Madrono, Vol. 46, No. 4, pp. 187-198, 1999 FIELD ASSESSMENT OF THE CALIFORNIA GAP ANALYSIS PROGRAM DATABASE FOR SAN DIEGO COUNTY Y Jae Chung1 and Arthur M. Winer2 Environmental Science and Engineering Program, School of Public Health, University of California, Los Angeles, CA 90095-1772 Abstract Given the key role played by biogenic hydrocarbons (BHC's) in photochemical smog formation and atmospheric chemistry, it is critical to generate accurate BHC emission inventories. Assembling such inventories requires reliable characterization ofthe areal coverage of important plant species in order to quantify the biomass ofBHC-emitting vegetation. A recent GIS-based description ofvegetation coverage in the natural areas of San Diego County is provided by the Gap Analysis Program (GAP) database. We conducted an assessment of this database through ground-based vegetation surveys prior to using the database to develop a BHC emission inventory for southern California. Quantitative vegetation surveys were conducted along belt transects in four polygons dominated by trees and along line transects in four polygons dominated by shrubs, in order to determine percent cover of major plant species. The species listed by GAP accounted for two-thirds to three-quarters ofthe relative cover in these selected polygons. About 60% ofthe species listed by GAP were found in high enough proportions in the field surveys to justify theirlisting. Summedoverall eightpolygons, BHC emission indicesbasedon GAPdatacorrelated with BHC emission indices generated with data from our field surveys. On balance, wejudge the GAP GIS database to be a useful source of species composition and dominance information for the purpose of assembling BHC emission inventories, provided supplementary data on crown volumes are available from the literature or can be obtained in the field. The emission of reactive hydrocarbons such as composition and dominance of the plant species in isoprene and monoterpenes by vegetation (i.e., bio- an airshed, green leaf biomass for the dominant genic hydrocarbons or BHC's) has been known for plant species, and quantitative rates of emission of several decades (Went 1960; Rasmussen 1972). individual organic compounds from these species. However, only in the last fifteen years has interest Plant species distributions, BHC emission rates, in the role of BHC's in photochemical smog for- and leaf mass constants have been developed for a mation and atmospheric chemistry expanded dra- substantial number ofspecies relevant to certain ar- matically, both in the scientific andregulatory com- eas of California (Winer et al. 1983, 1992; Miller munities (Winer et. al. 1983; Lamb et al. 1987; and Winer 1984; Horie et al. 1991; Karlik and Wi- Chamedies et al. 1988; NRC 1991; Corchnoy et al ner 1998). With the proposal of a taxonomic meth- 1992; Winer et al. 1992; Arey et al. 1995; Geron odology for assigning isoprene and monoterpene et al 1995; Sharkey and Singsaas 1995; Benjamin emission rates to unmeasured plant species (Ben- et al. 1996, 1997; Guenther 1997; Benjamin and jamin et al. 1996), emission rates can in principle Winer 1998). be estimated for many of the 6000 plant species in In the atmosphere, many BHC's are as reactive California without direct experimental measure- as, or more reactive than, volatile organic com- ments. Ofthe southern California airsheds, the veg- pounds (VOC) emitted from mobile or stationary etation spatial distribution and composition has anthropogenic sources (Carter 1994; Benjamin and been established for urban and natural areas within Winer 1998), and there is a growing body of re- Orange County and the non-desert portions of Los search suggesting BHC's can constitute a signifi- Angeles, Riverside, and San Bernardino Counties cVaOntCanidnveenvteonrydoimninbaontth croengtiroinbaultiaoinrsthoedtsheaonvdertahlel e(tWianl.er19et91a;l.B1e9n8j3a;miMnilleterala.nd19W97i)n,eran1d98a4;liHmoirtieed global atmosphere (Workshop on Biogenic Hydro- BHC emission study for Santa Barbara and Ventura casasrobcoinaste[dWwBiHt]h f1u9r9t7h)e.rGrievdeuncitnhge VenOoCrmoaunsd cNosOtsx C1o9u96n)t.ieHsowheavsera,lsoa bvaeleindarteepdoritnevdent(oCrhyinokfinveegtetaal-. emissions to meet state and federal air quality stan- tion species composition and spatial distribution, bdlaerdsr,eliitabilsecrBitHicCal teomiosbstiaoinn diantvaenntoereidees,d tioncalsusdeimn-g specifically to develop a BHC emissions inventory, has not been established for the San Diego County airshed. A 1 Currently at: U.S. Army Corps ofEngineers, Los An- potential source of information concerning geles District, Regulatory Branch, P.O. Box 532711, Los vegetation in the natural areas ofSan Diego County Angeles, CA 90053-2325. is the Gap Analysis Program database (GAP), 2Authorto whom correspondence shouldbe addressed. which is coordinated by the United States Geolog- MADRONO 188 [Vol. 46 ical Service-Biological Resources Division (for- turbance regime. Landscape boundaries were sub- merly the National Biological Service) to identify jectively determined through photointerpretationby the distribution and management status of plant expert personnel so that between-polygon variation communities. GAP compiled a geographic infor- was greaterthan within-polygon variation. Thefinal mation system (GIS) database (based primarily on result was a vegetation map with a 100 hectare remote-sensing data) describing vegetation type minimum mapping unit and a 1:100,000 mapping and dominance in terms of areal coverage (Davis scale (Davis et al. 1995). et al. 1994, 1995). Unlike other vegetation maps, For each polygon in the database, one primary which describe geographic areas using only plant and one secondary vegetation assemblage was list- communities, the California GAP describes vege- ed. Each assemblage consisted of three co-domi- j tation in given geographic areas using dominant nant overstory species, each covering a minimum plant species. Because BHC emissions inventories of20% ofthe relative