Defoliation Data Research Articles

Identifying significant scale-specific spatial boundaries using wavelets and null models: spruce budworm defoliation in Ontario, Canada as a case study

We combine wavelet analysis and multiple null models to identify significant spatial scales of pattern and spatial boundaries in historical spruce budworm defoliation in Ontario, Canada. Previous analyses of budworm defoliation in Ontario over the last two outbreaks have suggested three distinct zones of defoliation. We asked the following three questions: (1) is there statistical support for the existence of these three zones? (2) Are the locations of these boundaries consistent between outbreak periods? And (3) how does boundary identification depend on the spatial null model used? Defoliation data for the two outbreak periods (1941–1965 and 1966–2001), and the combined period (1941–2001) were analyzed using a 1D continuous wavelet transform. Boundaries were identified through comparison of wavelet power spectra of each outbreak period to reference distributions based on three different spatial null models: (1) a complete spatial randomness model, (2) an autoregressive model, and (3) a Gaussian random field model. The Gaussian random field model identified coarser scales of pattern than the autoregressive model. Locally, the Gaussian random field model found significant boundaries similar to those previously described, whereas the autoregressive model only did so for the first outbreak. These results indicate that the coarse scale spatial factors that influenced defoliation were more consistent between outbreaks relative to fine scale factors, and that previously described boundaries were strongly driven by the first outbreak. Wavelet analysis combined with spatial null models provides a powerful tool for identifying non-arbitrary scales of structure and significant spatial boundaries in non-stationary ecological data.

Geographic variation in North American gypsy moth cycles: subharmonics, generalist predators, and spatial coupling

Many defoliating forest lepidopterans cause predictable periodic deforestation. Several of these species exhibit geographical variation in both the strength of periodic behavior and the frequency of cycles. The mathematical models used to describe the population dynamics of such species commonly predict that gradual variation in the underlying ecological mechanisms may lead to punctuated (subharmonic) variation in outbreak cycles through period-doubling cascades. Gypsy moth, Lymantria dispar, in its recently established range in the northeastern United States may represent an unusually clear natural manifestation of this phenomenon. In this study we introduce a new statistical spatial-smoothing method for estimating outbreak periodicity from space-time defoliation data collected with spatial error. The method statistically confirms the existence of subharmonic variation in cyclicity among different forest types. Some xeric forest types exhibit a statistical 4-5 year period in outbreak dynamics, some mesic forest types a 9-10 year period, and some intermediate forest types a dominant 9-10 year period with a 4-5 year subdominant superharmonic. We then use a theoretical model involving gypsy moth, pathogens, and predators to investigate the possible role of geographical variation in generalist predator populations as the cause of this variation in dynamics. The model predicts that the period of gypsy moth oscillations should be positively associated with predator carrying capacity and that variation in the carrying capacity provides a parsimonious explanation of previous reports of geographical variation in gypsy moth periodicity. Furthermore, a two-patch spatial extension of the model shows that, in the presence of spatial coupling, subharmonic attractors can coexist whereas nonharmonic attractors (i.e., where the cycle lengths are not integer multiples of one another) cannot.

Legacy of Insect Defoliators: Increased Wind-Related Mortality Two Decades After a Spruce Budworm Outbreak

Abstract Effects of spruce budworm (Choristoneura fumiferana Clem.) outbreaks on growth and survival of balsam fir (Abies balsamea [L.] Mill.) and spruce (Picea spp.) are well documented, but few studies extend beyond 10 years after defoliation ceased. We used inventory data from 106 permanent sample plots in >50-year-old balsam fir stands in northern New Brunswick, Canada, to determine legacy effects of the 1969–1993 budworm outbreak on stand development up to 29 years after defoliation ceased. Defoliation data were from annual aerial surveys from 1945 to 1993 and plot ground sampling from 1985 to 1993. Plots were stratified into net stand volume development categories (decreasing, stable, and increasing 1985–2005 stemwood volume) and related to outbreak phases (outbreak, direct 1–10 years after defoliation ceased, and legacy >10 years), outbreak severity (1–4 [low], 5–8 [medium], and 9–12 [high] years of defoliation), and stand age (mature and overmature). Stand age was an important factor influencing outbreak severity (e.g., r2 = 0.383, P < 0.01). Trend and rate of volume development over time were related to past outbreak severity and increased rate of postoutbreak wind-related mortality, which peaked at 11 m3/ha/yr 11–15 years after defoliation ceased. Results indicate that aging postoutbreak stands are more vulnerable to wind disturbance events, effecting rapid stand decline.

