NC-4801, 2 /_OAvail, SPPI 4.11 .. m Staff: Code: VMPR 1.41 Problem: Reprints: 01 N Computers and electronics d. _o_ d "_"_ _ E LSEV [ ER Computers andElectronics Agriculture:7 {:000)7-:4in in agriculture _ w_ Modeling Missouri forest landscape elsevier.com, Ozarks under alternative practices USD.4 Anheuser-Busch Forest ,Vatural Sercice, Resources North Central Building, compag change in the management Stephen R. Shirley'-*, Frank R. Thompson David R. Larsen b William D. Dijak" 202 iocate, Research C.'n&'ersity III", Station. of Missouri, Columbia. MO 652I 1-7260, USA School "o,i" .Vatural Resources. 203 Anheuser-Busch ,Vatural Resources Building. Unirersity of Missouri, Coh_mbia, MO 652I I- 7280. USA Abstract We used a spatially exoiicit landscape model, LANDIS, to simulate the effects of five management alternatives on a 3216 ha forest landscape in southeast Missouri, USA. We compared management alternatives among two intensities of even-aged management with ctearcutting, uneven-aged management with _oup selection harvest, a mixture of even- and uneven-aged management, and no harvesting. Anticipated disturbances by windthrow and wildfire were included in the 100-year simulations across the landscape. The uneven-aged, even-aged long rotation, and mixed harvest regimes were similar to one another in total area in each forest size class, timber volume produced and volume of wood on the forest floor. However. they varied geatly in quantity of edge habitat and in the extent of the mature forest habitat tree from edge effects. The intensive even-aged harvest reNme and the no-harvest regime produced the greatest volume of timber and the greatest volume of down wood. respectively. This model provides a quantitative flamework to simultaneously explore multiple factors that affect landscape-scale management decisions, k_ 2000 Elsevier Science B.V. All rights reserved. Keywords: Simulation; Volume yield; Patch size; Edge; Landscape _ttern; Group selection: LANDIS Timber harvest; Clearcutting; * Corresponding author. 0168-1699/00/$ - see front matter _ 2000 Elsevier Science B.V. All rights reserved. PII: S0168- 1699(00)00087-9 ._ S.R. Shirley et ,.'i. Computersand Electronicsin Agriculture27 [2000) 7-24 I. Introduction Forests are the source of many things that people value: wood products, wildlife. water, recreation, scener2,', and solitude. Management of forest ecosystems t'or these and other values requires an understanding of how forests change over time in response to succession, natural disturbances, and management practices such as timber harvest. Inherent in virtually all forest management plans is the underlying requirement of sustainabdity -- that forests will be managed in a manner so that their future capacity to produce products, services, and amenities will not be diminished. However. many alternative sustainable management programs can be selected for a given forest. Some emphasize forest products, others emphasize dispersed recreation opportunities, and still others emphasize biodiversity. Every t'orest management activity, even the decision to do nothing, sets in motion patterns of forest vegetation change that persist for decades or even centuries. Sustainable forest management demands that managers learn as much as possible about the long term effects of management alternatives. This requires understanding how management activities and natural disturbances affect landscapes over time and space. Forest vegetation will change due to harvest and natural disturbances. and subsequent forest growth and succession will alter the size structure and species composition across a landscape. Ultimately. forest managers should be concerned with the short- and long-term effects of prescribed management activities and natural disturbances on the products and amenities that people vaiue. Methods to regulate and optimize the flow of forest products from a given area while still ensuring long-term product sustainability have been studied at length. However. optimizing forest values such as species diversity, forest aesthetics, or wildlife habitat quality is more difficult, particularly when the values are difficult to quantify numerically. The problem is further compounded by differences in relevant spatial scales for different forest resources. Management and inventor3, of forest products is usually implemented in individual forest stands, typically 2-50 ha in extent. In contrast, many wildlife species have habitats that span hundreds or thousands of hectares. For example, there are more than 180 species of neotropical migrant birds that nest in forests of the Midwestern United States each spring and summer (Probst and Thompson. 