BIOTROPICA 44(4): 449–453 2012

10.1111/j.1744-7429.2012.00892.x

INSIGHTS

Impact of Past Forest Fires on Bird Populations in Flooded Forests of the Cuini River in the Lowland Amazon Camila D. Ritter1, Christian B. Andretti, and Bruce W. Nelson Instituto Nacional de Pesquisas da Amazoˆnia, Avenida Andre´ Araujo 2936, 69060-001 Manaus, AM, Brazil

ABSTRACT Blackwater floodplain forests of the Rio Negro are susceptible to understory fires. Bird composition was distinct between burned and unburned floodplain forest but not between young (12–18 yr) and old burns (>25 yr), indicating low resilience after fire. Forest regeneration is slow, with open grassy areas persisting >80 yr. Abstract in Portuguese is available in the online version of this article. Key words: Amazon blackwater forest; birds; fires; point counts.

TROPICAL

FOREST FIRES HAVE RECENTLY RECEIVED ATTENTION DUE

(e.g., Cochrane et al. 1999, Nepstad et al. 1999). Humid tropical forests are not adapted to fires and may take many years to recover (Cochrane 2003). Fires damage soil structure and fragment the landscape (Brotons et al. 2004), modify vegetation structure (Cochrane et al. 1999, Barlow & Peres 2004), and impact fauna (Barlow & Peres 2008, Claveiro et al. 2011). Blackwater floodplain forests of the Rio Negro have burned during years of very low water level over the past century (Sternberg 1987, Williams et al. 2005). Understory fires in this region are a serious problem for four reasons. First, they occur despite high rainfall and low human population density (Borges et al. 2001, Oliveira et al. 2001). Fishermen’s campfires are the main ignition source. Second, tree mortality from ground fire is very high. For upland forest, Cochrane et al. (1999) found biomass loss of only 9 percent after a first ground fire, increasing to 47 and 81 percent after second and third burns. Data are lacking for reburning of blackwater floodplain forest, but all 36 floodplain burn scars examined by Flores et al. (unpublished) had tree mortality above 75 percent, suggesting that high mortality occurs after a single burn. Third, forest regeneration rates are extremely slow. Some open grassy areas are attributed by local informants to forest fires as long ago as 1925. Finally, even under the optimistic SRB1 emissions scenario of the IPCC, the HADCM3 global climate model shows >1000 mm drop in annual rainfall for the middle and lower Rio Negro by the 2080s (Intergovernmental Panel on Climate Change—IPCC 2012). Seasonally flooded woody vegetation comprises 12.1 percent of the Rio Negro basin and 11.5 percent of the entire Amazon basin (Melack & Hess 2010). Although upland forests have greater species richness, flooded forests have their often unique flora and fauna (Remsen & Parker 1983). In mist-netting studies TO THEIR FREQUENCY AND EXTENT

Received 25 August 2011; revision accepted 16 April 2012. 1

Corresponding author; e-mail: [email protected]

in the Jaú National Park, 14 percent of the bird species were restricted to flooded forests (Borges & Carvalhaes 2000). Birds are one of the best-studied animal groups in the Amazon (Cohn-Haft et al. 1997). They are good environmental indicators, in the light of the cost-benefits of sampling methods and their habitat sensitivity (Gardner et al. 2008). After fire disturbance, some bird species decline in abundance while others increase (e.g., Barlow et al. 2002, 2006, Barlow & Peres 2004, 2008, Claveiro et al. 2011). Here, we ask what has been the impact of past wildfires on the bird composition of blackwater floodplain, using both conventional and occupancy model methods. We compare post-fire recovery of the bird fauna between older (>25 yr) and more recent burns (12–18 yr) to give some indication of the resilience of the ecosystem in the face of fire. The study area is the blackwater floodplain of the lower Cuini River (0°58′ S, 62°56′ W), a tributary of the middle Rio Negro. The floodplain has many open areas with sparse standing trunks of dead trees, evidence of past forest fires (see Supporting Information). Most of these burned prior to 1985, the year of the earliest reliable satellite images. In 2010, the burn scars were highly variable in their vegetation structure, ranging from open grassy cover or open with many standing and fallen dead trunks to dense woody secondary vegetation. We chose 44 sampling points in the floodplain, half in unburned forest and half in nearby burned sites (Fig. S1). To reduce the chances of recording the same bird at two different sites or of recording upland species, all plots were separated from each other and from upland forest by at least 500 m (Barlow & Peres 2004, Sberze et al. 2009). We measured the fraction of forest area within a radius of 250 m. We weighted this fraction by distance from the listening point, starting with a weight of 100 percent at 0–25 m, dropping to 10 percent at 200–250 m, as birds more than 200 m away are inaudible to the recorder. We recorded bird vocalizations at high water and the early stages of receding water, between 27 June and 4 August, 2010.

