žafter Klinka and Carter, 1990.. 2.2. Experimental design A single fertilizer experiment, consisting of 6 plots, was established at each of the study sites. Each plot was 0.04 ha with 5 m treated buffers. Five dominant or elite codominant sample trees Žcrop
trees. were randomly selected in each plot. The treatments, randomly applied to two plots each, were control, 225 kg Nrha as urea and 225 kg Nrha as urea plus any additional nutrients identified as potentially deficient on the basis of a pre-treatment foliar analysis ŽCarter and Klinka, 1988.. The two fertilizer treatments will be referred to as N alone and blend, respectively. Rates of application of the additional nutrient elements are given in Table 2. The fertilizer was applied by hand during the spring ŽFebruary–
Table 2 Chemical form and application level for each nutrient element potentially added in the ‘blend’ treatment Nutrient element
Rate
Form
N P K Mg S Cu Zn Fe B
225 kg Nrha 40 kg Prha 50 kg Krha 40 kg Mgrha 44 kg Srha 10 kg Curha 20 kg Znrha 25 kg Ferha 2.0 kg Brha
Urea Triple super phosphate Potassium sulphate Magnesium sulphate Degra-sul Ž0-0-0-90. Copper-sulphate Zinc-sulphate wNaŽFeŽEDTA.3H 2 O.x Solubor w Ž20.5% B.
278
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
March.. Weather conditions were wet and cool during the period of fertilizer application. 2.3. Measurements The forest floor and surface 30 cm of the mineral soil were sampled prior to the application of fertilizer. At each experiment, one randomly located sample point was established in each of the 6 plots. At each of these 6 points, 3 subsamples were taken on the points of an equilateral triangle Ž2 m on each side. that was centered over the sample point. Forest floor samples were taken using a template of 625 cm2 and weighted by thickness. Mineral soil samples were collected on an equal volume basis. Analytical methods for forest floor and mineral samples are only provided for those measures presented and discussed in the results and discussion; methods for other analyses are available from the principal author. All analyses were done separately for the forest floor and mineral soil. Mineralizable-N was determined using the anaerobic incubation method of Waring and Bremner Ž1964. as modified by Powers Ž1980. with released NH 4 –N determined colorimetrically using a Technicon Autoanalyzer. Sulphate–sulphur ŽSO4 –S. was extracted with Morgan’s solution at pH 4.5 and S was determined by the method of Johnson and Nishita Ž1952.. Bulk density was determined following the procedure of Nuszdorfer Ž1981. allowing calculation of kilograms of mineralizable-N in the forest floor and surface 30 cm of the mineral soil. Measurements made on each sample tree prior to fertilization included: tree height ŽHT.; diameter at breast height ŽDBH.; distance to the nearest neighbour tree in each of four 908 quadrants ŽDNN.; length of live crown ŽLLC.; and width of live crown ŽWLC.. Percent live crown Ž%LC. and conical crown surface area ŽCA. were then calculated for each tree. One growing season after treatment, during October and November, samples of the current year’s foliage were taken at the base of the upper one-third of the live crown from each of the 30 sample trees at all experiments. Foliage samples for each plot were composited and oven-dried at 708C for 12 h before determining mass per 100 needles in triplicate and grinding in a Braun Type KSM-2 blender to pass a 20 mesh sieve. All samples were then analyzed for
total N, P, K, Ca, Mg, S, SO4 –S, Cu, Zn, ‘active’ Fe, B, and Mn, following the guidelines and procedures given in Ballard and Carter Ž1986.. Individual sample tree measurements taken 6 years after treatment included diameter at breast height, height, bark thickness and radial growth determined by taking two increment cores 908 apart at breast height. Inside bark basal areas and basal area growth were determined based on the quadratic means of the two radial increment measures. From these values total basal area growth 3 and 6 years after treatment as well as total basal area growth in each of the 5 years before treatment were calculated for use in subsequent analyses. Total without bark volumes at the time of treatment and 6 years after treatment were calculated using the volume equation for small coastal Douglas-fir by Omule Ž1987.. Total volume growth for the 6 years after treatment was then calculated. 2.4. Statistical analyses All analyses of fertilizer response were done in terms of total individual tree growth. The response variables analyzed were 3-year basal area growth and 6-year basal area, height and volume growth. Fertilizer response of a single tree was defined as the difference between growth after fertilization and the growth that would have occurred if the tree had not been fertilized. To predict how the fertilized trees would have grown had they not been fertilized, simple linear models of post-treatment growth as a function of pre-treatment growth were fit to control tree data from each experiment ŽMcWilliams and Burk, 1994.. For each experiment, the following equations were fit to control tree data: Bapost3s a q b Ž BApren. Bapost6s a q b Ž BApren. VOLpost6s a q b Ž BApren. Htpost6s a q b Ž BApren. Htpost6s a q b Ž HT0 .
