Annual runoff increases from 188 to 1382 mm yr^–1along the Pacific coast following a south-north gradient (Table S1; Fig. 1). According to the database, the 28 sampled rivers produced a total discharge of approximately 59 km³·yr⁻¹, which corresponds to ∼49% of the Ecuadorian and Peruvian Pacific rivers’ total discharge (∼89% of the Peruvian R. discharge and ∼32% of the Ecuadorian R. discharge) (Autoridad Nacional del Agua, 2012; Milliman and Farnsworth, 2011; Table S1). All gauging rivers respond to homogenous unimodal seasonality, with a higher discharge between December and May (approximately 80% of the annual river discharge) and a low water stage during the June–November period. The seasonality index (SI = (Q_m max−Q_m min)/Q_m mean, with Q_m the monthly discharge) varies from 1.7 to 2.8 in the five studied basins.

The 28 sampled rivers range from neutral to slightly alkaline, and exhibit pH values from 6.6 to 7.8. The annual average electrical conductivity and TDS concentration are highly variable throughout the studied area and range between 122 and 2940 μS·cm⁻¹ and between 107 and 1915 mg·L⁻¹, respectively. These values are equivalent to or higher than the global riverine concentration average (∼100 mg·l^–1; Milliman and Farnsworth, 2011). Conductivity is highly correlated to TDS concentration (R2 = 0.96; P < 0.001; TDS = 0.63 × cond. + 48 with TDS in mg·L⁻¹) (Table S3; Fig. S2). The two highest values (cond. > 2290 μS cm⁻¹; TDS > 1400 mg·L⁻¹) are recorded in the two southern basins (Locumba and Sama Rivers), which are the driest basins, while all the other values exhibit a conductivity and a TDS concentration less than 1000 μS cm⁻¹ and 600 mg·L⁻¹, respectively. The conductivity (and TDS concentration) followed a north–south gradient, a general trend inversely proportional to the annual runoff (Fig. 1, S3). Interestingly, the Cl⁻, SO₄²⁻, Na⁺, Ca²⁺, Mg²⁺ and K⁺ concentrations are highly correlated with the TDS concentration (R > 0.89; P > 0.001), while Si and HCO₃⁻ are significantly correlated, but at a lower level (For HCO₃^–,R = 0.59; for Si, R_Si = 0.72; P > 0.001). This observation implies that the concentrations of all solutes tend to increase together from north to south (Fig. S3 in supplementary material). Cl⁻ and SO₄²⁻concentrations vary by a magnitude > 100 (magnitude = maximum concentration/minimum concentration). TDS, Na⁺, Ca²⁺, Mg²⁺, and K⁺ vary by magnitudes of 14 to 32, and the HCO³⁻ and Si concentrations vary only by a magnitude of 6 and 7, respectively, while the runoff varies by a magnitude of 54 along the coast. However, the relative contribution of each element to the anionic or cationic charge is not homogenous along the climate gradient. The ternary diagram (Piper diagram) is often used as a diagnostic tool to characterize water (Fig. S4A in supplementary material). The sampled rivers exhibit contrasting water types. In the ternary cation diagram, Mg²⁺ fluctuates moderately, and represents 10%–28% of the cationic charge. K⁺ accounts for only 1%–6% of the cationic charge. Ca²⁺ dominates the cationic charge in northern basins, while Na⁺ dominates the cationic charge in southern basins. In the ternary anion diagram, bicarbonates dominate the anionic charge in northern basins, while the southern basins generally exhibit an SO₄–Cl type. Some central rivers (Santa and neighboring rivers) are of SO₄ type.

At the five monitored stations, the TDS concentration decreases with discharge during the hydrologic cycle (Fig. 2), showing a typical dilution/concentration pattern. This behavior is observed for all major elements concentrations in the five sampled rivers, except for the HCO₃⁻ and Si concentrations, which remain almost constant in the Santa and Cañete R., respectively. However, with variation coefficients ranging between 0.2 and 0.4, the magnitude of the TDS concentration is lower than that of the daily discharge (0.8–1.3). This implies that the TDS flux is mainly controlled by discharge variability during the hydrologic cycle. This dilution behavior is not homogenous for the major elements considered.