coverofthe assemblage. The rely on species-specific measurements of both leaf primary assemblage was defined as the assemblage mass and BHC emission rates (Benjamin et al. covering the majority of the polygon, and the sec- 1997), GAP offers the advantage ofproviding spe- ondary assemblage as covering theremainderofthe cies-specific vegetation distribution data. Moreover, polygon. Relative cover is the proportion of total the GAP GIS database is recent for southern Cali- vegetation cover occupied by a given plant species, fornia (Davis et al. 1995) and therefore more up to excluding certain vegetation such as plants below date than older vegetation databases employed in a pre-established height andbare ground. Overstory California, such as the vegetation type map (VTM) plants are those plants viewable directly from surveys conducted in the 1930's andCALVEGgen- above. In addition, GAP listed the percent crown erated in the 1970's (Sawyer and Keeler-Wolf cover of each assemblage in the polygon in four 1995). Although large-area, small-scale GIS data- classes. bofafseersabpaosteedntoianllryemionteex-pseennssiivnegandadtas,imspulcehaapsprGoAaPc,h basAec.quTihsietiGonAPanddatparbeapsaeraftoirotnheofsotuhtehwGeAstPedcaotrae-- to characterizing the distribution and species iden- gion was obtained in August 1997 from the De- tity of natural vegetation within an airshed, use of such GIS databases in BHC emissions inventory partment of Geography at the University of Cali- development requires evaluation of their accuracy fornia at Santa Barbara. The southwest ecoregion and reliability through ground-based observations. covers all or portions of Santa Barbara, Ventura, We report here the results of a ground-based as- Kern, Los Angeles, San Bernardino, Orange, and sessment of the GAP GIS database for San Diego Riverside Counties and the western two-thirds of County using vegetation surveys of representative San Diego County. The remaining eastern third of GIS polygons. The surveys employed a modified San Diego County is located in the Sonoran ecore- stratified random sampling approach and a survey gion, which was not obtained for use in the present protocol based in part on the recommendations of study because it is located far from the major ur- the developers of the GAP database (Stoms et al. banized centers of San Diego County, and because 1994). Data gathered from field surveys conducted the ecoregion is composed mostly of deserts with from September 1997 to April 1998 in San Diego little biomass. From the 2014 original polygons for County were used to assess the utility of the GAP the southwest ecoregion, the San Diego County subset of 437 polygons was extracted using GIS database for predicting the distribution and ArcView 3.0a GIS software. species identity of vegetation, and for providing a quantitative description of plant species assem- Vegetation survey protocol and methods. The blages. vegetation survey protocol employed was initially based on recommendations foraGAPdatabase val- Methods idation study suggesting 1 square km sample units (Stoms et al. 1994). Because the GAP database is Gap analysis program database. As noted ear- a large-area land cover map, use of a large sample lier, GAP's purpose was to identify the distribution element (e.g., 1 km2) avoids quantifying heteroge- and management status of selected components of neity below the intended resolution of the map. biodiversity. The central tool of this program was Stoms et al. (1994) noted other issues affecting an ARC/INFO GIS database with plant species and vegetation surveying such as the need to obtain le- vegetation class attributes associated with polygons gal access from private land-owners, safety, and within a defined geographic region. This database proximity of sample sites to roads. The specific was generated from summer 1990 Landsat The- shape of the vegetation survey unit was left unre- matic Mapper satellite imagery, 1990 high altitude solved in the guidelines. VTM color infrared photography, surveys based on In the present study, the specific survey protocol field surveys conducted between 1928 and 1940, chosen depended on the type of vegetation being and miscellaneous vegetation maps and ground sur- assessed. Within the polygons dominated by trees, veys (Davis et al. 1995). Polygons were delimited surveys were performed in three sample elements m m based on climate, physiography, substrate, and dis- consisting of two 6 wide, 500 long belt tran- 1999] CHUNG AND WINER: SAN DIEGO GAP ANALYSIS 189 sects bisecting each other at right angles. Other re- Selection ofpolygons from the GAP database. m searchers have demonstrated 6 wide belt tran- Polygons were chosen for potential inclusion in the sects make the mechanics of sampling easier while present study based on an index estimating their not significantly compromising accuracy (Lindsey isoprene or monoterpene emissions. Use of these 1955). For these belt transects, the surveyors indices identifiedeighty polygons (Fig. 1) estimated walked 250 m north, south, east, orwest away from to have the largest biogenic hydrocarbons emis- the centerpoint, using a magnetic compass to main- sions based on the presence of high-emitting plant tain course. species (Benjamin et al. 1996) and their areal cov- In the present study, one person measured the erage within a polygon. crown radii and diameter at breast height of trees Further selection from among the eighty poly- and the crown height of shrubs (plants with more gons with the highest BHC emissions involved an than one stem), while another measured the crown iterative process accounting for representativeness height oftrees (plants with one stem) and recorded and feasibility. In considering representativeness, the field data. Crown radii in trees were measured roughly equal numbers of polygons were selected m with a 10 tape in four directions (north, south, with woodland/forest vegetation and scrub/chapar- east, and west). Crown radii in shrubs were mea- ral