1996). These avian species have differen't habitat ' requirements and assessing habitat quality requires a landscape perspective that accounts for the spatial arrangement of forest vegetation, open lands, and edge habitats (Thompson et al., 1996). Habitat assessment for many other wildlife species is equally complicated and requires an examination of conditions at a scale larger than the individual stand. Joint assessment of timber and other forest resources requires mechanisms to integrate information about anticipated ecosystem dist(trbance processes (e.g. harvest, wind and fire) and to evaluate their impact on a range of different forest values that are often measured at different spatial and/or temporal scales. LANDIS is one landscape simulation model that has been used to simulate long-term changes in forest age structure and species composition in the Lake States (He et al., 1996; Mladenoff et al., 1996 He and Mladenoff, 1999). LANDIS can be applied to mapped t: S.R. Shirley et al. Computers and Electronics in .4_rfc'.dture 27 (2000_ 7-'_-1 9 forest !andscapes from a few hundred ha to over 100000 ha in extent. Simulations of future forest conditions are spatially explicit and maps of simulated landscape conditions can be used in conjunction with a geographical information system _GIS) to analyze forest vegetation patterns at large spatial scales. Simulation of changes in forest vegetation across a mapped forest landscape provides a critical link that allows the assessment of a wide array of associated forest characteristics over time (e.g. timber, wildlife and aesthetic qualities). We use the LANDIS model to simulate the effects of five management reNmes on a forested landscape in the Missouri Ozarks. We briefly describe the capabilities and limitations of the LANDIS model. We also examine the assumptions and resource yield relationships that are required to simulate change in the temporal and spatial distribution of forest age classes, size classes and timber volume in the Missouri Ozarks. We then apply LANDIS tO simulate long-term change in vegetation on a 3261 ha forest landscape in the Missouri Ozarks. We compare the long-term impacts of five different management regimes on the simulated pattern of forest vegetation size and age classes. Comparisons include patch size and length of forest edge in addition to maps showing spatial arrangement of forest size/age classes. We also estimate volume of timber harwest, volume of residual standing timber, and volume of wood on the forest floor over a centu_" of simulated change. 2. Methods 2. !. The LA.VDIS model LANDIS is a spatially-explicit forest landscape model that simulates forest change over time with for without) the disturbance processes of fire, wind and harvest and their interactions (He et al., 1996; Mladenoff et al., 1996; Mladenoff and He, 1999; Gustafson et at.. 2000). LANDIS was designed to simulate forest change in 10-year increments over large landscapes and to track the spatial arrangement of the resulting forest conditions. LANDIS was originally developed for species and forest conditions in the Lake States. but recently it has been adapted for Missouri Ozark forests by recalibrating the species reproduction probabilities for species and ecological land types four_d in the region tShirley et al., 2000). Internally, LANDIS represents the forest landscape as a matrix of sites or cells, each corresponding to one square raster unit or pixel on a map. We used sites that were 30 x 30 m (0.09 ha) in size, but sites can be scaled as necessary, for different simulation objectives. Forest vegetation on each site is represented as the presence/ absence of trees by species at 10-year age class intervals. Information about tree species and age classes can be used to estimate species dominance for a given sit_. Data from multiple sites can be combined to estimate age structure or forest vegetation composition for a goup of sites (e.g. for a stand). An additional map layer required by LANDIS defines ecological land types or eco-regions within which environmental conditions that affect species establishment and historical fire distributions are expected to be similar. The size. frequency, and intensity of simulated - 10 S.R. Shirley et al., Computers and Electronics in Agriculture 27 (2000) 7-24 fire events can also be varied by land types. Fire damage to individual sites varies with :he fire tolerance of the species present on a site and with the relative fuel loads {a function of time since last fire and any recent wind damage) (He and Mladenoff, 1999). We based simulated fire size and frequency on wildfire patterns recorded over the last three decades by the Mark Twain National Forest and the Missouri Department of Conservation (Westin, 1992). Data on wind disturbance in the Missouri Ozarks came from 96 km of line transects in the Missouri Ozarks sampled by Man Rebertus in 1995-96 (personal communication). Other map layers define forest stands and _oups of stands called management areas. These are used to simulate timber harvest in LANDIS (Gustafson et al., 2000). Timber harvest practices can differ tbr each management area. Simulated harvests within a given management area can be restricted to stand boundaries or may be allowed to spread to adjacent stands until they cover a specified area. All common forest harvest practices can be simulated. Simulated disturbance events kill or harvest species in one or more age