Kammarrätten i Sundsvall. Skogsstyrelsen hävdar felaktigt att jag måste upprätthålla hög kubikmassa

1 Harald Holmberg 2009-03-31 Kammarrätten i Sundsvall Mål nr 868-08 Harald Holmberg / Skogsstyrelsen Skogsstyrelsen hävdar felaktigt att jag måste upprätthålla hög kubikmassa I det följande bevisar jag att jag måste hålla en låg stående kubikmassa av stamved per hektar (men hög bladyta) för att uppfylla 1 i skogsvårdslagen. Skogsstyrelsen kräver att jag skall hålla hög kubikmassa per hektar eftersom 5 och 10 föranstaltar detta. Jag följer de moderna vetenskapliga undersökningar, som utförts på ett korrekt sätt, medan skogsstyrelsen vid utformningen av 5 och 10 följde de undersökningar som senare visat sig vara felaktiga. Jag väljer att följa 1 men inte 5 och 10, eftersom de sistnämnda är oförenliga med 1. Claes Fries vid skogsstyrelsen har nu korrigerat sina felaktiga uppgifter, vilket innebär att myndigheten nu vet att de forskares resultat som ligger bakom utformningen av 5 och 10 är felaktiga. Myndighetens jurist anser att korrigeringen saknar betydelse, vilket jag uppfattar som om han inte alls förstår innebörden av korrigeringen. Skulle skogsstyrelsens chefsjurist förstå att den nuvarande lagens utformning leder till inoptimal volymproduktion i landets skogar och ekonomisk nettoförlust för skogsägarna, så skulle han genast be att detta mål avslutas med en uppmaning till mig att fortsätta mitt framgångsrika skogsbruk. Därefter skulle han sätta igång med en revision av skogsvårdslagens anvisningar. Myndigheten lämnar också som nytt bevis en nypublicerad debattartikel skriven av professorn Björn Elfving och docent Lars Lundqvist. Den lögn som framfördes under de muntliga förhandlingarna av Clas Fries vid skogsstyrelsen har en helt central betydelse för målets utfall, ty den gällde den vetenskapliga basen för den nuvarande skogsvårdslagens 5 och 10. I båda dessa paragrafer utgår myndigheten från att volymproduktionen i en skog ökar med dess täthet uttryckt i kubikmeter stamvirke per hektar. På denna felaktiga hypotes har virkesförrådsdiagrammet (bilaga 1 till 5 och 10 i skogsvårdslagens anvisningar) konstruerats. Skogsstyrelsens tjänsteman i Lycksele anförde i sitt skriftliga förbud till mig att min skog skulle bli för gles och att jag måste lämna 75 m3/ha för att få utföra min typ av gallring. För att inta denna ståndpunkt måste tjänstemannen ha använt virkesförrådsdiagrammet. Genom korrigeringen av Clas Fries har skogsstyrelsen klargjort att den nu litar på resultatet av analysen publicerad av Hagner och Holm (bilaga 1). Detta har en oerhörd betydelse eftersom det vänder värdet av skogsstyrelsens senast andragna bevis, som är författat av docenten Lars Lundqvist, från att vara positivt för skogsstyrelsen, till att bli ett bevis för att mitt skogsbruk är det bästa. Skälet därtill är följande. Statistikern Sören Holm och professor Mats Hagner gjorde en djupgående statistisk analys av det material som Lars Lundqvist publicerat i sin doktorsavhandling 1989. Deras nya analys visade att volymproduktionen minskade med ökad täthet uttryckt i kubikmeter stamvirke per hektar, dvs. tvärs emot det antagande som skogsvårdslagens 5 och 10 bygger på, och som

2 kommer till uttryck i virkesförrådsdiagrammet. Resultatet av arbetet gjort av Hagner och Holm innebär även att Lars Lundqvist i sin doktorsavhandling presenterade en slutsats som var omvänd den som han skulle ha presenterat om han gjort en korrekt analys av sitt stora material. Eftersom Lundqvist känner till att han gjort fel, och valt att försöka dölja detta genom att försöka få folk att misstro Mats Hagner, tar sig detta uttryck i den nya debattartikel som skogsstyrelsen lämnat som nytt bevis, samt i de artiklar av Lundqvist som skogsstyrelsen tidigare anfört som bevis. Professorn Björn Elfving, som är medförfattare till den inlämnade debattartikeln, har även egna skäl till att uttrycka ringaktning för den ekonomiska principen Naturkultur som Mats Hagner lanserat. Elfving har nämligen gett skogsstyrelsen det vetenskapliga underlaget till det nämnda virkesförrådsdiagrammet (Elfving 1993) i form av en stencil grundad på bearbetning av riksskogstaxeringens tillfälliga provytor. Den bearbetningen ledde fram till samma felaktiga hypotes som Lundqvist publicerat, dvs. att det föreligger ett positivt samband mellan stamvolym och virkesproduktion. Hagner har genom analyser av teoretiskt producerat material, liknande det i riksskogstaxeringens provytor, visat att det rent slumpmässigt uppstår sådana samband som Elving presenterat för skogsstyrelsen (bilaga 2). Elvings stencil gav därför skogsstyrelsen missvisande uppgifter. Det kan finnas prestigemässiga dvs. mänskliga skäl att skogsstyrelsen gärna vill dölja sitt misstag att lita på Elfvings analys, eftersom det lett till att lagens väktare tvingat skogsägarna till reducerade inkomster och orsakat en minskad råvarubas för skogsindustrin. Såsom Mats Hagner klargjort i bilaga 3 beror missuppfattningen att det finns ett positivt samband mellan kubikmassa stamvirke och volymproduktion på följande. Efter gallring uppstår det omedelbart ett samband mellan volymproduktion och intensiteten i gallringen. Ett stort uttag i gallring reducerar givetvis den stående kubikmassan stamvirke, men det reducerar samtidigt mängden klorofyll i form av bladyta, som är den egentliga grunden för all tillväxt. Klorofyllet fångar solenergin som senare omvandlas till biomassa. Misstaget som forskarna Elfving och Lundqvist gjort, grundar sig på att de blandar samman stamvirke, som är en tärande del av ekosystemet, med bladyta, som är den närande delen av ekosystemet. Efter ett uttag av träd i form av höggallring (uttag av stora träd) kan en full bladyta så småningom återställas på ett stort antal små träd som tillsammans har en liten stamvolym. Denna lilla stamvolym förbrukar en liten del av ekosystemets resurser för sitt uppehälle och det blir stort överskott av fotosyntesprodukter för nyproduktion av stamved, dvs. volymproduktionen blir hög. Efter ett lika stort uttag i form av låggallring (uttag av små träd) kan full bladyta utvecklas på ett fåtal stora träd som dock tillsammans har stor stamvolym. Denna stora stamvolym förbrukar en stor del av ekosystemets resurser för sitt uppehälle och det blir litet utrymme för nyproduktion av stamved, dvs. volymproduktionen blir låg.