Ž 1. Ž 2. Ž 3. Ž 4. Ž 5.
where: BApren is n years pre-treatment basal area growth Ž n s 1 to 5.; Bapost3 is 3 years post-treatment basal area growth; Bapost6 is 6 years posttreatment basal area growth; VOLpost6 is 6 years post-treatment volume growth; Htpost6 is 6 years
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
post-treatment height growth; HT0 is height at the time of treatment. The best fit equation for each of the response variables in each experiment was then used to predict how the fertilized trees would have grown in that dimension if they had not been fertilized. The range and average R 2 values for each response variable for the best fit equations are available on request from the principal author. No significant relationships were found between post-treatment height growth and either pre-treatment basal area growth or initial height. As a result, height growth response was simply analyzed as the difference between average height growth on the fertilized and control plots. To obtain estimates of absolute basal area response and absolute volume response to fertilization, the predicted unfertilized growth for each tree Žbased on the appropriate control growth equation. was subtracted from the observed post-treatment growth for that tree. To obtain estimates of relative fertilized growth, the observed post-treatment basal area and volume growth for each tree was expressed as a percentage of its respective predicted unfertilized growth. These estimates of absolute and relative fertilizer responses were then used as the dependent variables in all subsequent analyses. Responses are presented in both absolute and relative terms to allow for a complete analysis of the results. Absolute response provides a direct measure of the additional wood grown as a result of fertilization. This is the quantity in which those planning to harvest the trees are ultimately interested. However, absolute response is influenced by the initial size and past growth rate of the tree of interest, making correlations between site and stand characteristics across experiments poor. Relative response measures the increase in growth rate and thereby takes into account differences in initial tree sizes and past growth rates. As a result it generally shows a higher correlation with other site and stand variables than absolute response. However, relative response values do not provide information on the actual amount of additional wood produced unless unfertilized growth rates are also presented. Absolute and relative responses were also calculated for the control trees for use in subsequent analyses of variance ŽANOVAs.. Because the equa-
279
tions used to predict post-treatment growth were fit to control tree data in each experiment, the mean absolute and relative control responses for every experiment were 0 and 100, respectively. The absolute responses for individual control trees simply being the errors about the fitted regression lines. Using these values in analysis of variance to test for treatment effect results in the mean treatment responses being tested against average control responses of 0 or 100 with a full accounting of the variability about the equation used to predict control growth. This allows control tree response to be non constant rather than fixed at 0 or 100 with no variance. ANOVAs on absolute and relative basal area and volume responses were carried out for each individual experiment. ANOVAs on height growth were also carried out for each experiment. The experimental design specified for each experiment was a completely randomized design with subsampling. For those experiments showing a significant fertilizer effect Ž p F 0.10. two orthogonal post-ANOVA contrasts, one testing average control response vs. average fertilized response and the other testing average N alone response vs. average blend response were carried out. Linear regression analyses were used to examine trends between relative basal area response and site and stand characteristics across treatments and year of response. These were performed by first fitting separate slopes and intercepts for the two fertilizer treatments in each of the two years Žfull model, cf. Weisberg, 1985. and then fitting reduced models Že.g., common slope and intercept for each treatment, common slope and intercept for each year, common slope and intercept for all data. and testing the hypotheses that the reduced models explained as much of the observed variability as the full model. Average treatment responses from each experiment were used as the dependent variables. In some instances scatter and residual plots indicated nonlinear relationships that could be described by a power function Ž y s a P x b .. In these cases log transformations of both the response and predictor variables were used as a linear approximation of the power function and, if this provided the best fit, nonlinear regression was then used to fit a power function to the data.
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
280
3. Results 3.1. Fertilizer uptake An evaluation of foliar nutrient concentrations Žgrkg. and contents Žmg nutrient element per 100 needles. shows that the applied nutrients were taken up by the trees. Foliar N concentrations increased an average of 3.2 and 3.1 Ž26% and 25%. for the nitrogen alone and blend treatments, respectively ŽTable 3.. Phosphorus was added to the blend treatment in only five experiments as concentrations were deemed to be adequate in the remaining experiments. Phosphorus applications resulted in no apparent increase in foliar P concentration or content. Pretreatment reconnaissance showed that potassium concentrations were not particularly low. As a result, K was added to the blend treatment in only 16 experiments. Across all experiments, no dilution of K was evident following the N alone treatment. Additions of K in the blend treatment resulted in significantly higher foliar K concentrations and content Ž p - 0.01; Table 3.. Sulphur was included in the blend treatment for all experiments. No significance difference was observed between total sulphur concentrations for the two fertilizer treatments. However, total sulphur content was slightly, but significantly higher in the blend treatment than in the N alone treatment Ž p - 0.01; Table 3.. Sulphate–sulphur concentrations and con-
tents in both the N alone and blend treatments were considerably lower than those in the control treatments. This is likely the result of sulphate sulphur reserves being converted to organic S forms in the presence of the additional N applied. Sulphate– sulphur availability may have limited response to N in both treatments as evidenced by the low concentrations found ŽTable 3.. 3.2. Magnitude and significance of fertilizer responses Only five experiments showed a significant fertilizer effect on height growth 6 years after treatment, two of these being negative effects. These observations are likely due to small height growth responses, measurement errors associated with determining heights with a clinometer and tape, and the inability to find a covariate to explain any initial differences in height. For these reasons, little confidence was placed in the observed height growth responses and consequently the observed volume responses. As a result basal area responses were chosen to indicate fertilizer effects. The impact of different significance levels on the 3 year summary fertilizer response is given in Table 4. By showing the percentage of experiments with a significant positive fertilizer response by varying levels of significance, an indication of the potential impact of repeating the experiment with larger sam-
Table 3 Average concentrations Žgrkg. and contents Žmgr100 needles. of applied nutrient elements across all 48 experiments Nutrient
Concentrations Žgrkg. c
Nitrogen Phosphorusa Phosphorusb Potassiuma Potassiumb Sulphur d Sulphate Sulphur d,e a
Contents Žmgr100 needles.
