Figure 1.
Forest canopy dominated by mature Guadua sarcocarpa.
Sixteen year-old cohort of climbing bamboo forms a dense carpet weighing heavily on mid-canopy trees. Upper canopy trees are widely spaced. Location is 8° 58′ 56′′ S, 68° 42′ 34′′ W, date September 18, 2004.
Figure 2.
Reflectance spectra of three forest types.
Landsat Thematic Mapper band numbers and positions are above the graph. Each vertex is an average of 20 spatially separate samples.
Figure 3.
Three forest types in Landsat TM R-G-B composite of bands 5-4-3.
Image acquired July 26, 1988. From left to right along 50 km transect indicated by solid white line: forest with mature bamboo (light green), with recently dead bamboo (rust brown) and bamboo-free forest on low infertile plateau (dark green). Border between mature and dead bamboo populations was not evident in a 1986 Landsat image (not shown) when both populations were alive and mature, exhibiting a light green color. Two sets of soil samples compared in Table 1 were collected at the white asterisks.
Figure 4.
Greenness indicators and topography in the three forest types.
Top graph shows green vegetation fraction (dashed trace) and non-photosynthetic vegetation fraction (solid trace) along the 50 km transect shown in Figure 3. Contacts between forest types are at the gray vertical dashed lines. Bottom graph is the same transect across a digital elevation model, contrasting topography of the hilly eroded landscape having fertile, poorly drained clay soil supporting bamboo-dominated forest and the plateau with infertile sandy latosol that excludes bamboo.
Figure 5.
Full extent of southwest Amazon bamboo-dominated forests and study locations.
Bamboo extent is based on visual interpretation of two Landsat Geocover mosaics and one MODIS mosaic. Base-map is a hill-shaded Digital Elevation Model from the Shuttle Radar Topography Mission, with gray tones scaled from 0–300 m altitude. Line A-A’ is a topographic profile across the Fitzcarrald Arch, also evident as higher terrain in the inset map. Soils and spectral patterns of forest types were compared at a topographically defined edge of bamboo forest in the small white rectangle, enlarged in Figure 3. Life-cycle study was conducted in the large white square. White asterisks are five sites of flowering or fruiting specimens of sarmentose Guadua, where Landsat images confirmed a temporal change in forest canopy spectral patterns from mature to post-reproduction life stage. Flowering specimens from west to east are: Nelson 6026 (INPA), June 1995 (Guadua sarcocarpa); Daly et al. 9932 (NY), March 1999 (G. weberbaueri); Krukoff 5235 (NY), July 1933 (G. sarcocarpa); field observation, flowered in 1988 (G. sarcocarpa); Daly et al. 12144 (NY), October 2003 (G. sarcocarpa).
Table 1.
Soil attributes compared between two transects along asterisks in Figure 3.
Figure 6.
Bamboo-dominated forest (cross-hatch) coincides with eutrophic haplic cambisol and luvisol soils (gray).
Inset shows these two fertile soil types are uncommon elsewhere in the Brazilian Amazon. White areas in Brazil on the main map are infertile argisols on dissected surfaces, gleysols along rivers and infertile latosol on low plateaus. Soil map from IBGE and EMBRAPA.
Figure 7.
Bamboo life cycle of 27–28 y inferred from spatial congruence of mature-stage maps.
(A) Congruence as a function of temporal separation. Each point represents one pair of mapped years of the 33,100 km2 study area. (B) Four of the mapped years spanning one full life cycle. Overlying the 2004 map is a hachure showing highly congruent extent of mature bamboo 28 y earlier. Gray is forest with mature bamboo at time of mapping, white is forest with juvenile bamboo hidden in understory, black areas were masked due to clouds.
Table 2.
Dates and orbital sensors used for estimating bamboo life cycle length.
Figure 8.
Spatial-temporal clustering of flowering events.
Limits of 74 populations were detected in eight annual MODIS images (2001–2008). Gray area is full extent of bamboo-dominated forest in the southwest Amazon.
Figure 9.
Performance of three-band combinations for distinguishing mature bamboo, dead bamboo and no bamboo in upper canopy.
Equivalent Landsat TM bands were used to represent MSS and MODIS bands. Higher values of B-distance and of F-ratio indicate better target separation in false-color composites. Using a leave-one-out jackknife validation, 17 of the 20 combinations correctly classified 100% of all samples (solid symbols). Better combinations all include a near-infrared band (circles). Labels indicate false-color composites from sensors used in this study: Landsat MSS equivalent to TM 2-3-4 (maps for life cycle estimation); Landsat Geocover TM 2-4-7 (map of bamboo extent); Landsat TM 3-4-5 (life cycle) and MODIS equivalent to TM 3-4-5 (life cycle, bamboo extent, mortality patches).