3 Figur 1. Ovanstående figur med text är kopierad från Elfving (2005). Volymtillväxten är uppmätt under de tio år som följde efter gallring i en skiktad granskog på 500 möh i Jämtland. Skogen hade tidigare vårdats genom höggallring utförd ca 30-40 år före den nu företagna gallringen. Detta kan vara ett av skälen till skillnaden mellan den övre och nedre kurvan i figur 1. Resultatet i detta försök, placerat i en miljö liknande den jag har på min fastighet, dvs. en skiktad granskog på 500 möh i Jämtland, bevisar att min typ av skogsbruk leder till högsta möjliga volymproduktion. Av figuren framgår att volymproduktionen redan inom tio år efter gallringen låg på över 90 % av den i orörd skog, om gallringsuttaget låg på 30 %. Detta var dock fallet endast om uttaget gjordes i form av höggallring, dvs. uttag av stora mogna träd, såsom jag förordar. Gjordes uttaget i form av små träd, en gallringsform som skogsstyrelsen förordar, sänktes volymproduktionen till ca 50 % av den i orörd skog. Figur 1 är skapad av professorn Björn Elfving, efter bearbetning av material framställt av Lars Lundqvist. Den sistnämnde är den som anlagt gallringsförsöket som utgör underlaget för figuren. Trots att Elfving och Lundqvist själva har uppnått den kunskap som presenteras i figur 1, vilken väl stödjer, dels den ekonomiska princip som Mats Hagner presenterat såsom Naturkultur, dels det skogsbruk som jag bedriver, skriver de en debattartikel som är negativ mot det nämnda. Förklaringen måste sökas i att de två forskarna försöker dölja sina egna vetenskapliga misstag samt att de försöker förhindra att skogsmyndigheten och skogsnäringens ledarskikt förlorar allmänhetens förtroende. Skogsstyrelsen solidariserar sig med SLU när universitetet bryter mot lagen I det följande visar jag att SLU bryter mot grundlagen och att skogsstyrelsen inte misstycker. SLU Bröt mot grundlagen I Lantbrukets Affärstidning ATL 29 mars 2009 finns en artikel (bilaga 5) Där SLU får allvarlig kritik av justitieombudsmannen för två brott mot tryckfrihetsförordningen. SLU har hittat det perfekta juridiska medlet för att undgå granskning SLU har lyckats undanhålla allmän handling som Mats Hagner begärt i 8 år. Den allmänna handlingen är en datamatris som utgör underlag för en av Lars Lundqvist publicerad

4 regressionsanalys. Två oberoende statistiker garanterar att varje forskare som utfört en regressionsanalys vet vilken datamatris som utgjort underlaget för analysen (bilaga 4). Trots detta beslöt regeringsrätten att inte pröva SLU:s vägran grundad på chefsjuristens påstående att man inte förstod vilken allmän handling som avsågs. I denna långdragna konflikt ingick även moment av stöld. Mats Hagners forskningsdator stals från hans låsta tjänsterum på SLU. Den återfanns av polisen på Lars Lundqvists kontor, innan innehållet hade blivit raderat. Lundqvist och den tidigare dekanen och prefekten PO Bäckström påstod att man önskat ta tillbaka datorn eftersom den tillhörde universitetet, och att professor Mats Hagner gått i pension. Datorn återlämnades omedelbart efter polisens påhälsning. Slutsats om SLU:s och skogsstyrelsens juridiska agerande Det ovan nämnda bekräftar vad som jag tidigare visat, nämligen att SLU undviker kritik genom att vägra utomstående tillgång till allmän handling. SLU:s forskare kan fortsätta att publicera undermåliga vetenskapliga arbeten och de kan bemöta dem som upptäckt bristerna genom att offentligt publicera medvetna felaktigheter. Den av skogsstyrelsen insända debattartikeln är ett exempel på detta. Clas Fries lögnaktiga uttalande under den muntliga förhandlingen visar också att Fries, som tidigare disputerat på samma institution där Elfving och Lundqvist arbetar, tagit med sig denna tradition till sin nya arbetsgivare skogsstyrelsen. Skogsstyrelsen hävdar att jag inte använder en beprövad metod I det följande bevisar jag att den metod jag använder har tillämpats i praktisk skala under mycket lång tid och att den utförligt testats i många olika vetenskapliga försök. Resultaten är positiva. Vanligt kalhyggesbruk innebär att volymproduktionen aldrig når upp till idealboniteten. Denna definieras såsom högsta möjliga löpande volymproduktion, dvs. den produktion som en ogallrad skog har. Idealboniteten kan uppskattas indirekt, till exempel genom beskrivning av växterna på marken, ty artsammansättningen varierar med bördigheten. Att kalhyggesbruket inte kan nå upp till mer än ca 85 % av idealboniteten beror på att produktionen av stamved under kalhyggesfasen givetvis är mycket låg, samt på att fullvuxna träd under slutet av omloppstiden använder betydande energi åt reproduktion, dvs. produktion av kott och pollen. Gallring innebär att en del träd skördas och att skogens bladyta därmed reduceras. Eftersom detta minskar skogens totala uppfångande av solenergi går volymproduktionen av stambiomassa tillbaka efter en gallring. Träden känner dock att de fått mera plats och förlänger sina grenar, varpå de efter några år tillsammans återskapat lika stor bladyta som före gallringen. All gallring medför att den totala volymproduktionen understiger idealboniteten och att den löpande tillväxten är lägre än den som uppmäts i närliggande ogallrad skog. I det följande jämför jag volymproduktionen i olika försöksserier med idealboniteten, eller med den löpande tillväxten i orörd skog. Syftet är att bevisa att mitt sätt att gallra är vetenskapligt testat och att mitt sätt leder till hög volymproduktion. Jag vill poängtera att kalhyggesbruket, som skogsstyrelsen rekommenderar, troligen kan nå upp till 85 %.