vegetation, the two main classes ofnatural veg- sured using two diameters perpendicular to each etation in San Diego County. Polygons were se- other. Readings were taken to the nearest tenth of lected to insure that all geographic regions of the a meter. Crown height of trees was obtained from county, except the desert regions, were represented. a clinometer. From a distance ofapproximately 10- In considering feasibility, physical access and per- 20 meters, the observermeasuredtothe nearestme- mission to survey vegetation on private or military ter the distance from the tree to the observer using property were important. Polygons with a large an optical rangefinder. With a clinometer, the ob- public land component (e.g., California State Parks, server determined the crown height as a percentage San Diego County Parks, United States National of the observer's distance away from the tree. Forest System, Bureau of Land Management, and Within polygons dominated by scrub or chapar- local parks) were favored due to the relative ease ral, one individual performed surveys in four sam- of gaining permission to conduct surveys on such m ple elements, each consisting of two 300 long properties compared with privately-owned proper- line transects bisecting each other at right angles. ties. Line transects have been used to estimate relative The minimum square-shaped area needed to en- coverforchaparral (Bauer 1943) and for sage scrub compass a sample element within a polygon was (Kent and Coker 1992; Zippin and Vanderwier determined to be 62.5 acres for forests and wood- 1994). The individual surveyed along line transects lands and 22.5 acres for scrub and chaparral, and m using a 50 tape and collected data on the identity owners of parcels of land of these sizes within se- ofthe topmost plant species directly over the meter lected polygons were identified using information tape, was determinedthe numberof0.1 m segments from the San Diego County assessor's records. A occupied by that plant species, and recorded the letter was prepared requesting permission to con- height ofthe crown for each individual plant to the duct a vegetation survey, stating the goals of the nearest 0.1 m. The crowns were envisioned as rect- research, and enclosing a form to be returned of- m angular prisms and measured as such. The 150 fering or denying access. Out of 69 mailers, 10 transects running north, south, east, and west from owners agreed to participate in the study, 15 de- m the centerpoint were completed using three 50 clined and the rest were non-responders, for a suc- segments. cess rate of about 14%. Those polygons with a The survey team located the centerpoint ofa par- large public land component and/or with many pri- ticular sample element using a global positioning vate land owners responding positively to the mail- receiver (GPS) locked onto universal transmercator ers were included for further consideration. (UTM) coordinates gathered from the GAP data- Based on these criteria and the time and re- base. A Garmin 12XL handheld GPS unit, with an sources available for this research, eight polygons accuracy of ± 100 m 99% of the time, was em- were selected for the present study. Four polygons ployed. The survey team then recorded the species consisted primarily of woodland/forest vegetation, identity and related data as described above. For and four polygons consisted primarily of shrub/ forested polygons (areas where crowns of trees in- chaparral vegetation (Table 1). Ofthese eight poly- terlocked), only data from plants taller than waist gons, five were estimated to be dominated by iso- height were recorded. For woodland polygons (ar- prene emissions, two were estimated to be domi- eas where crowns of trees did not interlock), only nated by monoterpene emissions, and one exhibited plants taller than knee height were recorded. For both high isoprene and high monoterpene emis- scrub and chaparral, all plant species except for un- sions. Table 1 lists data foreach ofthese eight poly- derstory species and grasses were recorded. All gons according to the GAP database, including the plants were identified in the field, and samples of expected species assemblages and their relative unidentifiable plants were taken to the herbarium at proportion of the polygon and crown closure. In UCLA for identification. addition, Table 1 lists the polygon rank by the iso- MADRONO 190 [Vol. 46 Fig. 1. GAP polygons surveyed in San Diego County for plant species composition and dominance. prene or monoterpene emission index. Figure 1 Data analysis. As noted earlier, the GAP GIS shows the location of the polygons investigated in database provides semi-quantitative information on the present study. the abundance and distribution ofplant species. For each polygon, the GAP database lists species as- Selection of sample elements within a polygon. semblages and the estimated areal proportion (p) of After a polygon was chosen by the process de- that assemblage within a polygon. Each species in scribed above, sample elements were selected. The a listed assemblage is a co-dominant, providing centerpoints of the elements were located so all >20% relative cover. The expected relative cover transects were at least 100 meters away from the of a species listed in the GAP database for a pol- polygon boundary. If permission was obtained to ygon is then >0.2p. For example, in polygon Fl, access most of the polygon, sample eUleTmMents were Quercus agrifolia Nee is a co-dominant in an as- selected by overlaying a 500 meter grid on semblage that occupies 60-70% of the polygon. the polygon, assigning sequential numbers to every Using a mean value of65%, GAP predicts Q. agri- grid element within 1 km of a road, and randomly folia would cover > 13% of the polygon. selecting the needed number of 500 meter grid el- The relative cover ofplant species inferred from ements. This method was similar to the one em- the GAP database by this procedure was compared ployed in the Utah GAP validation project (Ed- with the cover data gathered from the field surveys wards et al. 1995). Only four polygons had enough in the eight selected polygons. First, the relative area accessible by roads or sufficient permission to cover of each species within each sample element be sampled. For the other four polygons, large por- ofa polygon was calculated. Then from the species tions were physically or legally inaccessible, and relative cover for each sample element, the mean sample elements were chosen from within the ac- relative cover and upper limit of the two standard cessible areas. To minimize bias in site selection error (SE) confidence interval for the polygons for these four polygons, the final selection of sam- were calculated, corresponding to an 85% confi- ple elements was decided before entry into the pol- dence interval (McClave and Dietrich 1985). ygon. In several cases, suitable survey sites were If the upper limit of the confidence interval for not available within the vicinity of a road, so hikes the relative cover of a measured species was less of up to two hours along a trail were needed to than its GAP-predicted relative cover, then the spe- reach the desired area within the polygon. cies was considered an "incorrect" listing as a co- 1999] CHUNG AND WINER: SAN DIEGO GAP ANALYSIS 191 TaaFbrloem1G.APPodlaytgaobnassef(rUonmivtehrseitGyAoPfDCaaltiafboransiea SatelSeacntteadBafrobraFriaeDldepSaurrtvmeenytooffSGpeecoigersapCohmypo1s99i7t)i.onbFan=dFAobruesntd,anWce=. Woodland, C = Chaparral, S = Sage Scrub. cIi = Isoprene emission index. dIm = Monoterpene emission index. Pri- mary or Area secon- Percentage of Crown Rank RanVk IDb (ha)a dary3 Species assemblage3 polygon3 closure3 by \{ by Fl 1990 P Quercus kelloggii, Quercus chryso- 60-70% 60-100% 10 159 lepis, dim \juetcus agrijuiia S Pinus lambertiana, Pinus coulteri, 30-40% 60-100% and Libocedrus decurrens F2 2317 P Quercus kelloggii and Pinusjeffreyi 50-60% 40-59% 29 136 s Quercus cornelius-mullerii, Cerco- 40-50% 60-100% ccirpus betuloides, and Adenosto- mafasciculatum Wl 1904 p Quercus agrifolia, Quercus engel- 60-70% 25-39% 9 170 mannii, and Quercus kelloggii c Ceanothus leucoderrnis, Adenostoma ou—1UU/c fasciculatum, and Quercus ber- beridifolia W2 1778 P Quercus kelloggii, Quercus agrifol- 60-70% 60-100% 16 249 ia, and Quercus engelmannii o rAxinA-toZiyltUniSlLfiilLl JidsiiScsL~*ifC/"*1j4j1Itniltluivfyfily /Ar/vsC^lfin/HcSltt/li- 30—40% 60—100% phylos tomentosa, and Cercocar- pus betuloides CI 6578 P Quercus berberidifolia and Ceano- 50-60% 60-100% 7 37 thus leucoderrnis Q rxtlt,rlUblt/tflLlJLIJSLIL,HILIIittft, LcAC(JCCir 60—100% pus betuloides, and Ceanothus sp. C2 3986 p Adenostomafasciculatum, Quercus 80-90% 60-10% 20 56 berberidifolia and Ceanothus leu- coderrnis s Ceanothus greggii and Arctostaphy- 10-20% 60-100% lospungens SI 3650 p Artemisia californica, Eriogonum 60-70% 40-59% 102 8 fasciculatum, and Salvia apiana s Adenostomafasciculatum, Ceano- 30-40% 40-59% thus oliganthus, and Quercus ber- beridifolia S2 2718 p Artemisia californica, Salvia melli- 80-90% 60-100% 251 2 fera, and Malosma laurina s Avena spp., Bromus spp., etc. and 10-20% 25-39% Baccharispilularis dominant in the GAP database. Otherwise, it was compared to the field data. Crown closure is equiv- considered a "correct" listing. An observed species alent to the percentcoverage by all overstory plants not listed by GAP as a co-dominant was considered within a polygon divided by the area of the poly- a "potential" co-dominant ifthe upper limit ofthe gon. A confidence interval within two SE was cal- uncertainty interval was greater than the predicted culated from these data and compared with the data relative cover value of any co-dominant in the sec- predicted from the GAP database. ondary assemblage in that polygon. For example, in polygon Fl, the smallest predicted relative cover Results is that of a co-dominant in the secondary assem- blage which provides 30-40% cover for that pol- Species composition and abundance within GAP ygon. Using amean value of35% forthe secondary polygons. Table 2 summarizes the overall results assemblage in polygon Fl and a relative cover val- fromthe field surveys, listing the six mostabundant ue of >20% for a co-dominant, then >7% of the overstory species observed for each polygon, the polygon is expected to be covered by a co-domi- percent composition predicted from the GAP data- nantofa secondary assemblage inpolygonFl. Any base, the percent composition determined by the plant species not listed by GAP but observed in field surveys, and the upper limits of a two SE in- polygon Fl with a relative cover > 7% was con- terval of the percent composition. In the forest and sidered a potential co-dominant in that polygon. wooded polygons, overstory plants accounted for Crown closure from the GAP database was also 87-92% of the relative crown cover according to 1 n J 91 1 MADRONO 192 [Vol. 46 Table 2. Measured Species Cover Composition ofthe Table 2. Continued. Six MostAbundant Plant Species ObservedinSelect- ed GAP Polygons. * Species listed in the GAP database Pre- as a co-dominant, but not ranked in the top 6 species dicted Sampled observed for the polygon. Poly- cover cover (s + 2 gon Species (%) (%) (s) SE) Pre- dicted Sampled \/f/j//->rill/I //;i/r/Vl/ A Q Pgoolyn-Quercus cShrpseycoileespis c(o>—%viI)ed^r (cZ%oj)ve(rs) (sS7E+f>)2 RT**ohCCtueeasaalnnOoooVfttCLhhstuuCiLssx hgliregeuhgcegosidtiermis >——11j7/ OJ4J1I 1// P1l^i1S1nI1A/u>cisf~tC•ljjtjeAtf"fKAr-t/e.jl/y1li/(\Jtg)gi)lil/ —>11-2 11QQ ZL7 EGrAiPogCoon-udmofmaisncainctuslatum —1J ZD /IQ AC^rucptyo-csitiakpnhoyvliofnslip/uingens >\3 I7\ 2147 A/fA\rC/tlJ/ceJfImTKSJi.