The evolution of Amazonian biota and its remarkable biodiversity are intricately linked to the evolution of the physical landscape, and the formation of the transcontinental drainage system is proposed as a main driver for biotic diversification within the Amazonian biome (Hoorn et al., 2010; Ribas et al., 2012; Boubli et al., 2015). In addition to the well-known closed upland rainforest, Amazonia also hosts isolated patches of open vegetation, such as savannas and white-sand ecosystems (WSE) (Anderson, 1981; Adeney et al., 2016), which are considered important environments that drove biodiversification in upland Amazonia (Haffer, 1969) and other ancestral regions (Frasier et al., 2008). Therefore, extended-range chronologies of sedimentary substrates supporting upland vegetation in Amazonia are crucial to understand the evolution of the landscape during the Quaternary and Neogene and its relation to the current biodiversity and biogeographical patterns of the biome. The Pliocene and Quaternary lowland fluvial record in Central Amazonia is composed of heavily weathered siliciclastic and non-fossiliferous deposits, suitable for the application of quartz-based dating techniques (e.g., Soares et al., 2010; Pupim et al., 2019; Bezerra et al., 2022), as feldspar is frequently lacking due to heavy weathering. Most of the reliable ages that constrain sedimentation in these sequences are based on optically stimulated luminescence (OSL) dating of quartz (e.g., Rossetti et al., 2015; Cremon et al., 2016; Pupim et al., 2019) and are restricted to the last 300 ka.

OSL is the light arising from charges trapped in crystal lattice defects when quartz is exposed to ionizing radiation and subsequently released and recombined when quartz is exposed to stimulation light. Measuring the natural OSL signal allows the determination of radiation dose absorbed by quartz during burial, and sediment depositional ages are estimated by dividing the radiation dose (Gy) by the environmental dose rate (Gy/ka). Several extended-range luminescence dating methods were proposed with the potential to date as far as the Early Pleistocene, including thermally transferred optically stimulated luminescence (TT-OSL), with reported equivalent dose (D_e) values up to 2,500 Gy (Wang et al., 2006a). The TT-OSL signal is measured by depleting the OSL signal, applying a heat treatment ≥260°C to transfer charges from TT-OSL traps into OSL traps, and remeasuring the OSL signal (Wang et al., 2006a; Adamiec et al., 2010). The charge transfer has been hypothesized as a double transfer, where electrons from OSL traps are released and stored in a refuge trap, before being thermally transferred back (Wang et al., 2006a) or alternatively as a single transfer, where electrons are thermally transferred from a source trap to an OSL trap (Adamiec et al., 2010). The TT-OSL signal is represented as 1) the recuperated OSL (Re-OSL), which makes up about 90% of the signal and is more sensitive to light, and 2) the basic-transferred OSL (BT-OSL) (Aitken, 1998; Duller and Wintle, 2012), which makes up the remaining 10%. Both multiple-aliquot (MAR) and single-aliquot (SAR) regenerative dose protocols have been proposed to estimate D_e using the TT-OSL signal (e.g., Wang et al., 2006a; Porat et al., 2009; Adamiec et al., 2010; Chapot et al., 2016).

Despite the routine use of TT-OSL dating in recent years, only few studies use this technique in geological settings of South America (e.g., Pupim et al., 2016; Bezerra et al., 2022). In addition to the understating of the evolution of the Amazonian landscape in the Pliocene and Quaternary, TT-OSL dating protocols still need to be tested in different geological contexts, including South America, as quartz OSL properties are not uniform and also vary at a grain-to-grain level (Preusser et al., 2009), and the same might be valid for the TT-OSL signal. In this study, we used both quartz OSL and TT-OSL for dating upland sandy deposits covered by open and closed vegetation in lowland Central and Eastern Amazonia. The formation chronology of these deposits allows us to understand how upland terrains bounding the Amazon River floodplains were assembled in periods beyond the late Pleistocene.

This study used two datasets from Central Amazonia and two from Eastern Amazonia (Figure 1 and Table 1). The Eastern Amazonia dataset comprises outcrop samples from sandy substrates of savannas exposed in pits and road cuttings, for which the depositional and/or weathering origin is still under debate. Eastern Amazonia sandy substrates are composed of massive, loose, medium, and poorly to moderately sorted light reddish to brownish sands (AVA01—Figures 2A,B). The Central Amazonia datasets comprise both outcrop samples from substrates of open vegetation ecosystems (i.e., WSE), composed of massive, loose, and coarse white sands (AVA22—Figures 2C,D), and drill core samples composed of heavily weathered fluvial sandstones and siltstones (Autazes Core: PBAT-15-43, Figures 2E,F). These sand substrates cover Pliocene-Miocene and late Cretaceous-Paleogene fluvial deposits from the Novo Remanso and Alter do Chão Formations (Gautheron et al., 2022).