5 1 Min gallring är praktiskt beprövad och bevisligen framgångsrik Jag tillämpar den typ av gallring som min far använde på denna fastighet, dvs. jag plockar ut mogna stora träd och småträd med dåliga egenskaper som står för tätt. Min fars gallring utfördes för många tiotals år sedan, och före honom gjorde hans far detsamma. Jag har tidigare gallrat sådan skog som min far skapade. Den innehöll nu mogna träd och den gav mig avsevärda inkomster. Detta bevisar att min fars sätt att gallra har gett mig och nationen hög avkastning, dvs. uppfyllt kraven skogsvårdslagens 1. Det är också ett bevis för att metoden är beprövad. Detta i synnerhet som omgivande fastigheters eftertraktade slutavverkningsbara skogar har uppkommit efter samma typ av gallring. Skogens värde är alltså hög, även utan att min far utförde berikande plantering i luckorna. Jag själv har lagt till denna dyrbara åtgärd för att garantera att eventuella luckor omedelbart blir fullt beklädda med återväxt. Denna extra åtgärd kan inte försämra resultatet, som enligt tidigare resonemang erfarenhetsmässigt gett ett acceptabelt resultat. 2 Min gallring är vetenskapligt beprövad och bevisligen framgångsrik Plockhuggning av mogna träd upprepades vart tionde år under mycket lång tid i en försöksserie anlagd och behandlad av lantbruksuniversitetet SLU. Namnet blädning användes för denna behandling och innebar gallring utan någon berikande plantering. De elva försöksytor, som behandlades på detta sätt, låg utspridda i Dalarna, Jämtland och Västerbotten. Antalet observationer blev sammanlagt 58, vilket innebär att det vetenskapliga materialet är mycket stort och att det täcker reaktionen hos skog under mycket lång tid. Den stora geografiska spridningen innebär att resultaten kan anses generaliserbara i södra och mellersta Norrland. Resultaten kan kortfattat sammanfattas i att blädningen resulterade i en volymproduktion som låg på 94 % av idealboniteten. Underlaget för denna siffra finns i Lars Lundqvists doktorsavhandling publicerad 1989. 3 Min gallring är vetenskapligt beprövad och bevisligen framgångsrik Skogsstyrelsen har sammanställt resultaten i ett mycket stort fältförsök med Naturkultur anlagt 1990 i tallskog nära polcirkeln på 375 möh (Rapport 23, 2008). Den visar att den plockhuggna (höggallrade) parcellen gav 85 % av volymtillväxten i orörd parcell, redan inom 5-12 år efter gallringen. Detta uppnåddes efter att ca 50 % av volymen skördats i gallringen. Under motsvarande period producerade den kalhuggna och planterade parcellen 0 % av tillväxten i orörd skog. 4 Min gallring är vetenskapligt beprövad och bevisligen framgångsrik I ett fältförsök, anlagt av SLU 1990 och reviderat tio år senare, blev volymproduktionen i en höggallrad parcell 93 % av volymproduktionen i orörd skog redan inom den första tioårsperioden efter gallring (se figur 1). Skälet till denna snabba reaktion var förmodligen att granskogen tidigare blivit vitaliserad genom plockhuggning (Chrimes 2004). Resultatet är troligen överförbart på min skog eftersom försöket ligger på samma höjd över havet i Jämtland. Referenser Anon 2008: Naturkultur. Skogsstyrelsen, Rapport 23 2008 Chrimes, D. (2004) Stand development in partially harvested uneven-aged Picea abies forests in boreal Sweden. Paper 2 in: Chrimes, D. Stand development and regeneration

6 dynamics of managed uneven-aged Picea abies forests in Boreal Sweden. Silvestria, ISSN 1401-6230, ISBN 91-576-6538-9.304, 1-9. Elfving, B. (1993) Volymtillväxtfunktioner för tall och gran, avsedda att belysa begreppet produktionsslutenhet. Skogsstyrelsen, Stencil nr 598/01009.1-10. Elfving B. 2005 Kalhyggesfritt skogsbruk- hur fungerar det. Stencil, Skogsskötsel, SLU Hagner, M., Holm, S. (2003) Effects of standing volume, harvest intensity, and stand structure on volume increment in plots managed with single tree selection over long time. Swedish University of Agricultural Sciences, Dept of Silviculture, Working Paper.187, 1-16. Lundqvist, L. (1989) Blädning i granskog. Strukturförändringar, volymtillväxt, inväxning och föryngring på försöksytor skötta med stamvis blädning. Sveriges Lantbruksuniversitet, Skogsskötsel, Avhandling, ISBN 91-576-3837-3.1-105. Bilagor 1 Rapport 2 2003 UBICON (tillgänglig i sin helhet som elektronisk resurs på SLU Bibliotek) 2 Rapport 21 2005 UBICON (tillgänglig i sin helhet som elektronisk resurs på SLU Bibliotek) 3 Mats Hagners replik på Elfvings och Lundqvists debattartikel som skogsstyrelsen nyligen lämnat som bevis. 4 Rapport 5 2008 UBICON (tillgänglig i sin helhet som elektronisk resurs på SLU Bibliotek) 5 SLU bröt mot grundlagen