Control
N alone
Blend
Control
N alone
Blend
12.5 1.7 2.3 7.0 7.2 1.2 319
15.7 1.6 1.9 7.0 7.2 1.1 64
15.6 1.6 2.0 7.5 7.1 1.1 81
6.37 0.91 1.13 3.44 3.66 0.621 0.155
9.06 1.00 1.11 3.91 4.15 0.625 0.036
9.10 0.98 1.14 4.33 4.11 0.48 0.045
Summary of experiments receiving the indicated nutrient in the blend treatment. Summary of installation not receiving the indicated nutrient in the blend treatment. c Nitrogen was applied to all experiments in both the N alone and blend treatments. d Sulphur was applied to all experiments in the blend treatment. e Concentration expressed in grMg. b
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
281
Table 4 Ranges of site and stand characteristics for experiments showing varying levels of significant positive 3 year relative basal area growth response Characteristic
p F 0.05
p F 0.10
p F 0.15
p F 0.20
% of experiments showing a significant response 3 year relative basal area growth Site index Žm at 50-yr bh. Pre-treatment foliar N Žgrkg. Pre-treatment foliar SO4 –S ŽgrMg. Pre-treatment foliar NrSO4 –S ratio Pre-treatment foliar P Žgrkg. Pre-treatment foliar NrP ratio Mineral soil min-N ŽgrMg. Total min-N Žkgrha.
25 1.11–1.74 16–35 9.7–15.1 202–630 20–59 1.6–2.9 4.2–7.3 3–30 5.9–39.7
42 1.11–2.04 16–37 9.5–15.1 158–630 20–78 1.6–3.4 3.6–7.3 3–30 5.9–39.7
54 1.11–2.04 16–37 9.5–15.1 158–630 20–78 1.6–3.4 3.6–7.3 3–30 5.9–39.7
68 1.11–2.04 16–37 9.5–15.1 158–630 17–78 1.6–3.4 3.6–7.3 3–30 5.9–40.1
ple sizes effect is also provided. The range of responses presented are based on the average of the two fertilizer treatments in each experiment. Also included in Table 4 are ranges for selected characteristics of the experiments showing significant responses. Lowering the significance level more than doubled the percentage of significantly responding experiments in both year 3 as well as 6 Žnot presented.. However, for year 3, lowering the significance level did little to change the summary of characteristics of responding sites. For year 6, the ranges of characteristics of responding sites using p F 0.05 or F 0.20 were very similar to those observed for year 3. For p F 0.05 or F 0.10, the ranges were narrower with the responding sites being on the lower end of the observed distributions of foliar and soil nutrient levels, particularly foliar N. The lower foliar N levels the more certain the response.
respectively. These were the lowest observed in any experiment. An examination of all other foliar nutrient concentrations and ratios did not indicate any other potential nutrient deficiencies or imbalances. 3.4. N alone Õs. blend treatment For experiments showing a significant positive fertilizer effect, results of the post-ANOVA contrasts showed that eight experiments had a significant difference between the N alone and blend treatments 3
3.3. NegatiÕe fertilizer effect Experiment 4 showed a significant negative absolute fertilizer effect both 3 and 6 years after treatment. There was also a significant difference between the two fertilizer treatments both 3 and 6 years after treatment with the blend treatment showing the more pronounced negative effect ŽFig. 1.. One potential reason for this observed negative fertilizer effect is an induced P deficiency. This experiment had one of the lowest pre-treatment foliar P levels Ž1.5 grkg. and one of the highest Ž9.1. foliar NrP ratios. One year after fertilization the foliar P levels in the N alone and blend treatments were 1.3 and 1.2 grkg,
Fig. 1. Average basal area growth per tree by treatment Žbased on the average of two plots, five trees per plot. over time for the experiment 4 which showed significant negative fertilizer effects. The solid line represents control growth, the large dash growth in the N alone plots and the small dash growth in the blend plots. The vertical dashed lines represent the time of fertilization dividing the graph into pre- and post-treatment growth periods.
282
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
ments showing a significant positive 3 year absolute basal area response, all but one displayed a decline in response during the second 3-year period with total 6 year responses being 1.1 to 1.9 times that of the total 3-year response. The one experiment showing an increased basal area response was on a lowproductivity site ŽS119. with an initial foliar N concentration of 10.9 grkg. 3.6. Relationships between basal area responses and site and stand characteristics
Fig. 2. Six year relative basal area responses for each treatment for each of the 6 site associations: Ž1. Lichens; Ž2. Salal; Ž3. Flat moss, slightly dry; Ž4. Flat moss, fresh; Ž5. Sword fern; Ž6. Foamflower. B Represents the N alone treatment and l represents the blend treatment.
years after treatment. Of these eight, four had a greater response to the N alone treatment and four had a greater response to the blend treatment. 6 years after treatment, only one experiment out of each group of four continued to show a significant difference between the two treatments. Evaluation of site and stand characteristics did not reveal any unique characteristics for these experiments that could explain the observed significant differences between the two fertilizer treatments.
