Riverine nitrate loads are usually estimated from sparse concentration samples and continuous discharge records with regression models, and machine learning (ML) is increasingly proposed as an alternative. This study compared the seven-parameter LOADEST regression, an extended multiple linear regression (MLR), random forest (RF), extreme gradient boosting (XGBoost), support vector regression (SVR) and a multilayer perceptron (MLP) for estimating monthly nitrate-plus-nitrite concentrations and loads in the Choptank River near Greensboro, Maryland (USGS 01491000; 292.7 km²). The data comprised 606 samples and daily discharge for water years 1980–2011, aggregated to 345 months with paired observations. Predictors were monthly discharge statistics, a digital-filter baseflow index (BFI), time and seasonal terms. Concentrations averaged 1.20 mg L⁻¹ as N, showed a weak negative relation with discharge and a winter maximum, and increased significantly over time (seasonal Kendall Z = 6.56, p < 0.001; about 1% per year, also after flow adjustment). In eight-fold blocked cross-validation, all models explained only a modest part of the concentration variance (NSE = 0.21–0.38; RMSE = 0.25–0.29 mg L⁻¹). SVR performed best, followed by MLP and XGBoost, and all ML models outperformed LOADEST. In a forward test (2005–2011), the tree ensembles underestimated the rising concentrations by 11–13% because they cannot extrapolate the time trend, whereas MLP (NSE = 0.30) and MLR (0.28) were the most robust. Permutation importance ranked seasonality and time above discharge, and partial dependence showed higher concentrations under baseflow-dominated conditions. Annual loads (mean 134 t N yr⁻¹, 4.6 kg ha⁻¹ yr⁻¹) from all models agreed closely with the WRTDS reference (NSE = 0.81–0.92) but were 6–8% higher because monthly aggregation ignores the within-month covariance of concentration and discharge. Monthly load estimates are thus controlled mainly by discharge, whereas concentration trends require models that can represent time explicitly.