Data

Input data format

The input data format must be in comma seperated values (.csv) file. There are two types of input data, and they must be in two files:

  • Dynamic (time series) input data file: This file is required (a MUST). A template of this file (e.g., can be found here. This file contains time series data for all catchments.

  • Static (catchment attributes) input data file: These are optional data. An example of such input data file is here. This file contains the attributes of all catchments.

Note

Both dynamic and static data files MUST have a column with a name object_id which could be the catchment name or id. This object_id is used to link the two files together, e.g., with a specific object_id, HydroEcoLSTM knows where are the corresponding dynamics and static data.

The dynamic data file MUST have a column time in format YYYY-MM-DD HH:MM (for example, 2024-12-13 11:30)

Inputs and target outputs MUST be in the dynamic data file.

Statics input data file is needed when you model, for example, streamflow for multiple catchments. For a single catchment, this data (file) does not required. However, a recent paper in HESS argues that we should not train LSTM for a single catchment.

Example data

CAMEL-CH data

The example data are only the subset of the CAMEL-CH data. This data contains dynamic (time_series.csv) and static (static_attributes.csv) data of 10 catchments from the CAMEL-CH data. The time_series.csv contains discharge_vol_m3_s, precipitation_mm_d, temperature_min_degC,``temperature_mean_degC``, temperature_max_degC, and rel_sun_dur. The input units does not matter for all dynamic input data, for example, you can used different unit for discharge (such as mm/day or cubic feet meter per second). This is becuase LSTMs does not based on the mass balance equations. What important is that the unit MUST be the same for all catchments (which I named the catchment ID as object_id) (e.g., you cannot use the unit of discharge is m3/s for the first catchments and mm/day for the second catchment). Same are applied for the units of catchment characteristics in the static_attributes.csv file.

Please refer to the Hoege et al. (2023) for a detailed description of the CAMEL-CH data.

Stable isotope data

The second dataset are the high frequency isotope data in precipitation and streamflow in the Alp and Erlenbach catchment in Switzerland.

Please refer to the von Freyberg et al. (2022) for a detailed description of the data.