cohorts. Recruitment of new trees is based on spatially explicit seed dispersal and seedling establishment defined by reproduction probabilities that are specific to an ecologcal land type. Reproduction probabilities consider species' longevity, shade tolerance, seed dispersal distances, and sprouting probability. In the absence of disturbance for a given simulation cycle, the vegetation present on a site moves to the next older ¢10-year) age cohort, and shade tolerant species may be recruited into the youngest age class. .At any given decade during a simulation, the species and age cohorts can be examined on individual sites, vegetation characteristics for individual sites can be a___re__ated and summarized (e.g. by stand, by species or by age classt, and the spatial distribution of vegetation characteristics can be analyzed. Ultimately, the condition and spatial arrangement of vegetation conditions on sites across the forest landscape can be used to estimate levels of forest products, amenities, or habitat characteristics of interest. LANDIS is not suitable for site specific planning; rather it is a tool to get a landscape-scale view of simulated future forest conditions (Mladenoff and He. 1999) 2.2. Stud)" area "3 The forest landscape we examined in this study is a ,_16 ha portion of the Mark Twain National Forest in northern Oregon county, southeast Missouri (Fig. 1). Missouri Hig2away 19 splits the landscape in two sections, and this linear feature is conspicuous by the absence of forest cover. This forested re,on was heavily logged between 1890 and 1920. The second-growth forests are a mixture of white oak (Quercus alba), post oak (Q. stelatta), black oak (Q. _'elutina), scarlet oak (Q. coccinea), hickory. (Carya spp.), and shortleaf pine (P_us echinata). Slopes typically range from 0 to 33 percent; half the sites had a slopes < 11 percent. Site quality is relatively low (site index _ 19 m at age 50) with the better sites found on the northeast slopes and in stream bottoms. Ecolo_cal land types, stand maps, and vegetation characteristics for each stand were provided by the Mark Twain National Forest. S.R. Shirley et al. ,' Computers and Electronics in Agriculture 27 (2000) 7-24 11 Z3. Initial conditions We assigned initial vegetation conditions for each site on the landscape based on the a__e class and forest type recorded during the most recent forest inventory. We grouped Miller's (1981) ecological landtypes (recorded for each site) into seven broader ecolog'ical landtypes for use with LANDIS: south and west slopes, north and east slopes, ridge tops or upland flats, upland waterways, floodplains or low terraces, side slopes on limestone, or _ades. We populated each 30 x 30 m space on the initial distal landscape with one of four species groups (white oak group, black oak group, shortleaf pine group, or maple group). We based initial species abundance by landtype on proportions observed in data collected for the Missouri Ozark Forest Ecosystem Project (Brookshire and Shirley, 1997). 2.4. Har_'est regimes We evaluated five harvest regimes (Table 1). They varied in the type of harvest, the area harvested each decade, and the distribution of harvest sites. The even-aged intensive har_'est regime clearcut _ t0% of the area each decade (on average a 100-year rotation) with the oldest stands harvested first. The even-aged long rotation harvest regime reduced the clearcut area to 5% per decade (on average a 200-year rotation) and extended the minimum rotation age to 80 years. The uneven-aged mana__ement regime har',ested 5_; of the area per decade with group openings averaging 0.2 ha in size• The mixed harvest regime included a combination of ciearcutting and group selection was on 5% of the area each decade. In the mixed 10_ I 20011 3000 N rodE slopes 4000 _ : Fig. 1. Ecological landscape types across the 326t-ha study region Simulated vegetation response to disturbance differs by landtype. . on the Mark Twain National Forest. 12 S.R. Shirley et ai. Computers and Electronics Z Z ".3 -:Z :S L_ Z ZZ_,_ _ _.: -._ _ i __ _,"" _ e. _ _ g 27 (2000) _ _-- 2 _. in Agriculture _._ _. °" ._e _. ._.° Z -e ._ '_ - ._ ._ _-_ _"-_. "_ o.. o _-_ _--_ _ .= "__ -'d _ _ _._ _ _ ._=._ .,,, 7-24 S.R. Shirley et aL /Computers and Electronics in Agriculture 27 (2000) 7-24 13 re,me, clearcutting was applied to forests on north and east slopes: group selection ha_ests were applied to forests on south and west slopes, ridge tops, and floodplains (Table 1). Other land units were not harvested. For all simulations we set 300 years as the mean period between repeat wildfires at a given site. We set 800 years as the mean period between repeat occurrences of intensive wind damage at a Wen site. The simulated wind and fire events were only those of sufficient size and intensity to kill overstory trees in an area > 0.09 ha. Blowdowns of individual trees were not included in the simulation. 