Relationships between elm spanworm, Ennomos subsignaria, juvenile density and defoliation on mature sycamore maple in an urban environment

Using field surveys, we established sampling procedures for estimating defoliation resulting from elm spanworm, Ennomos subsignaria (Hübner) (Lepidoptera: Geometridae), feeding on mature sycamore maple, Acer pseudoplatanus L. (Aceraceae), in St. John's, Newfoundland and Labrador. We also determined whether densities of E. subsignaria eggs, egg masses, early- or late-instar larvae could predict the amount of defoliation at the end of the larval feeding period. Defoliation estimates acquired by sampling branches from only the lower, mid or upper crown explained ≥80% of the variation in tree-level defoliation, suggesting that density–defoliation relationships established using defoliation data from any crown level would also be useful for predicting tree-level defoliation. In linear regressions, egg and egg mass densities explained ≤20% of variation in defoliation and thus only provide a crude relative estimate of the amount of defoliation that will occur. Early- and late-instar larval density in the lower crown explained 53 and 29%, respectively, of the variation in defoliation in the lower crown, where defoliation levels were highest. Thus monitoring early-instar density in the lower crown should provide pest managers with reliable information for decisions regarding whether to apply suppression tactics while allowing enough time to implement these tactics, if necessary.

Geoadditive Bayesian models for forestry defoliation data: a case study

AbstractWe analyze forestry defoliation data from the Emission Impact and Forest Nutrition (IWE) survey, which was carried out in Baden‐Württemberg, Germany in the year 1988. The survey contains information on individual trees such as the degree of defoliation, age, species and measurements on nutrients in the needles, as well as information on tree locations such as soil and geographical characteristics. Our goal is to find suitable predictors for tree defoliation from the above information, and to find a set of models which can explain the underlying biological and environmental processes. To model the spatial correlation in the data and possible nonlinear effects of the covariates we use a geoadditive hierarchical Bayesian model. Posterior inference and model comparison are computationally assessed via Markov Chain Monte Carlo (MCMC) techniques and deviance information criterion (DIC) respectively. Copyright © 2008 John Wiley & Sons, Ltd.

Forest Health Status in Europe

Forest health status in Europe is assessed by the International Cooperative Programme on Assessment and Monitoring of Air Pollution Effects on Forests (ICP Forests). Established by the Convention on Long-Range Transboundary Air Pollution (CLRTAP) under the United Nations Economic Commission for Europe (UNECE), the ICP Forests has been monitoring forest condition in close cooperation with the European Commission (EC) for 20 years. The present paper describes the latest results of the deposition measurements on permanent monitoring plots and of the extensive defoliation sample survey. The findings reveal marked spatial patterns in bulk and throughfall depositions of nitrate (N-NO3), ammonium (N-NH4), and sulfate (S-SO4), as well as an obvious decrease in bulk and throughfall deposition of sulfate. Latest analyses of defoliation data confirm previous results, indicating a high correlation with weather extremes.

Ozone exposure, defoliation of beech ( Fagus sylvatica L.) and visible foliar symptoms on native plants in selected plots of South-Western Europe

The relationships between crown defoliation of beech, visible foliar symptoms on native vegetation and ozone exposure were investigated on permanent monitoring sites in South-Western Europe in the years 2000–2002. Relationships between defoliation of beech and O 3 (seasonal mean, 2-week maximum, AOT40) were investigated by means of multiple regression models (11 plots, 1–3 years of data each) and a model based on temporal autocorrelation of defoliation data (14 plots, 1–3 years of data each). Different multiple regression techniques were used. The four models generated ( R 2 = 0.71–0.85, explained variance in cross-validation 61–78%) identified several significant predictors of defoliation, with AOT40 ( p = 0.008) and foliar content of phosphorous ( p = 0.0002–0.0004) being common to all models. The autocorrelation model ( R 2 = 0.55; p < 0.0001) was used to calculate expected defoliation on the basis of the previous year's defoliation, and model predictions were used as an estimate of expected defoliation under constant site and environmental condition. Residuals (predicted-measured) plotted against current AOT40 shows that a possible effect of ozone occurs only at very high AOT40 (>35,000 ppbh). O 3-like visible foliar symptoms were recorded on 65 species at 47% of the common monitoring sites in 2001 and 38% in 2002. No relationship was found between O 3 exposure, frequency of symptomatic sites and frequency of species with symptoms ( R 2 = 0.11; p > 0.05). A number of questions related to the ecological and methodological basis of the survey were identified. Inherent sampling and non-sampling errors and multicollinearity of the data suggest great caution when examining results obtained from mensurational, correlative studies.