7 Bilaga 1 Effects of standing volume, harvest intensity and stand structure on volume increment in plots managed with long-term, single-tree selection. Mats Hagner and Sören Holm, 2003-08-10 5 4 Annual Volume Increment, m3 / ha 3 2 1 0 0 50 100 150 200 250 Standing volume, m3 / ha UBICON Rapport 2, 2003 ISSN 1654-4455 ----------------------------------------------------------------------------------------------------------------- UBICON, Blåbärsvägen 19, 903 39 Umeå. Tel 090-141620. Epost mats.hagner@telia.com Organisationsnummer: 340827-8210. Postgiro: 130 75 17-1 ----------------------------------------------------------------------------------------------------------------- Publicerad även som: SLU, Institutionen Skogsskötsel, Arbetsrapport 187

8 Abstract Hagner, M. & Holm, S. 2003. Effects of standing volume, harvest intensity and stand structure on volume increment in plots managed with long-term, single-tree selection. Eleven plots in central Sweden were selectively cut 2-7 times in periods spanning 20-63 years, and in cutting intervals of 5-13 years. The average standing volume was 147 m 3 ha -1 (range, 41 to 287 m 3 ha -1 ). Norway spruce (Picea abies (L.) Karst.) dominated, and in some plots pine (Pinus sylvestris L.) and birch (Betula pendula Roth. + Betula pubescens Ehrh.) were also present. Vaccinium myrtillus L. dominated the field layer. Multiple regression analyses revealed that annual volume increment decreased with increasing harvest intensity, and with increasing standing volume. In plots where thinning was carried out in such a manner that a multi-storied structure was preserved, the volume increment was stable over time. In other plots, treated in such a way that the structure became more single-storied, the productivity decreased over time. The results were not consistent with those of other studies, probably because the data relate to plots that varied in standing volume over time, while previous studies compared productivity in plots with differences in standing volume. Hence this is a temporal comparison while most researchers have examined spatial differences. The regression functions suggest that substantial gains in volume production can be made if low standing volume is combined with frequent thinnings of low intensity. It also indicates the importance of preserving a multi-storied structure. Key words: polycyclic, uneven-aged, continuous cover, tree selection, stand structure, productivity, standing volume, frequent thinnings, multiple, regression. Authors address. SLU, 901 83 Umeå, Sweden. E-mail Mats.Hagner@telia.com and Soren.Holm@resgeom.slu.se ---------------------------------------------------------------------

9 1 Introduction A general goal of forest management is to optimise potential benefits in a wide sense, including economic returns, biodiversity and multiple uses. The Swedish government has expressed two major goals for Swedish forestry: one environmental and the other related to productivity. The Forest Act defines the second goal as being to keep the standing volume higher than a function, correlating standing volume and the average height of the trees, derived from a study by Elfving 1993, in which he analysed several thousand temporary plots enumerated by the National Forest Inventory. In a multiple regression, including site index as one of the variables, he found a statistically significant and positive correlation between standing volume and volume increment. Based on the same material, Fridman (1995) found a very similar, positive relationship indicating that high volume increment was maintained in the Swedish forests when standing volume was kept high. Hagner et al. (2001) proposed that an effective principle for forest management would be to maximize the present net value for every group of trees that compete for the same resources. The authors also presented a computer-aided model for choosing trees to be harvested. In practise, application of this principle led to stands with uneven-sized structure and low standing volume. Hence, there was a risk that forest owners adopting the principle would violate the rule stated by the Swedish Forest Act. In Sweden, even-aged forest management has been generally practiced since 1950 and, hence, the results presented by Elfving and Fridman were valid for forests managed under such a system. Accordingly, it was of great interest to study the volume increment and its relation to standing volume in plots with uneven-aged structure. Therefore, eleven plots with such a structure were studied in detail. They had been subjected to single-tree selection over very long periods: 20-60 years, and monitored by the Swedish University of Agricultural Sciences. The results were presented by Lundqvist (1989), who stated that the relation between volume increment and standing volume was positive and that The relative annual volume increment is approximately 3 %. (of standing volume). Andreassen (1994) also found a positive correlation between standing volume and volume increment in plots managed with a singletree selection system over long periods. Further corroboration was supplied by Lähde et al. (2002) and Lundqvist (1994), who also reported positive correlations for plots with unevenaged structure. However, since none of the studies cited above had considered thinning intensity, we believed there was an obvious risk that the low volume increment found in stands with low standing volume could be due to the low leaf area index that follows an intense thinning (O Hara et al 1999), rather than from low standing volume per se. We also believed that site fertility could influence the results, as standing volume is kept higher in plots on fertile sites. Variations in fertility might continue to have an impact even if site index is introduced as a covariate in multiple regression analyses, since site index is known to be an imprecise tool that leaves a large portion of the variance in fertility unexplained. The study presented here is based on the material mentioned above in the eleven Swedish plots subjected to single-tree selection and monitored over long periods. The statistical analyses extend evaluations of earlier studies, and the following hypothesis was tested: Volume increment increases with standing volume.

10 2 Material Eleven experimental plots in Sweden, averaging 0.63 ha in area (range, 0.25-1.0), at latitudes between 61 0 and 64 ºN, were subjected to single-tree selection over long periods: 20-63 years (Anon.1974). All plots were dominated by Norway spruce (Picea abies (L.) Karst.), while other species present included Scots pine (Pinus silvestris L.), silver birch (Betula pendula Roth) and white birch (Betula pubescens Ehrh). The ground vegetation was dominated by Vaccinium myrtillus L. Additional information is given in Table 1. Volume and increment refers only to trees with a diameter (dbh) of 8.5 cm or more. Data on annual increment were published by Lundqvist (1989 b), but unfortunately the results were presented simply as points in diagrams. As the exact figures were lost we digitised the 58 values from Fig. 1. We have been able to compare our data with 42 figures collected from other sources, and they correlated extremely well (r = 0.99). Hence, we are confident that the statistical results presented here would also be valid for the original material. For plot S4 the last cutting and the following period of growth were excluded, as there were clearly errors in the data. Some extremely small harvests at the S5, S6, V2 and J sites were not separately considered, but in each case the harvested volume was added to the following harvest. As Lundqvist did not present any values for mortality at each harvest within the plots, the loss of wood from mortality was not included in the following estimations. The volume lost from mortality was on average 10.4 % (Table 1). 3 Methods The meanings of the abbreviations used for the tested variables are given in Table 2. Standing volume, Standvol, was equivalent to the volume at the beginning of the growth period between two harvests. Annual volume increment, Volincr, was estimated from the difference between the standing volumes at the start and end of the growth period, divided by the length of the period in years. Harvest intensity, Harvint, was the volume, in m 3 ha -1, harvested before the growth period. The relative harvest intensity, Harvint%, was Harvint as a percentage of standing volume before harvest. Years from start of treatment, Yearsfst, is the number of years elapsed since the first harvest. The statistical analyses were based on the regression model y ij = β 0 j + β1x1ij + β 2x2ij +... + β p x pij + eij where yij is the value of the dependent variable for the ith observation in time of the jth plot, x for r = 1,... p are values of independent variables and the random effects associated with rij y ij, e ij, are assumed to be NID (0, σ ). 2 The intercept β for each plot, which reflects its general volume increment, was estimated by 0 j use of indicator (dummy) variables. Different sets of reasonable, independent variables were tested, as well as some interactions among the original variables. The final choice of variables was dictated by simplicity and logic rather than by the statistical residual sums of squares.

