Fish sampling and physicochemical data for aquatic habitats in the Santarem region of the Lower Amazon, Brazil

Explaining the mechanisms underlying spatial and temporal variation in community composition is a major challenge. Nevertheless, the processes controlling temporal variation at a site (i.e., temporal β-diversity, including its turnover and nestedness components) are less understood than those affecting variation among sites (i.e., spatial β-diversity). Short-term temporal turnover (e.g., throughout an annual cycle) is expected to correlate positively with seasonal environmental variability and landscape connectivity, but also species pool size (γ-diversity). We use the megadiverse Amazonian freshwater ichthyofauna as a model to ask whether seasonality and landscape connectivity drive variation in temporal species turnover among geomorphological habitat types, while accounting for between-habitat variation in γ-diversity. 11,397 fish representing 260 species were collected during a year-long sampling program from an area containing the lowland Amazon’s four major geomorphological habitat types: rivers, floodplains, terra firme streams, and shield streams. River-floodplain systems exhibit strong but predictable seasonality (via a high-amplitude annual flood pulse), high connectivity, and high species richness with many rare species. Terra firme and shield streams exhibit low seasonality, low connectivity, and low species richness with proportionally fewer rare species. Based on these parameters we predicted that river-floodplain systems should have higher temporal turnover than stream systems. Using a null model approach combined with β-deviation calculations, we confirmed that rivers and floodplains do exhibit higher turnover (but not nestedness) than terra firme and shield streams, even when controlling for the potentially confounding effect of higher species richness in river-floodplain systems. All habitats exhibit low temporal nestedness, indicating that short-term changes in community composition result primarily from temporal species turnover. Our results provide a timely reminder that efforts to conserve the Amazon’s threatened aquatic biodiversity should account for the distinct temporal dynamics of habitat types and variation in hydrological seasonality.

Methods
We studied one of the few Amazonian regions where all four geomorphological habitats occur in close proximity (near Santarém, Brazil, Fig. 1d). This lower Amazon region also includes river-floodplain systems belonging to the three distinct biogeochemical water types of the Amazon basin: nutrient-poor, humic-stained ‘blackwaters’  (BW); nutrient-poor ‘clearwaters’ (CW); and nutrient rich, sediment-laden ‘whitewaters’  (WW) (Sioli 1984, Bogotá-Gregory et al., 2020). We estimated temporal β-diversity for rivers-floodplain systems in all three water types, allowing us to evaluate whether water type, as well as geomorphological habitat type, influence rates of temporal turnover and nestedness.

We sampled 16 sites (Fig. 1d) every two months through a complete annual hydrological cycle, beginning October 2014: Rivers: two river-margin sites in the BW River (R.) Arapiuns (BR1, BR2), two in the CW R. Tapajós (CR1, CR2), and two in the WW R. Amazonas (WR1, WR2); Floodplains: two floodplain lake sites in the BW R. Arapiuns (BF1, BF2), two in the CW R. Tapajós (CF1, CF2), and two in the WW R. Amazonas (WF1, WF2); Terra firme streams: two sites in the lowland terra firme peneplain (Cretaceous Alter do Chão formation) (TS1, TS2); Shield streams: two sites in the upland Paleozoic shield (Devonian Ererê formation) (SS1, SS2). Hence our sampling design comprised eight geomorphological habitat/water type combinations, each replicated at two sites, and sampled six times through the year for a planned total of 96 sampling events. TS1 and TS2 were sampled only five times due to logistical challenges during the first sampling event, and BF2 and WR2 were sampled only five times due to severe weather at the first event for BF2 and fifth for WR2. Thus, the actual number of sampling events was 92.
The river-floodplain habitats exhibited a flood cycle of ca. 5 m amplitude, with an April-July high-water period and October-December low-water period. Year-long variation in water levels in the shield and terra firme streams is < 0.5 m (except briefly after heavy rain) in response to local rainfall – with a December-June wet season followed by a dry season. We selected permanent floodplain lakes of similar size (BF1, 0.17 km2; BF2, 0.25 km2; CF1, 0.18 km2; CF2, 0.21 km2; WF1, 0.18 km2, WF2, 0.26 km2; Fig. 1d) and permanent terra firme and shield streams of similar sizes (3-6 m wide, to 1 m deep). All 16 sites were surrounded by relatively well-protected natural forest.