Crown condition within integrated evaluations of Level II monitoring data at the German level

The assessment of crown condition is an annually repeated core activity within the ‘Intensive Forest Monitoring Programme’ (Level II monitoring) at selected plots. As the intensive monitoring focuses on a better understanding of direct and indirect effects of air pollution at the ecosystem level, a considerable number of parameters are collected. This allows the analysis of numerous relationships which may determine tree-crown condition with respect to different ecological situations and their temporal variation. This article describes possibilities for integrated evaluations of crown condition data within the German Level II programme. Besides descriptions of defoliation and mortality rates over time, relationships among defoliation, and stand and site characteristics and sampling circumstances are explored. Autocorrelative properties of defoliation data are also described. The results reveal that the date of the field assessment is influential in three out of four species, while the dominance structure has effects in Scots pine and oak. Temporal variation of defoliation reveals strong plot-specific characteristics with limited possibilities for generalisations. Species-specific temporal autocorrelation structures were detected. Within integrated approaches, the ‘federal state’ variable gains high coefficients of determination; however, alternative models incorporating significant ecological factors also reveal plausible explanations. Interactions between predictors complicate the statistical approaches. This is also true for models that include temporal aspects of defoliation and indicators of climatic drought stress. Methodological limitations and significant interpretations are discussed. Statistical evaluations are restricted mainly by the low number of cases, limiting the number of predictors within multiple models. In the future, both a consequent processing of the collected data according to relevant questions and the application of advanced statistical models will be needed.

Field Evaluation of Cold Hardy Citrus in Coastal Georgia

Sixteen cultivars of citrus (Citrus spp.) and close citrus relatives were planted in Savannah, Georgia to evaluate their potential as fruiting landscape trees in an area that routinely experiences minimum temperatures of 15 to 20 °F (-9.4 to -6.7 °C) during winter. Three to six trees of each cultivar were planted in 1998, and stem dieback and defoliation data were collected in 1999, 2001, and 2002. During the 4 years of the study, air temperatures fell below 32 °F (0.0 °C) 27 to 62 times per season, with absolute minima ranging from 13 to 18 °F (-10.6 to -7.8 °C), depending on year. In general, kumquats (Fortunella spp.), represented by `Meiwa', `Nagami', and `Longevity', were completely killed (or nearly so) in their first year in the field after air temperature minima of 13.5 °F (-10.28 °C). Others experiencing 100% dieback were `Meyer' lemon (Citrus limon × C. reticulata) and `Eustis' limequat (C. aurantifolia × Fortunella japonica), which were tested twice during the study. Kumquat hybrids, including procimequat [(C. aurantifolia × F. japonica) × F. hindsii), `Sinton' citrangequat [(C. sinensis × Poncirus trifoliata) × unknown kumquat], `Mr John's Longevity' citrangequat [(C. sinensis × P. trifoliata) × F. obovat], razzlequat (Eremocitrus glauca × unknown kumquat), and `Nippon' orangequat (C. unshiu × F. crassifolia) survived freezing, but all experienced at least some defoliation and stem dieback. `Owari' satsuma (C. unshiu), `Changsha' mandarin (C. reticulata), nansho daidai (C. taiwanica) and ichang papeda (C. ichangensis) experienced only minor stem dieback but substantial defoliation in most years, except that ichang papeda was substantially damaged in the last year of the study. Seven cultivars produced fruit at least once during their first 4 years: nansho daidai, ichang papeda, `Nippon' orangequat, `Mr John's Longevity' citrangequat, `Owari' satsuma, `Changsha' mandarin, and procimequat. Based on cold hardiness, fruiting, and growth characteristics, `Owari' satsuma, `Changsha' mandarin, `Mr John's Longevity' citrangequat, and `Nippon' orangequat provided the hardiest, most precocious and desirable fruiting landscape trees in this study.