The slope b of the Cⁱ = a Q^b regression indicates the dilution behavior of the element i, with C and Q representing the concentration and daily discharge, respectively, and a and b the regression parameters. When b = −1, the element i is totally diluted and the flux is consequently constant through time. When b = 0, the concentration of the element i remains constant and the flux variability of the element is totally controlled by discharge variability. With values ranging between −0.72 and 0.05, the value of b is very variable according to both the element and the monitored station (Table S4). With values ranging between −0.20 and 0.05, the Tumbes R. exhibits values of b less variable than the four other monitored stations. For these four stations, a similar pattern is recorded, and the values of b increase according to the following hierarchy: Cl⁻(−0.72 to –0.50) < Na⁺ (−0.62 to −0.35) < SO₄²⁻(−0.48 to −0.40) ∼ Mg²⁺ (−0.50 to−0.35) < K⁺ (−0.31 to −0.20) ∼ Ca²⁺ (−0.36 to−0.21) < HCO₃⁻(−0.41 to −0.09) ∼ Si (−0.17 to −0.01).

The relative proportion of the major elements with respect to the TDS concentration also varies along the hydrologic cycle (Fig. S4B). As observed for the Napo and Pastaza basins along the Amazonian slope (Moquet et al., 2011), Ca²⁺, Mg²⁺ and Na⁺ + K⁺ co-vary proportionally and represent approximately the same proportion as the cationic charge in Esmeraldas R. In the Santa and Cañete R., the cationic charge is dominated by Ca²⁺ during the rainy season, and the relative proportion of Na⁺, K⁺ and Mg²⁺ increase during the dry season. In the Ocoña and Tumbes R., the relative proportion of Mg²⁺ remains relatively constant during the hydrologic cycle (Mg²⁺ Ocoña ∼ 10%; Mg²⁺ Santa ∼ 20%). Ca²⁺ dominates the cationic charge during the rainy season, and the Na⁺ + K⁺ proportion increases during the dry period. Concerning the anionic charge, HCO_3–dominates the anionic charge in the Esmeraldas and Tumbes R. However, the relative contribution of SO₄²⁻ and Cl⁻ increase during the dry period. SO₄²⁻ dominates the anionic charge of the Santa R., but the relative contribution of HCO₃⁻ increases during the rainy period. In the Cañete and Ocoña R., HCO₃⁻ dominates the anionic charge during the rainy period and the contributions of the SO₄²⁻ and Cl⁻ increase during the dry period.

The distribution of b generally responds to a typical dilution/concentration pattern, as recorded in numerous rivers in mountainous environments (e.g., Ollivier et al., 2010; Zakharova et al., 2005). This behavior shows that the reactions are either kinetically limited, when the minerals involved are present over the entire watershed (as silicate minerals, for example), or when they are limited by the availability of weathered minerals, which dominate the solutes production. The values of b measured in this study are similar compared to the observations performed on the eastern slope of the Andes (Torres et al., 2015–for the catchment MLC, which is the same surface area range as the studied basins; Bouchez et al., 2017; Moquet et al., 2016). Indeed, with a slope b generally lower than –0.4, the elements characteristic of evaporite inputs (Cl⁻, Na⁺, and SO₄²⁻) are highly diluted during the hydrologic cycle. As observed for the Marañón and Ucayali basins in the upper Amazon (Moquet et al., 2016), this implies that evaporite sources tend to produce a constant flux independently of discharge variability. This is consistent with the fact that evaporites (for these elements) or pyrites (for SO₄²⁻) are generally local formations and easily dissolved. Thus, their production depends on local weathering processes and on whether the flux is diluted by the runoff contribution from the remaining part of the basin.

TDS and the other solutes (HCO₃⁻, Mg²⁺, K⁺, Ca²⁺, SiO₂) also exhibit a dilutional pattern, but with values of b generally higher than –0.5, as generally recorded in most rivers at the global scale (Gislason et al., 2009; Walling and Webb, 1986). This shows that the dissolved flux produced by other weathering endmembers (carbonates, silicates) mainly responds to discharge variability in the basin. As observed by Baronas et al. (2017) under the Andean Amazon catchment of the Madre de Dios, this observation highlights that major elements exhibit significant C–Q relationships and depend on various degrees of dilution with increasing runoff under Andean catchment conditions. At the large river basin scale (Bouchez et al., 2017; Moquet et al., 2016), or at a smaller basin scale (Madre de Dios R.; Torres et al., 2015), the Andean basins generally exhibit a value of b smaller than those downstream, because less concentrated lowland water would dilute more chemostatic mountain inputs. Interestingly, the values of b recorded over the Pacific basins are commensurate to those recorded over the rivers monitored near the Amazonian Andes outlet. However, even though the values of b recorded in both Pacific basins and Amazonian Andean basins are all generally higher than –0.5 (Baronas et al., 2017; Bouchez et al., 2017; Moquet et al., 2016; Torres et al., 2015; 2017), they are variable. The C–Q relationship (and the related value of b) can either depend on conservative water mixing (e.g., a tributary confluence or groundwater vs. surface mixing; Baronas et al., 2017; Bouchez et al., 2017; Kirchner, 2016; Torres et al., 2015, 2017) and parameters that control the mean reaction rate, such as temperature (Li et al., 2016), lithological weatherability (Ibarra et al., 2016), water availability (West, 2012) and flow path length (Maher and Chamberlain, 2014). These studies were generally conducted over monolithological basins to identify the model governing the C–Q relationship. In this case, considering all Andean data available (Batonas et al., 2017; Bouchez et al., 2017; Moquet et al., 2016; Torres et al., 2015, 2017), no quantifiable relationship between b and either annual runoff, elevation or lithology is observed. The absence of relationship between b and the basin characteristics can be attributed to the fact that the shape of the C–Q relationship for each solute depends on a mixing of various sources. Considering the lithological heterogeneities over all these areas, their relative contribution to each solute concentration and their influence on the respective values of b can vary from one sub-basin to another. This result shows that even under a given geomorphological domain (an active mountain, in this case), contrasting spatial runoff variability does not control the variability of b for these solutes (HCO₃⁻, Mg²⁺, K⁺, Ca²⁺). Therefore, under orogenic context, water availability alone does not control the shape of the C–Q relationship. The mineral weatherability and flow path of each individual source (lithologies) need to be considered.

