DELWAVE 1.0: deep learning surrogate model of surface wave climate in the Adriatic Basin

We propose a new point-prediction model, the DEep Learning WAVe Emulating model (DELWAVE), which successfully emulates the behaviour of a numerical surface ocean wave model (Simulating WAves Nearshore, SWAN) at a sparse set of locations, thus enabling numerically cheap large-ensemble prediction over synoptic to climate timescales. DELWAVE was trained on COSMO-CLM (Cli- mate Limited-area Model) and SWAN input data during the period of 1971–1998, tested during 1998–2000, and cross- evaluated over the far-future climate time window of 2071– 2100. It is constructed from a convolutional atmospheric en- coder block, followed by a temporal collapse block and, finally, a regression block. DELWAVE reproduces SWAN model significant wave heights with a mean absolute error (MAE) of between 5 and 10 cm, mean wave directions with a MAE of 10–25°, and a mean wave period with a MAE of 0.2s. DELWAVE is able to accurately emulate multi- modal mean wave direction distributions related to domi- nant wind regimes in the basin. We use wave power anal- ysis from linearised wave theory to explain prediction er- rors in the long-period limit during southeasterly condi- tions. We present a storm analysis of DELWAVE, employ- ing threshold-based metrics of precision and recall to show that DELWAVE reaches a very high score (both metrics over 95 %) of storm detection. SWAN and DELWAVE time se- ries are compared to each other in the end-of-century sce- nario (2071–2100) and compared to the control conditions in the 1971–2000 period. Good agreement between DEL- WAVE and SWAN is found when considering climatologi- cal statistics, with a small (≤ 5 %), though systematic, un- derestimate of 99th-percentile values. Compared to control climatology over all wind directions, the mismatch between DELWAVE and SWAN is generally small compared to the difference between scenario and control conditions, suggest- ing that the noise introduced by surrogate modelling is sub- stantially weaker than the climate change signal.