Spatial Ecology Applied to Identify Differential Uses of Habitat Types and Hotspots of Amazon River Dolphin Activity

Context: Identifying the heterogeneity in the habitat use and hotspots of Amazon River dolphin activity is essential to design effective strategies for the management and conservation of these cetaceans and their habitats in the Colombian Amazonas River and Orinoco basin. Objectives: Quantify the differential use of habitat that Amazon River dolphins exhibit and to identify hotspot activity during seasons of rising waters in the Colombian Amazonas River and other major rivers in the Orinoco basin. Methods: Based on processed satellite images Landsat 8 (2018-2022), we classied habitat types used by I. geoffrensis and reported in the literature: (1) main river, (2) conuences, (3) tributaries, (4) channels, (5) islands, (6) bays, and (7) lagoons. We combined this dataset with GPS location data obtained from 17 tagged Amazon River dolphins to quantify the proportion of habitat types used, and we used GAMLSS to explore the relationship between the number of locations per habitat as the predictor variable and nine response variables. Kernel density estimate (KDE) analysis was used to identify both areas used (K 95 ) and hotspot activity associated with core areas (K 50 ) in the habitat types for the individuals monitored in the Amazonas River and Orinoco basin. Results: Satellite tracking of I. geoffrensis individuals reported 16.098 locations (Amazonas River, n = 2.934 locations and Orinoco basin, n = 13.164 locations) classied into seven habitat types. For the analyzed period (rising waters) main habitat types used were as follow: in the Amazonas River (1) main river (n = 1.346, 46%), and (2) lagoons (n = 1.158, 39%); and in the Orinoco basin (1) main river (n = 7.798, 59.2%), and (2) channels (n = 1.535, 11.7%). The best tting GAMLSS for the I. geoffrensis monitored individuals in the Amazonas River showed that body length, sex, and habitat types (lagoons and main river) were signicant predictors of Amazon River dolphins; and in the Orinoco basin they showed that body length, longest distance, K 50 , and habitat types (conuences, lagoons, main river, and tributaries) were signicant. Individuals monitored in the areas evaluated in the Amazonas River reported K 95 ranged: 30.7–105.5 km 2 (mean = 62 ± 32), and the K 50 or hotspots of activity were recorded in the main river, conuences, lagoons, and tributaries. In the ve rivers of the Orinoco basin, monitored individuals registered K 95 ranged: 6–116 km 2 (mean = 32.8 ± 31.53), and hotspots of activity were recorded in the main river, islands, channels, conuences, bays, tributaries, and lagoons. Conclusions: Evaluation of differences in habitat type use and identication of hotspot activity will contribute to improving management and conservation efforts for the populations of these cetaceans in the Colombian Amazonas and Orinoco basins.


Introduction
The space animals use to live and reproduce, commonly known as the 'home range', is considered a fundamental metric in animal ecology ( Home range studies have been focused primarily on terrestrial species, analyzing habitat selection and use; and these studies have allowed us to gain important insights into the spatiotemporal distribution of resources and the use of core area and hotspot sizes (Relyea et al. 2000;Di Bitetti 2001). The intrinsic di culties in analyzing the spatial ecology of aquatic fauna have prevented us from understanding those factors determining the spatial-temporal patterns of resource availability and how they are used by aquatic animal communities (Mosquera-Guerra et al. 2021). This is particularly true for riverine ecosystems that are characterized by greater dynamics and are in uenced by ood pulses and other climatic, geographic, and ecological determinants (Barthem et al. 2017).
A major challenge for aquatic species is to e ciently, continuously, and quanti ably monitor behaviors In the current context, biodiversity in freshwater ecosystems is in rapid decline. It is considered even more threatened compared to that of terrestrial and marine ecosystems (Revenga et al. 2000;Vörösmarty et al. 2010). Decline in many large mammal migratory populations on regional and global scales has been reported (Berger 2004 In this paper, we identify the types of hábitats used by 17 Amazon River dolphins using deployed satellite tracking transmitters receiving individual locations. Locations reported by the monitored individuals were classi ed by habitat types, allowing for the quanti cation of habitat use and for determining the core areas or hotspots of activity and their size. The identi cation of hotspots can offer a reference for scienti cally de ning prioritized areas for conservation, an aspect that is fundamental when allocating limited resources to the process of ecosystem management (Reyers et al. 2009;Zhang and Fu 2014).