2.5. Characteri:ation of landscape conditions The LANDIS model reports the 10-year age class for each site on the landscape at each decade of simulation. Simulations included up to 17 different age classes (age 0-170). To improve clarity of figures and tables and to aid interpretation of results, _e summarized age class data into the following forest size classes: seedling (age 0-9). sapling (age i0-29), pole (age 30-59), and sawlog (age >60). Reported values t'or the old-growth size class correspond to that portion of the sawlog size ciass, that is > [00 years Did. We used a variety of techniques to summarize the forest landscape conditions for each har_,est regime at each decade of the simulation. We computed age and size c!ass frequencies with commercially available GIS software. We computed patch size and edge statistics using FRAGS-rATS (McGarigai and Marks. 1995). We defined patch boundaries and edges by size classes (seedling. pole. sapling, and sawlog). We defined sawlog core area as the total area in the sawiog size class that was > 100m from an edge with any other size class. We estimated volume yields by age class for each site based on average yields by age class observed during the 1989 inventory. of the Mark Twain National Forest (Kingsley and Law. 1991; Hansen et al., 1992). We estimated volume of down wood by age class as a function of time since last disturbance based on equations and data from Spetich et al. (1999), Jenkins and Parker (1997) (Table 2). 3. Results 3. I. Vegetation age structure The forest size class distributions that result from the five harvest regimes produce notably different spatial patterns (Fig. 2). The even-aged intensive and long-rotation harvest re_mes resulted in relatively large patches of forest in a : variety of size classes. The mixed and uneven-aged harvest regimes show the small" patch structure associated with numerous 0. 1-0.3 ha _oup selection regeneration openings. The no harvest management alternative produced large expanses of mature forest (predominantly > 100 years old) with small patches of younger forest that result from fire and wind disturbance. [4 S.R. Shirley ¢t aL / Computers and Electronics in Agriculture 27 (2000) 7-24 Table 2 Estimated volume of timber and coarse wood debris by age class on the Mark Twain National Age class Growing 1-10 11-20 207 1350 9 32 79 64 21-30 31--,0 41-50 51--60 61-70 71--80 81-9O 91-100 101-110 111-120 121-130 131-140 141-150 151-160 161-170 171-180 181-190 > 190 2984 4585 6174 7323 8122 8630 8927 9087 9168 9206 9223 9230 9232 9233 9234 9234 99_34 9234 50 61 67 70 72_ 72 _ 73 73 73 73 73 73 73 73 73 73 73 52 42 34 29 26 25 26 29 33 39 46 55 65 75 87 100 113 127 .-MIages 7486 65 57 Timber volumes stock Growing Bd.ft'ha) are the smoothed average for 331 plots Forest as part of the 1989 inventory. (Kingsley and Law. from Spetich et al. (1999) and Jenkins and Parker (1997). , stock (mZ:ha_ measured Down wood on the Mark 1991). Down wood volumes Forest _ (rn_:ha) Twain National were estimated The number of sites by size class is similar for the even-aged long rotation harvest regime, the uneven-aged harvest regime, and the mixed harvest regime (Fig. 3). Each of these three regimes harvest _ 5% of the forest area each decade, but the distribution of that harvest across the landscape results in conspicuously different spatial arrangements on the landscape (Fig. 2). The no harvest regime and the even-aged intensive harvest regimes are characterized by the presence (95% of area) and absence (1% of area) of old-growth forest, respectively. The even-aged intensive harvest re mme maintains relatively constant proportions among size classes after 60 years of simulation, but for the other harvest regimes the proportion of area by size class changes continuously over the 100"years of simulation (Fig. 3). 3.2. Landscape patterns The forest size class patterns on the landscape can _e characterized by the patch sizes (Table 3) and length of edge (Table 4). These statistics change at every decade in the simulation, but the values for the final landscapes after 100 years of simulation highlight the relatively small size of the sawlog patches associated with the two even-aged harvest regimes and the relatively small patch sizes for seedling, ,R S ::qe:' e: ._:'. Computers .'r._ "..-.'ec:r'Jn:c: :.: 4__-..:,,:are 2" t2000) 7-24 !.5 sapiing ar.d -tie size classes that resu2 from "he c:?-.er three re_mes. The sawlog core are_ ,_<a'._ic,_area > t00 m from -,_'_ od-,e: _C;acen[ size class) is greatest for the _.o?.ar'.e_<- -e_irr.e. The sa_dcg core _-e.- :s :.-iv half as large under the e_en-zze_: ..c_ ::.._Uor., harvest :egi.-..e. _,nd :: 2::7<_ _.-.c_her 50% to 332 ha for the Initial Even-aged Even-aged, long rotation Uneven-aged SExed. , No harvest 0 1 Seedling Sapling intensive 2__00 1 1 Fob Sawl.og 50<)0meters 1 Fig. 2. Seat:ai pa_.:e."n, or" irutmt tbrest size classes and ?a::e._.,s simulation !'c: :a,:h ::" :he _'.e harvest re mmes. Associated landscape Old growth . :hat result following 100 y_am of statistics are found in Tables 3 and 4. size classes '._ere su.."n,mar:zed from age for each site: seed',!.-._, iage ')-9). 30-59). sa'._iog ,.,.