Scatter plots of 3 and 6 year absolute and relative basal area response vs. site and stand variables indicated that relative responses were more strongly related to site and stand variables than absolute responses. A weak overall trend of declining average absolute basal area responses Žboth 3 and 6 year. was observed with increasing site index. In addition, the variation in absolute basal area response also increased with increasing site index. This increased variation was due in part to the increased variation in unfertilized growth on better sites reflecting the age range of the experimental sites. Some large signifi-
3.5. Relationships between 3 and 6 year responses While the number of experiments showing a significant fertilizer effect declined between 3 and 6 years, all experiments showing a fertilizer effect at year 6 also showed a significant effect at year 3. Greater variation in response measures and declining responses were the two primary reasons why fewer significant fertilizer effects were observed after 6 years. On average, across all experiments and both treatments, the fertilized trees continued to grow at a faster rate than the control trees in the second 3-year period but only at approximately half the rate of the first 3-year response period. Of a total of 24 experi-
Fig. 3. Relative basal area response vs. site index. Hollow plot symbols indicate year 3 responses, shaded plot symbols indicate year 6 responses. B Symbols represent the N alone treatment and l symbols represent the blend treatment. The dashed lines represent the regressions fit to each year’s data. For year 3, the equation is relative responses 2.326y0.034 Ø SI Ž R 2 s 0.46, SE E s 0.20.. For year 6, the equation is relative responses 2.099y 0.030 Ø SI Ž R 2 s 0.46, SE E s 0.18..
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
Fig. 4. Relative basal area response vs. mineral soil mineralizable N ŽgrMg.. Hollow plot symbols indicate year 3 responses, shaded plot symbols indicate year 6 responses. B Symbols represent the N alone treatment and l symbols represent the blend treatment. The dashed lines represent the regressions fit to each year’s data. For year 3, the equation is relative responses 2.282 Ø Žmin N.y0 .198 Žcorrected R 2 s 0.50.. For year 6, the equation is relative responses 2.020 Ø Žmin N.y0 .182 Žcorrected R 2 s 0.46.. Note: Removing the experiment with mineralizable Ns93 ŽgrMg. only slightly altered the fit of the equations, for year 3 relative response s 2.286 Ø Žmin N.y0 .199 Žcorrected R 2 s 0.48. and for year 6, relative responses 2.033 Ø Žmin N.y0 .184 Žcorrected R 2 s 0.44..
cant positive absolute responses were observed at high-site index sites ŽSI G 35.. These sites were characterized by the absence of growing season water deficits, relatively low foliar N concentrations compared to other high sites Ž12 and 13 grkg., and adequate foliar P concentrations Ž2.0 grkg.. A trend of declining average relative response with increasing site quality Ži.e., decreasing water deficit and increasing levels of plant-available soil N. was evident ŽFig. 2.. Significant differences in 6 year relative basal area response were observed for both fertilizer treatments between site association 1 and the five other site associations, and between site associations 2 and 5, and 2 and 6 ŽTukey’s honestly significant difference, p F 0.1.. From scatter plots, relationships between relative response and site index, control foliar N Žgrkg., control foliar sulphate–sulphur Žgrkg., mineral soil mineralizable N ŽgrMg., and total mineralizable-N Žkgrha. were identified and examined using regres-
283
sion analyses. No other site or stand factors showed any potential relationship with average relative or absolute response. In all cases a single equation fit to each year’s data provided the best fit Že.g., there were no significant differences between equations fit to each treatment for each of the two response periods.. Site index, mineral soil mineralizable-N and total mineralizable-N ŽFigs. 3–5. were found individually to be the best predictors of relative response, each explaining approximately half of the observed variance. Of these, site index was considered the most useful predictor based on its relative ease of measurement. Due to problems of multicollinearity between the above predictor variables alternative models of relative response as a function of two or more independent variables only marginally accounted for any additional variation in relative response. This, combined with the difficulties and costs of measuring soil and foliar nutrient levels, resulted in no other
Fig. 5. Relative basal area response vs. total mineralizable N Žkgrha.. Hollow plot symbols indicate year 3 responses, shaded plot symbols indicate year 6 responses. B Symbols represent the N alone treatment and l symbols represent the blend treatment. The dashed lines represent the regressions fit to each year’s data. For year 3 the equation is relative responses 2.342 Ø Žmin N.y0 .191 Žcorrected R 2 s 0.47.. For year 6, the equation is relative responses 2.149 Ø Žmin N.y0 .188 Žcorrected R 2 s 0.50.. Note: Removing the experiment with mineralizable Ns176 ŽgrMg. only slightly altered the fit of the equations, for year 3 relative response s 2.367 Ø Žmin N.y0 .195 Žcorrected R 2 s 0.45. and for year 6 relative responses 2.193 Ø Žmin N.y0 .196 Žcorrected R 2 s 0.49..
