Steven A. Sader

Detection of forest harvest type using multiple dates of Landsat TM imagery

Abstract A simple and relatively accurate technique for classifying time-series Landsat Thematic Mapper (TM) imagery to detect levels of forest harvest is the topic of this research. The accuracy of multidate classification of the normalized difference vegetation index (NDVI) and the normalized difference moisture index (NDMI) were compared and the effect of the number of years (1–3, 3–4, 5–6 years) between image acquisition on forest change accuracy was evaluated. When Landsat image acquisitions were only 1–3 years apart, forest clearcuts were detected with producers accuracy ranging from 79% to 96% using the RGB-NDMI classification method. Partial harvests were detected with lower producers accuracy (55–80%) accuracy. The accuracy of both clearcut and partial harvests decreased as time between image acquisition increased. In all classification trials, the RGB-NDMI method produced significantly higher accuracies, compared to the RGB-NDVI. These results are interesting because the less common NDMI (using the reflected middle infrared band) outperformed the more popular NDVI. In northern Maine, industrial forest landowners have shifted from clearcutting to partial harvest systems in recent years. The RGB-NDMI change detection classification applied to Landsat TM imagery collected every 2–3 years appears to be a promising technique for monitoring forest harvesting and other disturbances that do not remove the entire overstory canopy.

Deforestation rates and trends in Costa Rica, 1940 to 1983

Forest area change associated with life zones, slope gradients, and transportation networks was examined within the framework of a geographically referenced data base for Costa Rica. Locations of forest boundaries and other landscape attributes were digitized from available map sources. Differential rates of primary forest clearing associated with these variables were derived for four reference periods between 1940 and 1983. Deforestation occurred predominantly in tropical dry and moist life zones during the early reference periods; in intermediate periods, tropical and premontane moist and wet zones were affected. By 1983, only the less accessible high-rainfall zones in rugged terrain retained relatively undisturbed forest. The relationship between total primary forest cleared and slope gradient began as inverse and did not approach linearity until the last reference period, when improved transportation routes had penetrated the northeastern lowland Atlantic region. Road development that provided access to the forest was an important agent of change in all reference periods. By 1977, all major regions of the country had been penetrated by roads, and only high mountain forests were relatively inaccessible. Significant gaps exist in the data base because forest maps represent only broad zones, and locations of regenerating forest were not available. The historical data will be used to direct satellite monitoring toward landscapes of predicted change to quantitatively assess forest change dynamics. SEVERAL REPORTS HAVE BEEN COMPILED to describe impacts and propose strategies to slow the advance of tropical forest destruction (U.S. Department of State 1978, 1980; World Bank 1978; Barney 1980; Zerbe et al. 1980; Office of Technology Assessment 1984). Although the seriousness of tropical deforestation at the global level is subject to disagreement (Myers 1980, Lanly 1982, Sedjo & Clawson 1983, Brown & Lugo 1984), concensus exists among authors that the problems caused by deforestation are severe in many areas of the tropics. Many uncertainties persist concerning the rates and trends of tropical forest clearing. Current estimates of tropical moist and wet forest clearing rates differ by a factor of six (Woodwell et al. 1983). Not all forests are equally susceptible to clearing because some may be protected or less suitable for agricultural use. Deforestation studies would be more meaningful with better estimates of how forest clearing rates are associated with land systems and processes (Gwynne et al. 1983). Estimates of the extent and rate of tropical forest clearing, as well as how the rate changes over time, have been hindered by insufficient data (Myers 1980, Grainger 1982, Gwynne et al. 1983, Houghton et al. 1983, Woodwell et cal, 1983, Buschbacher 1986). The potential effects of the clearing and burning of tropical forests on atmospheric CO2 levels have been the subject of great concern and debate in recent years (Brown & Lugo 1980, 1982; Revelle 1982; Houghton et al. 1983; Seidel & Keyes 1983). The range of global carbon flux estimates could be reduced by almost 60 percent with more reliable data on the rate and permanence of tropical deforestation (Houghton et al. 1983). Permanence of deforestation relates to how long the area stays in nonforest cover (i.e., agriculture) before it reverts back to successional forest following disturbance or land abandonment. Knowing the locations of areas susceptible to deforestation and agricultural areas more likely to come back in succession forest would help to direct sampling efforts and satellite monitoring to the landscapes of interest. Written records are often incomplete and do not include maps showing locations of forest coverage that are accurate enough to establish a baseline. Aerial photography and other remote-sensing techniques can provide a means by which present and past forest conditions can be compared (Williams & Miller 1979, Sader 1980, Green 1982, Woodwell et al. 1983, Tucker et al. 1984, Sader & Joyce 1985). Spatially and temporally coregistered data can be used to show geographically specific changes and trends for the time frames selected (Joyce et al. 1984).