__,e._,)_,a,a!. old growth (age > 100t. sapling (age 10-29), !:)o1¢(age 16 S.R. S,'_tT"t'v_': _.. tOO C_mputers Even-aged and Electrontcs intensive ................ 20 n '. .......... _-O 60 80 long 8040 20 100 0 20 Year :f simulation 40 60 80 100 Year of simulation Uneven-aged Mixed ;00- 27 (2000) 7-24 Even-aged, 100- 80 -_ ,...,...................-.....-.........%-...°..., a. 40 """"" """""""'"'"'"" 20 in Agriculture 80 - 80 - 60 _ 6o- o. 40 a. 40- 20 20 - 0 O O 23 -'3 _0 80 100 0 Year :f simulation 20 40 60 100 Year of simulation No harvest too [] Seealing [] Saming o. 40 [] Pole 20 • Sawlog • Old growth 80 = 0 6O oo 2_ -to 60 80 100 Year 3f simulation , Fig. 3. Change in ?r,__?<_r::,::_ :;" -he forest size classes o_er t:me t'or ,_,le har'_'es: re_mes: even,aged intensive, even-aged iong :_,ca::cn. mixed, uneven-aged and no harvest. See Fig. 2 and Table 1 for additional details about har'.e__: regimes and spatial arrangement of size classes across the landscape. Size classes were summarized :'-ore age for each site: seedling _age 0-gL sapling lage 10-29), pole (age 30-59). sawtog tage 6 _-_, _,J -,rowth (age > |00). mixed harvest regime. The uneven-aged harvest regimes, bv virtue of many large and widely dispersed h_rvest patches, results in the smallest sawlog core area. The uneven-aged har_est regime produced the greatest length of edge between size classes (Table 41. As expected, the no harvest management regime produced the least edge. The total length of edge for each of the four size classes was the most evenly balanced for the intensive even-aged harvest regime. S.R. Table Shifle.v et at. ,' Computers and Electronics 27 (20001 17 7-24 3 Mean patch size Ihal by harvest re_-._e and forest ,_( simulation Har,-est regime Even-aged Even-aged Mi.xed intensive long rotation Uneven-aged No harvest Core area size class. Table in Agriculture size class for the study area at the end of 100 years Seedling Sapling Pole Sawlog 1.2 1.6 0.3 2.2 1.8 0.3 2.3 1.6 0.3 13.7 63.7 103.9 275 663 332 0.2 0.2 0.2 0.3 0.2 0.7 103.2 389.8 179 1340 is the total area of t'orest in the sawlog size class that Core area _ is at least 100 m from any smaller 4 Length of edge lkm) Hat-,,est between regime and size class Etch-aged Seedling Sapling Pole Total edge Length of edge with size class Sapling Pole Sawlog 173 2,3 18-4. IIS "00 266 82 15 76 109 l 21 197 55 22 61 204 306 322 ,.ntens_re 3_'-4 556 2,92 Sawiog Eren-aged Seedling Sapling Pole size classes for the study area at the end of I00 years of simulation 5_4 [on_ rotatton Sawlog 207 279 288 2,27 Mb:ed Seedling Savlin_ • _ Pole Sawlog 282 "_ 4__ 2,O6 833 Uneven -aged Seedling 349 Sapling Pole 521 489 Sawlog 1117 38 30 281 53 430 406 0 0 7 16 47 ,Vo harvest Seedling Sapling Pole 9 17 47 Sawlog 70 2 18 Table S.R. Shirley et al. ' Computers in Agriculture wood by decade and harvest 27 (2000) 7-24 6 Estimated volume (thousand m3_ of down Year of simuiation Even-aged intensive Even-aged rotation 0 I0 20 30 40 50 60 70 80 90 I(30 t25 [26 131 136 138 140 Ial I39 I38 138 137 Mean 135 3.3. Harvest and residual long re,me _ Mixed Uneven-aged No harvest t25 I09 Ill 115 118 122 127 132 140 i48 158 125 it6 lla 115 116 ll9 t26 13o. 145 157 170 125 tl5 113 114 116 120 I28 138 150 164 180 125 109 102 101 101 107 119 i35 155 180 209 123 131 i33 131 * Values are totals for the 3261 ha study , and Electronics area at the end of each decade. rolurnes The estimated area harvested and volume removed was nearly constant by decade for the even-aged intensive, uneven-aged, and mixed harvest re_mes (Table 5). For the even-aged long rotation harvest re,me, no stands were harvested during the first decade because during that period no stands on the forest landscape met the minimum age requirement of 80 years (Table 1). For each of the subsequent decades the harvest for the even-aged long rotation harvest regime was _ 165 ha and 1.5 million board feet. Area and volume harvested under the even-aged intensive regime was more than double the amounts under the other harvest regimes. The remaining harvest regimes in order of decreasing area and volume harvested are even-aged long rotation, mixed, uneven-aged and no harvest. The combined cumulative harvest plus residual standing volume at the end of the simulation period followed the same order (Table 5). The estimated volume of standing (residual) timber at the end of the 100-year simulation period varied inversely with the total harvest volume and ranged from 18 to 24 million board feet per ha or 186-226 thousand m 3. Mean volume of down woody debris was similar for all harvest regimes, but values differed substantially by decade (Table 6). After moving through a period with relatively low down wood volumes in decades 2-4, the no-harvest management regime eventually produced the greatest volume of down wood. 3.4. Fire and wind disturbance In the harvested landscapes, the extent of fire and wind disturbances was less than one-third the extent of area disturbed by harvest. Wind disturbance was S.R. i Shirley et al. / Computers and E!eetronicz i ._ _._':_ _. _'_ _:'_'_ _ _: _. " Z "- "_ "" =.o 19 in Agric+alture 27 (2000) 7-24 .... -- _ -',d _ _ _ _ #'# e_l ._+ . , _: . _"1 ¢'q ..... ._ . _1 ¢_1 ¢_I ¢_# e'! ¢'1 _ _ _ _ _ _ _ _ <"1 : ,-, N_ NNN N _,,..