\syVlitl/aJMIillcl/Ha1lJTi/C~fltjoCrInfCilLicltal/CiltLlU1lWl1l >>—17T/ zu 4J^jo1 Pinus coulteri >7 6 18 \/f/I//)1'1VI/1 //(11»~/11/"1 11A4 zo *Calocedrus decurrens >7 4 7 Xylococcus bicolov O 11A4 *Pinus lambertinana >1 0.0 Ceanothus oligcinthus >"7 j1 11oU Total of six highest 85 *Quercus berberidifolia i1 5 GAP Co-dominants 64 Salvia apiana >13 1 ^tiC/Ciij iKClUJggll >\\ j^t4 ^7 Total ofsix highest 90 Pinuv ieffrevi >H 18 39 GAP Co-dominants 72 lC^^lh*pCy/ttiMiJk (h/pCr.h/t(-/{f_y/riIrLll1i1fInsliiIti 13 23 Artemisia californica >17 41 68 i -pnvintViu^ TisilvYipyi 26 Salvia mellifera >17 16 31 Cercocarpus betuloides >9 7 14 Malosoma laurina >17 14 24 Adenostomafasciculatum >9 5 12 Rhus integrifolia 6 15 *Quercus cornelius-mulleri >9 0.0 Baccharispilularis >3 4 12 Total ofsix highest 88 Malacothamnusfascicula- 4 10 vJrVI V_^U LlUlllllldlllo UJ tus Quevcus engelmannii >13 39 56 Total of six highest 85 Quercus agrifolia >13 26 41 GAP Co-dominants 75 Arctostaphylos gladulosa 10 27 Adenostomafasciculatum >1 8 25 lJUIVItl tipitltltl 4 1 L/*-i*l-/H/J)^(L)f/rililUufflyliJfC/i*ij/C'iIt/,'iljAIl/tiltliUJfii'ln 4 1 our data. In the polygons dominated by scrub and **QCueeanroctuhsuksellleougcgoiidermis >>173 02.2 32 chaparral, there was little or no understory. Quercus berberidifolia >7 0.0 Most of the relative cover was attributable to a TVfQJ.JoruV/-t'eAlaLlCrPlIXcPVoKuJnfslV-'qJeHVil.nn1\AVg7m1hle1ili1lnJlc1mgaIlltanilllnpWt1^dn1ct.li>i >13 976126 121 fdinaeanwnttsss)ppeecwcieieresse. (rFmeosarpnoyanlsloifpboltelhyefgomornlsoi,svteetrdhea8s0siG%xAmoPfostcthoe-adrboeulmna--- f^iipmiv Ifpllnooii >\3 17 46 tive cover (Table 2). In general, the GAP co-dom- Quercus berberidifolia 1u 9Z7/ inants providedroughly two-thirds tothree-quarters tyi^l/tU/C)lyC/U1;Cyit' Cj1i1g1f1"I1J{1U1l1l1C/l >i^ j o of the relative cover observed in the field surveys. 1 /A/ifA L-UUlLCll z. f. For polygons Fl, F2, Wl, W2, CI, C2, SI, and S2, AA*^rrAlcc^d-ttCeP/oonYtssr(ottn/saa-lppotUnihhom/yym/illHiaooJcssftraSiggis.plltc1iaaIiAinn1lc1nddu/iuu1flLtlliaztooxt.?ss>uaam >—>>—777// u9zI1.05 4A9 7Gco8Av%eP,r,c5or9e-%sdp,oemc7it2niv%aen,ltysa.npdro7v5id%edof64th%e,o6b5s%e,rv7e6d%r,el8a5ti%v,e Totnl of C1Y Vllolivet For both forest/woodland and chaparral/scrub GAP Co-dominants 85 polygons, the observed relative cover ofcertain co- Quercus berberidifolia >11 39 67 dominants in GAP polygons often substantially ex- Adenostomafasciculatum >9 36 54 ceeded the minimum predicted values (Table 2). Eriogonumfasciculatum 5 15 For example, in polygon F2, Quereus kelloggi Quercus engelmannii 4 12 Newb. provided 34% of the relative cover when Ceanothus crassifolius 2 4 >11% was predicted, and in polygon Wl, Q. en- H**eCCteeerarcnooomcteahlruepssulsaerbubecutotuidlfeoorlimidiaess >>119 022.4 441 gceolvmearnwnihienE.>G1r3ee%newapsropvrieddeidct3ed9.%InofpothleygroenlatCiIv,e *Buckbrush >9 0.0 Q. berberidifolia Liebm. and Adenostomafascicu- Total of six highest 88 latum Hook. & Am. provided 39% and 36% ofthe GAP Co-dominants 78 relative cover, respectively, when >11% and >9% C2 >17 51 73 were predicted by the GAP database. Thus, al- Adenostomafasciculatum though a lower limit for species relative cover can Xylococcus bicolor 13 25 be inferred from the data provided by the GAP da- EQruierocguosnubmerfbaesrciidcifuollaitaum >17 58 165 tabase, an upper limit for species relative cover is not available from GAP. 1999] CHUNG AND WINER: SAN DIEGO GAP ANALYSIS 193 Table 3. Species Listed Correctly and Incorrectly as Co-Dominants within Surveyed GAP Polygon Ordered by Decreasing Mean Relative Cover, (p) GAP primary assemblage species, (s) GAP secondary assemblage species. * Potential co-dominant listed as a co-dominant in an adjacent GAP polygon, f See test. Poly- GAP species observed in GAP species not observed in gon significantly large quantitiesf significantly large quantitiest Potential co-dominants Fl Quercus chrysolepis (p) Pinus lambertiana (s) Pinus jeffreyi* Quercus kelloggii (p) Arctostaphylospungens Quercus agrifolia (p) Quercus berberidifolia* Pinus coulteri (s) Calocedrus decurrens (s) F2 Quercus kelloggii (p) Quercus cornelius-mulleri (s) Quercus berberidifolia Pinus jeffreyi (v) Ceanothus palmeri* Cercocarpus betuloides (s) Pinus coulteri Adenostomafasciculatum (s) Wl Quercus engelmannii (p) Quercus kelloggii (p) Arctostaphylos glandulosa* Quercus agrifolia (p) Ceanothus leucodermis (s) Eriogonumfasciculatum* Adenostomafasciculatum (s) Quercus berberidifolia (s) Salvia apiana* W2 Quercus engelmannii (p) Quercus agrifolia (p) Quercus berberidifolia* Quercus kelloggii (p) Adenostomafasciculatum (s) Arctostaphylos tomentosa (s) Cercocarpus betuloides (s) CI Quercus berberidifolia (p) Cercocarpus betuloides (s) Eriogonumfasciculatum* Adenostomafasciculatum (s) Ceanothus leucodermis (p) Quercus engelmannii* Ceanothus cuneatus (s) C2 Adenostomafasciculatum (p) Quercus berberidifolia (p) Xyloccocus bicolor* Ceanothus greggii (s) Ceanothus leucodermis (p) Eriogonumfasciculatum* Arctostaphylospungens (s) Malosma laurina* Rhus ovata Cneoridium dumosum Ceanothus oliganthus Arctostaphylos glandulosa* SI Eriogonumfasciculatum (p) Quercus berberidifolia (s) Malosma laurina Adenostomafasciculatum (p) Salvia apiana (p) Xylococcus bicolor* Artemisia californica (s) Salvia mellifera Ceanothus oliganthus (s) S2 Artemisia californica (p) Rhus integrifolia Salvia mellifera (p) Malacothamnusfasciculatus Malosma laurina (p) Eriogonumfasciculatum* Baccharispilularis (s) Lotus scoparius Avena spp., Bromus spp., etc. (s) For each polygon, Table 3 shows the number of enough proportions to justify their co-dominant species listed as GAP co-dominants which agreed designation, or 64% and 50%, respectively among with our field observations, the species listed by primary and secondary assemblages. GAP but not observed in significant amounts in the There were several instances where species listed field, and species that could have been listed as co- in either the primary or secondary assemblages dominants based on their observed abundance. were not observed at all in the polygon in our field For all the polygons taken together, 59% of the surveys. In some cases, a taxonomically similar GAP co-dominants were observed in the field sur- species was found instead. For example, we did not vey in large enough proportions tojustify their co- observe