According to the atmospheric correction model applied to the monitored station data, atmospheric inputs contribute to less than 6% at all sampled stations. Atmospheric inputs generally contribute less than 5% of the individual solutes Cl⁻, SO₄²⁻, Na⁺, Ca²⁺ and Mg²⁺, except for the three northern basins, where this proportion can be higher than 15% (Table S5). K⁺ is more sensitive to atmospheric inputs, but generally atmospheric inputs contribute to less than 15% of the K⁺ concentration. Indeed, these atmospheric inputs are generally negligible for solute production in these basins. Throughout the 28 sampled rivers, the identification of the main lithological sources controlling dissolved load concentrations is not straightforward. For example, the Ca²⁺ concentration is significantly correlated (P < 0.001) with both HCO₃⁻ and SO₄²⁻ along the Pacific coast, and respectively follow the CaSO₄ (gypsum) and CaCO₃ dissolution lines (Fig. S5A and S5B). Consequently, it is not possible to quantify the relative part of Ca²⁺ released by the evaporites and the carbonate weathering based only on major element concentration variability. Moreover, the lithological map is not precise enough to report all the lithological formations that contribute to dissolving loads, especially for carbonates and evaporites. However, considering that Cl⁻ is a qualitative index of halite (NaCl) dissolutions, it can be observed that evaporite inputs tend to increase from north to south (Fig. S5C; S6). Indeed, the relative mass abundance of Cl^– increases from 4% to 29% following the north–south runoff gradient and is particularly high for basins located south of Cañete basins, where Cl⁻ represents more than 15% of the TDS (Fig. S5). This result highlights that in the context of dry basins, evaporite inputs are related to climate conditions, i.e. a semi-arid context would favor evaporite formation and particularly affect the riverine hydrochemistry. Silicon is only released by silicate minerals, so this element can be considered as a good proxy of silicate weathering distributions (e.g., Maher and Chamberlain, 2014). Interestingly, in a fashion similar to TDS concentrations, Si concentrations follow a global north–south increase, according to the runoff gradient (Table S3; Fig. S3c, f). This result is interesting because it suggests that, in semi-arid conditions, the silicate weathering rate is less dependent on water availability and, consequently, silicates in this environment are highly reactive. However, drier basins in the southern area (from Ocoña to Sama) are also characterized by the presence of active volcanos (Fig. 1). Hydrothermal inputs associated with volcanism can therefore explain the high Si concentration recorded in these rivers (Hurwitz et al., 2010). The combination of low discharge with high hydrothermal inputs can consequently explain the high Si concentrations recorded in this area (Fig. S3c, f). The highest HCO₃⁻ and Ca²⁺ concentrations are recorded in three regions along the Pacific coast, which corresponds to 13 rivers: between the Chira and Reque basins, between the Lacramarca and Chancay Rivers and in the southern basins from the Camana to the Sama Rivers (Fig. S3a, b, d, e). HCO₃⁻ and Ca²⁺ are both released in natural water by silicate and carbonate weathering. In some cases, the lithological information can help to estimate carbonate inputs (e.g., Li and Bush, 2015; Sun et al., 2010). Indeed, higher Ca²⁺ and HCO₃⁻ concentrations would be expected in basins where carbonates are present because of their higher sensitivity to weathering processes (e.g., karst formations in the Cañete R.; Guyot et al., 2014). However, no clear relationship between HCO₃⁻ and Ca²⁺ concentrations and carbonate outcrop cover (extracted from lithological maps) is recorded for the studied basins (Tables S2 and S1).

