A Unified Python Interface for Water Resource Dataset Acquisition and Harmonization

AquaFetch is a Python package designed for the automated downloading, parsing, cleaning, and harmonization of freely available water resource datasets related to rainfall-runoff processes, surface water quality, and wastewater treatment. The package currently supports approximately 70 datasets, each containing between 1 to hundreds of parameters. It facilitates the downloading and transformation of raw data into consistent, easy-to-use, analysis-ready formats. This allows users to directly access and utilize the data without labor-intensive and time-consuming preprocessing.

The package comprises three submodules, each representing a different type of water resource data: rr for rainfall-runoff processes, wq for surface water quality, and wwt for wastewater treatment. The rr submodule offers data for 47,291 catchments worldwide, encompassing both dynamic and static features for each catchment. The dynamic features consist of observed streamflow and meteorological time series, averaged over the catchment area, available at daily and/or hourly time steps. Static features include constant parameters such as land use, soil, topography, and other physiographical characteristics, along with catchment boundaries. This submodule not only provides access to established rainfall-runoff datasets such as CAMELS and LamaH but also introduces new datasets compiled for the first time from publicly accessible online data sources. The wq submodule offers access to 17 surface water quality datasets, each containing various water quality parameters measured across different spaces and times. The wwt submodule provides access to over 20,000 experimental measurements related to wastewater treatment techniques such as adsorption, photocatalysis, membrane filtration, and sonolysis.

The development of AquaFetch was inspired by the growing availability of diverse water resource datasets in recent years. As a community-driven project, the codebase is structured to allow contributors to easily add new datasets, ensuring the package continues to expand and evolve to meet future needs.

Installation

You can install AquaFetch using pip

pip install aqua-fetch

The package can be installed using GitHub link from the master branch

python -m pip install git+https://github.com/hyex-research/AquaFetch.git

To install from a specific branch such as dev branch which contains more recent code

python -m pip install git+https://github.com/hyex-research/AquaFetch.git@dev

The above code will install minimal depencies required to use the library which include
numpy, pandas and requests. To install the library with full list of dependencies use the
all option during installation.

python -m pip install "aqua-fetch[all] @ git+https://github.com/hyex-research/AquaFetch.git"

This will install addtional optional depencdies which include xarray, fiona, netCDF4 and easy_mpl.

Usage

The following sections describe brief usage of datasets from each of the three submodules i.e. rr, wq and wwt.
For detailed usage examples see docs

The core of rr sub-module is the RainfallRunoff class. This class
fetches dynamic features (catchment averaged hydrometeorological data at daily or sub-daily timesteps),
static features (catchment characteristics related to topography, soil, land use-land cover, or hydrological indices that have constant values over time)
and the catchment boundary. The following example demonstrates how to fetch data for CAMELS_SE. However, the method is the same for all available rainfall-runoff datasets.

from aqua_fetch import RainfallRunoff
dataset = RainfallRunoff('CAMELS_SE')  # instead of CAMELS_SE, you can provide any other dataset name

# get data by station id
_, dynamic = dataset.fetch(stations='5', as_dataframe=True)
df = dynamic['5'] # dynamic is a dictionary of with keys as station names and values as DataFrames
df.shape   # ->    (21915, 4)

# get name of all stations as list
stns = dataset.stations()
len(stns)  # -> 50

# get data of 10 % of stations as dataframe
_, dynamic = dataset.fetch(0.1, as_dataframe=True)
len(dynamic)  # 5

# dynamic is a dictionary whose values are dataframes of dynamic features
[df.shape for df in dynamic.values()]   # [(21915, 4), (21915, 4), (21915, 4), (21915, 4), (21915, 4)]

# get the data of a single (randomly selected) station
_, dynamic = dataset.fetch(stations=1, as_dataframe=True)
len(dynamic)  # 1

# get names of available dynamic features
dataset.dynamic_features

# get only selected dynamic features
_, dynamic = dataset.fetch('5', as_dataframe=True,
...  dynamic_features=['pcp_mm', 'airtemp_C_mean', 'q_cms_obs'])
dynamic['5'].shape  # (21915, 3)

# get names of available static features
dataset.static_features

# get data of 10 random stations
_, dynamic = dataset.fetch(10, as_dataframe=True)
len(dynamic)  # 10

# If we want to get both static and dynamic data
static, dynamic = dataset.fetch(stations='5', static_features="all", as_dataframe=True)
static.shape, len(dynamic), dynamic['5'].shape   # ((1, 76), 1, (21915, 4))

# If we don't set as_dataframe=True and have xarray installed then the returned data will be a xarray Dataset
_, dynamic = dataset.fetch(10)
type(dynamic)   # -> xarray.core.dataset.Dataset
	
dynamic.dims   # -> FrozenMappingWarningOnValuesAccess({'time': 21915, 'dynamic_features': 4})

len(dynamic.data_vars)   # -> 10

# get coordinates of all stations
coords = dataset.stn_coords()
coords.shape  #     (50, 2)
# get coordinates of station whose id is 5
dataset.stn_coords('5')       # 68.035599	21.9758
# get coordinates of two stations
dataset.stn_coords(['5', '736'])

# get area of a single station
dataset.area('5')
# get coordinates of two stations
dataset.area(['5', '736'])

# if fiona library is installed we can get the boundary as fiona Geometry
dataset.get_boundary('5')

The datasets related to surface water quality are available using functional or objected-oriented API
depending upon the complexity of the dataset. The following example shows usage of two surface water
quality related datasets. For complete name of Python functions and classes see documentation

from aqua_fetch import busan_beach
dataframe = busan_beach()
dataframe.shape  # (1446, 14)

dataframe = busan_beach(target=['tetx_coppml', 'sul1_coppml'])
dataframe.shape  # (1446, 15)

from aqua_fetch import GRQA
ds = GRQA(path="/path/to/data")
print(ds.parameters)

len(ds.parameters)    # 42
country = "Pakistan"
len(ds.fetch_parameter('TEMP', country=country))

The datasets for wastewater treatment are all available in function API design. These datasets consist of experimental conducted
to remove certain pollutants from wastewater. For complete list of functions, see documentation

from aqua_fetch import ec_removal_biochar
data, *_ = ec_removal_biochar()
data.shape  # -> (3757, 27)

data, encoders = ec_removal_biochar(encoding="le")
data.shape  # -> (3757, 27)


from aqua_fetch import mg_degradation
mg_data, encoders = mg_degradation()
mg_data.shape  # -> (1200, 12)

# the default encoding is None, but if we want to use one hot encoder
mg_data_ohe, encoders = mg_degradation(encoding="ohe")
mg_data_ohe.shape  # -> (1200, 31)

Summary of rainfall runoff Datasets

Summary of Water Quality Datasets

Summary of datasets related to wastewater treatment