Quercus cornelius-mulleri K. Nixon & K. dominant designation. Of the species listed as co- Steele (a scrub oak with tomentose hairs on the dominants in the primary assemblages, this per- underside ofthe leaves) in polygon F2, but did ob- centage was 73%, whereas for species listed within serve Q. berberidifolia. In another case, the species the secondary assemblages, only45% had sufficient listed in GAP did not even exist within San Diego abundance to match the predictions. The "correct" County according to botanical experts, although a listing of GAP co-dominants was more common closely-related species was found instead in our (61%) in forested or wooded polygons, with pri- study. For example, Arctostaphylos tomentosa mary and secondary assemblage co-dominants con- (Pursh) Lindley was not recognized by Beauchamp firmed in the field 82% and 42% of the time, re- (1986) as a species found in the county, but A. spectively. Overall, agreement with GAP was less glandulosa Eastw., a similar hairy manzanita with common in the chaparral and scrub polygons, with a basal burl, was found in our surveys for polygon 57% of the GAP co-dominants observed in large W2. In two other cases, a species listed by GAP MADRONO 194 [Vol. 46 was not observed in any ofthe sample elements we Table4. PredictedandMeasuredCrownClosurefor surveyed. Pinus lambertiana Douglas was not Selected GAP Polygons. found in the sample elements in polygon Fl, al- Primary Measured though California State Park personnel indicated or Predicted crown they thrived at higher elevations within the poly- Poly- second- crown closure (c - 2SE, c + gon, away from roads and in areas close to the pol- gon ary closure (%) (%) (c) 2SE) ygon boundary. In the other case, Ceanothus cu- nweitahtiuns wpoalsyngootnfCoIu.nd in any ofthe sample elements rPO1 prp OAA4AoCnnUUU\———j11IiynnUUnnUU ~^/7AZ0 VU) Similar to the experience with Pinus lamberti- Q ou—iuu ana, observations of polygon areas away from the W1 pr 0^ 1Q 4AZO (ZU, vj) chosen sample elements suggested that in some oc An inn cases the elements chosen did not encompass rep- wwoz p oAnu—iinunu J^OA /OJ resentative vegetation found elsewhere in the pol- S 60-100 ygon. For example, in polygon SI, large portions CI p 60-100 81 (63, 99) ofroadside areas inthe northernpartofthepolygon S 60-100 were covered with Salvia apiana Jepson. However, C2 P 60-100 74 (66, 83) s 60-100 permission to access those areas was not granted SI P 40-59 54 (36, 72) and no sampling could be performed. In polygon s 40-59 C2, large portions of north-facing slopes in the S2 p 60-100 63 (45, 80) southern part ofthe polygon were covered by con- s 25-39 tinuous stands of Quercus berberidifolia, but per- mission was not granted to access those areas. Al- though unavoidable, these experiences indicate lim- er for scrub/chaparral species. As noted earlier, the its to the representativeness of our sampling pro- accuracy of the GAP database is linked to the ac- tocol. curacy of VTM and soil vegetation maps, and to Conversely as noted above, numerous species more recent remote sensing data used to create this within the polygons were observed in high enough database. The original VTM and soil-vegetation abundance to warrant possible designation as a co- maps may have been quite accurate for forests and dominant although they were not listed in the GAP woodlands because of the ease of conducting stud- database (Table 3). The abundance ofmost ofthese ies in those relatively open areas, combined with species was modest, butgiventhe SEinterval about the economic incentive of producing accurate data the mean sampled composition (Table 2), these spe- for these potential timber sources. Scrub and chap- cies could be designated as co-dominants. Most of arral have little or no commercial value and the these species omitted from GAP were shrubs (e.g., effort required to maneuver through dense thickets Arctostaphylos glandulosa, Quercus berberidifolia, discourages data collection. Ceanothus palmeri Trel., Eriogonum fasciculatum For certain chaparral species, successional & (Benth.) Torrey A. Gray), except for Pinusjef- changes may explain some of the discrepancies in freyi Grev. & Balf. in polygon Fl and P. coulteri species composition between the GAPdatabase and D. Don in polygon F2. Many (15 of 27) of these the present field study. Although previous studies potential co-dominants were listed as co-dominants estimating natural cover of vegetation for BHC in- in neighboring GAP polygons (see Table 3), sug- ventory generation assumed little change in chap- gesting the influence of adjacent polygons on spe- arral vegetation over time (Winer et al. 1983), this cies composition of the surveyed polygons. assumption may not be appropriate. Chaparral fol- Crown closure. Table4 summarizes thepredicted lows certain successional trends after fires (Hanes and measured crown closure for the GAP polygons 1971; Keeley 1975; Hanes 1977; Barbour and Ma- studied. When the GAP-predicted crown closure of jor 1977; Gordon and White 1994). Ceanothus both primary and secondary assemblages were the chaparral and coastal sage scrub may emerge im- same, the measured crown closure was within both mediately after a fire depending on the elevation, ranges (polygons Fl, CI, and SI). When the GAP- aspect, and antecedent vegetative conditions, but predicted crown closure of both primary and sec- may be displaced by chamise or scrub oak chap- ondary assemblages were different, the measured arral. Ceanothus cuneatus (Hook.) Nott. is one spe- crown closure was within the crown closure range cies which within 50 years can die out completely, ofone ofthe assemblages (polygon F2) orbetween and C. leucodermis E. Greene is eliminated after the ranges of both assemblages (polygon Wl). 