284
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
Table 5 Percentage of experiments showing significant basal area responses by control foliar N and SO4 –S classes
Foliar N F 12 grkg Foliar N ) 12 grkg Foliar SO4 –S F 400 grMg Foliar SO4 –S ) 400 grMg Foliar N F 12 grkg and Foliar SO4 –S G 400 grMg
Total number of experiments
% Showing significant basal area responses 3 year absolute 3 year relative 6 year absolute
6 year relative
19 29 35 13 8
79 34 48 62 75
37 10 20 23 38
equations being considered more useful than relative basal area response as a function of site index. Past research has suggested that nitrogen fertilizer-induced sulphur deficiencies may limit responses in coastal Douglas-fir stands of the Pacific Northwest ŽTurner et al., 1979; Turner and Gessel, 1988.. It has been suggested that foliar sulphate– sulphur provides an indication of both the sulphur and nitrogen status of the tree and site and therefore is a more sensitive indicator of N fertilization response than foliar total S or N alone. Turner et al. Ž1979., using a different chemical analytical technique that was likely somewhat less rigorous, suggested that critical limits of 400 grMg for foliar sulphate–sulphur and 12 grkg for foliar N could be used to separate N-fertilization responding from non-responding stands. These criteria did improve identification of N-fertilization responding and nonresponding stands in the current study ŽTable 5.. For the sites observed in this study foliar N tended to be a better predictor of response than foliar sulphate– sulphur. Conversely, an examination of stands with low foliar N concentrations Že.g., F 12 grkg. and high in foliar sulphate–sulphur concentration Že.g., G 400 grMg. did not find these stands to be more responsive than the general population of stands with foliar N concentrations F 12 grkg ŽTable 5.. All previous analyses used experiment treatment means in assessing the relative utility of individual site and stand characteristics as response predictors. While the use of average crop tree measures diminishes the effects of stand structure on the fertilizer response, variations in individual crop tree characteristics may still play a significant role in determining the response. To examine this possibility, relationships between individual crop tree attributes Žage, DBH, HT, LLC, WLC, %LC, CA, DNN. and fertil-
74 21 40 46 63
37 28 31 31 25
izer response were also examined. Scatter plots of absolute response vs. these variables did not indicate any linear or nonlinear relationships. Scatter plots of relative response vs. individual tree measures showed weak negative linear relationships between relative response and LLC, WLC, CA, DBH, and HT. However, all of these individual tree measures showed a positive linear relationship with site index and the addition of any of these variables to the regression of relative response as a function of site index did not result in a better predictive equation.
4. Discussion With only 4 of 48 experiments showing a greater response to the blend than to the N alone treatment after 3 years and only 1 of 48 showing a greater response after 6 years, the results of this study suggests that nitrogen is the major limiting nutrient in immature stands of Douglas-fir in coastal British Columbia. Further, sites showing a greater response to the blend treatment had no distinct identifying feature indicating that there was no other common nutrient deficiency on these sites. This is in contrast to recent findings in interior locations that has indicated potential potassium and magnesium limitations ŽMika and Moore, 1990; Mika et al., 1992; Velazquez-Martinez et al., 1992. and in western Washington and Oregon that has indicated potential sulphur deficiencies ŽBlake et al., 1988.. Site index and mineralizable-N were found individually to be the best predictors of relative response while no strong relationships were found between absolute response and any of the site and stand characteristics examined. Given the wide range of sites, it could be expected that site index, a measure
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
that represents an integration of growing conditions, would be one of the better predictive variables. With the confirmation of N as the major limiting nutrient across the study sites it was expected that mineralizable-N would be one of the best predictors of fertilizer response. These results are supported by those found by Shumway and Atkinson Ž1978., Powers Ž1980., and Radwan and Shumway Ž1984.. CrN ratio, found to be a useful response predictor by Peterson et al. Ž1984. and Miller et al. Ž1989., was not identified as a potential response predictor in this study. Mineral soil-extractable sulphate–sulphur recognized as a Douglas-fir nitrogen fertilization response predictor by Blake et al. Ž1988. was not identified as a potential response predictor in this study. However, extractable sulphate–sulphur levels in the current study were generally higher than those cited by Blake et al. Ž1988.. This may be due to differences in parent material mineralogy between the soils in the present study and those of western Washington and Oregon studied by Blake et al. Ž1988.. Foliar sulphate–sulphur concentrations also failed to offer the useful prediction of N-fertilizer responses identified by Turner et al. Ž1979., who suggested that foliar sulphate–sulphur indicates both the sulphur and nitrogen status of the tree and site and is, therefore, a more sensitive indicator of N fertilization response than foliar total S or N alone. This statement was based in part on the number of correctly predicted responses to N fertilization based on comparisons of foliar nutrient levels to accepted deficiencyrsufficiency levels. Turner et al. Ž1979. found foliar sulphate–sulphur to be a correct predictor of response in 17 of 19 Ž89.5%. stands and foliar N to be a correct predictor in only 12 of 19 Ž63.2%. stands. Similar results were not found in this study. Following somewhat different chemical analytical procedures with the same critical levels of 12 grkg for N and 400 grMg for sulphate–sulphur, correct predictions of 3 year absolute and relative, and 6 year absolute and relative basal area responses would have been made for 71%, 77%, 58%, and 69% of the experiments respectively based on N levels and 54%, 56%, 58%, and 64% of the experiments, respectively, based on sulphate–sulphur levels. In addition there were no apparent response-limiting critical concentrations of sulphate–sulphur nor were there
285
consistently improved responses when sulphur was added in the blend treatment. No individual tree characteristics added significantly to the prediction of relative basal area response. This affirmed the choice of well-thinned stands and dominant or elite codominant sample trees to minimize the effect of stand and tree characteristics on response and allow for a more precise comparison of responses across a wide range of sites. Candidate stands were required to have been thinned to a level that allowed essentially ‘free-to-grow’ above-ground conditions. All study trees had been dominants and elite-codominants prior to thinning and had well developed crowns ŽTable 1.. This emphasis of the study design on minimizing variation in individual tree and stand structural characteristics resulted in reduced opportunities for using these characteristics to explain variation in fertilizer response. Comparisons of relationships between fertilizer response and site and stand parameters found in this study, and those identified by other researchers brings into question the portability of research results of this kind. Given the large number of potential variables that can influence fertilizer response most well designed fertilization experiments are constructed to minimize the variation in specific variables so that the role of other variables may be more easily identified. Unfortunately, this unavoidably limits the extrapolation of results to other sites and stand types. In addition, specific individual measures Že.g., CrN ratio, extractable sulphate–sulphur. would only be expected to be good predictors of fertilizer response across a relatively uniform range of sites where they are the most variable characteristics determining the response. Measures such as site index that represent an integration of growing conditions can be expected to be the best predictors of responses across a wide range of sites. Results of the present study, showing site index and mineralizable N to be the best predictors of relative response, should be widely applicable within coastal British Columbia given the range of sites studied. However, unknown influences of past stand management and existing stand structure on stand level fertilizer responses need to be acknowledged. Ideally, for operational purposes, response predictors should be efficient to measure as well as widely
