Thousands of watersheds drain from the perhumid coastal temperate rainforest (PCTR) of southeast Alaska into the nearshore estuary. This region is characterized by high rates of precipitation which interacts with terrestrial ecosystems and transports nutrients downstream. Modeling N and P fluxes from land to sea in the PCTR is needed to 1) incorporate nutrient losses into regional terrestrial productivity models and 2) determine the magnitude of this subsidy to downstream aquatic ecosystems. To develop estimates of these fluxes, models of stream water N and P concentrations are needed which can be coupled with hydrologic models across the landscape. In this study, we develop statistical models for both organic and inorganic forms of N and P using biophysical watershed characteristics and stream chemistry data from a large-scale synoptic survey including 57 independent watersheds with varying degrees of wetland extent, topography, and natural and anthropogenic disturbance legacies. We also take into account the influence of salmon spawning events and seasonal controls on stream chemistry. Overall, dissolved organic nitrogen (DON) comprised 81% of total dissolved N in fall and 69% in spring, while dissolved organic phosphorus (DOP) comprised 59% of total dissolved phosphorus in fall and 69% in spring. The overall dominance of DON and DOP over inorganic forms emphasizes the importance of soil development and leaching of organic matter on stream chemistry in the PCTR, and the significance of variables that indicate organic rich soils in the landscape in our final models supports this finding. Our final statistical models show that the presence of salmon spawning in the fall has a positive effect on concentrations of inorganic forms on N and P, as well as DOP. Interestingly, neither salmon spawning nor disturbance variables related to successional nitrogen fixing vegetation were retained in our final DON model. The proportion of the watershed with slopes <5° was an important predictor of stream DON concentrations. Overall, these models were able to explain 65% of the variance in DON, 32% DOP, 56% DIN and 38% SRP. Our results suggest that season, salmon spawning activity, slope variables indicative of organic rich soils, disturbance legacy and successional vegetation are useful predictors of stream N and P.