Temporal relations between defoliation caused by spruce budworm (Choristoneura fumiferana Clem.) and growth of balsam fir (Abies balsamea (L.) Mill.)

Defoliation by spruce budworm (Choristoneura fumiferana Clem.) greatly affects balsam fir (Abies balsamea (L.) Mill.) by reducing needle mass and ring width and eventually resulting in tree death. We analyzed temporal patterns of radial growth in various stem sections of twenty balsam fir growing in two mature stands, one intermediate-aged, and one immature stand in New Brunswick, Canada. Growth was compared to annual defoliation data from aerial surveys to determine the timing of growth reduction throughout the stem. Results indicated that there was a delay between the first radial growth reduction registered in the crown and reduction in the lower part of the stem. The delay was variable, depending upon stand age. In immature trees, ring width reduction occurred almost simultaneously along the entire stem, whereas in mature fir, the growth reduction occurred one to three years earlier in the crown than at the stem base. In mature trees, the reduction in the crown occurred immediately in the first year following severe defoliation (70–100% of current year foliage), whereas in severely defoliated immature trees, growth reduction occurred 1 to 2 years later.

The Spruce Budworm Decision Support System: forest protection planning to sustain long-term wood supply

The Spruce Budworm Decision Support System (SBW DSS) quantifies the marginal timber supply (m3/ha) benefits of protecting stands against spruce budworm (Choristoneura fumiferana (Clem.)) defoliation. It allows the user to quantify the volume benefit of protecting alternative areas and determine effects on forest development and annual allowable cut. Implementing the SBW DSS on a land base involves seven steps: (i) defining the base defoliation, or an explicit forecast of the defoliation level included in yield forecasts; (ii) compiling historical defoliation; (iii) defining base volume yields; (iv) obtaining the harvest schedule from the land base management plan; (v) building the stand impact matrix, which quantifies direct impacts of defoliation; (vi) building the forest impact matrix, which quantifies indirect impacts of defoliation on harvest schedules; and (vii) building the stand-history file, which contains all stand-level and defoliation data. These tasks are usually completed every 5 years. The remaining aspects of the planning methodology are implemented annually, including (i) recording the previous year's defoliation, (ii) estimating potential defoliation in the current year from budworm survey data, (iii) calculating volume loss or protection priority, (iv) generating budworm-caused volume loss maps, (v) digitizing potential spray blocks, and (vi) evaluating the protection program. Using ARC/INFO® and ArcView® geographic information system programs, the Protection Planning System component (PROPS) generates volume loss maps that can be used to help design and analyze costs and benefits of insecticide spray programs. Implementation of PROPS for the 450 000 ha Upper Miramichi Crown License in New Brunswick is described. Under "normal" and "severe" budworm outbreak scenarios, defined based on predictions of 1999–2008 defoliation, losses of 6.6 × 106 and 16.7 × 106 m3 of spruce (Picea sp.) – balsam fir (Abies balsamea (L.) Mill.) volume were projected to occur on this land base.

Forest Type Affects Predation on Gypsy Moth (Lepidoptera: Lymantriidae) Pupae in Japan

A compilation of historical defoliation data from Hokkaido, Japan confirms previous reports that plantations of Larix leptolepis Gordon are more susceptible to defoliation by the gypsy moth, Lymantria dispar (L.), than are natural forests dominated by Quercus crispula Thumb. In this study, we compared levels of predation on laboratory-reared pupae deployed at a natural Quercus stand, a plantation of Betula platyphylla Hara, and a plantation of L. leptolepis . Predation was highest in the natural Quercus stand, lowest in the Larix plantation and intermediate in the Betula plantation. Counts of small mammals in traps indicated that populations of Apodemus argenteus, Apodemus speciosus ainu , and Clethrionomys rufocanus bedfordiae were most abundant in the oak forest, least abundant in the Larix plantation, and intermediate in the Betula plantation. We hypothesize that variation in the abundance of small mammals and their predation on gypsy moth pupae during periods of low gypsy-moth densities is a partial cause of the variation among stand types in their susceptibility to defoliation by gypsy moth. We also argue that predation on gypsy moth pupae by small mammals in natural oak forests is more important to gypsy moth dynamics than previously considered and that considerable similarities may exist between gypsy moth dynamics in Japan and North America.