Due to habitat heterogeneity, it was impossible to use the same gear and sample technique in all habitat types. Therefore, for each habitat we used the gear considered to obtain a representative sampling and then standardized for effort within a given habitat type across all sampling events in the annual cycle. To maximize sampled species diversity in river and floodplain sites, we sampled with gill nets deployed ca. 30 m from the shoreline in batteries of four (25 x 3 m, 15, 30, 45, 60 mm mesh) from 6 am to 9 am and from 6 pm to 9 pm. In the much smaller habitat volumes of terra firme and shield streams, we used three gear types, each with a timed effort of 2 hours/sampling event during daylight hours: a 1.5 x 6 m seine net (5 mm mesh) in deeper pools; a 2.0 x 1.1 m bag-seine (2 mm mesh) in riffles and margins; a dipnet (Expedition Hex-Trapnet, Duraframe Dipnet, Viola WI, with 3 mm mesh and 30 cm bag depth), used with the aid of an electric fish finder (Haag et al. 2019), in leaf litter and marginal root mats; the dipnet-electric fish finder combination sampled gymnotiform electric fish and other nocturnally active fish hiding in the substrate. For each stream, all gear types were used along a 100 m stretch in a downstream-upstream direction with the ends of the section blocked with net panels to minimize escape. Fish were euthanized with 600 mg L-1 eugenol, preserved, and deposited at the biodiversity collections listed in Supporting Information Appendix S1. We identified specimens using species descriptions and keys from the taxonomic literature. 

Measuremens of physicochemical water parameters follow: 
Bogotá-Gregory, J.D., Lima, F.C.T., Correa, S.B., Silva-Oliveira, C., Jenkins, D., Ribeiro, F.R., Lovejoy, N.R., Reis, R.E. & Crampton, W.G.R. (2020) Biogeochemical water type influences community composition, species richness, and biomass in megadiverse Amazonian fish assemblages. Scientific Reports, 10, 15349.

解释群落组成时空变异的内在机制是生态学领域的核心挑战之一。尽管如此，相较于调控生境间变异的过程（即空间β多样性（spatial β-diversity）），人们对单一位点的时间变异调控过程（即时间β多样性（temporal β-diversity），包含其周转（turnover）与嵌套性（nestedness）组分）的认知仍较为匮乏。短期时间物种周转（如年度周期内的周转）被预测与季节环境变异、景观连通性呈正相关，同时也与物种库大小（γ多样性（γ-diversity））相关。本研究以物种极丰富的亚马逊淡水鱼类区系为模式系统，旨在探究季节动态与景观连通性是否驱动不同地貌生境类型间的时间物种周转差异，同时控制生境间γ多样性的潜在差异。

本研究在为期一年的采样计划中，于包含亚马逊低地四大典型地貌生境类型的区域采集到11397尾鱼类，隶属于260个物种：河流、泛滥平原、未淹水陆地溪流与盾地溪流。河-泛滥平原系统具有强烈且可预测的季节动态（依托高振幅的年度洪水脉冲）、较高的景观连通性，以及物种丰富度高且稀有种较多的群落特征；未淹水陆地溪流与盾地溪流则表现为季节变异弱、连通性低、物种丰富度低且稀有种占比更少的特点。基于上述参数，我们预测河-泛滥平原系统的时间周转速率应高于溪流系统。通过结合零模型（null model）方法与β偏差（β-deviation）计算，本研究证实：即便控制河-泛滥平原系统更高的物种丰富度这一潜在混杂效应，河流与泛滥平原的时间周转速率仍显著高于未淹水陆地溪流与盾地溪流（但嵌套性无显著差异）。所有生境均表现出较低的时间嵌套性，表明群落组成的短期变化主要由时间物种周转驱动。本研究结果及时提醒相关保护工作：针对亚马逊受威胁水生生物多样性的保护举措，应充分考虑不同生境类型独特的时间动态特征与水文季节变异规律。

## 研究方法
本研究选取亚马逊流域少数四大地貌生境毗邻分布的区域（巴西圣塔伦附近，图1d）开展研究。该下亚马逊区域同时涵盖属于亚马逊盆地三类典型生物地球化学水型的河-泛滥平原系统：营养匮乏、腐殖质染色的“黑水（blackwaters）”、营养匮乏的“清水（clearwaters）”，以及营养丰富、携带有大量沉积物的“白水（whitewaters）”（Sioli 1984, Bogotá-Gregory et al., 2020）。我们针对三类水型的河-泛滥平原系统分别估算时间β多样性，以此评估水型与地貌生境类型是否共同影响时间周转与嵌套性的速率。