-,_,m =. : > = y. ._ L_ ¢ _ p _ -..7_. ........ _ . -., r.+-....... mm .+,.+ _ _ _ ++-, ...... - _ >_ _ -W _ _ - _;__= = >> .... ."m".m __ _ _. ,.,_ +, _'-_ _:_ ,_ + 20 S.R. Sh(ttey et at'. Computers in .4gricutture and Electronics 27 (gO00) 7-24 neglig-ible: fire disturbed up to 50 ha per decade (Table 7). The greatest total area of fire damage occurred under the even-aged intensive management regime. That re,me produced the greatest area of the seedling and sapling size classes that were especially susceptible to fire damage. The least fire damage occurred under the no harvest re,me. The mature forests that dominate under that regime are resistant to damage from the simulated low severity fires. Wind damage, however, was greatest under the no harvest regime and least under the intensive even-aged har,est regime. 4. Discussion The images of the final forest size class distribution under the five different disturbance regimes visually illustrate the long-term consequences of the management alternatives (Fig. 2). However. it is the ability to quantitatively analyze the patterns displayed on the maps that makes it possible to link the visual display of size classes to factors that are rele,_'ant to humans and. to wildlife. For example, the uneven-aged management regime produced seven times as much edge habitat as the no harvest regime, but only 13% as much core area. The other harvest re_mes varied among these extremes and provided a range of alternative forest structures across the landscape. The best management regime for a landscape will depend on the desired future condition of the landscape and the desired products or values. For simplicity, we concentrated on the presentation of the landscapes at the end or" i00 _ears o( simulation. However. because of fire and harvest events, t?le simulated landscape conditions constantly change over the course ef a centurF. With the exception of the intensive even-aged harwest regime, which maintained a relatively constant proportion of area by size class after 60 years or"simulation, the forest area by size class changed throughout the entire simulation period (Fig. 3). We could, in tact. produce images and statistics comparable to Fig. 3 and Tables 3 and 4 for each decade of the simulation. Although it becomes cumbersome here to report all statistics by decade, t'orest managers would certainly find many of these temporal trends of interest. The uneven-aged, the long rotation even-aged, and the mixed ha_est regimes each cut _ 5% of the forest area each decade. The corresponding t'orest area by size class and decade of simulation is similar for all three of these harvest regimes (Fig. Table 7 Mean area (ha decade Harvest remme Even-aged Even-aged Mixed Uneven-aged No harvest intensive long rotation disturbed by fire. wind. Mean and harvest fire damage under Mean 50 47 36 0. I 0.2 0.2 35 27 0.3 0.4 five alternative wind damage 't harvest Mean 329 l_¢0 132 119 0 re_mes harvest S.R. Shirley et al. Computers _tnd E.:ectronics in AgricMture 27 (2000) 7-24 21 3). However. the spatial distribution of forest size classes is vastly different among those three alternatives (Fig. 2. Table 31. Assumptions about the suitability of these atternative forest size and age structures for meeting wildlife habitat needs, aesthetic goals, forest biodiversity goals, efficient scheduling or forest operations, timber production and other forest values wiI1 help further distinguish these three alternatives. The even-aged intensive harvest repine produced more than twice as much harvest volume (board ft or cubic m measure) as any other alternative (Table 5). The combined standing volume at the end of the simulation and the cumulative harvest over the simulation period was also greatest for the even-aged intensive harvest regime. At 47.2 million board feet. the total volume for the even-aged intensive harvest regime was 27% greater that the re,me with the next highest volume production and 77% higher than the standing volume under the no-harvest regime. For any given harvest regime the harvested volume varied only slightly from one decade to the next. The residual standing volume increased over the simulation period for every regime except the even-aged intensive harvest. Under that regime the standing volume on the landscape t'ell an estimated 0.8 million board feet or 12 000 m 3 until the fifth decade where the standing volume appeared to reach an equilibrium. The volume of down wood is an indicator of habitat quality that is often measured in old-growth forests. Down wood volume is usually greatest in very. old forests tdue to iarge trees that fall to the forest floor) or immediately following forest regeneration (due to lo_mn_ residue). We estimated that the greatest mean voiume of down wood over the simulation period would occur under the even-aged intensive harvest regime, but mean down wood