11 The model was applied not only to the entire data set obtained for all eleven plots, but also separately to the two groups of four plots mentioned below. F-tests were used to test equality between sets of β r values for the two groups. In cases of rejection, group-specific parameter values were obtained by use of group indicators multiplied by the variables in question. The assumptions about the deviations e ij were checked by inspecting the residuals, both visually and by computation. The independence assumption should be scrutinised especially carefully, since the within-plot observations constitute time series. This was checked by estimating the pooled within-plot autocorrelation coefficient and the estimated value was found to be low and even negative (around -0.25, depending on the model and data set), so it should be valid to use the formal p-values for tests. A cross-validation of the final regression model was performed. The calculations were carried out, and the presented figures were prepared, using the Minitab package (Minitab 13). The SAS package (SAS 8, Proc Mixed) was used for estimating the autocorrelation, and for more general models, with random regression coefficients. Skewness in Harvint% was eliminated by natural logarithm transformation, but in the analyses this transformation did not increase the efficiency. Hence, the original values were used. The same was also true for arcsine p transformation of the percentages. Insignificant variables were eliminated by backward, stepwise selection. Two groups of four plots were formed after a study of diameter distributions at the start and end of the treatment period (Lundqvist 1989a). Group 1 (G1) "Well" included plots in which the treatment preserved the negative exponential distribution and kept the stand multi-storied (Plots S1, V2, V3, J). Group 2 (G2) "Wrong" comprised plots in which the treatment altered the negative exponential distribution, causing the stand to be more single-storied (Plots S3, S5, S6, V1). The three remaining, intermediate plots were placed in a third group, G3. For one regression it seemed logical to combine the two groups G2 and G3 to form a larger group, G23 (see below).

12 4 Results A simple linear regression analysis of Volincr on Standvol covering all of the material revealed a statistically significant positive correlation between the two variables (Table 3). However, when the same analysis was carried out on each plot separately, the correlation was negative in nine out of eleven plots (Table 3), although only one of the eleven regression coefficients was statistically significant (p < 0.039). A second multiple regression analysis was carried out to check whether including SI as a covariate would alter the positive correlation initially found for the whole material. However, a positive correlation between Volincr and Standvol remained (Tables 4 and 5). Due to correlations between residuals, the p-values are only approximate. According to these regressions, Volincr peaked close to a Standvol of 200 m 3 ha -1. Multiple regression analyses with the whole material, and with plots as dummy variables, revealed a statistically significant negative partial correlation between Volincr and Standvol (Fig. 1 and Table 6, Regression 1). The regressions also showed that Volincr significantly declined as either Harvint% or Yearsfst increased (Table 6; Fig. 2). Interactions with groups were found (Table 6 Regression 1). The variable Yearsfst was only significant for G23, showing that volume increment was reduced over time for this group. The variable Standvol was only significant for G1, showing that volume increment declined as standing volume increased for the group of plots treated well. To make sure that a detected correlation between Harvint% and Standvol (-0.323*) did not interact with the results of the multiple regression, an analysis like Regression 1 (Table 6) was carried out without Harvint%. In this analysis, the partial correlation between Volincr and Standvol was negative, but statistically insignificant. This indicated that the two variables could not replace one another in the analyses. Regression analyses revealed that Volincr was reduced (p < 0.000) with increasing Yearfst in group G2, but for group G1 Volincr was stable over time (Table 6, Regressions 2 and 3; Fig. 3). Standvol decreased in group G1, while it increased in group G2 over time (Table 7), but these trends were not statistically significant. G1 had lower Volincr and lower Standvol values, on average, than G2 (Table 8) A sensitivity analysis on plots in group G1 revealed that different combinations of the two variables Standvol and Harvint% had a strong influence on Volincr and, thus, productivity (Table 6 Regression 1, G1). With Standvol = 50 and Harvint% = 10, Volincr was 5.34 m 3 ha -1 a -1. If Standvol was changed to 200 and Harvint% to 50 %, Volincr fell to 2.62 m 3 ha -1 a -1. Hence, high stand density combined with high harvest intensity decreased the volume production to 49 % of that obtained with low standing volume and low harvest intensity.

