Source code for pastas.recharge

"""This module contains the classes for recharge models.

This module contains the different classes that can be used to simulate the
effect of precipitation and evapotranspiration on groundwater levels.
Depending on the mathematical formulation this effect may be interpreted as:

1. seepage to the groundwater
2. precipitation excess,
3. groundwater recharge.

For the implementation of each model we refer to the references listed in
the documentation of each recharge model.

The classes defined here are designed to be used in conjunction with the
stressmodel "RechargeModel", which requires an instance of one of the
classes defined here.

.. codeauthor:: R.A. Collenteur, University of Graz

See Also
--------
pastas.stressmodels.RechargeModel
    The recharge models listed above are provided to a RechargeModel

Examples
--------
Using the recharge models is as follows:

>>> rch = ps.rch.FlexModel()
>>> sm = ps.RechargeModel(prec, evap, recharge=rch, rfunc=ps.Gamma, name="rch")
>>> ml.add_stressmodel(sm)

After solving a model, the simulated recharge flux can be obtained:

>>> rch_sim = ml.get_stress("rch")

"""

from logging import getLogger
from numpy import add, float64, multiply, exp, zeros, nan_to_num, vstack, \
    where, power
from pandas import DataFrame

from pastas.decorators import njit
from .utils import check_numba

logger = getLogger(__name__)