Spread of Gypsy Moth (Lepidoptera: Lymantriidae) in the Central Appalachians: Comparison of Population Boundaries Obtained from Male Moth Capture, Egg Mass Counts, and Defoliation Records

Since its introduction near Boston in 1869, the gypsy moth, Lymantria dispar (L.), has spread over 1,000 km to the southwest. To measure its current rate of spread through the central Appalachian Mountains we estimated population boundaries, which are lines that separate areas with population densities generally above and below a specific threshold. Population boundaries were estimated in northwestern Virginia and southeastern West Virginia using male moth counts (1988-1994), egg mass counts (1988-1991), and defoliation maps (1988-1994). The boundary of 1 moth per trap was on average 110 km from the boundary of defoliation, and the male moth capture rate increased 10 times per 29 km perpendicular to the population front. Approximately 11 yr separated the time when traps caught 1 moth per trap until defoliation first occurred in the same area. The rates of gypsy moth spread estimated using different census methods and from different population thresholds changed almost synchronously from year to year. Only spread rates estimated from defoliation data had a different temporal pattern. Gypsy moth spread rate declined from 1988 to 1994 by 35%, as measured from time series of spread rates and boundary compression (reduction of the distance between adjacent boundaries). Reduction in gypsy moth spread rate may have been the result of intensive population management in the area.

Control of Colorado Potato Beetle in the Red River Valley, Grand Forks, ND, 1994

Abstract Potatoes were planted 17 May at the Research Farm of the Red River Valley Potato Growers Association. Plots were four rows wide, 60 ft long, with 38 in rows and 12 in between plants. Plots were separated on all sides by 6.3 ft of cultivated soil. The experiment had a complete randomized plot design with four replications. Egg masses were flagged 15 Jun and checked daily. Thirty percent egg hatch was estimated to have occurred 22 Jun, but heavy rains and wet field conditions delayed first application of insecticidal sprays until 24 Jun. Subsequent spray applications were made 5, 12 Jul, and 8 Aug. Jun and Jul sprays were targeted against first generation larvae, Aug application against summer generation adults. All applications were made using a tractor-mounted boom sprayer, with a single Teejet nozzle over each row. Sprays were applied in 40 gal of water/acre at 40 psi. Egg masses, small larvae (1st and 2nd instar), large larvae (3rd and 4th instar), and adults were counted 27 Jun, 7, 14 Jul, and 10 Aug on 20 randomly selected plants from the two middle rows of each plot. Tuber yields were estimated by mechanical harvesting one row per plot 2 Sept. Defoliation data were recorded at each evaluation and expressed as mean percent of defoliation across sampling dates. The average number of insect counts across sampling dates are presented in the table. Best control resulted from the use of Furadan, Imidan, and Capture. Yield differences among treatments were not significantly correlated with any measure of insect numbers. Parts of the field were saturated for an extended period of time by heavy Jun rains, which adversely affected yields in all plots. Tuber yields ranged from 136.7 to 207.4 cwt/acre.

Predicting effects of defoliation on spruce—fir stand development: a management-oriented growth and yield model

A management-oriented growth and yield model is described that forescasts the development of immature to overmature spruce ( Picea sp.) and balsam fir ( Abies balsamea (L.) Mill.) stands having various histories of spruce budworm ( Choristoneura fumiferana (Clem.)) defoliation. The model was calibrated using data from 60 permanent sample plots located within spruce-fir stands. Input data are structured in 2-cm DBH classes for each species, and the model has a 5 year growth period. Given an initial description of stand structure, defoliation status, and a future defoliation trend, the model predicts mortality and survivor growth based upon species-specific, defoliation- and competition-based functions. Model validation using a reserved set of 58 plots indicated that predicted 5-year volume changes were within 15% of measured volume for 86% and 78% of the plots, using individual-tree and aerial defoliation data, respectively. Model usage in yield curve formulation, forest protection planning, and inventory updating, and the prospective use of remote sensing technology for model inputs are discussed.

