Exploring Connectivity Between Amazonian Forest Types

Habitat Connectivity

Globally, conservationists emphasise the importance of connectivity between habitats.

“Particular emphasis is placed on the creation of ecological corridors connecting existing protected areas, thereby avoiding the risk of fragmentation. This greatly assists the effectiveness of the reserves by enabling animal species to seasonally migrate from one area to another and for natural seed dispersal to continue.” – Rainforest Concern

In this context emphasis is often placed on connectivity being essential for the survival of those migratory species, often large wide-ranging carnivores such as jaguars (Photo 1). In Amazonia, it is possible that inter-habitat movement is essential for the integrity of the forest itself through another invisible mechanism; functional connectivity.

The Amazonian flood pulse

Annually, Amazonian lowland forests (várzeas) are transformed by the yearly flood pulse (Photo 2). Across the várzea, this flood deposits minerals (e.g. phosphorus) washed down from the Andes. These fertilise várzea soils and are incorporated into várzea plants. Functional connectivity is the hypothetical movement of these minerals and additional nutrients from várzea uphill into the never-flooding terra firme forest, carried and deposited by animals that consume them in the várzea. Absorbed into the terra firme soil, they would then boost terra firme plant growth. Terra firme soils are generally depleted, so functional connectivity could be essential for what limited soil fertility there is.

My study uses camera trap data on mammal distributions in pristine várzea and terra firme to infer the seasonal movement of mammals between forest types. Thus I suggest species that may be key functional connectivity contributors.

Camera traps

Camera traps (CTs) automatically take photo(s)/video when an animal is in view (Photo 3), using infra-red illumination at night (Photo 4). This makes them extremely useful in Amazonian ecology studies; Amazonian forests are some of the densest and most impenetrable jungle on Earth, so deploying CTs to passively work is far preferable to manually surveying (on top of the fact that camera traps disrupt animal behaviour much less than people). In my project each CT was fixed to a tree (Photo 5) and retrieved two or three months later (Photo 6), its contents then downloaded.

I deployed my CTs in a 5×5 grid in each forest type with 500m between adjacent CTs (Figure 1). This ensured sampling repeatability and comparability, the many CTs of each grid helping mitigate biases within each environment. From late-July to early-November, just before the 2024 annual flood, the terra firme grid (TFG) and várzea grid (VZG) were deployed. From April to early-June 2025 I redeployed the terra firme grid (Wet TFG; WTFG) so as to sample the terra firme before the floodwaters started receding. I also deployed some CTs at tree trunks that had fallen over streams or narrow valleys to record animals using them as bridges (Video 1, Video 2), although this material was not used in my quantitative analyses.

River channels & lakes in black. Terra firme cameras - yellow points, várzea cameras - blue points. Nearest large settlement is Carauari, approx 45km south-west of the várzea grid.

A collaborative project

Amazonian fieldwork is impossible without local collaborators. I was extremely lucky to have the logistical support of the brilliant Dr Andressa Scabin, Hugo Costa, Almira Nascimento and Julian Santiago at Instituto Juruá and fantastic local field assistants; Antonio-José, Antonio, Maciel and Joachim of Juruá Communities São João and Lago Serrado.

Some of my local friends spoke about discontent against Western scientists who arrive in Amazonia, collect data and leave as soon as possible; with little regard for local custom or formal acknowledgement of contribution, barely engaging beyond greetings and gestures. In the scientific community this is termed “Parachute Science”. My colleagues and I completely disagree with this way of working. I think that any who have the immense privilege to work in foreign enviornments must do better. Therefore I taught myself a firm basis in Brazilian Portuguese, and to learn about local culture and tangibly express my gratitude I spent several weeks between CT deployments living in the Juruá forest community of Pupuaii with my dear friend Akilles, working on his manioc plantations and subsistence fishing into the night (Photos 7 & 8). This was physically demanding but immensely rewarding; my friends proudly taught me about their ways of life and relationship with the forest. I was greatly moved by how warm and inclusive the communities were. It was an extraordinary privilege to be welcomed with so much kindness and warmth; I am both lucky and grateful. Despite the Juruá being a hotbed of Amazonian ecological research, community members said that no foreigner had done this before.

Processing camera trap data

The most time consuming part of categorising CT photographs was thoroughly confirming that “Blank” photos truly were blank, which took weeks. This could be challenging (Figure 2), but I was helped by my CTs photographing in bursts; flicking between consecutive photographs helped identify changes and therefore locate trap-triggering creatures.

Once I categorised each of my 43720 images, I extracted a detection dataframe. For each CT, all photos of a species within 30 mins of that species’ first image were recorded as one detection to account for some animals spending longer in front of CTs and thereby triggering them multiple times in quick succession.

Species-level analyses

From my dataframe I calculated each species’ detection rates in TFG, VZG and WTFG, transforming these to get a comparable index of relative abundance for each species in each grid. Using detection rate as the basis for a relative abundance index is controversial, so I also used Single Species Occupancy Models (SSOMs) to estimate occupancy probabilities for species of suspected functional connectivity relevance or without significant differences in detection rates between grids.

For all but seven of the species detected, detection rate differed significantly between TFG and VZG (Figure 3), indicating that most had a ‘habitat preference’ in my pre-flood sampling window. Of the seven exceptions Coati, Common opossum, Spiny rat and Jaguar had enough data for SSOMs. None had a statistically significant difference in occupancy between forest types, complementing the abundance index comparison; for these species my data does not conclusively show a habitat preference. From this we can infer that throughout the low water phase some individuals of these species move between forest types, likely carrying minerals and nutrients with them. Any still inhabiting várzea as the low water season ends are also likely pushed back to terra firme by the rising waters.