[docs]class RechargeBase: """Base class for classes that calculate the recharge. """
[docs] def __init__(self, **kwargs): self.snow = False self.nparam = 0 self.kwargs = kwargs
[docs] @staticmethod def get_init_parameters(name="recharge"): """Method to obtain the initial parameters. Parameters ---------- name: str, optional String with the name that is used as prefix for the parameters. Returns ------- parameters: pandas.DataFrame Pandas DataFrame with the parameters. """ parameters = DataFrame( columns=["initial", "pmin", "pmax", "vary", "name"]) return parameters
[docs] def simulate(self, prec, evap, p, dt=1.0, return_full=False, **kwargs): pass
[docs]class Linear(RechargeBase): """Linear model for precipitation excess according to :cite:t:`von_asmuth_transfer_2002`. Notes ----- The precipitation excess is calculated as: .. math:: R = P - f * E """ _name = "Linear"
[docs] def __init__(self): RechargeBase.__init__(self) self.nparam = 1
[docs] def get_init_parameters(self, name="recharge"): parameters = DataFrame( columns=["initial", "pmin", "pmax", "vary", "name"]) parameters.loc[name + "_f"] = (-1.0, -2.0, 0.0, True, name) return parameters
[docs] def simulate(self, prec, evap, p, **kwargs): """Simulate the precipitation excess flux. Parameters ---------- prec, evap: array_like array with the precipitation and evapotranspiration values. These arrays must be of the same length and at the same time steps. p: array_like array_like object with the values as floats representing the model parameters. Returns ------- recharge: array_like array with the recharge series. """ return add(prec, multiply(evap, p))
[docs] def get_water_balance(self, prec, evap, p, **kwargs): ea = multiply(evap, p) r = add(prec, multiply(evap, p)) data = DataFrame(data=vstack((prec, ea, -r)).T, columns=["P", "Ea", "R"]) return data
[docs]class FlexModel(RechargeBase): """Recharge to the groundwater calculated according to :cite:t:`collenteur_estimation_2021`. Parameters ---------- interception: bool, optional Use an interception reservoir in the model or not. snow: bool, optional Account for snowfall and snowmelt in the model. If True, a temperature series should be provided to the RechargeModel. gw_uptake: bool, optional If True, the potential evaporation that is left after evaporation from the interception reservoir and the root zone reservoir is subtracted from the recharge flux. An additional parameter can be used to scale the excess evaporation. Note that this is an EXPERIMENTAL FEATURE that may be removed in the future! Notes ----- For a detailed description of the recharge model and parameters we refer to :cite:t:`collenteur_estimation_2021`. The water balance for the unsaturated zone reservoir is written as: .. math:: \\frac{dS}{dt} = P_e - E_a - R where the recharge is calculated as: .. math:: R = K_s \\left( \\frac{S}{S_u}\\right) ^\\gamma If snow=True, a snow reservoir is added on top. For a detailed description of the degree-day snow model and parameters we refer to :cite:t:`kavetski_model_2007`. The water balance for the snow reservoir is written as: .. math:: \\frac{dSs}{dt} = Ps - M Note that the preferred unit of the precipitation and evaporation is mm/d and the temperature is degree celsius. """ _name = "FlexModel"
[docs] def __init__(self, interception=True, snow=False, gw_uptake=False): check_numba() RechargeBase.__init__(self, interception=interception, snow=snow, gw_uptake=gw_uptake) self.snow = snow self.interception = interception self.gw_uptake = gw_uptake self.nparam = 5 if self.interception: self.nparam += 1 if self.gw_uptake: self.nparam += 1 if self.snow: self.nparam += 2
[docs] def get_init_parameters(self, name="recharge"): parameters = DataFrame( columns=["initial", "pmin", "pmax", "vary", "name"]) parameters.loc[name + "_srmax"] = (250.0, 1e-5, 1e3, True, name) parameters.loc[name + "_lp"] = (0.25, 1e-5, 1, False, name) parameters.loc[name + "_ks"] = (100.0, 1e-5, 1e4, True, name) parameters.loc[name + "_gamma"] = (2.0, 1e-5, 20.0, True, name) parameters.loc[name + "_kv"] = (1.0, 0.25, 2.0, False, name) if self.interception: parameters.loc[name + "_simax"] = (2.0, 0.0, 10.0, False, name) if self.gw_uptake: parameters.loc[name + "_gf"] = (1.0, 0.0, 1.0, True, name) if self.snow: parameters.loc[name + "_tt"] = (0.0, -10.0, 10.0, True, name) parameters.loc[name + "_k"] = (2.0, 1.0, 20.0, True, name) return parameters
