How network structure can affect nitrogen removal by streams

Streams and rivers can be highly reactive sites for nitrogen (N) transformation and removal. Empirical and model-based research show how location in a stream network affects rates of N removal. Because the structure of stream networks can vary widely and N cycling in headwater streams may affect N cycling in downstream reaches, we hypothesised that network structure may affect whole stream network processing of N. We generated three stream networks with the same catchment area but differing shapes, based on optimal channel network theory. We applied a model of nitrate ((Formula presented.)) transport and denitrification, and implemented model scenarios to examine how network shape affects (Formula presented.) removal with (1) increased (Formula presented.) loading from the catchment, (2) altered spatial distributions of (Formula presented.) loading and (3) decreased drainage density (i.e. loss of headwater streams). For all stream networks, the fraction of total (Formula presented.) removed decreased with increasing (Formula presented.) loading from the catchment. Stream networks in narrow catchments removed a higher fraction of (Formula presented.), particularly at intermediate (Formula presented.) loading rates. Network shape also controlled the distribution of removal in small versus large streams, with larger streams removing a higher fraction of the total (Formula presented.) load in narrower networks. The effects of network shape on (Formula presented.) removal when the spatial distribution of (Formula presented.) loading was altered varied with the magnitude of (Formula presented.) loading. At low loads, (Formula presented.) was entirely removed when added to distal parts of the stream network, and about 50% removed when added near the outlet; there was no effect of network shape. At intermediate and high loads, the fraction of total (Formula presented.) load removed by the narrow stream network was 1.5× higher than the rectangular and square networks when (Formula presented.) was added to distal parts of the networks. Network shape did not have an effect when (Formula presented.) load occurred near the outlet, regardless of the magnitude of the (Formula presented.) load. The fraction of total (Formula presented.) removed by the stream network was up to 5% lower when drainage density was reduced from 1.0 to 0.74 km⁻¹, with the least change for the narrow network. Reducing the drainage density also altered the role of small relative to large streams, with the net effect of moving the location of (Formula presented.) removal downstream. Overall, effects of network shape contributed up to 20% of the variation in the fraction of (Formula presented.) removed by stream networks. Network shape was most important at intermediate to high (Formula presented.) loads and when (Formula presented.) was loaded to distal parts of the catchment. The narrow network removed more (Formula presented.) across model scenarios, with elevated removal in larger streams explaining most of the difference. We suggest the shape of the catchment may modulate the degree to which large streams contribute to whole network (Formula presented.) removal.