co2_transpiration

CO₂ influence on crop transpiration.

Elevated atmospheric CO₂ reduces stomatal conductance, and therefore crop transpiration. This module scales the potential transpiration PTRAN by a linear reduction factor before it enters the water balance, so the effect propagates into the water-stress factor TRANRF.

This is a distinct mechanism from the CO₂ correction applied to reference ET in torchcrop.processes.evapotranspiration: that one rescales the reference evapotranspiration ET0 via an interpolation table, whereas this module rescales the potential transpiration demand that the water balance compares against soil water supply.

Equations

Linear reduction factor as a function of the atmospheric CO₂ concentration:

\[ f(\text{CO}_2) = m \cdot \text{CO}_2 + b \]

\[ \text{PTRAN}_{\text{reduced}} = \text{PTRAN} \cdot f(\text{CO}_2) \]

With the defaults m = -0.0003 ppm⁻¹ and b = 1.1, the factor is ≈ 1 at the ≈ 350 ppm reference and falls below 1 for higher concentrations (less transpiration under elevated CO₂).

Co2Transpiration (Module)

Reduce potential transpiration as a linear function of CO₂.

Parameters:

Name	Type	Description	Default
transpiration_m	float	Slope cTranspiration_m of the linear reduction factor [ppm⁻¹] (default -0.0003).	-0.0003
transpiration_b	float	Intercept cTranspiration_b of the linear reduction factor [-] (default 1.1).	1.1
clamp_nonnegative	bool	If True (default), clamp the reduction factor to ≥ 0 so an extreme CO₂ value can never flip the sign of transpiration.	True

Source code in torchcrop/processes/co2_transpiration.py

class Co2Transpiration(nn.Module):
    r"""Reduce potential transpiration as a linear function of CO₂.

    Args:
        transpiration_m: Slope ``cTranspiration_m`` of the linear reduction
            factor [ppm⁻¹] (default ``-0.0003``).
        transpiration_b: Intercept ``cTranspiration_b`` of the linear
            reduction factor [-] (default ``1.1``).
        clamp_nonnegative: If ``True`` (default), clamp the reduction
            factor to ``≥ 0`` so an extreme CO₂ value can never
            flip the sign of transpiration.
    """

    def __init__(
        self,
        transpiration_m: float = -0.0003,
        transpiration_b: float = 1.1,
        clamp_nonnegative: bool = True,
    ) -> None:
        super().__init__()
        self.transpiration_m = transpiration_m
        self.transpiration_b = transpiration_b
        self.clamp_nonnegative = clamp_nonnegative

    def factor(self, co2: torch.Tensor) -> torch.Tensor:
        r"""Compute the CO₂ transpiration-reduction factor.

        Args:
            co2: Atmospheric CO₂ concentration [ppm], broadcastable
                to the transpiration tensor.

        Returns:
            Reduction factor ``f(CO₂) = m·CO₂ + b`` [-]; clamped to ``≥ 0``
            when ``clamp_nonnegative`` is set.
        """
        f = self.transpiration_m * co2 + self.transpiration_b
        if self.clamp_nonnegative:
            f = torch.clamp(f, min=0.0)
        return f

    def forward(
        self,
        ptran: torch.Tensor,
        co2: torch.Tensor,
    ) -> dict[str, torch.Tensor]:
        r"""Apply the CO₂ reduction to potential transpiration.

        Args:
            ptran: Potential canopy transpiration [mm d⁻¹], shape ``[B]``.
            co2: Atmospheric CO₂ concentration [ppm], a scalar
                tensor or shape broadcastable to ``[B]``.

        Returns:
            Dict of ``[B]`` tensors:

            * ``ptran`` [mm d⁻¹] — CO₂-reduced potential
              transpiration ``PTRAN · f(CO₂)``.
            * ``co2_factor`` [-] — The applied reduction factor ``f(CO₂)``.
        """
        co2_factor = self.factor(co2)
        return {
            "ptran": ptran * co2_factor,
            "co2_factor": torch.broadcast_to(co2_factor, ptran.shape),
        }

factor(self, co2)

Compute the CO₂ transpiration-reduction factor.

Parameters:

Name	Type	Description	Default
co2	torch.Tensor	Atmospheric CO₂ concentration [ppm], broadcastable to the transpiration tensor.	required

Returns:

Type	Description
torch.Tensor	Reduction factor f(CO₂) = m·CO₂ + b [-]; clamped to ≥ 0 when clamp_nonnegative is set.

Source code in torchcrop/processes/co2_transpiration.py

def factor(self, co2: torch.Tensor) -> torch.Tensor:
    r"""Compute the CO₂ transpiration-reduction factor.

    Args:
        co2: Atmospheric CO₂ concentration [ppm], broadcastable
            to the transpiration tensor.

    Returns:
        Reduction factor ``f(CO₂) = m·CO₂ + b`` [-]; clamped to ``≥ 0``
        when ``clamp_nonnegative`` is set.
    """
    f = self.transpiration_m * co2 + self.transpiration_b
    if self.clamp_nonnegative:
        f = torch.clamp(f, min=0.0)
    return f

forward(self, ptran, co2)

Apply the CO₂ reduction to potential transpiration.

Parameters:

Name	Type	Description	Default
ptran	torch.Tensor	Potential canopy transpiration [mm d⁻¹], shape [B].	required
co2	torch.Tensor	Atmospheric CO₂ concentration [ppm], a scalar tensor or shape broadcastable to [B].	required

Returns:

Type	Description
Dict of ``[B]`` tensors	ptran [mm d⁻¹] — CO₂-reduced potential transpiration PTRAN · f(CO₂). co2_factor [-] — The applied reduction factor f(CO₂).

Source code in torchcrop/processes/co2_transpiration.py

def forward(
    self,
    ptran: torch.Tensor,
    co2: torch.Tensor,
) -> dict[str, torch.Tensor]:
    r"""Apply the CO₂ reduction to potential transpiration.

    Args:
        ptran: Potential canopy transpiration [mm d⁻¹], shape ``[B]``.
        co2: Atmospheric CO₂ concentration [ppm], a scalar
            tensor or shape broadcastable to ``[B]``.

    Returns:
        Dict of ``[B]`` tensors:

        * ``ptran`` [mm d⁻¹] — CO₂-reduced potential
          transpiration ``PTRAN · f(CO₂)``.
        * ``co2_factor`` [-] — The applied reduction factor ``f(CO₂)``.
    """
    co2_factor = self.factor(co2)
    return {
        "ptran": ptran * co2_factor,
        "co2_factor": torch.broadcast_to(co2_factor, ptran.shape),
    }