[docs] def simulate(self, prec, evap, temp, p, dt=1.0, return_full=False, **kwargs): """Simulate the soil water balance model. Parameters ---------- prec: numpy.Array Precipitation flux in mm/d. Must have the same length as evap. evap: numpy.Array Potential evaporation flux in mm/d. temp: numpy.Array Temperature in degrees Celcius. p: array_like array_like object with the values as floats representing the model parameters. Must be length self.nparam. dt: float, optional time step for the calculation of the recharge. Only dt=1 is possible now. return_full: bool return all fluxes and states as Numpy arrays. Returns ------- r: numpy.Array Recharge flux calculated by the model. """ ep = evap * p[4] if self.snow: ss, ps, m = self.get_snow_balance(prec=prec, temp=temp, tt=p[-2], k=p[-1]) pr = prec - ps # Remove snowfall from precipitation else: pr = prec # All precipitation is rainfall and melt is zero m = 0.0 if self.interception: si, ei, pi = self.get_interception_balance(pr=pr, ep=ep, simax=p[5]) ep = ep + ei # Update potential evaporation after interception pe = pr - pi # Update rainfall after interception else: pe = pr sr, r, ea, q, _ = self.get_root_zone_balance(pe=pe - m, ep=ep, srmax=p[0], lp=p[1], ks=p[2], gamma=p[3], dt=dt) # report big water balance errors (error > 0.1%.) error = (sr[0] - sr[-1] + (pe - m + r + ea + q).sum()) / \ (pe.sum() + 1e-10) # avoid division by zero if abs(error) > 0.1: logger.info("Water balance error: %s %% of the total pe flux. " "Parameters: %s", error.round(2), p.astype(float).round(2)) if self.gw_uptake: # Compute leftover potential evaporation if self.interception: gf = p[6] else: gf = p[5] eg = ep + ea # positive flux r = r + gf * eg if return_full: data = (sr, r, ea, q, pe) if self.interception: data += (si, ei, pi) if self.snow: data += (ss, ps, m) return data else: return -r
[docs] @staticmethod @njit def get_root_zone_balance(pe, ep, srmax=250.0, lp=0.25, ks=100.0, gamma=4.0, dt=1.0): """Method to compute the water balance of the root zone reservoir. Parameters ---------- pe: numpy.Array Effective precipitation flux in mm/d. ep: numpy.Array Potential evaporation flux in mm/d. srmax: float, optional Maximum storage capacity of the root zone. lp: float, optional Parameter determining when actual evaporation equals potential. ks: float, optional Saturated hydraulic conductivity in mm/d. gamma: float, optional Parameter determining the nonlinearity of outflow / recharge. dt: float, optional time step for the calculation of the recharge. Only dt=1 is possible now. Returns ------- sr: numpy.Array Storage in the root zone reservoir. r: numpy.Array Recharge flux in mm/d ea: numpy.Array Evaporation flux in mm/d. Consists of transpiration and soil evaporation. Does not include interception evaporation. q: numpy.Array surface runoff flux in mm/d. pe: numpy.Array Incoming infiltration flux in mm/d. Notes ----- If Numba is available, this method is significantly faster. """ n = pe.size # Create empty arrays to store the fluxes and states sr = zeros(n + 1, dtype=float64) # Root Zone Storage State sr[0] = 0.5 * srmax # Set the initial system state to half-full ea = zeros(n, dtype=float64) # Actual evaporation Flux r = zeros(n, dtype=float64) # Recharge Flux q = zeros(n, dtype=float64) # Surface runoff Flux lp = lp * srmax # Do this here outside the for-loop for efficiency for t in range(n): # Make sure the solution is larger then 0.0 and smaller than sr if sr[t] > srmax: q[t] = sr[t] - srmax # Surface runoff sr[t] = srmax elif sr[t] < 0.0: sr[t] = 0.0 # Calculate evaporation from the root zone reservoir if sr[t] / lp < 1.0: ea[t] = ep[t] * sr[t] / lp else: ea[t] = ep[t] # Calculate the recharge flux r[t] = min(ks * (sr[t] / srmax) ** gamma, sr[t]) # Update storage in the root zone sr[t + 1] = sr[t] + dt * (pe[t] - r[t] - ea[t]) return sr[:-1], -r, -ea, -q, pe
[docs] @staticmethod @njit def get_interception_balance(pr, ep, simax=2.0, dt=1.0): """Method to compute the water balance of the interception reservoir. Parameters ---------- pr: numpy.Array Numpy Array with rainfall in mm/day. ep: numpy.Array Numpy Array with potential evaporation in mm/day. simax: float, optional storage capacity of the interception reservoir. dt: float time step used for computation. Only dt=1.0 is possible now. Returns ------- si: numpy.Array Interception storage. ei: numpy.Array Interception evaporation. pi: numpy.Array Incoming rainfall that is intercepted. Notes ----- The water balance for the snow storage reservoir is defined as follows: .. math:: \\frac{dS_i}{dt} = P_r - E_i - P_e where $S_i$ [L] is the interception storage, $P_r$ [L/T] is the incoming rainfall, $E_i$ [L/T] the interception evaporation, and $P_e$ [L/T] the overflow from the interception reservoir. """ n = pr.size si = zeros(n + 1, dtype=float64) # Interception Storage State pe = zeros(n, dtype=float64) # Effective precipitation Flux ei = zeros(n, dtype=float64) # Interception evaporation Flux for t in range(n): # Interception bucket ei[t] = min(ep[t], si[t]) si[t + 1] = si[t] + dt * (pr[t] - ei[t]) pe[t] = max(si[t + 1] - simax, 0.0) si[t + 1] = si[t + 1] - pe[t] pi = pr - pe # Compute intercepted precipitation return si[:-1], -ei, pi