ª 2012 The Author(s) Journal compilation ª 2012 by The Association for Tropical Biology and Conservation

449

450

Ritter, Andretti, and Nelson

We recorded from 0540 to 0840 h for five continuous minutes, repeated four times during the study, during consecutive days. We did not repeat the census at the stable low water period because some burned sites become terrestrial and others remain flooded at a normal low-water stage. Recordings were identified by Christian Andretti. When a vocalization was doubtful, we conferred with Mario Cohn-Haft, with more than 20 yr experience in bird vocalization. For each bird species at each plot, a quantitative detection score was derived, ranging from zero to four, based on its presence or absence in each of the four visits. We analyzed the degree of separation of birds found in the unburned forest and the burned sites of different ages using different methods and compared the results. We first employed ANOSIM—a non-parametric analysis of similarity between the two groups of inventory sites in unreduced species-space—using a matrix of Bray-Curtis quantitative distances. We also ran a twodimensional Non-metric Multidimensional Scaling (NMDS) ordination of all plots (Legendre & Legendre 1998). These two analyses used the raw detection scores and were made with VEGAN Package (Oksanen et al. 2010), run in the R environment (R Development Core Team 2009). To obtain detectability-corrected occurrence probabilities, we ran an occupancy model adapted from Kéry (2010). We estimated the model parameters using the R-package R2WinBUGS (Sturtz et al. 2005), which runs an iterative Markov Chain Monte Carlo in WinBUGS 1.4 (Spiegelhalter et al. 1999). These probabilities were used as an indicator of species’ preferences for forest and were compared to a conventional measure of forest preference by two direct ordinations of species along the habitat gradient ‘forest fraction’. The conventional measure was the raw detection score (zero to four) of each species. Stacked bar graphs illustrating these two direct ordinations were constructed using the PONCHO function (C. Dambros, pers. comm.) in the R program. See Supporting Information for further details of methods. Of the 126 species recognized in the recordings, 23 were encountered with a frequency (>10) that permitted further analysis (Table S1). All analyses—ANOSIM, NMDS, direct gradient analysis and the occupancy model—were performed using these 23 species. Raw detections were higher in the burned sites than in unburned forest plots, by a factor of nearly two to one. Only 4 of the 23 species were found exclusively in burned or in unburned sites. Five of the 23 species, however, were at least twice as abundant in forest as in burned sites, whereas 14 of the 23 species showed a preference for burned areas by this criterion, leaving only four species indifferent to vegetation type. ANOSIM found the bird compositions of the two groups of sites to be highly distinct in the unreduced species-space (R = 0.45, P < 0.001). The two-axis NMDS ordination (Fig. 1) explained 84 percent of the variation in species composition. The first axis clearly separated the burned areas from the unburned forests. Younger (12–18 yr) and older burns (>25 yr), however, did not separate. A single axis NMDS explained 76 percent of the variation in species detections. These NMDS scores were moderately well explained by the fraction of forest at each site (R2 = 0.41, F = 30.7, P < 0.001), using a weighted least squares regression.

FIGURE 1. Two-axis non-metric multidimensional scaling NMDS ordination of all 44 seasonally flooded sites, using raw frequencies of 23 bird species. Symbol size indicates the amount of forest within 250 m of the sample point.

