Untangling soil‐weather drivers of daily N <sub>2</sub> O emissions and fertilizer management mitigation strategies in no‐till corn
Leonardo M. Bastos, Charles W. Rice, Peter J. Tomlinson, David Mengel
Abstract
Abstract Fertilizer N management can mitigate N 2 O emissions but complex soil‐weather conditions modulate the mitigation potential. Conditional inference tree (CIT) is a machine learning method able to untangle complex interactions while providing an interpretable model. The goals of this study were (a) to assess the effect of N fertilizer on N 2 O emissions, and to use CIT to identify (b) the main soil‐weather drivers of daily N 2 O hot moments and (c) fertilizer management options to mitigate them. The study was conducted in 2 yr in no‐till corn ( Zea mays L.) with seven combinations of N source and placement tested. Daily N 2 O emissions were measured with vented chambers, and soil temperature and water‐filled pore space (WFPS) were measured near the chambers on the same days of gas sampling. Overall, 2013 was drier with lower N 2 O emissions than 2014. Cumulative N 2 O losses differed across treatments and years, with broadcast emitting more in 2014 than in 2013, and only subsurface‐banded fertilizer with a nitrification inhibitor (NI) consistently abated N 2 O losses. The main hot moment conditions were within ∼80 d of fertilizer application when soil temperature >15 °C and WFPS >57%. Under these conditions, NI abated losses by 50% compared with fertilizer alone. The machine learning approach used here could be used in larger datasets to elucidate environment‐specific drivers of N 2 O hot moments and potential fertilizer mitigation practices under different soil, weather, and management conditions.