286
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
applicable. It is critical that distinctions are made between response predictors that are valuable for research purposes vs. those that are valuable for operational practice. With the overall goal of this study being the development of site-specific guidelines for fertilization of coastal Douglas-fir for operational use, potential response predictors were examined not only in terms of their empirical relationship to fertilizer but also in terms of their ease of measurement. In this regard, site index ranked highest in terms of having one of the best empirical relationships to fertilizer response and also being one of the most efficient to measure accurately. Mineralizable N, while having one of the better empirical relationships to relative fertilizer response is expensive and requires considerable expertise to measure consistently and accurately. This is also true of foliar chemical measures. Miller et al. Ž1989. also concluded that the most practical method for predicting response of Douglas-fir to fertilization is to use variables commonly collected, i.e., age, site index and density. They stated that tests for N or organic matter in the soil were not cost-effective measures for improving predictions of fertilizer response. Shumway and Olson Ž1992. also concluded that site index is the single best diagnostic tool for selecting Douglas-fir stands suitable for nitrogen fertilization. In designing an experiment, the aim is to have sufficient replication to observe treatment differences of practical significance. In this experiment, in order to cover the desired wide range of sites, the number of experimental units per treatment had to be limited to two with a subsample of five trees in each. When, as was the case in this experiment, there is a failure to observe significant treatment effects despite strong trends indicating a response to treatment, the inevitable questions arise ‘would more replication have resulted in observation of additional statistically significant treatment responses, and if so, would this alter the interpretation of the results?’ These questions were particularly important in this experiment given the goal of developing operational guidelines for site selection. In an attempt to address the impact of changing sample size summaries of responding experiments characteristics were completed for varying levels of statistical significance ŽTable 4.. The results show that an interpretation of the characteristics of re-
sponding sites 3 years after treatment was not very sensitive to the chosen level of significance. This suggests that although increased replication may have increased the number of significantly responding sites this in turn would not likely change a summary of the characteristics of responding sites. In contrast to the 3 year relative basal area growth responses ŽTable 4., interpretation of the characteristics of responding experiments 6 years after treatment was sensitive to the chosen level of significance. As a result of the declining response rates over the second 3 year period and the increased variation in growth measurements due to the longer time period, raising the significance level resulted in the exclusion of experiments with smaller relative responses. Experiments that continued to show a significant response Ž p F 0.05. support the relationships identified between relative response and site index, mineral soil mineralizable N, and control foliar N. It is important to consider the question of observing statistically and practically significant responses over time. The total response to fertilization at any point in time is the difference between the total growth of the stand since treatment and the total growth that would have occurred without treatment. If, as is often the case, the difference between fertilized and unfertilized growth diminishes over time, the total response becomes relatively smaller with respect to total stand growth and the variability of growth within the stand. As a result, it becomes increasingly difficult to observe significant responses over time. On sites with pronounced responses this will not likely pose a large problem but on sites with very short lived responses it will. If a response is short lived, it will be less likely to be statistically significant over the rotation; however, it still may be a practically significant response. A financial analysis carried out by McWilliams and Carter Ž1994. showed that even a small response Žmeasured as a 2-year reduction in rotation length. on high-productivity sites ŽSI ) 35. fertilized at age 20 provided a return on investment at discount rates as high as 8%. Therefore it is important not to disregard short-lived responses, particularly on high-productivity sites where unfertilized rotations lengths are the shortest, reducing the length of time the fertilizer investment must be carried.
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
An equal absolute fertilizer response across a range of sites will be represented as an increasingly larger relative response as unfertilized growth rates decrease. Large relative responses can be misleading as the same absolute response will have a greater economic value Žin terms of sawlog production. on larger faster growing trees rather than on smaller, slower growing ones. At the stand level the greatest financial returns from fertilization comes from maximizing the production of additional wood or viewed in alternate terms, from maximizing the reduction in rotation age. Absolute response showed a weak decline with increasing site index but the fact that some of the higher-productivity sites showed similar absolute responses to the low-productivity sites is worth noting. In this study, high-productivity sites responding to fertilization were characterized by the absence of growing season water deficits and relatively low foliar N and soil mineralizable-N concentrations. Brix Ž1972. who observed a highly significant positive interaction between summer irrigation and N fertilization of Douglas-fir supports this observation. Brix Ž1991. also stated that this significant interaction between moisture and nutrients indicates that growth response to N fertilizer would be best in years and on sites with favourable water conditions. If in the future high productivity sites that respond to fertilization can be consistently characterized, they will represent the greatest opportunities for economic gains from fertilization. 5. Summary The results of an extensive Douglas-fir fertilization experiments have shown significant increases in basal area growth 3 and 6 after the addition of nitrogen. They have also offered a basis for predicting a differential response between sites. Site index, site associations, and concentrations of mineral soil mineralizable-N, foliar N, and foliar sulphate–sulphur appeared to have some utility as response predictors. Site index was found to be the single best predictor of relative response in terms of practical utility. Site associations did not perform significantly better than site index as a response predictor. No individual tree or stand structural parameters were identified that improved the explained variation
287
in treatment response. This was attributed to the study design rather than a definitive conclusion regarding the role of tree and stand conditions in determining treatment response. Average relative response was shown to decline with increasing site productivity. This indicated that the probability of obtaining a fertilizer response will be highest on low-productivity sites. However, this does not mean that fertilizer operations should only focus on low-productivity poor sites. Significant absolute and relative responses were observed across the full range of sites. This indicates on opportunity to improve on average returns from fertilization by not only treating the low productivity sites but by being able to select those medium- and high-productivity sites showing the best responses to fertilization. Since the greatest financial returns will be realized by fertilizing high-productivity sites that respond to fertilization it appears that more costly, intensive pre-treatment analyses provide the best returns on better sites. On less productive sites less intensive response estimators such as site index likely represent the most cost-effective approach.