本研究在完整的年度水文周期内（2014年10月启动），每两个月对16个采样点（图1d）开展一次采样：
1. 河流生境：在黑水河流阿拉皮昂斯河（R. Arapiuns）设置2个河岸样点（BR1、BR2），在清水河流塔帕若斯河（R. Tapajós）设置2个河岸样点（CR1、CR2），在白水河流亚马逊河（R. Amazonas）设置2个河岸样点（WR1、WR2）；
2. 泛滥平原生境：在阿拉皮昂斯河设置2个泛滥平原湖泊样点（BF1、BF2），在塔帕若斯河设置2个泛滥平原湖泊样点（CF1、CF2），在亚马逊河设置2个泛滥平原湖泊样点（WF1、WF2）；
3. 未淹水陆地溪流：在低地未淹水陆地准平原（白垩纪阿尔特杜尚组）设置2个样点（TS1、TS2）；
4. 盾地溪流：在古生代盾地高地（泥盆纪埃雷雷组）设置2个样点（SS1、SS2）。

本研究的采样设计共包含8种地貌生境-水型组合，每种组合设置2个重复样点，全年共计采样6次，计划总采样事件数为96次。但受后勤限制，TS1与TS2仅完成5次采样；BF2与WR2分别因首次采样遭遇恶劣天气、第五次采样遭遇恶劣天气，仅完成5次采样。最终实际有效采样事件为92次。

河-泛滥平原生境的洪水周期振幅约为5米，高水位期为4-7月，低水位期为10-12月。盾地溪流与未淹水陆地溪流的水位年变化幅度小于0.5米（强降雨后短暂例外），响应局地降雨规律：12-6月为湿季，其余时段为干季。

本研究选取的泛滥平原湖泊规模相近（BF1：0.17 km²；BF2：0.25 km²；CF1：0.18 km²；CF2：0.21 km²；WF1：0.18 km²；WF2：0.26 km²；图1d），未淹水陆地溪流与盾地溪流的规模也较为一致（宽度3-6米，最大深度1米）。所有16个采样点均被保护良好的天然森林环绕。

由于不同生境存在异质性，无法在所有生境中使用统一的采样装备与方法。因此，本研究针对每类生境选用能代表性覆盖其物种多样性的采样装备，并在年度周期内的所有采样事件中，统一同一生境类型内的采样努力量。

为最大化河流与泛滥平原生境的采样物种多样性，我们采用刺网（gill nets）采样：在距离岸线约30米处布设4张一组的刺网（规格25×3 m，网目尺寸15、30、45、60 mm），采样时段为早6:00-9:00与晚6:00-9:00。针对体积更小的未淹水陆地溪流与盾地溪流，我们采用三类采样装备，每类装备在日间的采样时长为2小时/次采样事件：① 1.5×6 m围网（seine net，网目5 mm）用于较深的水潭；② 2.0×1.1 m袋形拉网（bag-seine，网目2 mm）用于急流与岸带区域；③ 抄网（Expedition Hex-Trapnet，Duraframe Dipnet，美国Viola WI公司，网目3 mm，网袋深度30 cm）配合电鱼探测器（electric fish finder，Haag et al. 2019）用于落叶层与岸带根系垫区域，该组合可有效采样裸背电鳗目鱼类与其他隐匿于底质中的夜行性鱼类。对于每条溪流，所有采样装备均沿100米的河段上下游方向布设，并在河段两端设置网栏以减少鱼类逃逸。

鱼类样本采用600 mg L⁻¹丁香油酚（eugenol）进行麻醉致死，固定后保存，并存放于支持信息附录S1中列出的生物多样性馆藏机构。本研究依据分类学文献中的物种描述与检索表对标本进行鉴定。

水体理化参数的测量方法参见：
Bogotá-Gregory, J.D., Lima, F.C.T., Correa, S.B., Silva-Oliveira, C., Jenkins, D., Ribeiro, F.R., Lovejoy, N.R., Reis, R.E. & Crampton, W.G.R. (2020) 生物地球化学水型对亚马逊物种极丰富鱼类群落的组成、物种丰富度与生物量的影响. 《科学报告》, 10, 15349.