volumes differed by no more than 5" _, among the harvest regimes (Table 6). Over the course of the simulation the no-harvest regime had the lowest and the highest volume of down wood. After reaching a low of I01000 m 3 of down wood. the no harvest reNme produced 209000 m J by the end of the simulation, more than the other harvest retries. Under the even-aged intensive harvest reNme the majority of the down wood would occur in forest openings following ha_'est. In contrast, under the no harvest reNme the down wood volume would occur almost exclusively beneath a mature forest overstory. Consequently. even though the mean volume of down wood over the simulation period is similar for all reNmes, there are substantial differences in the type of habitat it provides and the decades when it occurs. Differences in the landscape indices are indicative of differences in habitat quality for forest wildlife, such as neotropical migrant songbirds. For example, some migrant songbirds nest in mature forest while others nest in seedling and sapling stands (Thompson et al., 1996). Individual species will be the most abundant in landscapes that provide the greatest area of desirable habitat. Species that nest in seedling and sapling stands will be most abundant under the even-aged intensive/ harvest regime, and species that nest in mature forest will be most abundant under the no harvest option. Some species may be sensitive to stand size or the presence of edge; these species would do best under the no harvest or even-aged management options. Species that require habitat heterogeneity, such as ruffed grouse will be 22 S.R. Shirley e: _'l. Computers and Electronics in Agnc'.dture 27 (2000) 7-24 more abundant in the even-aged management, uneven-aged management, or mixed management options, but if they require large habitat patches even-aged management is superior (Thompson and Dessecker. 1997). In general, management options that provide high spatial diversity, such as the mixed management option, will result in the highest wildlife species richness unless the landscape becomes so heterogeneous that small patch sizes and associated edge habitats dominate the entire landscape. Fire is a relatively rare event (mean fire-free interval of 300 years), but over a century, of simulation up to one third of the landscape can be affected by fire. The pattern of simulated fire events is most visible in the landscape image for the no harvest re,me (Fig. 2): the irregularly shaped areas that do not show up as otd growth were disturbed by fire sometime during the simulation period. However. a similar pattern of fire events is also visible for the other harvest regimes. With or without timber har_'est the fire and wind disturbances add diversity to the forest size and age structure. The LANDIS model operates on age and species cohorts on a 0.09 ha site. We make a number of assumptions about the average conditions associated with a forest age class on a given site. This is a reasonable approach across the landscape, but the simulated conditions on any given site may differ greatly from what will exist on the ground in the future. Typically. estimates of forest conditions by age class have a large variance. The analyses presented here provide reasonable simulations of future landscape patterns, but the results are not suitable for site-s_ecific pianning. 5. Conclusions , The t.ANDIS model is a useful tool for exploring the long-term, large-scale consequences of forest management alternatives. LANDIS is not the only way to produce estimates of t'orest landscape change over time, but it is a method that affords considerable flexibility in exploring management alternatives. The examples we presented illustrate the complexity of long-term, large-scale forest management decisions. For individual stands, long-term changes in t'orest structure are easy to visualize. For landscapes comprised of numerous forest stands, the_ spatial and temooral patterns of forest vegetation that result from management are virtually impossible to visualize without the aid of simulation models and maps. The use of a spatially explicit model allows quantification of future landscape characteristics, including spatial and temporal distribution of forest size classes, core areas of mature forest, patch sizes, down wood volume, harvest volume, residual volume, and area of wind or fire disturbance. This model is not desimaed to predict the precise time and location of individual harvest eve_s or wildfires, but it is suitable for comparing the long-term impacts of various management alternatives on landscape patterns and species composition. Among the alternatives we analyzed, the area of edge habitat, the mature forest core area, and the patch size by forest size class all varied by a factor of four or more. The cumulative harvest volumes S.R. Shirley et al. , Computers and Electronics in .4gric'alture 27 (2000) 7-24 and standing volumes at the end of the rotation varied by 21 and 13 million board feet. respectively. This model provides a quantitative framework for simultaneously exploring multiple factors associated with management decisions. The quantitative framework and the distal maps produced by the model also provide the basis for subsequent analysis of other forest and landscape characteristics. Acknowledgements Eric Gustafson and Kevin Nimerfro, North Central Research Station. desi__ned and programmed the LANDIS harvest simulation software. David Mladenoff and Hong He. University of Wisconsin, supplied the LANDIS software and helped us learn to use it. Mike Shanta, Mark Twain National Forest. supplied GIS coverages and inventory data, we thank them all. References Brookshire, B.L. and Shirley. S.R. (Eds). 