40 years. Other species of Ceanothus tend to thin with time because recruitment of new individuals Discussion does not occur in the absence of fire. Some of the more underrepresented species in our field obser- Species composition within the GAPdatabase. In vations which were predicted by the GAP database our study, the apparent accuracy of the GAP data- were members of the genus Ceanothus (C. leucod- base was high forforest/woodland species and low- ermis, C. oliganthus Nott., C. greggii, A. Gray and 1999] CHUNG AND WINER: SAN DIEGO GAP ANALYSIS 195 C. cuneatus). A successional process could explain were not present or were present at lower levels the absence of Ceanothus species from some ofthe than expected for a co-dominant species, these dis- field data even though they were predicted in the crepancies will not necessarily translate into signif- GAP database. icant errors in BHC emission inventories. For the However, explaining discrepancies in the chap- purpose ofthe development ofsuch BHC emissions arral andcoastal sage scrubpolygons is still atpres- inventories, species composition errors have ad- ent speculative. Direct evidence for successional verse effects only when a plant species listed in trends in chaparral and coastal sage scrub species GAP as a co-dominant for a polygon should actu- is not readily available in the literature for San Di- ally be a species with a significantly different mea- ego County, nor can it be easily determined. Estab- sured BHC emission rate. For example, if a co- lishing successional trends requires knowing the dominant listed by GAP is a high isoprene-emitting species composition at historical times (as well as species, but the actual plant species which occurs at present) and such historic information is not in the polygon is a low- or non-emitting species, readily available for specific locations within San the resulting BHC emission flux calculated for that Diego County. It would be useful to re-examine the polygon will over-estimate isoprene emissions. In VTM plots forchaparral and coastal sage scrub and contrast, forcases where a low-emitting species oc- compare them to the VTM plots forblue oakwood- curs in place of another low-emitting species (for lands evaluatedby Allen-Diaz (1993) andtothe fire example,Adenostomafasciculatum in place ofCer- maps compiled by the California Department of cocarpus betuloides Torrey & A. Gray), the error Forestry and Fire Protection. However, such eval- may be significant with respect to correct species uation was beyond the scope of this project. assignment in GAP, but have minimal effect on the Within forests and woodlands, studies have eval- resulting BHC emissions inventory. VTM uated the accuracy of plots over time. For In order to evaluate the significance of species example, Allen-Diaz (1993) found little natural discrepancies between the GAP listings and the VTM change in species composition in plots within field surveys, the indices identifying the polygons blue oak woodlands in the Central Valley but some with the largest biogenic hydrocarbon emissions increase in the size and number of oak species in- based on GAP (see above) were recalculated forthe dividuals over a period of50-60 years. Minnich et eight polygons investigated in this study using the VTM al. (1995) reported gradual species change in observed percent covers of plant species within plots within the San Bernardino Mountains from each polygon. Values ofthe indices were calculated Pinusponderosa Laws, andP. lambertianatoAbies by summing the isoprene or monoterpene emission concolor (Gordon & Glend.) Lindley and Caloced- rates of the plant species in a given polygon, rus decurrens (Torrey) Florin attributable to fire weighted by their percent relative cover, and mul- suppression. Some of these changes were attribut- tiplying this sum by the area of the polygon. For able to effects of air pollution on beetle-induced all eight polygons, total isoprene emission and total mortality and seedling recruitment of pine species monoterpene emissions were calculated based on (Miller et al. 1997). However, these studies did not the survey data, and these were compared with the report wholesale replacement of species within for- totals generated during the polygon selection phase ested and wooded VTM plots. of the study based on the GAP data. Although for Other discrepancies observed between GAP pre- some individual polygons the discrepancy between dictions and data from the surveyed sample ele- the indices obtained for GAP species vs observed ments could be attributed to the GAP database. For species was quite large, when summed over all example, the GAPdatabasepredicted species which eight polygons the differences in the total emission are not recognized by local botanists as present in indices were negligible. Thus, the total isoprene the county. Some of these discrepancies appear to emission forthe eightpolygons basedon the survey be due to taxonomic distinctions (Arctostaphylos data was only 7% greater than the total calculated tomentosas listed by GAP in place ofA. glandulosa using the GAP data, while the total monoterpene or Quercus cornelius-mulleri listed by GAP in emission for the eight polygons based on the field place of Q. berberidifolia). Fortunately, these dis- survey data was just 2% lower than the GAP de- crepancies may have little impact on the utility of rived total. Clearly, the GAP GIS database can be the GAP database foruse in assembling BHC emis- a useful source of species composition and domi- sion inventories because taxonomically-related spe- nance information for the purpose of assembling cies were found instead. We have shown that tax- BHC emission inventories when a sufficiently large onomy can be a strong predictor of BHC emission number of polygons are averaged. rates, especially at the genus level (Benjamin et al. Finally, the observation of numerous species in 1996). high-enough abundance to be designated as co- Despite the discrepancies between predicted and dominants but not listed in the GAP database is to observed plant species cover, on average the utility be expected given that GAP designates only six co- ofthe GAP database for developing BHC emission dominants perpolygon, three in