13 5 Discussion Cross validation was performed here, using the Minitab Press option, by comparing sum of squares of predictions (PRESS) with the sum of squares of residuals (SSRES) to avoid spurious effects of single observations. In our case, the PRESS value considerably exceeded SSRES, but this was mostly an uninteresting consequence of the high variability of the estimate for plot effects, due to the low number of observations per plot. To study the reliability of the parameter estimates of the true explanatory variables, the cross validation procedure (deleting one observation at a time) was carried out for the two groups of four plots (Regressions 2 and 3 in Table 6). Of the 23 and 22 regressions involved, respectively, the estimates of all parameters (all except for that of the insignificant Yearsfst in G1) had the same sign and about the same size. In non-experimental data, correlations between independent variables cannot be avoided. In the present case it was found that the variables Harvint% and Yearsfst are negatively correlated ( 0.36). This means that the effects of the two variables on Volincr cannot be completely separated. In addition, the indicators of each plot are correlated to the site quality index. However, the results of an F-test applied to check whether the indicators could be exchanged for site quality strongly rejected this hypothesis. Less straightforward is the interpretation of the effects of the variables Standvol and Harvint% on Volincr since a high value for either one of them will generally imply a low value for the other. However, as reported in the Results section, the two variables were found to cover different variance components. The functions reported in Table 6 were not the best found, according to the respective standard deviations, S, of their residuals. In fact, the interaction variable Yearsfst*Standvol proved to explain a considerable amount of the residual variation, giving S values of 0.668, 0.610 and 0.603 for the regressions 1, 2 and 3 (Table 6) if inserted. However, we were unable to detect any causal link explaining this correlation. An attempt to exchange Yearsfst in the product here for the correlated variable Harvint% did not improve the S values as compared to those listed in Table 6. In the traditional regression model, the effects of the independent variables are expressed by constant parameters. The effects probably vary between plots, which would generate a more realistic model with random parameters, with a fixed common population mean. An attempt to test such a complete model (using the SAS package) failed, likely due to the limited number of observations. However, when testing random parameters, one at a time, the general results of Table 6 seemed to hold for such models too, even if single estimated values differed from those derived using the traditional model. Many scientists have carried out spatial studies in which productivity and structure have been compared in a large number of plots that were observed on a single occasion. A problem with such studies is that it is difficult to extract information on a number of potentially important interactions from such data. Stand density in natural forests is higher on fertile ground than on infertile sites, and lower in the far north and close to the tundra (irrespective of fertility) than further south. These interactions also occur in managed forests. Elfving (1993) and Fridman (1995) used thousands of temporary plots measured by the Swedish National Forest Survey, in attempts to identify stand features that had an influence on Volincr. They added SI and

14 latitude as covariates, and still found a positive correlation with Standvol. When SI was introduced in the data studied here (Table 5), the positive correlation with Standvol still remained. It was not until the true productivity was introduced as a dummy variable, that the negative correlations were discovered (Table 6). A possible reason for this is that SI, measured from ground vegetation, is imprecise, as previously found by Hägglund and Lundmark (1977). Hence, SI reflects just a fraction of the true variance in site productivity. The remainder of the variance component is probably sufficient to give a positive correlation with stand density. Site productivity is probably the common basis for the correlation between Volincr and Standvol found in superficial statistical analyses, like those summarised in Table 4 and the top row of Table 3. The statement by Lundqvist (1989 b) that The relative annual volume increment is approximately 3 % (of the standing volume), is thus statistically correct, but rather misleading. A similar drawback applies to an analysis by Andreassen (1994), who studied 16 permanent plots in Norway subjected to single-tree selection over long periods. He concluded that the Volincr was equal to 3.7 % of Standvol, but did not extract plot effects from his analyses. The volume increment was found to be strongly correlated with harvesting intensity (Table 6, Fig. 1), indicating that frequent and low intensity thinnings are best, if the aim is high productivity. The partial correlation shows that if the harvest intensity was increased from 8 % to 50 % of the volume, the volume increment in the following period was reduced by 35 %. Removal of a large portion of the volume is equivalent to removing a large portion of the assimilating leaf area. This naturally leads to decreased productivity until leaf area is restored. Selective cutting successively reduced the volume increment in the studied material (Table 6). More narrow analyses of groups of plots (Table 6) showed that the plots managed in such a way that a multi-storied structure was preserved did not show any decline in productivity over time. In the other group, in which a build-up of standing volume was combined with a transformation of stand structure towards a single-storied pattern, the reduction of productivity over time was strong (Fig. 3). For even-aged stands there seems to be a general agreement that stands with full density that are not thinned have higher volume increments over the rotation period than more open stands. However, opinions differ about the size of the decrease in volume increment that reductions in density cause. Möller (1954) found that a 50 % reduction in density did not lower the total yield, and stands could still produce 85 % of maximum yield even after a 70 % reduction. Braastad (1975) found the total yield decreased with decreasing density, while Carbonnier (1957) and Assman (1970) reported that lowering stand density by thinning had differing impacts on young and older spruce stands. Assman describes this as being due to the tendency of growth to accelerate in young stands that have been heavily thinned, since in periods following such thinning productivity was 10 % higher in young stands than in corresponding, unthinned stands. The same degree of thinning carried out at a later stage of the rotation reduced volume increment by 20 %. According to Assman, this pattern was observed in many different thinning trials. He believed that the accelerated growth observed following thinning in young stands represented merely an earlier peak of growth, that would be followed by an earlier retardation, hence no extra volume would be gained overall. However, if the accelerated growth he referred to was due to a shift in the allocation patterns