1997. Proceedings of the Missouri Ozark Forest Ecosystem Project Symposium: An Experimental Approach to Landscape Research. General Technical Report. NC-193. US Department of Agriculture. Forest Service. Nor':._ Central Forest Experiment Station. St Paul. MN. 3-8 pp. Gustat'son. E.L. Shidey, SR.. Mladeno(f. D.J.. Nimerfro. K.K.. 29(',0. Simulating timber har',est and (orest succession using t.._,NDtS. Can. J. Forest Res. 30. 32--S3. Hansen. M.H.. Fries;_yk, T.. Glo_er. J.F.. Kelly. J.F.. 1992. The eastwide forest inventor': data base: users manual. General Technical Report NC-151. USDA Forest Service. North Central Forest Experiment Station. St Paul. MN. p. 48. H.S.. Mladenoff. D.J.. I999. Spatially explicit and stochastic disturbance and succession. Ecology SO. 81-99. He. H.S.. Mladenoff. D.J.. Boeder. J.E.. 1996. t.,_NDtS: a spatially He. change, t-_,_DtS 2.0 User,s Jenkins. M.A.. Parker. G.R., simulation explicit Guide. Department of Forestry,. University I997. Changes in down deadwood volume of forest-landscape model of forest fire landscape of Wisconsin. MA. across a chronosequence of silvicultural openings in southern Indiana forests. In: Pallardy, S.G.. Cecich. R.A.. Garrett. H.G., Johnson. P.S. (Eds.). Proceedings of the t lth Central Hardwood Forest Conference. General Technical Report NC-188. USDA Forest Service. North Central Forest Experiment Station. St Paul. MN. pp. 162-t69. Kingsley. N.P., Law, J.R.. 1991.,Timber resource of the Mark Twain National Forest. Resource Bulletin NC-129. USDA Forest Service. North Central Forest Experiment Station, St Paul. MN. p. 31. McGarigal K.. Marks. B., 1995. F_,.,Gsr._,TS. spatial pattern analysis program for quantifying structure. General Technical Report PNW-351. USDA Forest Service. Pacific Northwest landscape Forest and Range Experiment Station. Miller. M.R. 1981. Ecological system Portland. OR. p. 122. land classification terrestrial subsystem, a basic inventory, planning and mana_ment on the Mark Twain National Forest. USDA Forest Region. Rolla, Missouri. 87 pp. Mladenoff. D.J., He, H., 1999. Design and behavior of L-x._ots: an object-oriented Service for Eastern /" model of forest landscape disturbance and succession. In: Mladenoff, D.J., Baker, W.L. (Eds.), Spatial Modeling of Forest Landscape Change: Approaches and Applications. Cambridge University, Cambridge, UK. Mladenoff. D.L, Host, G.E.. Boeder, J.E.. Crow. T.R., 1996. landis: A spatial model of forest landscape disturbance, succession, GIS and Environmental and management. In: Goodchild, M.F.. Steyaert. L.T.. Parks, B.O. (Eds.), Modeling: Progress and Research Ideas. GIS World Books, pp. 175-179. 24 S.R. Shirley et aL Computers and Electronics in Agriculture 27 (2000) 7-24 Probst. J.R.. _aompson, F.R.. III. 1996. A multi-scale assessment of the geographic and ecolo_cal dist_bution of midwestern neotroptcal migratory, birds. In: Thompson. F R.. II[ (Ed.). Management of midwestern landscapes for :2ae conservation of neotropical migrator., birds. 1995. General Technical Report NC-187. L'SDA Forest Service. North Central Forest Experiment Station. St Paul. MN. p. 207. Shirley. S.R.. Thompson F.R.. llI. Larsen. D.R.. Mladenoff, D.J.. Gustafson. E.J., 2000. Utilizing inventory, information to calibrate a landscape simulation model. In: Proc,--edings of Intem'ated Tools for Natural Resources Inventories in the 21st Century. 1998. General Technical Report NC-212, USDA Forest Service. North Central Forest Experiment Station. St Paul. MN. Spetich. M.A., Shirley, S.R., Parker. G.R., 2000. Regional distribution and dynamics of coarse woody debris in temperate deciduous old-_,rowth forests. Forest Science 45_2_ 302-313. Thompson. F.R.. IIL Dessecker. D.R.. 1997. Management hardwood forests. General Technical Report NC-195. Experiment Station, St Paul. MN. p. 33. of early-successional communities in central USDA Forest Service. North Central Forest Thompson. F.R.. III. Robinson. S.K.. Whitehead. D.L., Brawn. J.D.. 1996. Management of central hardwood landscapes for the conservation of migratory, birds. In: Thompson. F.R.. III (Ed.). Management of midwestem landscapes for the conser'vation of neotropical mim'atory birds. 1995. General Technical Report NC-187. USDA Forest Service. North Central Forest Expertrnent Station. St Paul, MN. p. 207. Westin. S.. 1992. Wildfire p. 161. in Missouri. Missouri Department of Conservation. f" Jefferson City. MO.