[docs] @staticmethod @njit def get_snow_balance(prec, temp, tt=0.0, k=2.0): """Method to compute the water balance of the snow reservoir. Parameters ---------- prec: numpy.Array Numpy Array with precipitation in mm/day. temp: numpy.Array Numpy Array with the mean daily temperature in degree Celsius. tt: float, optional k: float, optional Returns ------- ss: numpy.Array storage in the snow reservoir. ps: numpy.Array snowfall flux in mm/d. m: numpy.Array snow melt flux in mm/d. Notes ----- The water balance from the snow reservoir is as follows: .. math:: \\frac{dS_s}{dt} = P_s - M where $S_s$ [L] is the snow storage, $P_s$ [L/T] the snowfall, and $M$ [L/T] the snow melt from the snow reservoir. """ n = prec.size # Create empty arrays to store the fluxes and states ss = zeros(n + 1, dtype=float64) # Snow Storage ps = where(temp <= tt, prec, 0.0) # Snowfall m = where(temp > tt, k * (temp - tt), 0.0) # Potential Snowmelt # Snow bucket for t in range(n): if temp[t] > tt: smoothing_factor = 1.0 - exp(-(ss[t] / 1.5)) m[t] = min(m[t] * smoothing_factor, ss[t]) ss[t + 1] = ss[t] + ps[t] - m[t] return ss[:-1], ps, -m
[docs] def get_water_balance(self, prec, evap, temp, p, dt=1.0, **kwargs): data = self.simulate(prec=prec, evap=evap, temp=temp, p=p, dt=dt, return_full=True, **kwargs) columns = ["State Root zone (Sr)", "Recharge (R)", "Actual evaporation (Ea)", "Surface Runoff (Q)", "Effective precipitation (Pe)"] if self.interception: columns += ["State Interception (Si)", "Interception evaporation (Ei)", "Intercepted precipitation (Pi)"] if self.snow: columns += ["State Snow (Ss)", "Snowfall (Ps)", "Snowmelt (M)", ] data = DataFrame(data=vstack(data).T, columns=columns) return data
[docs] def check_snow_balance(self, prec, temp, **kwargs): ss, ps, m = self.get_snow_balance(prec, temp, **kwargs) error = (ss[0] - ss[-1] + (ps + m).sum()) return error
[docs] def check_interception_balance(self, prec, evap, **kwargs): si, ei, pi = self.get_interception_balance(prec, evap, **kwargs) error = (si[0] - si[-1] + (pi + ei).sum()) return error
[docs] def check_root_zone_balance(self, prec, evap, **kwargs): sr, r, ea, q, pe = self.get_root_zone_balance(prec, evap, **kwargs) error = (sr[0] - sr[-1] + (r + ea + q + pe).sum()) return error
[docs]class Berendrecht(RechargeBase): """Recharge to the groundwater calculated according to :cite:t:`berendrecht_non-linear_2006`. Notes ----- Note that the preferred unit of the precipitation and evaporation is mm/d. The water balance for the unsaturated zone reservoir is written as: .. math:: \\frac{dS_e}{dt} = \\frac{1}{D_e}(f_iP - E_a - R) where the recharge is calculated as: .. math:: R(S_e) = K_sS_e^\\lambda(1-(1-S_e^{1/m})^m)^2 For a detailed description of the recharge model and parameters we refer to the original publication. """ _name = "Berendrecht"
[docs] def __init__(self): check_numba() RechargeBase.__init__(self) self.nparam = 7
[docs] def get_init_parameters(self, name="recharge"): parameters = DataFrame( columns=["initial", "pmin", "pmax", "vary", "name"]) parameters.loc[name + "_fi"] = (0.9, 0.7, 1.3, False, name) parameters.loc[name + "_fc"] = (1.0, 0.7, 1.3, False, name) parameters.loc[name + "_sr"] = (0.25, 1e-5, 1.0, False, name) parameters.loc[name + "_de"] = (250.0, 20, 1e3, True, name) parameters.loc[name + "_l"] = (2.0, -4, 50, True, name) parameters.loc[name + "_m"] = (0.5, 1e-5, 0.5, False, name) parameters.loc[name + "_ks"] = (100.0, 1, 1e4, True, name) return parameters
[docs] def simulate(self, prec, evap, p, dt=1.0, return_full=False, **kwargs): """Simulate the recharge flux. Parameters ---------- prec: numpy.Array Precipitation flux in mm/d. Has to have the same length as evap. evap: numpy.Array Potential evapotranspiration flux in mm/d. p: array_like array_like object with the values as floats representing the model parameters. dt: float, optional time step for the calculation of the recharge. Only dt=1 is possible now. Returns ------- r: numpy.Array Recharge flux calculated by the model. """ r, s, ea, pe = self.get_recharge(prec, evap, fi=p[0], fc=p[1], sr=p[2], de=p[3], l=p[4], m=p[5], ks=p[6], dt=dt) if return_full: return r, s, ea, pe else: return nan_to_num(r)