Tropical forest biomass and successional age class relationships to a vegetation index derived from landsat TM data

Abstract Forest stand structure and biomass data were collected using conventional forest inventory techniques in tropical, subtropical, and warm temperate forest biomes. The feasibility of detecting tropical forest successional age class and total biomass differences using Landsat-Thematic mapper (TM) data, was evaluated. The Normalized Difference Vegetation Index (NDVI) calculated from Landsat-TM data were not significantly correlated with forest regeneration age classes in the mountain terrain of the Luquillo Experimental Forest, Puerto Rico. The low sun angle and shadows cast on steep north and west facing slopes reduced spectral reflectance values recorded by TM at orbital altitude. The NDVI, calculated from low altitude aircraft scanner data, was significantly correlated with forest age classes. However, analysis of variance suggested that NDVI differences were not detectable for successional forests older than approximately 15–20 years. Also, biomass differences in young successional tropical forest were not detectable using the NDVI. The vegetation index does not appear to be a good predictor of stand structure variables (e.g., height, diameter of main stem) or total biomass in uneven age, mixed broadleaf forest. Good correlation between the vegetation index and low biomass in even age pine plantations were achieved for a warm temperate study site. The implications of the study for the use of NDVI for forest structure and biomass estimation are discussed.

Accuracy of landsat-TM and GIS rule-based methods for forest wetland classification in Maine☆

An investigation was undertaken to compare satellite image classification techniques to delineate forest wetlands in Maine. Four classification techniques were compared, including a GIS rule-based model. Accuracy assessments of the four methods on two study sites, Orono and Acadia, revealed very similar results. Overall accuracy for four super groups (forest wetland, other wetland, forest upland, other upland) ranged from 72% to 81% at Orono and 74% to 82% at Acadia. Pairwise significance tests indicated that the GIS model was significantly better than unsupervised classification at both study sites, and significantly better than tasseled cap (Acadia) in classifying the four super groups. Although Kappa coefficients were slightly higher for the GIS model compared to hybrid classification, there was no significant difference between the two methods at either study site. Forest wetland users and producers accuracy was in the 80% range for the highest accuracy achieved either by the GIS model or hybrid classification. Hydric soils, National Wetland Inventory data, and slope percentage were the most important variables in the GIS model. From this study, it appears that a combination of hybrid and GIS rule-based classification methods are the most promising for further investigations of forest wetland delineation.

RGB-NDVI colour composites for visualizing forest change dynamics

Abstract A simple and logical technique was developed to display and quantify forest change using three dates of satellite imagery. The normalized difference vegetation index (NDVI) was computed for each date of imagery to define high and low vegetation biomass. Colour composites were generated by combining each date of NDVI with either the red, green, or blue (RGB) image planes in an image display monitor. Additive colour logic was used to interpret forest change (forest harvest and regeneration) across the landscape on the three dale NDVI colour composite. Harvest and regeneration area were quantified by applying a modified parallelepiped classification creating an RGB-NDVI image with 27 classes that were grouped into nine major forest change categories. Road construction, harvest and regeneration status on old clearcuts can be monitored by interpretation of the additive colour observed at any site. Aerial photographs and stand history maps obtained from a major forest industrial landowner were compared...

Migratory bird habitat monitoring through remote sensing

Abstract Unsupervised classification of Landsat-TM data was employed to identify habitats important for migratory birds in Costa Rica. The overall habitat classification accuracy was 70 per cent (Kappa correction). Mature forest could be identified with high accuracy (93 per cent) but Landsat-TM classification accuracy for major successional stages was low. Habitat availability and conversion rates from 1976 to 1986 were derived from multidate Landsat imagery supplemented with interpretation of historical air photos to document the specific types of habitat change. The major trend in habitat conversion between 1976 and 1984 was forest clearing followed by establishment of permanent pasture. Some of the pasture land was converted to perennial tree crops by 1986. The implication of habitat modification on groups and species of migrant land birds are discussed.

Analyzing a forest conversion history database to explore the spatial and temporal characteristics of land cover change in Guatemala's Maya Biosphere Reserve