the primary assem- inventories appears to be adequate. Even though blage and three in the secondary assemblage. A some plant species predicted by the GAP database species assemblage with only three co-dominants MADRONO 196 [Vol. 46 may not necessarily capture the species composi- public lands in these areas. Nevertheless not being tion within a polygon. able to survey more private lands resulted in po- tential sampling bias from this constraint on se- Limitations in thepresent GAPfield assessment. lecting random sample elements. For example in Our survey data indicated a difference in the ac- polygons C2 and SI, certain areas were removed \ curacy ofprimary versus secondary assemblages in from the random selection process due to private GAP. Primary assemblage co-dominants were cor- ownership. If such areas had been available for rectly listed by the GAP database more often than sampling and could have been included in the ran- secondary assemblage species. In a GAP polygon, dom selection process, better correlation between primary assemblages by definition are more prev- survey data and the GAP database might have been alent. In the present study, a sample element was observed. more likely to be placed in the more prevalent as- Additional potential biases exist due to sampling semblage, resulting in the gathering of more data only eight ofthe original 437 GAPpolygons within on primary assemblage species and higher repre- western San Diego County. However, the sample sentation by those species. With use of only three size for our purposes was relatively large, since the or four elements, there was a smaller chance of eight polygons studied were selected from a subset sampling a species from a secondary assemblage of 40 polygons believed to be the highest isoprene with a frequency proportional to the area occupied or monoterpene emitting polygons, respectively. by that assemblage. In a study with more resources, Thus, high-emitting polygons were well represent- and hence a larger number of randomly-placed ed. Although a bias existed for undersampling the sample elements, representation by species from ei- low-emitting polygons, the likelihood of low-emit- ther assemblage should be proportional to that as- ting polygons actually being high-emitting poly- semblage's cover within the polygon. gons was not significant since a relatively small Limitations in siting the sample elements may number of plant species are high-emitting (Benja- have accounted for other discrepancies as well. By min et al. 1996). sampling only near roads, away from polygon Given the effort needed to gather the field data, boundaries, and only with the permission of land it was necessary to reduce the area sampled. More- owners, large areas were removed from inclusion over, the sample area required to estimate the true in the study. These limitations were recognized and relative cover of individual species in a polygon accepted as a condition to performing this type of was not known. The literature suggested surveying survey. Observations from a distance and input 7% of a forested area using parallel belt transects from individuals knowledgeable about local botany provided a 65% chance the sample mean of the were helpful in identifying additional plant species basal area of the trees would be within 10% ofthe outside the selected sample elements but did not true mean for more common species (Bormann add to the quantitative characterization of vegeta- 1953). The effortneededto obtain an accuratemea- tion cover reported here. sure ofrelative cover may be similar. In the present As noted earlier, the GAP assessment in San Di- study, the belt transects directly sampled 1.8 hec- ego County posed special problems in terms of tares within polygons of about 1800 to 2300 hec- sampling representative areas within privately- tares, or about 0.1% of the polygon area. The line owned parts of a polygon. In the Utah GAP vali- transects approximating 3 m belt transects directly dation project, 42% ofthe state was under the con- sampled 0.72 hectares in polygons ranging in size trol of the US Bureau of Land Management, with from 3600 to 6600 hectares, or about 0.01%. On private interests owning only 21% (Edwards et al. the other hand, the effective size of our samples 1995). In San Diego County, the San Diego County may be larger. The vegetation cover composition Association ofGovernment (SANDAG) 1990 own- within the transects may approximate the cover ership database indicated private interests owned composition ofa square which immediately bounds 41% ofcounty land (SANDAG 1997). Private land the ends of the perpendicular transects. If this was owners typically purchase land in accessible areas the case, the three sample elements in each forest/ within the vicinity of roads, and therefore, suitable woodland polygon may have effectively sampled public lands within the vicinity ofroads forthe pur- 75 hectares or about 4% ofthe polygon area, while poses of conducting a GAP assessment were lim- the four sample elements in the chaparral/scrub ited. Although a 14% success rate for our mailers polygons may have effectively sampled 36 hectares seeking property access was high by the standards or about 1% of the polygon area. of some surveys, obtaining permission to access A more intensive sample design could have al- private property was a limiting factor in being able lowed the surveying of plants that were missed in to site sample elements. the current surveys. The anecdotal comments about These access limitations present possible biases the existence ofP. lambertiana athigherelevations related to not being able to survey private lands. in polygon Fl suggest a more intensive sampling Observations from roadsides and hilltops suggested effort could address these omissions. Although that land use on private lands in the areas investi- large sample elements avoids quantifying hetero- gated did not differ appreciably from land use on geneity below the intended resolution of the map,