Direct ordination of the 44 sites also shows that the fraction of forest within 250 m of the recording station has a clear effect on bird composition. This is true whether the indicator of each species’ importance is ‘raw detection score’ (Fig. 2, left stack) or ‘occurrence probability’ (Fig. 2, right stack). Because occurrence probabilities are adjusted for the detectability of each species, some changes in species order along the environmental gradient are expected. For the two species orders in Figure 2, the Spearman rank correlation is 0.63. In the occupancy model, confidence intervals were narrow for the estimate of average detectability of each species, but were broad for the final estimate of occurrence probability of each species in each vegetation type. Corrections for detectability were very large for some species. Average detectability of the 23 species ranged from 0.1 to 0.9. Six species were less detectable in forest, so their frequencies were effectively raised there when estimating occurrence probability. Four species were less detectable in burned areas. Using a confidence interval of ±1 SD for the occurrence probability (Fig. S3; Table S2), four species showed a significant detectability-corrected preference for burned areas: Camptostoma obsoletum, Coereba flaveola, Phaomyias murina, and Todirostrum maculatum. Myiopagis gaimardii preferred forest. Similarity analysis, ordination by composition and direct ordination by forest fraction all show that the bird species of the blackwater floodplain are influenced by past burns, most of which occurred at least 25 yr before this study. Contrary to expectation, the bird composition of burned sites does not become more similar to that of unburned forest as burn age increases. There are three possible reasons for this decoupling between age and recovery. First, the one site dominated by secondary forest did not have a composition more similar to that

Amazon Birds and Floodplain Fires

451

FIGURE 2. Comparison of detection-corrected and non-corrected habitat preferences, using direct ordination of species by habitat gradient or category. The 44 sites are ordered from left to right by increasing forest fraction and the 23 species are ordered from bottom to top by their increasing preference for forest. Forest preference indicators are uncorrected raw frequencies (left) and detection-corrected occurrence probabilities (right, sites grouped into categories). Vertical lines separate burned plots from unburned forest plots.

of undisturbed forest, suggesting that birds found in mature forest are not attracted to secondary forest, but birds preferring open areas are more tolerant of differences in habitat, occurring also in young secondary forest. Second, the age of the burn had no relationship with the recovery stage of the forest. The only plot covered with secondary forest burned more recently than the 15 plots >25 yr old, which all remained open. This plot was burned by a low intensity fire, evidenced by more patches of surviving trees. Recovery rate of the forest can be determined by the

intensity of the first fire (Barlow & Peres 2004, Smucker et al. 2005). Finally, recovery can be set back by reburning, so that older fire scars may have reburned after the formation of some younger fire scars. Forest succession in blackwater floodplains proceeds much more slowly than in upland areas (Oliveira et al. 2001). The extended grassy stage of succession keeps the area susceptible to repeated burning (Frelich 2002). Reburn events in floodplains are not easily detected with Landsat TM.

452

Ritter, Andretti, and Nelson

The more rigorous occupancy model found only five species significantly affected by past fires. One cause for high uncertainty of the occupancy model is the low number of vocalizations in this study: an average of 4.5 species detected in each 5 min session. Birders accustomed to the noisy mix of morning bird song in upland Amazon forests are surprised at the near silence of blackwater flooded forests. To compensate for low detections, Kéry (2010) suggests including up to 150 sites, more replicated visits per site and longer sessions. Although detection-corrected occurrence probabilities require much greater sampling effort, such corrections are worth pursuing. Corrections varied by an order of magnitude among the species in this study. Detectability at any one site will be highest for those species that: (1) vocalize loudly; (2) vocalize frequently; (3) are more abundant; and (4) are territorial (Kéry 2010). Malhi et al. (2009) concluded that the higher and less variable rainfall of the western Amazon will afford protection against habitat shifts in the face of future drier climate and fire. But this may not be true for blackwater floodplain forests. Post-fire habitat shifts are already taking place and appear to be permanent. Areas which, according to local inhabitants, were burned over 80 yr ago remain dominated by grassy vegetation. The area faces a clear threat of more frequent fires as local population increases and as climate changes. We suggest that wildfires are an imminent threat to these flooded forests and their avian assemblages. Future studies should focus on less common and more sensitive species to better quantify the importance of this threat.

ACKNOWLEDGMENTS We thank the ‘Ponta da Terra’ community for their kindness and support, INPA for the research opportunity, CNPq for a graduate fellowship, Mario Cohn-Haft for his help in identification, and Gonçalo Ferraz for discussion and help with the occurrence probability model. Cintia Cornelius, Gonçalo Ferraz, Jos Barlow, Luiza Magalli Henriques, Paul Kina, Sergio Borges, and two anonymous reviewers provided suggestions to improve the manuscript.

SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article: TABLE S1. The 23 bird species used in this study. TABLE S2. Mean posterior parameters for the species and their credibility intervals of 95% in parentheses. FIGURE S1. Map of the study area, showing location of the sampling plots along the lower Cuini River. FIGURE S2. Hierarchical structure of the local occupancy model modified from Kéry (2010), and flow chart explaining the model. FIGURE S3. Mean ±1 standard deviation for the posterior occurrence probability of each species in forested and burned areas.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

LITERATURE CITED BARLOW, J., T. HAUGAASEN, AND C. A. PERES. 2002. Effects of ground fires on understorey bird assemblages in Amazonian forests. Biol. Conserv. 105: 157–169. BARLOW, J., AND C. A. PERES. 2004. Avifaunal responses to single and recurrent wildfires in Amazonian forests. Ecol. Appl. 14: 1358–1373. BARLOW, J., AND C. A. PERES. 2008. Fire-mediated dieback and compositional cascade in an Amazonian forest. Phil. Trans. R. Soc. B 363: 1787– 1794. BARLOW, J., C. A. PERES, L. M. P. HENRIQUES, P. C. STOUFFER, AND J. M. WUNDERLE. 2006. The responses of understorey birds to forest fragmentation, logging and wildfires: An Amazonian synthesis. Biol. Conserv. 28: 82–92. BORGES, S. H., AND A. CARVALHAES. 2000. Bird species of black water inundation forests in the Jaú National Park (Amazonas state, Brazil): Their contribution to regional species richness. Biodivers. Conserv. 9: 201– 214. BORGES, S. H., M. PINHEIRO, A. MURCHIE, AND C. DURIGAN. 2001. Preservação do Rio Negro: as unidades de conservação. In A. Oliveira, and D. Daly (Eds.). As florestas do Rio Negro, pp. 303–329. Universidade Paulista Press, São Paulo, Brazil. BROTONS, L., S. HERRANDO, AND J. L. MARTIN. 2004. Bird assemblages in forest fragments within Mediterranean mosaics created by wild fires. Landsc. Ecol. 19: 663–675. CLAVEIRO, M., L. BROTONS, AND S. HERRANDO. 2011. Bird community specialization, bird conservation and disturbance: The role of wildfires. J. Anim. Ecol. 80: 128–136. COCHRANE, M. A. 2003. Fire science for rainforests. Nature 421: 913–919. COCHRANE, M. A., A. ALENCAR, M. D. SCHULZE, C. M. SOUZA, D. C. NEPSTAD, P. LEFEBVRE, AND E. A. DAVIDSON. 1999. Positive feedbacks in the fire dynamic of closed canopy tropical forests. Science 284: 1832–1835. COHN-HAFT, M., A. WHITTAKER, AND P. C. STOUFFER. 1997. A new look at the “species-poor” Central Amazon: The avifauna north of Manaus, Brazil. Ornithol. Monogr. 48: 205–235. FRELICH, L. E. 2002. The disturbance regime and its components. In L. E. Frelich (Ed.). Forest dynamics and disturbance regimes – studies from temperate evergreen-deciduous forests, pp. 15–44. Cambridge University Press, Cambridge, U.K. GARDNER, T. A., J. BARLOW, I. S. ARAUJO, T. C. AVILA-PIRES, A. B. BONALDO, J. E. COSTA, M. C. ESPOSITO, L. V. FERREIRA, J. HAWES, M. I. HERNANDEZ, M. S. HOOGMOED, R. N. LEITE, N. F. LO-MAN-HUNG, J. R. MALCOLM, M. B. MARTINS, L. A. MESTRE, R. MIRANDA-SANTOS, W. L. OVERAL, L. PARRY, S. L. PETERS, M. A. RIBEIRO-JUNIOR, M. N. DA SILVA, C. DA SILVA MOTTA, AND C. A. PERES. 2008. The cost-effectiveness of biodiversity surveys in tropical forests. Ecol. Lett. 11: 139–50. INTERGOVERNMENTAL PANEL on CLIMATE CHANGE – IPCC. 2012. Precipitation change anomaly for period 2070-2099 for the HADCM3 model under SRB1 emissions scenario. Available at: http://www.ipcc-data.org/cgibin/downl/ar4_nc/pr/HADCM3_SRB1_1_pr-change_2070-2099.nc (accessed 29 May 2012). KÉRY, M. 2010. Non-standard GLMM’s 1: site-occupancy distribution model. In M. Kéry (Ed.). Introduction to WinBUGS for ecologists: Bayesian approach to regression, ANOVA, mixed models and related analyses, pp. 156–167. Elsevier, Amsterdam, The Netherlands. LEGENDRE, P., AND L. LEGENDRE. 1998. Numerical ecology (2nd Edition). Elsevier, Amsterdam, The Netherlands. MALHI, Y., L. E. O. C. ARAGÃO, D. GALBRAITH, C. HUNTINGFORD, R. FISHER, P. ZELAZOWSKI, S. SITCH, C. MCSWEENEY, AND P. MEIR. 2009. Explor-