Study area
We conducted this study from October 2017 to May 2021 across six Colombian rivers located in the following localities: (1) Amazonas, (2) Arauca, (3) Bita, (4) Guayabero, (5) Guaviare, and (6) Orinoco (see Fig. 1a-f). Transects evaluated in the rivers where the different habitat types were classi ed covered 247 km 2 for the Amazonas River and 428 km 2 for the Orinoco basin. Transects were de ned through the Amazon River dolphin home range sizes, satellite-monitored in the following localities: (1) Amazonas River, from the con uence of the Atacuari and the Amazonas River in Colombia to Zaragoza Creek, including the Tarapoto wetland complex, and the con uence of the Loretoyacu and the Amazonas rivers (see Fig. 1a); (2) Orinoco River, from the mid-basin of the Orinoco River in Colombia, including its con uences with the Bita and Meta rivers, Negro Creek, El Pañuelo Island, and San José, Angela, and Guaripa creeks (see Fig. 1b); (3) Arauca River, from mid basin of the Arauca River, including El Martinero Creek, to Jesús Creek (see Fig. 1c); (iv) Bita River, from the lower basin of the Bita River from La Pinzonera Creek to the con uence of the Bita and Orinoco rivers and El Pañuelo Island in the Orinoco River (see Fig.   1d); (v) Guayabero River, from Temblón Creek, including Yarumales Creek to Yuca Creek (see Fig. 1e), and (vi) Guavire River, from La Laguna Ciega, including the con uence of the Guaviare and Inírida rivers, to El Carbón Creek (see Fig. 1f).

Location data
River dolphin tagging was carried out during the rising water season for the Colombian Amazonas River and Orinoco basins. A total of 17 ( = 8 and = 9) Amazon River dolphins were tagged with SPOT-299A and SPOT6-F single-point n mounted satellite tags (Wildlife Computers, Redmond, WA). Only adult individuals were selected for tagging, and their age class was estimated based on body length, following the methods of da Silva (2009) and Martin and da  and avoiding females with calves (see Table S1). The tags were programmed for continuous transmission (24 hours) up to 250 times a day from an ARGOS satellite constellation, providing location and tag status data, and yielding up to 240 tracking days based on battery life (Mosquera-Guerra et al. 2021).
Locations were classi ed by the ARGOS system into one of six location classes (LCs). Based on the level of accuracy measured in kilometers of uncertainty for latitude and longitude. ARGOS classi es location quality relative to an estimated error radius in the following location classes: 3 (accurate to <250 m), 2 (accurate to 250-500 m), 1 (accurate to 500-1500 m), and A and B (1-2 messages received but no accuracy estimation). In our study, we used only the most accurate data, LCs 3 and 2, after ltering the Habitat type spatial classi cations A set of satellite images of the study areas during the periods of satellite monitoring of Amazon River dolphins were downloaded from the USGS (The United States Geological Survey) platform and provided by the Landsat 8 satellite (USGS 2021). The satellite images used had cloudiness that was lower than 30% and had su cient resolution for the identi cation of the different types of habitats used by river dolphins, such as (1) main river, (2) con uences, (3) tributaries, (4) channels, (5) islands, (6) bays, and (7) lagoons, as proposed by Gómez-Salazar et al. (2011a); see Table 1). The satellite images were processed using ArcGIS 10.8.1 software; and two combinations were selected to exhibit the spatial dynamics of the river stretches evaluated, such as (1)   Main river Whitewater rivers of Andean and Guianan shield origin, typically turbid, brown-yellow in color with low transparency, basic pH, and sedimentrich (Sioli 1984). At least 400 m in width.
Amazonas and Orinoco rivers.
Con uences Intersection areas of the main channel with other channels or rivers. Con uences maintain connections during all hydrologic seasons and may or may not show a mix of white and blackwaters. Approximately 250 m width.

Tributaries
Small and medium size rivers no more than 400 m in width. Water in tributaries is usually black and clear, originate from the ooded forest plains with few suspended sediments and relatively acidic, high in tannins and particulate organic matter (Sioli 1984).

Channels
Watercourses not exceding 300 m wide and generally associated with island and main river systems. Navigability is limited depending on the rainy seasons.
Channels from Amazonas River to the Caballo Cocha Lake, between the El Correo and Tarapoto lakes, and Cacao Island. Channels from Orinoco River beteween the con uence with the Bita River to the Ventanas Rapids, and nally from the Guaviare River to Carbon Creek.

Islands
Waters around land bodies in the watercourse of main rivers with vegetation that may appear or disappear due to hydrologic dynamics.
Bugeo, Cacao, Mocagua, Patrullero, San Salvador, and Vamos Islands in the Amazonas River. El Pañuelo and Ventanas in the Orinoco River.

Bays
Meanders or curves along the main river course with longer water retention times than the main channel and lower current velocity.
All sections in the assessed rivers had bays.

