<p>Photosynthesis (gross primary production, GPP) and evapo-transpiration (ET) are ecosystem processes with global significance for the carbon cycle, climate, hydrology and a range of ecosystem services. The mechanisms governing these processes are complex but well understood. There is strong coupling between these processes, mediated directly by stomatal conductance and indirectly by root zone soil moisture. This coupling must be effectively modelled for robust predictions of earth system responses to global change. It is highly demanding to model cellular processes, like stomatal conductance or electron transport, with responses times of minutes, over decadal and global domains. computational demand means models resolving this level of complexity cannot be fully evaluated for their parameter sensitivity, nor calibrated using earth observation data through data assimilation approaches requiring large ensembles. To resolve this problem, here we describe a coupled photosynthesis evapo-transpiration model of intermediate complexity. The model reduces computational load and parameter numbers by operating at canopy scale and daily time steps. But by including simplified representation of key process interactions it retains sensitivity to variation in climate, leaf traits, soil states and atmospheric CO<sub>2</sub>. The new model is calibrated to match the biophysical responses of a complex terrestrial ecosystem model (TEM) of GPP and ET through a Bayesian model-data fusion process. The calibrated ACM-GPP-ET generates unbiased estimates of TEM GPP and ET, and captures 80–95 % percent of the sensitivity of carbon and water fluxes by the complex TEM. The ACM-GPP-ET model operates ∼ 2200 times faster than the complex TEM. Independent evaluation of ACM-GPP-ET at FLUXNET sites, using a single global parameterisation, shows good agreement with typical R<sup>2</sup> ∼ 0.60 for both GPP and ET. This intermediate complexity modelling approach allows full Monte Carlo based quantification of model parameter and structural uncertainties, global scale sensitivity analyses for these processes, and is fast enough for use within terrestrial ecosystem model-data fusion frameworks requiring large ensembles.</p>