Spatially, we observed a north–south increase in TDS concentration with decreasing discharge value (Fig. 1 and S3). This gradient may be attributed to an increase in evaporites (either neogenous or unreported by the lithological map), because the SO₄²⁻ + Cl⁻ concentration increase together with the TDS concentration (Fig. S5). However, the magnitude of the TDS concentration is lower than that of the annual discharge (and annual runoff) of sampled rivers. Spatial discharge (and spatial runoff) distribution is consequently the first-order parameter that controls the TDS flux (and spatial specific flux) variability along the Pacific coast (Fig. 3, Fig. S5A). The range of the specific TDS flux is large and varies from 5 to 182 t·km^–2·yr⁻¹. This range is comparable to the values reported by Milliman and Farnworth (2011; based on the Meybeck, 1994 database) for wet young tropical mountainous rivers. The specific flux values are generally lower than those measured over the eastern Andes, but follows the same general trend as a function of the annual runoff (Fig. 3). The TDS specific flux follows a power-law relationship with the runoff and exhibits a slope of 0.46, which shows that the increase in runoff does not proportionally affect the TDS specific fluxes due to the diversity of the individual solute behaviors.

The specific fluxes of individual solutes exhibit a significant (P < 0.01) power-law relationship with runoff for most of the elements. These relationships exhibit a slope higher than 0.65 for SiO₂ and HCO₃⁻ and a slope between 0.48 and 0.6 for K⁺, Mg²⁺ and Ca²⁺ (Fig. 3). Interestingly, as observed for TDS, these specific flux values are lower than those observed in the eastern Andes, but follow the same general trend as a function of runoff (Fig. 3). The SiO₂ specific flux vs. runoff relationship is particularly close to the regression performed over the wet young tropical mountainous rivers and, therefore, confirms this empirical relationship (Milliman and Farnworth, 2011). Conversely, Cl⁻, SO₄²⁻ and Na⁺ (Fig. 3) do not exhibit a significant relationship, which shows that the evaporite dissolution budget is not sensitive to runoff distributions throughout the studied area. This confirms previous observations made based on a C–Q relationship for individual catchments; i.e. these lithologies, which are particularly sensitive to dissolution, are not homogenously distributed throughout the studied area. When they are present, the solute production is independent of the runoff intensity received by the basin. The respective response for each basin, in terms of solute export, can be contrasted according to the occurrence and/or absence of these lithologies. The specific fluxes for elements purely released by silicates (SiO₂) or by silicates and carbonates (HCO₃⁻) increased northward almost proportionally to runoff. This observation highlights that these lithologies are particularly sensitive to hydroclimatological conditions. This region is affected by drastic climate variability from interannual (ENSO events; Morera et al., 2017; Sanabria et al., 2017) to multi-millennium timescales (Carré et al., 2012; Jorneli et al., 2009). Together with the C–Q relationship analyses, this result suggests that the Si and HCO₃⁻ fluxes respond almost proportionally to rainfall variability, following both the rainfall distribution changes along the Pacific coast and the annual to inter-millennial intensity variations on the scale for each individual catchment. This is explained by the fact that, under an active mountain context, the weathering processes are controlled by thermodynamic limit conditions (Maher and Chamberlain, 2014) and, more generally, that runoff is a dominant process that controls continental weathering (e.g., Hartmann et al., 2009; Meybeck, 2003). The fact that K⁺, Mg²⁺ and Ca²⁺ exhibit intermediate slope values is probably due to the diversity of their sources, including carbonates, silicates and evaporites which contrastingly respond to the runoff spatial variability.