15 of photosynthetic products, typical of small trees following high degrees of release, it could explain why the plots in this material combined high volume increment with low standing volume. O Hara et al. (1999) presented data concerning volume increment per unit sapwood area (which is strongly correlated with leaf area) for individual trees in a multi-storied spruce forest. Overstory trees showed a larger volume increment per cm 2 sapwood than understory trees. This does not support the presented hypothesis. However, low standing volume was combined with high volume increment in the group of plots treated well, G1, but not in the other plots. This supports the hypothesis as small trees were liberated in G1. Öyen and Nilsen (2002) presented results from 16 permanent plots selectively cut in mountain forests of Norway. The thinnings were heavy, with 55-81 % of the volume removed. Only a slight reduction in volume increment with increasing volume removed was noted, and no correlation was found between volume increment, standing volume or basal area. The plot with the lowest basal area after thinning, 3 m 2, was one of the four highest-producing plots over a 25- year period. These results support the theory that the volume increment tends to remain constant within a wide range of densities. Lundqvist (1994) found a positive relationship between standing volume and volume increment in three stands assessed 15 and 10 years after selective cutting. The analysis was carried out without any consideration of variations in site productivity, harvest intensity or stand structure. In addition, observations were done in small plots that were influenced by competition from surrounding trees. Trees on plots with small trees are, of course, generally exposed to competition from bigger trees outside the plot, and vice versa. Lundqvist (1989 b) refers to Barth (1929) and Böhmer (1957) and states that they also found positive correlations between volume increment and standing volume. However, these authors do not appear to us to have found any positive relation. On the contrary, Barth (1929) clearly expresses, in his summary, that the standing volume could be heavily reduced without loss of productivity. Lähde et al (2002) presented a diagram showing that volume increment increases with standing volume in 23 plots subjected to single-tree selection in Finland. As site index and harvest intensity were not included in the simple linear regression, the author s inference that growth is better in plots with higher density could be wrong. As it is very difficult to finance repeated measurements on permanent plots over very long periods, few such temporal studies have been published globally, and studies describing the long-term effects of single-tree selection are especially scarce. One exception is a study of 16 Norwegian plots subjected to single-tree selection over longer periods than the plots analysed here. The Norwegian plots were laid out on sites similar to those in Sweden. Hence, the results should be comparable. Andreassen and Öyen (2002) found a positive partial correlation between basal area and volume increment when they used site index and average tree height as covariates. The R 2 value was 0.72. Their function gives similar estimates of Volincr to our function in Table 5. This shows that the Norwegian material might show very different results, if site index was replaced by plot indicators and Harvint% was introduced as a covariate. For a long time it has been considered important for the efficiency of selective management for stands to be kept open and multi-storied. A number of observations have indicated that such a structure promotes sustainability by affluent recruitment (Schütz 1989, Andreassen 1994, Lähde 1992). It is satisfying that this empirical study now confirms the validity of these hypotheses, by finding high, long-term volume productivity in plots with such structure.

16 5.1 Conclusions Analyses of volume increment in uneven-aged forests should include variance components such as actual site productivity, relative harvest intensity (percent of volume), and stand structure. Site index seems to be such an imprecise tool that the remaining variance still leads to the misleading conclusion that high standing volume is a prerequisite for high productivity. Temporal, long-term studies with repeated assessments in each plot are needed to obtain a measure of true productivity. Hence, spatio-temporal analyses of volume increment should not be based on data gathered by observing large numbers of plots on single occasions. With respect to correlation between volume increment and standing volume, these results differ from the findings of many earlier studies. The hypothesis tested must be rejected, as volume increment decreased with standing volume in plots with preserved structure. In the other plots, standing volume had no impact on productivity. These results suggest that the Swedish Forest Act should be revised, and its focus shifted away from stand density. The results show that single-tree management was most successful in terms of volume production if management concentrated on low intensity thinnings carried out in a manner promoting multi-storied structure and low standing volume. In other words, high volume production was obtained where the forest had widely spaced dominants over a large number of small trees. References Andreassen, K. 1994. Development and yield in a selection forest (Utvikling og produksjon i bledningskog). Meddelelser fra Skogforsk ISBN 82-7169-697-1.47,5: 1-37. (In Norwegian). Andreassen, K.& Höyen, B.-H. 2002. Nye tillvekstmodeller for granskog behandlet med bledningshogst. In: Öyen, B-H (red) Modellering av skogsproduksjon for ökologisk och ökonomisk forvaltning. Aktuelt, Skogforsk, NLH.02: 10-12. (In Norwegian ). Anon. 1974. Redovisning av fasta försöksytor. Skogshögskolan. Rapporter och uppsatser, 32:A, B1-B3. (In Swedish). Assman, E. 1970. The principles of forest yield study. Pergamon Press, New York. 506 p. Barth, A. 1929. Skjermforyngelsen i produktionsökonomisk belysning. Acta Forestalia Fennica.34,15, 33 p. (In Norwegian). Böhmer, J., G. 1957. Bledningsskog II. Forét Jardinée. Tidskrift for skogbruk.4: 203-247. (In Norwegian). Braastad, H. 1975. Produktionstabeller og tillvekstmodeller for gran. Meddelelser fra Det Norske Skogsförsöksvesen.31,9: 356-537. (In Norwegian). Carbonnier, C. 1957. Ett gallringsförsök i planterad granskog. Svenska Skogsvårdsföreningens Tidskrift.55: 463-476. (In Swedish). Elfving, B. 1993. Volymtillväxtfunktioner för tall och gran, avsedda att belysa begreppet produktionsslutenhet. Skogsstyrelsen, Stencil nr 598/01009: 1-10. (In Swedish). Fridman, J. 1995. Volymtillväxtprocent enligt Riksskogstaxeringen. Sveriges Lantbruksuniversitet, Institutionen för skoglig resurshushållning och geomatik, Rapport.1: 1-95. (In Swedish). Hägglund, B.& Lundmark, J.-E. 1977. Site index estimation by means of site properties, Scots pine and Norway spruce in Sweden. Studia Forestalia Suecica.138, 38 p.