Acknowledgements Financial support for this study was provided by the Canadian Forest Service and the British Columbia Ministry of Forests under the Canada–British Columbia Forest Resource Development Agreement, Extension and Demonstration Subprogram Ž1985– 1990. and the Canadian Forest Service under the Canada–British Columbia Forest Resource Development Agreement, Research Program Ž1991–1994.. This support is gratefully acknowledged.
References Allen, H.L., 1983. Forest soils shortcourse. North Carolina State Forest Nutrition Cooperative, Raleigh, NC. Ballard, T.M., Carter, R.E., 1986. Evaluating forest stand nutrient status. Land Management Report No. 20, B.C. Ministry of Forests, Victoria, British Columbia. Beaufils, E.R., 1973. Diagnosis and Recommendation Integrated System ŽDRIS.. Soil Sci. Bull. 1, University of Natal, Pietermaritzburg, South Africa. Blake, J.I., 1985. Characterization of soil nitrogen and sulfur availability in relation to volume response of Douglas-fir in
288
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289
western Oregon and Washington. PhD dissertation, University of Washington, Seattle, WA. Blake, J.I., Webster, S.R., Gessel, S.P., 1988. Soil sulfate–sulfur and growth responses of nitrogen fertilized Douglas-fir to sulfur. Soil Sci. Soc. Am. J. 52, 1141–1147. Brix, H., 1972. Nitrogen fertilization and water effects on photosynthesis and earlywood–latewood production in Douglas-fir. Can. J. For. Res. 2, 467–478. Brix, H., 1991. Mechanisms of response to fertilization: II. Utilization by trees and stands. In: Lousier, J.D., Brix, H., Brockely, R., Carter, R., Marshall, V.G. ŽEds... Improving forest fertilization decision-making in British Columbia. B.C. Ministry of Forests, Victoria, British Columbia. Bruce, D., 1981. Constant height-growth and growth-rate estimates for remeasured plots. For. Sci. 27, 711–725. Cajander, A.K., 1926. The theory of forest types. Acta For. Fenn. 2 Ž3., 11–108. Carter, R.E., Klinka, K., 1988. Douglas-fir fertilization decisionmaking for industrial use: an establishment report. FRDA Report No. 33. Forestry Canada, Pacific and Yukon Region, Victoria, British Columbia. Comerford, N.B., Fisher, R.F., 1982. Use of discriminant analysis for classification of fertilizer responsive sites. Soil Sci. Soc. Am. J. 46, 1093–1096. Daubenmire, R.F., 1968. Plant communities. Harper and Row, New York. Edmonds, R.L., Hsiang, T., 1987. Forest floor and soil influence on response of Douglas-fir to urea. Soil Sci. Soc. Am. J. 51, 1332–1337. Green, R.N., Klinka, K., 1994. A field guide to site identification and interpretation for the Vancouver Forest Region. Land Manage. Handbook No. 28, B.C. Ministry of Forests, Victoria, British Columbia. Hockman, J.N., Allen, H.L., 1990. Nutritional diagnoses in Loblolly pine stands using a DRIS approach. In: Gessel, S.P., Lacate, D.S., Weetan, G.F., Powers, R.F., ŽEds... Sustained productivity of forest soils. Proceedings of the 7th North American Forest Soils Conference, University of British Columbia, Vancouver, British Columbia, pp. 500–514. Ingestad, T., 1971. The concept of nutrient requirement in plants. In: Samish, R.M. ŽEd... Recent Advances in Plant Nutrition. Gordon Breach Science Publishers, New York. Johnson, C.M., Nishita, H., 1952. Microestimation of sulphur in plant materials, soils and irrigation waters. Anal. Chem. 24, 736–742. Klinka, K., Carter, R.E., 1990. Relationships between site index and synoptic environmental factors in immature coastal Douglas-fir stands. For. Sci. 36 Ž3., 815–830. Krajina, V.J., 1972. Ecosystem perspectives of forestry. H.R. MacMillan Forestry Lecture Series, Faculty of Forestry, University of British Columbia, Vancouver, British Columbia. Klinka et al., 1991. Lipas, E., 1985. Assessment of site productivity and fertilizer requirements by means of soil properties. Folia For. 618. McWilliams, E.R.G., Burk, T.E., 1994. Evaluation of eight forest fertilizer response estimators by means of a simulation study. Can. J. For. Res. 24, 107–119.