Concentration values, available over the sampled Peruvian and Ecuadorian rivers, allow the estimation of total TDS and individual solute fluxes exported annually by the Ecuadorian and Peruvian coasts to the Pacific Ocean. This estimate is performed for basins between northern Ecuador (Santiago R.) and southern Peru (Caplina R.), and is representative of the annual budget not influenced by an ENSO event. The sampled rivers produce ∼15 Mt·yr⁻¹ TDS for a total discharge of ∼59 km³·yr⁻¹ and a total area of 157 10³ km². Discharge from the sampled rivers correspond to approximately 49% of the total discharge from the Ecuadorian and Peruvian Pacific rivers (ANA, 2012; Milliman and Farnsworth, 2011; Table S1). Considering that the sampled rivers are representative of the Pacific rivers, this area could produce ∼30 Mt·yr⁻¹ over ∼400 10³ km² (Table S1). Interestingly, this estimate is equal to Milliman and Farnsworth's (2011) estimate for the same area, which was based only on a first order estimate from a discharge vs. flux relationship interpolation. According to global budget estimates, the studied area contributes to less than 1% of the TDS flux annually exported to all oceans (3250 and 3800 Mt·yr⁻¹; Meybeck, 1976 and Milliman and Farnsworth, 2011) and to less than 2% of the Indian and the Pacific oceans inputs (1775 Mt·yr⁻¹; Milliman and Farnsworth, 2011). In terms of TDS specific flux, this area produces ∼70 t·km^–2·yr⁻¹, which is about two times higher than the global mean value (30 to 50 t·km⁻²·yr⁻¹; Meybeck, 1976, 2003; Milliman and Farnworth, 2011). This shows that, even under low rainfall conditions, this region is active in terms of solute production. Individual solute fluxes (and specific fluxes) vary from 0.3 Mt·yr⁻¹ (0.8 t·km⁻²·yr⁻¹) to 11 Mt·yr⁻¹ (27 t·km⁻²·yr⁻¹) according to the following hierarchy: HCO_3– > SO₄²⁻ > Ca²⁺ > Na⁺ > Cl⁻ > SiO₂ >  Mg²⁺ > K⁺ (Table 1; Fig. 3C). This budget is representative of the annual budget not influenced by an ENSO event. During El Niño events, the rainfall and discharge of the northern rivers can increase by one order of magnitude with respect to the average (e.g., Tumbes and Piura basins; Morera et al., 2017), whereas the southern basins (south of 7°S) exhibit a very low sensitivity to these events (Morera et al., 2017; Rau et al., 2017a). If the C vs. Q relationships for individual rivers and the annual specific flux vs. runoff relationships along the coast are stationary during these events, HCO₃⁻ and SiO₂ fluxes (and specific fluxes) would increase almost proportionally to the increase in discharge. The associated K⁺, Mg²⁺, and Ca²⁺ fluxes would also increase with discharge, but to a lower relative extent in comparison, and Cl^–and SO₄²⁻ exports would be almost insensitive to these climate events.

The Andean Amazon basins located at the same latitude (Solimões Andean tributaries: Napo, Maranon and Ucayali Rivers) produce a TDS flux of approximately 123 Mt year^–1, with a total annual discharge of 695 km³·yr⁻¹, over a surface of 581 km² (budget extracted from Moquet et al., 2016). Therefore, the Andean Pacific rivers produce approximately 20% of the TDS flux and 14% of the discharge over an area that represents approximately 41% of the total western and eastern Andean area located between ∼1°N and ∼18°S (Fig. 4B). This budget is valuable in the absence of ENSO events. This result shows that even if the Pacific slope of the Andes receives a low amount of rainfall compared with Amazonian basins, this area will significantly contribute to the TDS annual production of the Andes, almost proportionally to the relative discharge (Fig. 4A). The flux and the specific flux of each individual element is also systematically lower in the Pacific basin budget than in the Atlantic basin budget (Fig. 4C; Table S3). The solute flux ratios between eastern and western slopes vary from 2 to 6 (ratio = Fx _Amazon/Fx _Pacific, where Fx is the flux of the element x). Higher ratios are observed for HCO₃⁻ and Ca²⁺ due to the high contribution of carbonate weathering to TDS export over the Amazonian slope (Moquet et al., 2011). This ratio is lower for SO₄²⁻, which shows that Pacific basins are particularly active in term of SO₄²⁻ production linked to evaporite and/or sulfide weathering. TDS exports also differ between the two Andean sides, in terms of mean chemical composition. On both side of the Andes, the HCO₃⁻ dominated the TDS composition; however, this element contributes to more than 50% of the TDS over the Amazonian Andes, while it contributed to approximately 40% in the Pacific basin. This is probably due to higher contribution of carbonate weathering over the Amazonian Andean watersheds. In Pacific basins, the SO₄²⁻ largely contributes to TDS production (20% of the TDS), while it contributes to only 8% of the TDS in the Andean Amazon basins. This observation shows that processes that release SO₄²⁻ are highly active over Pacific basins. Over the Amazonian slopes, Ca²⁺ contributes to 17% of the TDS, while it contributes to only 12% over the Pacific basins, illustrating the dominance of carbonate weathering over the Amazonian basins. With 7% of the TDS, SiO₂ contributes commensurately to TDS over both sides of the Andes. Finally, K⁺ and Mg²⁺ contribute less than 2% of the TDS over both sides of the Andes.