Amazon Birds and Floodplain Fires

ing the likelihood and mechanisms of a climate-change induced dieback of the Amazon rainforest. Proc. Natl. Acad. Sci. U.S.A. 106: 20610–20615. MELACK, J. M., AND L. L. HESS. 2010. Remote sensing of the distribution and extent of wetlands in the Amazon basin. In W. J. Junk, M. T. F. Piedade, F. Wittmann, J. Schöngart, and P. Parolin (Eds.). Amazonian floodplain forests: ecophysiology, biodiversity and sustainable management, pp. 43–59. Springer, Berlin, Heidelberg, Germany/New York, NY. NEPSTAD, D. C., A. VERISSIMO, A. ALENCAR, C. NOBRE, E. LIMA, P. LEFEBVRE, P. SCHLESINGER, C. POTTER, P. MOUTINHO, E. MENDOZA, M. COCHRANE, AND V. BROOKS. 1999. Large-scale impoverishment of Amazonian forests by logging and fire. Nature 398: 505–508. OKSANEN, J., F. G. BLANCHET, R. KINDT, P. LEGENDRE, R. B. O’HARA, G. L. SIMPSON, P. SOLYMOS, M. H. H. STEVENS, AND H. WAGNER. 2010. Vegan: Community ecology package. R Package. Version 1.17-4. Available at http://cran.r-project.org/ (accessed 19 October 2011). OLIVEIRA, A. A., D. C. DALY, A. VICENTINI, AND M. COHN-HAFT. 2001. Florestas sobre areia: campinaranas e igapós. In A. A. Oliveira, and D. C. Daly (Eds.). Florestas do Rio Negro, pp. 89–117. Companhia das Letras, UNIP, São Paulo, Brazil. R DEVELOPMENT CORE TEAM. 2009. R: A language environment for statistical computing. R Foundation for Statistical Computing, Vienna,

453

Austria. Available at http://www.r-project.org (accessed 18 January 2010). REMSEN, J. V., AND T. A. PARKER. 1983. Contribution of river-created habitats to bird species richness in Amazonia. Biotropica 15: 223–231. SBERZE, M., M. COHN-HAFT, AND G. FERRAZ. 2009. Old growth and secondary forest site occupancy by nocturnal birds in a Neotropical landscape. Anim. Conserv. 13: 3–11. SMUCKER, K. M., R. L. HUTTO, AND B. M. STEELE. 2005. Changes in bird abundance after wildfire: Importance of fire severity and time since fire. Ecol. Appl. 15: 1535–1549. SPIEGELHALTER, D. J., A. THOMAS, AND N. G. BEST. 1999. Winbugs version 1.4. MRC Biostatistics Unit, Cambridge, U.K. Available at http:// www.mrc-bsu.cam.ac.uk/bugs/winbugs/contents.shtml (accessed 03 July 2009). STERNBERG, H. O. R. 1987. Aggravation of floods in the Amazon River as a consequence of deforestation? Geogr. Ann. 69A: 201–219. STURTZ, S., U. LIGGES, AND A. GELMAN. 2005. R2WinBUGS: A package for running WinBUGS from R. J. Stat. Softw. 12: 1–16. WILLIAMS, E., A. DALL’ANTONIA, V. DALL’ANTONIA, J. M. ALMEIDA, F. SUAREZ, B. LIEBERMANN, AND A. C. M. MALHADO. 2005. The drought of the century in the Amazon Basin: An analysis of the regional variation of rainfall in South America in 1926. Acta Amazon. 35: 231–238.