Outcrop samples were collected in aluminum or opaque PVC tubes to avoid sunlight exposure. Drill core samples were wrapped in aluminum foil, and only the inner sections were used for luminescence analysis while the outer section exposed to light was used for determination of radionuclides concentrations needed for dose rate estimation. Sediments within a radius of 30 cm from the luminescence sampling point in the outcrops were collected separately to estimate radiation dose rates.

Sample preparation and measurements were carried out under subdued orange-red light. The preparation procedures of quartz concentrates included wet sieving to separate the 180–250 μm fraction; dissolving organic matter with 35% H₂O₂ and carbonates with 10% HCl solutions; density separation at 2.75 g/cm³ and 2.62 g/cm³ with lithium metatungstate (LMT) solution to isolate quartz grains from heavy minerals and feldspar grains; and etching in concentrated 40% HF solution for 40 min to eliminate the outer rind of quartz grains affected by alpha radiation and remnant feldspar grains. Density separation showed that little to no feldspar was present in the studied samples.

All measurements were performed using two Risø DA-20 TL/OSL readers and a Lexsyg Smart reader, all equipped with ⁹⁰Sr/⁹⁰Y beta source (dose rates of 0.07, 0.12 and 0.11 Gy/s respectively), blue and infrared LEDs for stimulation, and a 7.5 mm Hoya U-340 filter for light detection in the ultraviolet band. Aliquots of 4–9 mm diameter were prepared using adhesive silicone spray on stainless steel discs or cups. OSL-IRSL depletion ratio was used to confirm that there was no feldspar contamination in the quartz aliquots (Duller, 2003).

The SAR protocols used to determine the D_e using the OSL (Murray and Wintle, 2000) and TT-OSL (Porat et al., 2009) signals are described in Table 2. The OSL SAR protocol was also used to estimate the characteristic dose (D₀) of dose response curves, and 2D₀ was considered to appraise the upper limit for OSL dating in the studied sediment samples (Wintle and Murray, 2006). The intervals used to derive both OSL and TT-OSL signals were the integral of the first 1 s of light emission, minus the last 10 s as background. Dose-response curves (DRC) were fitted using a linear, single saturating exponential, or an exponential plus linear functions. D_e was calculated using the unweighted arithmetic mean (empirical average) of individual dose estimates, as suggested by (Guérin et al., 2017), accepting recycling ratios between 0.9–1.1 and 0.8–1.2 and recuperation up to 5% and 10%, for OSL and TT-OSL, respectively. The suitability of both OSL and TT-OSL SAR protocols for estimation of D_e under laboratory conditions was evaluated by dose recovery tests (data summary of dose recovery tests on Supplementary Tables S1, S2). The dose recovery tests performed by Bezerra et al. (2022) on samples from the same study area were also used for evaluation of the suitability of the TT-OSL protocol. OSL average calculated-to-given dose ratio is 0.96, and TT-OSL average calculated-to-given dose ratio is 1.5 (Supplementary Table S2).

D1 < D2 < D3 < D4, D5 = 0 Gy, D6 = D1, D7 =D1 (with infrared stimulation before blue led stimulation).

The concentrations of U, Th, and K for determination of dose rates were measured using high-resolution gamma spectrometry with a high-purity germanium detector (Canberra Instruments, relative efficiency of 55% and energy resolution of 2.1 keV at 1,332 keV) encased in an ultralow background shield. Inductively coupled plasma mass spectrometry (ICP-MS) was also used for determination of U and Th concentrations and inductively coupled plasma atomic emission spectroscopy (ICP-OES) was used to determine K concentration. U, Th, and K concentration uncertainties from gamma spectrometry are 1σ. ICP-MS U and Th uncertainties are 5 % and 10%, respectively. ICP-OES K uncertainty is 3%. The concentrations of U (ppm), Th (ppm), and K (%) were converted into dose rates (Gy/ka) using the conversion factors proposed by Guérin et al. (2011). Cosmic rays’ contribution to the dose rate was calculated according to Prescott and Hutton (1994), considering the latitude, longitude, altitude, current depth below the surface, and density of each sample. The internal dose rate was assumed as 0.01 Gy/ka, as this value is considered the upper limit to the internal alpha dose rate in quartz (Vandenberghe et al., 2008).