McWilliams, E.R.G., Carter, R.E., 1994. Soo TSA forest fertilization 10-year plan. Contract Report ŽContract No. SA92V03-51. to B.C. Ministry of Forests, Victoria, British Columbia. Mika, P.G., Moore, J.A., 1990. Foliar potassium status explains Douglas-fir response to nitrogen fertilization in the Inland Northwest, USA. Water, Air and Soil Pollution 54, 477–491. Mika, P.G., Moore, J.A., Brockley, R.P., Powers, R.F., 1992. Fertilization response by interior forests: When, where and how much? In: Chappell, H.N., Weetman, G.F., Miller, R.E. ŽEds.. Forest fertilization: sustaining and improving nutrition and growth of western forests. Contribution No. 73, College of Forest Resources, University of Washington, Seattle, WA. Miller, R.E., Fight, R.D., 1979. Fertilizing Douglas-fir forests. United States Department of Agriculture, For. Serv. Gen. Tech. Report PNW-83, Corvallis, OR. Miller, R.E., McNabb, D.H., Hazard, J., 1989. Predicting Douglas-fir growth response to nitrogen fertilization in western Oregon. Soil Sci. Soc. Am. J. 53, 1552–1560. Mitchell, K.J., Cameron, I.R., 1985. Managed stand yield tables for coastal Douglas-fir: Initial density and precommercial thinning. Land Manage. Report No 31, B.C. Ministry of Forests, Victoria, British Columbia. Morrison, I.K., 1974. Mineral nutrition of conifers with special reference to nutrient status interpretation: A review of literature. Can. For. Serv. Publication No. 1343, Ottawa, Ontario. Nuszdorfer, F.C., 1981. Bulk density. In: Klinka, K., Green, R.N., Trowbridge, R.L., Lowe, L.E. ŽEds... Taxonomic classification of humus forms in ecosystems of British Columbia. Land Manage. Report No. 8, B.C. Ministry of Forests, Victoria, British Columbia. Omule, S.A.Y., 1987. Total and merchantable volume equations for small coastal Douglas-fir. FRDA Report No. 10, B.C. Ministry of Forests, Victoria, British Columbia. Peterson, C., Ryan, P.J., Gessel, S.P., 1984. Response of northwest stands of Douglas-fir to urea: correlations with forest soil properties. Soil Sci. Am. J. 48, 162–169. Pojar, J., Klinka, K., Meidinger, D.V., 1987. Biogeoclimatic ecosystem classification in British Columbia. For. Ecol. Manage. 22, 119–154. Powers, R.F., 1980. Mineralizable soil nitrogen as an index of nitrogen availability to forest trees. Soil Sci. Soc. Am. J. 44, 1314–1320. Radwan, M.A., Shumway, J., 1984. Site index and selected soil properties in relation to response of Douglas-fir and western hemlock to nitrogen fertilizer. In: Stone, E.L. ŽEd.., Forest soils and treatment impacts. Proceedings of the 6th North American Forest Soils Conference, University of Tennessee Press, Knoxville, TN. Shumway, J., Atkinson, W.A., 1978. Predicting nitrogen fertilizer response in unthinned stands of Douglas-fir. Comm. Soil Sci. and Plant Anal. 9, 529–539. Shumway, J., Olson, J., 1992. Stand selection criteria for nitrogen fertilization: current practices and future needs. In: Chappell, H.N., Weetman, G.F., Miller, R.E. ŽEds.. Forest fertilization: Sustaining and improving nutrition and growth of western forests. Contribution No. 73, College of Forest Resources, University of Washington, Seattle, WA.
R.E. Carter et al.r Forest Ecology and Management 107 (1998) 275–289 Timmer, V.R., Ray, P.N., 1988. Evaluating soil nutrient regime for black spruce in the Ontario Claybelt by fertilization. For. Chron. 64, 40–46. Turner, J., Gessel, S.P., 1988. Forest productivity in the southern hemisphere with particular emphasis on managed forests. In: Gessel, S.P., Lacate, D.S., Weetman, G.F., Powers, R.F. ŽEds... Proceedings of the 7th North American Forest Soils Conference, University of British Columbia, Vancouver, British Columbia, pp. 23–39. Turner, J., Lambert, M.J., Gessel, S.P., 1977. Use of foliage sulphate concentration to predict response to urea application by Douglas-fir. Can. J. For. Res. 7, 476–480. Turner, J., Lambert, M.J., Gessel, S.P., 1979. Sulphur requirements of nitrogen fertilized Douglas-fir. For. Sci. 25 Ž3., 461–467. Turner, J., Lambert, M.J., Gessel, S.P., 1988. Nitrogen requirements in young Douglas-fir of the Pacific Northwest. Fert. Res. 15, 173–179.
289
van den Driessche, R.N., 1974. Prediction of mineral nutrient status of trees by foliar analysis. Bot. Rev. 40 Ž3., 347–394. van den Driessche, R.N., 1979. Estimating potential response to fertilizer based on tree tissue and litter analysis. In: Gessel, S.P., Kenady, R.M., Atkinson, W.A. ŽEds... Proceedings of Forest Fertilization Conference, Contribution No. 40, College of Forest Resources, University of Washington, Seattle, WA. Velazquez-Martinez, A., Perry, D.A., Bell, T.E., 1992. Response of aboveground biomass increment, growth efficiency and foliar nutrients to thinning, fertilization and pruning in young Douglas-fir plantations in the central Oregon Cascades. Can. J. For. Res. 22, 1278–1289. Waring, S.A., Bremner, J.M., 1964. Ammonium production in soil under water-logged conditions as an index of nitrogen availability. Nature 201, 951–952. Weisberg, S., 1985. Applied Linear Regression. Wiley, New York.