Search

The importance of snowmelt and spring rainfall to water and nutrient exports from macro-scale watersheds (>1000 km²) is not well established. Data collected from the Androscoggin River watershed (Maine and New Hampshire) between February 1999 and March 2002 show that the 90-day spring melt period accounted for 39-57% of total annual discharge and is likely driven both by snowpack melting and spring rainfall. While large loads of dissolved inorganic nitrogen (DIN) are delivered to the watershed from snowmelt and rain (from 1.16× 106 to 1.61× 106 kg N over the study years), only one third of this N load is exported from the basin during the snowmelt period (0.40× 106-0.48 × 106 kg N). Despite reduced residence time and temperature limitations on biological N retention, there is a poor mass balance between DIN input to the watershed and the nitrogen exported from mouth of the river. Inferences from a geochemical hydrograph separation suggests that approximately 51-63% of the water leaving the mouth of the Androscoggin river is from these 'new' water sources (rain and snowmelt) while 37-49% is from DIN depleted soil and groundwater. Mixing of water from different sources, as well as nutrient retention by dams in the upper watershed, may account for the large discrepancy between DIN inputs and exports from this watershed. [ABSTRACT FROM AUTHOR]

With high-frequency data of nitrate (NO 3 –N) concentrations in waters becoming increasingly important for understanding of watershed system behaviors and ecosystem managements, the accurate and economic acquisition of high-frequency NO 3 –N concentration data has become a key point. This study attempted to use coupled deep learning neural networks and routine monitored data to predict hourly NO 3 –N concentrations in a river. The hourly NO 3 –N concentration at the outlet of the Oyster River watershed in New Hampshire, USA, was predicted through neural networks with a hybrid model architecture coupling the Convolutional Neural Networks and the Long Short-Term Memory model (CNN-LSTM). The routine monitored data (the river depth, water temperature, air temperature, precipitation, specific conductivity, pH and dissolved oxygen concentrations) for model training were collected from a nested high-frequency monitoring network, while the high-frequency NO 3 –N concentration data obtained at the outlet were not included as inputs. The whole dataset was separated into training, validation, and testing processes according to the ratio of 5:3:2, respectively. The hybrid CNN-LSTM model with different input lengths (1d, 3d, 7d, 15d, 30d) displayed comparable even better performance than other studies with lower frequencies, showing mean values of the Nash-Sutcliffe Efficiency 0.60–0.83. Models with shorter input lengths demonstrated both the higher modeling accuracy and stability. The water level, water temperature and pH values at monitoring sites were main controlling factors for forecasting performances. This study provided a new insight of using deep learning networks with a coupled architecture and routine monitored data for high-frequency riverine NO 3 –N concentration forecasting and suggestions about strategies about variable and input length selection during preprocessing of input data. • Hybrid deep learning architectures showed improved prediction performances. • Only accessible routine monitoring variables were included in the input dataset. • Scenarios with shorter input lengths displayed better prediction results. • Variables impacting NO 3 –N forecasting were assessed through modeling performances. [ABSTRACT FROM AUTHOR]