Home range and hotspot activity
We estimated the home range and core areas were calculated for monitored individuals using the kernel density estimate ( The GAMLSS approach is a semi-parametric method that allows exploring the relationship between the predictor variables and the response variables to be modeled either parametrically with linear or nonlinear predictors or non-parametrically with smooth nonparametric terms (e.g., cubic splines or loess) and random effects (
The best-tting GAMLSS for the I. geoffrensis monitored individuals in the Amazonas River showed that body length, sex, and habitat types (lagoons and main river) were signi cant predictors of Amazon River dolphin habitat use (see Tables 3 and 4). The model predicted that habitat use of river dolphins depends on the body length and sex of the individuals and the speci c use of habitat types such as lagoons and the main river (see Fig. 6a-c). Note: Those signi cant for the explanatory predictors are marked with an asterisk (*). The best-tting GAMLSS for the I. geoffrensis monitored individuals in the Orinoco basin showed that body length, longest distance, K 50 , and habitat type (con uences, lagoons, main river, and tributaries) were signi cant predictors of Amazon River dolphin habitat uses (see Tables 5 and 6). The model predicted that habitat use of river dolphins depends on the longest distance and core area (K 50 ) of the individuals as well as the speci c use of habitat types such as con uences, lagoons, main river, and tributaries (see Fig. 7a-c). Table 5 GAMLSS models for the I. geoffrensis monitored individuals in the Orinoco basins with df (degree of freedom) and GAIC for the response variable numbered locations per habitat types and explanatory predictors as follow: (1) body length (cm), (2) body mass (kg), (3) sex (F: female and M: male), (4) tracking duration (days), (5) habitat types, (6) K 95 (km 2 ), (7) K 50 (km 2 ), (8) longest distance between locations (km), and (9)  Home range and hotspot activity Home range size (K 95 ) for the four ( = 1; = 3) I. geoffrensis individuals monitored in the section evaluated in the Amazonas River ranged from 30.7-105.5 km 2 (mean = 62 ± 32), and the largest core areas (K 50 ) or hotspots of activity were recorded in the main river, con uences, lagoons, and tributaries (see Table 7). In ve rivers of the Orinoco basin, 13 monitored individuals ( = 7; = 6) registered K 95 : 6-116 km 2 (mean = 32.8 ± 31.53), and the largest core areas (K 50 ) or hotspots of activity were recorded in the main river, islands, channels, con uences, bays, tributaries, and lagoons (see Table 7). Table 7 Range (mean ± SE) of the size of hotspots of Amazon River dolphin activity or core area ( Arauca River monitored a female of I. geoffrensis (AR 181018 F) and her two central areas of use were associated with the main river in the vicinity of the border city of Arauca (see Fig. 8c). In the Bita River the female (BI 181017 F) had a hotspot in the Orinoco River in the vicinity of the Bojonawi Nature Reserve (see Fig. 8d). In the Guayabero River, two males were monitored (see Fig. 8e) and one individual (GYO the con uence of the habitat type between Guayabero River with the Yarumales and Claro creeks and Guaviare River. The female (GVE 181016 F) had a hotspot in the tributary in the vicinity of the con uence between the Guaviare and Inirida rivers (see Fig. 8f). Our results show strong sexual segregation for habitat use among satellite-monitored individuals in the Amazonas River. The documented differential behaviors in the intensity of habitat use between males and females coincide with observations made by Trujillo (2000) in the lakes of Tarapoto and the Colombian Amazonas and Martin and da Silva (1998; who analyzed data from 24 individuals monitored with radio telemetry in the Mamiraua Sustainable Development Reserve. They report that during the period of rising water adult females and calves use lagoons, oodplain and channels more intensively due to factors such as distribution patterns and abundance of small prey (gymnotids, cichlids, and cat sh) associated with oating vegetation (Crampton 1999 Hot spot activity size may be explained by the fact, that some individuals may remain residents in relatively small areas (Bjørge 2002). Although these individuals generally have well-de ned home ranges, their distribution and the use of space within these home ranges refers to individual movements rather than to an entire population (Morales et al. 2004;2010). This is particularly true in ssion-fusion societies, such as those seen in many cetacean species including Amazon River dolphins (Gómez et al.

Conclusions
This study demonstrated the usefulness of special ecological analyses in determining the intensity of habitat use, home range, and the identi cation of hotspots of Amazon River dolphin activity. This information contributes in a relevant way to the construction of conservation strategies for the species and its habitats, such as the designation of three wetlands of international importance or Ramsar sites in Colombia: (1) Tarapoto wetland complex in the Amazonas River, (2) Estrella Fluvial de Inirida, and (3) Bita River in the Orinoquia. Additionally, I. geoffrensis as an ecological indicator of freshwater ecosystems was prioritized in the management plans of Ramsar sites through actions such as conservation agreements for sustainable shing with local communities and responsible observation programs for these cetaceans.
We highlight the in uence of the sexual condition of the monitored individuals, ecological dynamics of prey such as abundance and movements, and the ood pulse on the intensity of use of the different habitats occupied by Amazon River dolphins. We also underline the vulnerability of I. geoffrensis to drives such as (1) the regulation of the ood pulses by climatic variability events or the construction of hydroelectric dams in the upper parts of the watersheds hindering river dolphin access to lagoons and tributaries during the low water period and reducing the extent of oodplains during the high water period; (2) sedimentation of channels, bays, and tributaries resulting from altered uvial geomorphology due to bank erosion; and (3)    Smoother estimates for the predictors (a) body length, (b) sex (F: female and M: male), and (c) habitat types, obtained by the GAMLSS predicting the probability of I. geoffrensis individual habitat use in the Amazonas River. Shaded circles represent partial residuals; shaded areas, 95% con dence intervals around mean partial effect. Shaded circles on the x-axis are sampled data points Figure 7 Smoother estimates for the predictors (a) longest distance (km), (b) K50 (km2), and (c) habitat types, obtained by the GAMLSS predicting the probability of I. geoffrensis individuals habitat use in the Orinoco basin. Shaded circles represent partial residuals; shaded areas, 95% con dence intervals around mean partial effect. Shaded circles on the x-axis are sampled data points