Deforestation effects on Attalea palms and their resident Rhodnius, vectors of Chagas disease, in eastern Amazonia

Abstract

Attalea palms provide primary habitat to Rhodnius spp., vectors of Trypanosoma cruzi. Flying from palms, these blood-sucking bugs often invade houses and can infect people directly or via food contamination. Chagas disease (CD) risk may therefore increase when Attalea palms thrive near houses. For example, Attalea dominate many deforested landscapes of eastern Amazonia, where acute-CD outbreaks are disturbingly frequent. Despite this possible link between deforestation and CD risk, the population-level responses of Amazonian Attalea and their resident Rhodnius to anthropogenic landscape disturbance remain largely uncharted. We studied adult Attalea palms in old-growth forest (OGF), young secondary forest (YSF), and cattle pasture (CP) in two localities of eastern Amazonia. We recorded 1856 Attalea along 10 transects (153.6 ha), and detected infestation by Rhodnius spp. in 18 of 280 systematically-sampled palms (33 bugs caught). Distance-sampling models suggest that, relative to OGF, adult Attalea density declined by 70-80% in CP and then recovered in YSF. Site-occupancy models estimate a strong positive effect of deforestation on palm-infestation odds (βCP-infestation = 4.82±1.14 SE), with a moderate decline in recovering YSF (βYSF-infestation = 2.66±1.10 SE). Similarly, N-mixture models suggest that, relative to OGF, mean vector density sharply increased in CP palms (βCP-density = 3.20±0.62 SE) and then tapered in YSF (βYSF-density = 1.61±0.76 SE). Together, these results indicate that disturbed landscapes may support between ~2.5 (YSF) and ~5.1 (CP) times more Attalea-dwelling Rhodnius spp. per unit area than OGF. We provide evidence that deforestation may favor palm-dwelling CD vectors in eastern Amazonia. Importantly, our landscape-disturbance effect estimates explicitly take account of (i) imperfect palm and bug detection and (ii) the uncertainties about infestation and vector density arising from sparse bug data. These results suggest that incorporating landscape-disturbance metrics into the spatial stratification of transmission risk could help enhance CD surveillance and prevention in Amazonia.

Fig 1. Field-sampling design, main data features, and data-analytical strategy.

We sampled 10 transects by walking a central trail and measuring the perpendicular distance (red lines) between each adult Attalea palm seen (green star-like shapes) and the center of the trail. Palm detection was assumed certain for palms at zero distance (black-outlined palm). A systematic sample of evenly-spaced palms (darker green) was selected for triatomine-bug sampling in each transect. The right-hand-side picture illustrates bug searches, which included one direct catch on the palm plus a second, independent catch in palm debris collected in a bag; the white arrow points to a palm-fruit cluster. Palm-trail distances, bug detections/non-detections and repeated bug counts were used for descriptive/exploratory analyses and for inferential statistical modeling–in which we estimated covariate-adjusted landscape-disturbance effects after taking palm and bug detection failures into account. Joint analyses of palm and bug density allowed us to derive estimates of bug density per unit area in three landscape classes–old-growth forest, cattle pasture, and young secondary forest. We thus were able to make inferences about whether and how anthropogenic landscape disturbance is associated with three putative proxies for Chagas disease risk–palm density, palm infestation, and vector density.

Fig 2. Attalea palm detections.

The histograms show the relative frequencies of adult Attalea-palm detections (vertical axes, all scaled to a 30% maximum) as a function of perpendicular distance (horizontal axes; 0–40 m) from linear trails established in 10 transects representing three landscape classes in two eastern-Amazonian localities (“L1” and “L2”). The upper row shows overall and locality-specific results. Landscape class-specific results (middle row) show how detections clustered at shorter distances in young secondary forest (“YSF”), but were spread over all distances in deforested cattle pasture (“CP”); the pattern was somewhat intermediate for old-growth forest (“OGF”). The lower row shows results stratified by landscape class and locality. The number of palm detections used to build each graph is given in parentheses.

Fig 3. Estimated decline of Attalea palm detection probabilities as a function of distance.

Results are shown by locality (“L1” and “L2”) and by landscape class (“OGF”, old growth forest; “CP”, cattle pasture; “YSF”, young secondary forest) within localities. Gray bars are overall palm-detection frequencies.

Fig 4. Effects of deforestation on adult Attalea palm occupancy (or infestation) by Rhodnius spp.

(A) and on the number of Rhodnius spp. per palm (B). The graphs show model-predicted probabilities of palm occupancy (A) and model-predicted mean number of bugs per palm (B) for three landscape classes in two localities (“L1” and “L2”) of eastern Amazonia. “Avg.” values (diamonds) are average predictions derived from simplified models without locality effects. All predictions are for “typical” palms with mean stem height (4.75 m) and mean organic score (0.73 units); note that the y-axes are on log10 scale. Error bars are 95% confidence intervals.

Fig 5. Effects of anthropogenic landscape disturbance on Attalea and Rhodnius spp.: A summary of results and hypotheses.

The original old-growth forest (green box) is our “baseline” or “reference class”, and has therefore 1.0 values for all putative risk proxies (palms per hectare, palm infestation, and bug density per palm and per hectare; we could not quantify disturbance effects on vector-infection frequency). The right-hand-side grey box (“Stage I”) summarizes changes associated with transformation of old-growth forest into cattle pasture; note that some palms are usually spared during clearcut (hence the asterisk). The orange box summarizes changes, relative to old-growth forest, of our putative risk proxies. Cattle pasture is maintained through slash-and-burn; when abandoned, it recovers (“Stage II” box) into secondary forest. The yellow box again summarizes changes, relative to old-growth forest, of our putative risk proxies. If left undisturbed, secondary forests mature (“Stage III” box) into old-growth forests and the cycle is eventually completed. Note that the grey boxes show hypothetical processes (involving availability of arboreal habitats, vertebrate-host responses, and food-web features) that we suggest should be the focus of future research; the green, orange, and yellow boxes summarize the main results (presented as ratios–relative to old-growth forest) of the present study. See S1 Text for details.