Species detected more frequently in terra firme below dashed line; those more frequently detected in várzea above. Point labels correspond to the "ID" column. In linear regression species highlighted green or blue had significantly higher detection rate by TFG or VZG cameras respectively, grey-highlighted species had no significant difference in detection rate between forest types.

Both Amazonian big cat species, jaguars and pumas (Photo 9), were detected in both forest types. All primates on the forest floor had significantly higher detection rates in várzea. Of the strictly terrestrial mammals, only White lipped peccaries (WLPs) had a significantly higher detection rate in VZG. SSOM confirmed a parallel difference in WLP TFG/VZG occupancy. As they are fully terrestrial, I infer that in this region of Amazonia WLPs move between terra firme and várzea seasonally, entering várzea for seasonally available food and being forced out months later by rising floodwaters. WLPs weigh up to 50kg, can travel in 400-strong herds and carry large quantities of biomass in their gut contents; these factors and their inferred inter-forest movement make them the prime candidate to transport meaningful quantities of nutrients and minerals to terra firme from várzea, greatly contributing to functional connectivity.

A paired Wilcoxon signed-rank test was used to compare detection rates between TFG CT stations and their WTFG repeats for each species, to test if relative abundance in terra firme changed with the change in water level. Although in terra firme relative abundances for some mammals were higher in the high-water season than the low-water season (Figure 4), for no species was this difference statistically significant. This indicates that terra firme terrestrial mammal assemblages are relatively stable year-round, contrary to the expectation that terra firme relative abundances would increase with rising floodwaters as terrestrial animals are forced from várzea. WLPs not having a significantly higher detection rate in terra firme in the wet season over the dry season suggests that after being pushed from the várzea they migrated further into terra firme than WTFG could detect.

Species more frequently detected in WTFG than TFG fall above y=x, the inverse below. Point labels correspond to the "ID" column. There were no statistically significant differences in detection rate for any species in terra firme between flood seasons, despite some deviance from y=x.

Community-level analyses

As a measure of biodiversity, Species richness is over-influenced by rare species (Photo 10) and does not incorporate relative abundances. Therefore I use Principle Coordinate Analysis (PCoA) and SIMPER analysis to compare terrestrial mammal communities between grids and determine which species contributed most to overall dissimilarity.

PCoA showed strong differentiation between TFG and VZG communities (Figure 5). The TFG/WTFG ellipses’ near-perfect overlap complements Figure 4 in indicating that terra firme terrestrial mammal assemblage is relatively stable between várzea flood phases, suggesting that most mammal species remain in terra firme through the flood cycle. SIMPER revealed that the greatest contributors to TFG-VZG community dissimilarity were Four eyed opossums, which contributed approximately 15% dissimilarity, Brown capuchins (~12%), Black agouti (~10%) and Common squirrel monkeys (~9%).

Green points represent TFG cameras, Blue WTFG cameras and orange points VZG cameras. Dashed ellipses represent the 95% confidence intervals of each group’s centroid.

Greater than the sum of its parts

My work adds to the body of evidence that flooding and unflooding Amazonian forests are interdependent, but only through some species migrating between them. Of these, WLPs (Photo 11) could majorly contribute in moving fertilising nutrients and minerals from the floodplains into upland forest. This conclusion is supported by published scientific literature.

WLPs are sensitive to habitat fragmentation and overhunting, due to which presumed local extinctions have occurred. If functional connectivity works as hypothesized, the loss of WLPs from Amazonian forests or disruption of their migratory behaviour could have unpredictable and cascading damaging effects through long-term diminishing of terra firme soil fertility. Future research should aim to confirm this and conservation efforts should react accordingly, giving the species the necessary protections to flourish. Protected areas that are majority terra firme should be expanded to include adjacent várzea, supporting the WLPs and other species that straddle the inter-forest-type boundary.

Although this project identifies a possible major animal vector, the movement of minerals and nutrients through food webs is immensely complicated and this study illuminates just part of a much larger picture. For example due to methodological constraints this project focussed entirely on terrestrial mammals, but it should be recognised that other animals such as fishing birds (Photo 12) likely also move minerals and nutrients between forest types as the hydrology of the landscape changes. I believe that we will never fully understand the glorious complexity of the forest, but the key mechanisms of this ancient process must be understood if we are to best protect it.

Impact

The project and data referred to in this blog has not yet been published in a scientific journal, this is the next step. Any and all publications will fully acknowledge my collaborators’ essential contributions.

I return to Amazonia with Rainforest Concern Founder and Director Peter Bennett in October 2025 and I will present my findings to Juruá community members. On this visit Peter and I will work to grow the partnership that was set up through this work up between Rainforest Concern and Instituto Juruá, focussed on biodiversity conservation and supporting River Juruá communities’ sustainable natural resource management. Rainforest Concern will not dictate the spending of 2026’s possible grant; Instituto Juruá will present ideas for its best use, all decisions and implementation will then be a collaborative process between the NGOs and Juruá community members.

Acknowledgements and thanks

As mentioned, this project was highly collaborative and would have been impossible without the belief, encouragement and support of many people and several organisations.

I thank my supervisor, Prof Carlos Peres, for his support and immeasurable insight.

I thank Dr Andressa Scabin, Hugo Costa, Almira Nascimento and Julian Santiago of Instituto Juruá, and Antonio-Jose, Antonio, Maciel and Joachim of Juruá Communities São João and Lago Serrado, without whom this study’s fieldwork would have been impossible.

For funding support I thank Joan Borinstein and Dr Gary Gartsman, Instituto Juruá Programa de ciências, the University of East Anglia and Rainforest Concern.

Finally I thank my family and my wonderful girlfriend, Rianna, for their love and support through this long and challenging process.

Xavier Tobin, September 2025