INTRODUCTION
Gross primary productivity (GPP) is the rate at which solar energy is utilised by plants to capture the greenhouse gas carbon dioxide (CO2) in the atmosphere. Knowledge of GPP in diverse terrestrial ecosystems is critical for quantifying the feedbacks of natural ecosystems to climate change and predicting future climatic changes (IPCC, 2013; Falkowski et al., 2000; Campbell et al., 2017; Quere et al., 2000).
Eddy covariance (EC) flux towers provide continuous and integrated measurements of net ecosystem CO2 exchange from which the ecosystem GPP can be estimated (Baldocchi et al., 2001; Baldocchi, 2003; Reichstein et al., 2005). Currently, over 900 flux sites across the globe employ the EC technique to study carbon cycling in various ecosystems, and much of this data can be found on the FLUXNET database (fluxdata.org). Although the number of sites with an EC system has been rapidly increasing in recent decades, the monitored sites still account for only a small fraction of the global ecosystems. This is largely due to: (1) the relatively small footprint area of EC (i.e. usually within a radius of 1 km) and (2) the lack of accessibility in many remote areas. Therefore, there is a need for approaches that estimate GPP over extensive areas without EC systems.
Process-based ecological models simulate GPP depending on environmental variables that constrain plant photosynthesis (such as solar radiation, temperature, precipitation, nutrient limits, etc.) and plant-specific parameters (such as the maximum photosynthetic rate). However, these models tend to be complex and need many parameters when compared with empirical models (Loarie et al., 2011; Gustafson, 2013; Gitelson et al., 2016; Jiang & Ryu, 2016; Lin et al., 2020). The different parameterisations used in these models often result in large differences in their estimated carbon dynamics (Cramer & Field, 1999; Lin et al., 2017; Running et al., 2000).
Net radiation, which is the balance of total radiation input/output for an ecosystem, indicates the energy that is retained within the system over a given time period. In theory, the energy retained is partially converted into heat that is transmitted throughout the system with the rest being mostly fixed by plants through photosynthesis (Running et al., 2004; Wu et al., 2010; Xiao et al., 2010). Therefore, we hypothesised that a radiation index (RI) that reflects the proportion of radiation energy fixed through photosynthesis is indicative of GPP. To test this hypothesis, we developed an RI based on the radiation balance and investigated its relationship with GPP at different temporal scales using EC data from 54 sites of diverse terrestrial ecosystems. Due to the easy access of radiation data globally through satellite-based remote-sensing products (Monteith, 1972; Heinsch et al., 2003; Sjöström et al., 2013; Xin et al., 2015), this radiation-based approach would be an important alternative for GPP estimation, upscaling and prediction and has the potential to be implemented globally.