Hagner, M., Lohmander, P.& Lundgren, M. 2001. Computer-aided choice of trees for felling. Forest Ecology and Management.151: 151-161. Lähde, E. 1992. Natural regeneration of all sized spruce dominated stands treated by single tree selection. In Hagner, M. ed: Silvicultural Alternatives. Proceedings from an inter- Nordic workshop June 22.25 1992. Swedish University of Agriculture Sciences, Dept of Silviculture, Reports.35: 117-123. Lähde, E., Laiho, O., Norokorpi, Y.& Saksa, T. 2002. Development of Norway spruce dominated stands after single-tree selection and low thinning. Canadian Journal of Forest Research.32: 1577-1584. Langsaeter, A. 1941. Om tynning i enaldret gran- og furuskog. Meddelelser fra Det Norske Skogsforsöksvesen.27, 8: 131-216. (In Nowegian). Lundqvist, L. 1989a. Changes in the stand structure on experimental plots managed with single-tree selection. Uppsats 1 i: Blädning i granskog. Sveriges Lantbruksuniversitet, Skogsskötsel, Avhandling, ISBN 91-576-3837-3: 1-25. -- 1989b. Volume increment on experimental plots managed with single-tree selection Paper 2 in: Blädning i granskog. Sveriges Lantbruksuniversitet, Skogsskötsel, Avhandling, ISBN 91-576-3837-3: 1-21. -- 1994. Growth and competition in partially cut sub-alpine Norway spruce forests in northern Sweden. Forest Ecology and Management.65: 115-122. Möller, C., M. 1954. The influence of thinning on volume increment. Results of investigations. In: Thinning. Problems and Practices in Denmark. State University of New York, Coll Forestry, Tech Pub.76: 5-32. Nilsen, P. 1988. Fjellskogshogst i granskog - gjenvekst og produksjon etter tidligere hogster. Norsk Institutt for Skogforskning Rapport.2/88: 1-26. (In Norwegian). O Hara, K., Lähde, E., Laiho, O., Norokorpi, Y.& Saksa, T. 1999. Leaf area and tree increment dynamics on a fertile mixed-conifer site in southern Finland. Ann. For. Sci. 56: 237-247. Öyen, B.-H.& Nilsen, P. 2002. Growth effects after mountain forest selective cutting in southeast Norway. Forestry 75, 4: 401-410. Schütz, J.-P. 1989. Der Plenterbetrieb. Fachbereich Waldbau, ETH, Zurich, 54 p.(in German). 17

18 Table 1. Data from Lundqvist 1989b. Plot name Numbe r of obs. Observati on Period Years Site index m 3 ha -1 year - 1 Accumulat ed Cut m 3 ha -1 Annual vol.incr. m 3 ha -1 Mortalit y % 1) S1 7 62 5.3 157 2.78 8.1 S2 2 20 6.1 58 4.37 3.1 S3 6 57 6.1 265 6.19 12.5 S4 5 57 6.1 228 5.92 19.6 S5 5 30 6.1 89 3.68 6.9 S6 6 38 6.1 232 7.59 3.3 S7 6 30 6.1 205 5.75 23.2 V1 5 63 3.6 146 4.40 16.5 V2 5 63 3.6 132 4.23 19.8 V3 6 49 3.3 144 3.52 1.4 J 5 39 4.9 301 4.09 4.4 Average 10.4 1) Site index measured from site properties (Hägglund and Lundmark 1977) 2) Accumulated mortality/accumulated cut 2) Table 2. Variables tested in multiple regressions with Volincr as dependent variable. Name Description Volincr Annual volume increment, m3 ha -1 Volincr% Volincr/Standvol in percent Harvint Harvesting intensity before the period in which Volincr was measured, m 3 ha - 1 Harvint% Harvint in percent of standing volume before harvest Yearsfst Years from start of selective cutting in the plot Standvol Standing volume at start of period in which Volincr was measured, m 3 ha -1 SI Site index, m 3 ha -1 year -1 Table 3. Simple linear regression of Volincr on Standvol at the start of the study period Plot name Number of obs. Constant = a Regr. coeff. = b p for regr.coeff. S1-J 58 2.08 +0.015 0.000 S1 7 3.63-0.021 0.462 S2 2 (5.68) (-0.013) (1.000) S3 6 10.74-0.026 0.604 S4 5 8.64-0.012 0.400 S5 5 7.79-0.663 0.222 S6 6 6.34 +0.004 0.925 S7 6 1.63 +0.414 0.414

19 V1 5 6.87-0.018 0.496 V2 5 6.66-0.022 0.262 V3 6 4.03-0.005 0.892 J 5 4.64-0.897 0.039 Table 4. Regression analysis with Volincr as dependent and Standvol as independent variables. All experimental plots taken together. Dependent Volincr Indep. Coeff Sig Const 0.262 0.802 Standvol 0.0445 0.004 Standvol 2-1.01E-04 0.046 N 58 F 11.870 0.000 R Sqr adj 0.276 Culminati on Standvol 220 Table 5. Regression analysis with Volincr as dependent and Standvol and SI as independent variables. All experimental plots taken together. Dependent Volincr Indep. Coeff Sig Const -2.87 0.032 Standvol 0.052 0.000 Standvol 2-1.33E-04 0.005 SI 0.554 0.001 N 58 F 13.6 R Sqr adj 0.398 Culminati on Standvol 194

20 Table 6. Regression analyses with all the plots (regression 1) and with two groups of plots (regressions 2 and 3): those with a negative exponential diameter distribution after treatment (Group treated well, S1, V2, V3, J), and those with a tendency to display a normal diameter distribution (Group treated wrong, S3, S5, S6, V1). For consistency, the same variables are used for both groups. For regression 1 the variable Yearsfst*G23 means Yearsfst if plot does not belong to group 1 (indicator) and 0 otherwise, and in the same way Standvol*G1 means Standvol if plot belongs to group 1 and 0 otherwise. Regression no 1 Regressions no 2 and 3 Dependent Volincr Dependent Volincr Group treated Well Indep. Coeff p Indep. Coeff p Const 4.102 0.000 Const 3.682 0.000 S2 1.590 0.012 V2 1.757 0.004 S3 3.538 0.000 V3 1.324 0.028 S4 3.561 0.000 J 2.012 0.002 S5 0.602 0.221 Harvint% -2.04E-02 0.126 S6 4.585 0.000 Yearsfst 1.5E-04 0.990 S7 1.241 0.012 Standvol -9.27E-03 0.034 V1 1.896 0.001 N 23 V2 1.649 0.002 S 0.718 V3 1.177 0.030 F 3.53 0.020 J 2.034 0.000 R sqr adj 0.408 Harvint% -3.21E-02 0.000 Group treated Wrong Yearsfst*G2 3-5.740E-02 0.000 Const 9.763 0.000 Standvol*G1-9.57E-03 0.016 S5-3.576 0.002 N 58 S6 1.073 0.009 S 0.707 V1-1.846 0.001 F 21.13 0.000 Harvint% -6.18E-02 0.001 R sqr adj 0.821 Yearsfst -6.11E-02 0.000 Standvol -7.47E-03 0.477 N 22 S 0.608 F 28.88 0.000 R sqr adj 0.888