We analyzed forest clearing and regrowth over a 23-year time period for 21 forest concession and management units within the Maya Biosphere Reserve(MBR), Guatemala. The study area as a whole experienced a clearing rate of0.16%/year from 1974 through 1997. The overall clearing rate appears rather low when averaged over the entire study area over 23 years because most of the reserve was inaccessible. However, despite the granting of legal protection to the MBR in 1990, clearing rates continued to rise, with the highest rates occurring in the most recent time period in the analysis, 1995 to1997. Higher rates of clearing relative to regrowth occurred in newly established communities and in the Reserves buffer zone, where the clearing of high forest was preferred for pasture development. Exploratory models were built and analyzed to examine the effects of various landscape variables on forest clearing. The different units in the analysis showed different relationships of forest clearing with variables such as forest cover type and distance to access(roads and river corridors). Where available, socio-economic household survey data helped to explain patterns and trends observed in the time series Landsat imagery. A strong relationship between forest clearing and distance to access was demonstrated. More clearing occurred further from roads during later time periods as farmers moved deeper into the forest to find land to clear. Communities inside the MBR that were less dependent on farming had forest clearing to regrowth ratios of one:one or less. These communities used fallow fields in greater proportions than communities in the Reserves buffer zone. General trends in clearing by forest cover type suggest a preference for clearing high forest (bosque alto) types found on the higher elevation, better-drained soils, and fallow fields,and an avoidance of low-lying, seasonally flooded terrain(bajos). Satellite remote sensing observations of forest clearing and regrowth patterns can provide an objective source of information to complement socio-economic studies of the human driving forces in land cover and land use change.

The forests of Maine: 2003

In 1999, the Maine Forest Service and USDA Forest Services Forest Inventory and Analysis program implemented a new system for inventorying and monitoring Maines forests. The effects of the spruce budworm epidemic continue to affect the composition, structure, and distribution of Maines forested ecosystems. The area of forest land in Maine has remained stable since the 1970s. Although relatively small acreages of forest are converted to other land uses, these conversions often remove highly valued forests such as white pine. The total inventory volume of live trees increased slightly, indicating the beginning of a response of Maines forest to the tremendous devastation from spruce budworm.

Spectral analysis and classification accuracy of coffee crops using Landsat and a topographic-environmental model

Coffee is an extremely important cash crop, yet previous work indicates that satellite mapping of coffee has produced low classification accuracy. This research examines spectral band combinations and ancillary data for evaluating the classification accuracy and the nature of spectral confusion between coffee and other cover types in a Costa Rican study area. Supervised classification using Landsat Enhanced Thematic Mapper (ETM+) with only red, near‐infrared, and mid‐infrared bands had significantly lower classification accuracy compared to datasets that included more spectral bands and ancillary data. The highest overall accuracy achieved was 65%, including a coffee environmental stratification model (CESM). Producers and users accuracy was highest for shade coffee plantations (91.8 and 61.1%) and sun coffee (86.2 and 68.4%) with band combination ETM+ 34567, NDVI, cos (i), and including the use of the CESM. Post‐classification stratification of the optimal coffee growing zone based on elevation and precipitation data did not show significant improvement in land cover classification accuracy when band combinations included both the thermal band and NDVI. A forward stepwise discriminant analysis indicated that ETM+ 5 (mid‐infrared band) had the highest discriminatory power. The best discriminatory subset for all woody cover types including coffee excluded ETM+ 3 and 7; however, the land cover accuracy assessment indicated that overall accuracy, as well as producers and users accuracy of shade and sun coffee, were slightly improved with the inclusion of these bands. Although spectral separation between coffee crops and woodland areas was only moderately successful in the Costa Rica study, the overall accuracy, as well as the sun and shade coffee producers and users accuracy, were higher than reported in previous research.

Landsat imagery reveals declining clarity of Maine's lakes during 1995–2010

Abstract.  Water clarity is a strong indicator of regional water quality. Unlike other common water-quality metrics, such as chlorophyll a, total P, or trophic status, clarity can be accurately and efficiently estimated remotely on a regional scale. Satellite-based remote sensing is useful in regions with many lakes where traditional field-sampling techniques may be prohibitively expensive. Repeated sampling of easily accessed lakes can lead to spatially irregular, nonrandom samples of a region. Remote sensing remedies this problem. We applied a remote monitoring protocol we had previously developed for Maine lakes >8 ha based on Landsat satellite data recorded during 1995–2010 to identify spatial and temporal patterns in Maine lake clarity. We focused on the overlapping region of Landsat paths 11 and 12 to increase availability of cloud-free images in August and early September, a period of relative lake stability and seasonal poor-clarity conditions well suited for annual monitoring. We divided Maine into 3 regions (northeastern, south-central, western) based on morphometric and chemical lake features. We found a general decrease in average statewide lake clarity from 4.94 to 4.38 m during 1995–2010. Water clarity ranged from 4 to 6 m during 1995–2010, but it decreased consistently during 2005–2010. Clarity in both the northeastern and western lake regions has decreased from 5.22 m in 1995 to 4.36 and 4.21 m, respectively, in 2010, whereas lake clarity in the south-central lake region (4.50 m) has not changed since 1995. Climate change, timber harvesting, or watershed morphometry may be responsible for regional water-clarity decline. Remote sensing of regional water clarity provides a more complete spatial perspective of lake water quality than existing, interest-based sampling. However, field sampling done under existing monitoring programs can be used to calibrate accurate models designed to estimate water clarity remotely.