[docs] @staticmethod @njit def get_recharge(prec, evap, fi=1.0, fc=1.0, sr=0.5, de=250.0, l=-2.0, m=0.5, ks=50.0, dt=1.0): """ Internal method used for the recharge calculation. If Numba is available, this method is significantly faster. """ n = prec.size # Create an empty arrays to store the fluxes and states pe = fi * prec # Effective precipitation flux ep = fc * evap # Potential evaporation flux s = zeros(n, dtype=float64) # Root zone storage state s[0] = 0.5 # Set the initial system state r = zeros(n, dtype=float64) # Recharge flux ea = zeros(n, dtype=float64) # Actual evaporation flux for t in range(n - 1): # Make sure the reservoir is not too full or empty. if s[t] < 0.05: s[t] = 0.05 * exp(20.0 * s[t] - 1.0) elif s[t] > 0.95: s[t] = 1 - (0.05 * exp(19.0 - 20.0 * s[t])) # Calculate the actual evaporation ea[t] = (1.0 - exp(-3 * s[t] / sr)) * ep[t] # Calculate the recharge flux r[t] = ks * s[t] ** l * (1.0 - (1.0 - s[t] ** (1.0 / m)) ** m) ** 2 # Calculate the s[t + 1] = s[t] + dt / de * (pe[t] - ea[t] - r[t]) return r, s, ea, pe
[docs] def get_water_balance(self, prec, evap, p, dt=1.0, **kwargs): r, s, ea, pe = self.simulate(prec, evap, p=p, dt=dt, return_full=True, **kwargs) s = s * p[3] # Because S is computed dimensionless in this model data = DataFrame(data=vstack((s, pe, ea, r)).T, columns=["S", "Pe", "Ea", "R"]) return data
[docs]class Peterson(RechargeBase): """Recharge to the groundwater calculated based on :cite:t:`peterson_nonlinear_2014`. The water balance for the unsaturated zone reservoir is written as: .. math:: \\frac{dS}{dt} = P_e - E_a - R where the fluxes $P_e$, $E_a$ and $R$ are calculated as: .. math:: P_e = P \left(1 - \frac{S}{\hat{S_{cap}}}\right)^\alpha .. math:: E_a = E_p \left(\frac{S}{\hat{S_{cap}}}\right)^\gamma .. math:: R = \hat{k_{sat}}\left(\frac{S}{\hat{S_{cap}}}\right)^{\hat{\beta}} with the parameters: .. math:: \hat{S_{cap}} = 10^{S_{cap}}; \hat{k_{sat}} = 10^{k_{sat}}; \hat{\beta} = 10^{\beta} Note that the method currently uses forward Euler method to solve the ODE so significant water balance errors can occur. """ _name = "Peterson"
[docs] def __init__(self): check_numba() RechargeBase.__init__(self) self.nparam = 5
[docs] def get_init_parameters(self, name="recharge"): parameters = DataFrame( columns=["initial", "pmin", "pmax", "vary", "name"]) parameters.loc[name + "_scap"] = (1.5, 0.5, 3.0, True, name) parameters.loc[name + "_alpha"] = (1.0, 0.0, 1.5, True, name) parameters.loc[name + "_ksat"] = (1.0, 0.0, 3.0, True, name) parameters.loc[name + "_beta"] = (0.5, 0.0, 1.5, True, name) parameters.loc[name + "_gamma"] = (1.0, 0.0, 2.0, True, name) return parameters
[docs] def simulate(self, prec, evap, p, dt=1.0, return_full=False, **kwargs): """Simulate the recharge flux. Parameters ---------- prec: numpy.Array Precipitation flux in mm/d. Must have the same length as evap. evap: numpy.Array Potential evapotranspiration flux in mm/d. p: array_like array_like object with the values as floats representing the model parameters. dt: float, optional time step for the calculation of the recharge. Returns ------- r: numpy.Array Recharge flux calculated by the model. """ r, s, ea, pe = self.get_recharge(prec, evap, scap=p[0], alpha=p[1], ksat=p[2], beta=p[3], gamma=p[4], dt=dt) if return_full: return r, s, ea, pe else: return nan_to_num(r)
[docs] @staticmethod @njit def get_recharge(prec, evap, scap=1.0, alpha=1.0, ksat=1.0, beta=0.5, gamma=1.0, dt=1.0): """ Internal method used for the recharge calculation. If Numba is available, this method is significantly faster. """ n = len(prec) # Create an empty arrays to store the fluxes and states pe = zeros(n, dtype=float64) # Effective precipitation flux sm = zeros(n + 1, dtype=float64) # Root zone storage state r = zeros(n, dtype=float64) # Recharge flux ea = zeros(n, dtype=float64) # Actual evaporation flux # Update params smsc = power(10, scap) ksat = power(10, ksat) beta = power(10, beta) # Set the initial system state sm[0] = smsc / 2 for t in range(n): sm_frac = sm[t] / smsc pe[t] = prec[t] * power(1 - sm_frac, alpha) ea[t] = max(sm[t + 1], evap[t] * power(sm_frac, gamma)) r[t] = max(sm[t + 1], ksat * power(sm_frac, beta)) sm[t + 1] = min(smsc, max(0.0, sm[t] + (pe[t] - ea[t] - r[t]) * dt)) return r, sm[1:], ea, pe
[docs] def get_water_balance(self, prec, evap, p, dt=1.0, **kwargs): r, s, ea, pe = self.simulate(prec, evap, p=p, dt=dt, return_full=True, **kwargs) data = DataFrame(data=vstack((s, pe, ea, r)).T, columns=["S", "Pe", "Ea", "R"]) return data