Representative OSL and TT-OSL decay curves and DRC of Eastern and Central Amazonia study sites are shown in Figure 3. All samples have the OSL curve dominated by the fast component in the first 1 s, with the OSL signal decaying slightly slower in Central Amazonia samples. The TT-OSL signal also decays slower in Central Amazonia samples. All samples have saturated natural OSL signals, except for the two shallowest samples of the Autazes Core, at 0.2 and 1.8 m depth, which have D_e of 4.6 ± 0.5 Gy and 94.6 ± 4.7 Gy, respectively (Table 3), and for sample AVA08, with D_e of 52.3 ± 3.5 Gy. The OSL 2D₀ of all samples ranges from 90 to 180 Gy (Table 3), indicating the limits of OSL dating. TT-OSL D_e ranges from 92 ± 9 Gy to 1,692 ± 229 Gy in Central Amazonia samples and from 161 ± 10 Gy to 747 ± 74 Gy in Eastern Amazonia samples (Table 3 and Figure 5A). Overdispersion of the TT-OSL D_e distribution varies from 16% to 61% (Table 3 and exemplary D_e distribution on Figures 4B,D,F). All samples have high sensitivity changes meaning that the test dose signal increases with SAR cycles as a response to the same beta dose, with more significant increases occurring on the AVA22 samples (Figures 4A,C,E). The difference between TT-OSL D_e and OSL D_e for samples AVA08B, 155812, and 155813 is 109 Gy, 87 Gy, and 142 Gy, respectively.

The Autazes core has a trend of increasing D_e with depth, and the Eastern and Central Amazonia outcrop depth profiles (AVA01 and AVA22) present statistically indistinguishable values. Dose rates are lower in the outcrops in Central and Eastern Amazonia (∼0.6 Gy/ka) than in the Autazes core (∼3.6 Gy/ka) (Figure 5A). All samples have less than 1% of K content, and U is usually higher in Central Amazonia samples (both AVA22 and Autazes core) than in Eastern Amazonia (Figures 5B–D; Supplementary Material S1). The Th content in the Autazes core is ∼20x higher than in the outcrop profiles (AVA01 and AVA22) (Figure 5D and Supplementary Material S1). Ages in the Eastern Amazonia AVA01 profile range from 1927 ± 246 ka to 1,239 ± 294 ka and ages in the Central Amazonia AVA22 profile range from 1,058 ± 198 ka to 923 ± 142 ka (Table 3; Figure 6). In the Autazes core, the TT-OSL ages obtained are younger than the ages from other outcrop profiles, ranging from 585 ± 85 ka to 23 ± 3 ka (Table 3; Figure 6).

Quartz grains representative of lowland Amazonia sandy substrates colonized by open and closed vegetation ecosystems were used for SAR OSL and TT-OSL dating. The ages obtained in study sites of Central and Eastern Amazonia can represent the time of: 1) sediment deposition, 2) soil mixing processes, or 3) the minimum age of the weathered parent rock. Comparison between TT-OSL ages and (U-Th)/He ages from underlying iron crusts suggests that the sandy substrates can be formed by weathering of fluvial sediments accumulated during the Pleistocene, despite other interpretations cannot be ruled out. In Eastern Amazonia savanna sandy substrate, TT-OSL ages ranged from 1927 to 1,239 ka. The Autazes core in Central Amazonia comprises heavily weathered fluvial sandstones and siltstones, covered by upland closed rainforest, with OSL ages from ∼25 to 1 ka in the upper 2 m of the record. The OSL signal is saturated in downcore samples, where TT-OSL ages range from 505 to 23 ka until 14 m depth. TT-OSL ages obtained for the sandy substrate of WSE in Central Amazonia are in the 1,167-923 ka range. The difference between OSL and TT-OSL ages for the same samples suggests that TT-OSL ages have an overestimation of ∼20 to 117 ka. Assuming that the TT-OSL ages represent sediment deposition ages, we can interpret fluvial systems with a higher base level and wider floodplains in Central and Eastern Amazonia during the mid-Pleistocene. In this case, savanna and WSE in the study areas expanded over fluvial deposits during the Pleistocene.