Benchmarking Data-Driven Rainfall-Runoff Models in Great Britain: A comparison of LSTM-based models with four lumped conceptual models

Data for our benchmarking study of 2 LSTM based models compared against four traditional (lumped-conceptual) hydrological models (Lane et al 2019) for Great Britain. These models were trained using data from CAMELS GB (Coxon et al 2020) and using the model training and inference structure at [neuralhydrology](https://github.com/neuralhydrology/neuralhydrology). References: Coxon, Gemma, et al. "CAMELS-GB: hydrometeorological time series and landscape attributes for 671 catchments in Great Britain." Earth System Science Data 12.4 (2020): 2459-2483.APA Lane, Rosanna A., et al. "Benchmarking the predictive capability of hydrological models for river flow and flood peak predictions across over 1000 catchments in Great Britain." Hydrology and Earth System Sciences 23.10 (2019): 4011-4032.APALong short-term memory models (LSTMs) are recurrent neural networks from the emerging field of Deep Learning (DL), which have shown recent promise when predicting time-series especially when data are abundant. Rainfall-runoff modelling presents such a challenge, yet it is vital to have accurate hydrological models for flood forecasting, hazard impact assessment, and to assess the potential effects of climate change on floods and water resources. In this study, we compare the performance of two DL-based models, the LSTM and the Entity Aware LSTM (EA LSTM), in which static catchment attributes are hypothesised to aid the transferability of model structure to ungauged basins. The DL models were trained using a newly published data set, CAMELS-GB for a sample of 518 catchments across Great Britain to identify the spatial and seasonal patterns in model performance, We compare the DL models against benchmarks from four lumped conceptual models recently configured for rainfall-runoff modelling in Great Britain. Our findings show that the LSTM models can simulating discharge with high model performance scores consistently for a large sample (518) of catchments across Great Britain, including in catchments typically considered difficult to model. The LSTM achieves a mean catchment NSE of 0.88 ( 0.86 for the EALSTM), which represents a performance improvement of 10\% -- 16\% compared with the benchmark conceptual models. Seasonal and spatial patterns of model performance indicate that that the largest performance improvement compared to our benchmark is in the summer months and in more arid conditions. By comparing LSTMs with conceptual models, we diagnose possible physical reasons for their different performance. We suggest that LSTMs offer useful predictive capability for rainfall-runoff modelling in Great Britain and elsewhere and note their value to stimulate further research in locations where perceptual understanding is uncertain