Control of Colorado Potato Beetle in the Red River Valley, Grand Forks, Nd, 1993

Abstract Potatoes were planted 3 May at the Research Farm of the Red River Valley Potato Growers Association, Grand Forks, ND. Plots were 4 rows wide, 60 ft (18.3 m) long, with 38 in (97 cm) spacing between rows and 12 in (30 cm) between plants. Plots were separated on all sides by 6.3 ft (1.9 m) of cultivated soil. The field was divided into three experiments each consisting of a complete randomized plot design with 5 replications. Few Colorado potato beetle adults were found in our field before mid-June. Egg masses were flagged 23 Jun and checked daily. Nine percent egg hatch was estimated to have occurred on 1 Jul, but heavy rains and wet field conditions prevented first application of insecticidal sprays until 7 Jul. Subsequent spray applications were made on 14, 21 Jul, 11 and 18 Aug. The July sprays were targeted against first generation larvae, the August applications against second generation larvae. All applications were made using a tractor-mounted boom sprayer, with a single Teejet nozzle over each row. Sprays were applied in 40 gal of water/acre (150 liters/ha) at 40 psi (275 kPa). Control evaluations were made on 16, 28 Jul, 13 and 27 Aug. Because Colorado potato beetle population pressure was very low and not uniformly distributed across plants, we did not count larvae directly. Instead, we scored 20 plants per plot for small larvae (1st and 2nd instar) and large larvae (3rd and 4th instar), for each using a \-4 scale: 1 = 0 insects, 2 = 1-5 insects, 3 = 5-9 insects, 4 = 10 or more insects. Adults were counted on a per plant basis. Data were also collected 9 Jul, but these data are not presented because population pressure was very low and not uniform across plots. Defoliation data were not recorded as treatment differences were not evident. For purposes of analysis of data on larvae, we used mean scores across sampling dates. Tuber yields were estimated by mechanical harvesting one row per plot on 9 Sep or 20 Sep.

Colorado Potato Beetle Control with Foliar Sprays and Soil Applied Systemics, 1980

Abstract Uncut Norchip potatoes (Size B) were planted, on 3 ft centers at Wooster, OH on May 15, 1980. Each plot consisted of 3, 25-ft long rows with 12 ft between plots. Between each replicate was a 15 ft alley. The soil type was silt loam with O.M. 2.5% and pH 6.6. The herbicide Eptam 7EC at 4.0 lbs ai/A was applied May 6 and the preemergence herbicide Sencor 50W was applied May 27. The fungicide Dithane M-45 at 1 lb ai/acre was applied periodically throughout the season for control of late blight. Granular Temik was applied as insecticide for two treatments with a Planet Jr. in-furrow at planting. The foliar insecticides were applied at a walking pace of 2 mph with a hand-held CO2 sprayer at 40 psi. Spray volume was 2 gal of water per treatment (50 GPA). The decision whether to spray in a given week was based on weekly counts of Colorado potato beetle larvae and adults. The thresholds were: (1) greater than 2 overwintered adults/plant; (2) greater than 5 larvae/plant, and (3) greater than 5 first generation adults/plant. Larvae were classed as small (l-ll instar) or large (MI-IV instar). The insect count data were supplemented by percent defoliation data taken Jul 11 and Aug 1 for individual plants within each treatment. By late July, the check plots were completely defoliated. Potatoes were harvested and yields measured on Sep 16.

Estimating Defoliation of Peanuts From Spectral Data1

Abstract Analysis of spectral reflectance potentially can be used to determine the condition of a crop, e.g., drought stress and nutritional deficiencies. Since there is a paucity of information on the feasibility of quantifying disease severity in Arachis hypogaea (L.) with spectral reflectance measurements of the canopy, the objective of this study was to determine if spectral reflectance values of the peanut canopy could be used to estimate the amount of defoliation attributable to epidemics of Cercospora leafspot. Defoliation data and spectral reflectance values (Exotech 100-A radiometer) were periodically obtained in fungicide evaluation tests with varying amounts of defoliation, depending on the efficacy of the fungicide treatments. Combination parameters were developed from reflectance values and statistically compared with appropriate defoliation data. During this two-year study, defoliation percentages were accurately predicted (R2=0.89) by subjecting the combination parameters to regression analyses. Therefore, these results indicate that there is a definite relationship between canopy reflectance and defoliation resulting from Cercospora leafspot epidemics. Since full sunlight near solar noon is the only requirement for reliable reflectance measurements, this radiometric technique can probably be used to monitor development of leafspot epidemics and quantify yield losses on a large scale with either aerial or satellite measurements.

AAAS Board Reviews Defoliation Data

AAAS Board Reviews Defoliation Data