Integrating Radiative Transfer with Ecosystem Models to Reflect Litter Dynamic and Optical Vegetation Properties in Semi-Arid Grasslands

Savannas, characterized by scattered trees with a grass layer, are key ecosystems in semi-arid regions. They profoundly influence global carbon (C) and water fluxes through high seasonal and inter-annual variations. Understanding these dynamics at the ecosystem scale is essential for better representing their impacts on the Earth’s climate system. Terrestrial ecosystem models (TEM) such as QUINCY (QUantifying Interactions between terrestrial Nutrient CYcles and the climate system), is a new generation TEM that integrates C, nitrogen (N), and phosphorus (P) cycles, are essential for assessing ecosystem responses to climate variability and extremes. However, these models' complexity and reliance on site-specific parameters can limit predictive accuracy, especially in complex ecosystems such as savannas.

Remote sensing (RS) images can be leveraged to improve TEM predictions (e.g., assimilation of RS data) when radiative transfer models (RTM) are coupled with TEMs. In semi-arid grasslands and savannas, the mixture of green and senescent vegetation challenges RS-based vegetation property retrieval. To address this, we integrated senSCOPE, an advanced version of the Soil-Canopy Observation of Photosynthesis and Energy fluxes (SCOPE) RTM that separately simulates green and senescent leaves, with QUINCY to improve the representation of absorbed photosynthetically active radiation (aPAR) and therefore photosynthesis and ecosystem dynamics using RS data.

To leverage the high computational demands of complex RTMs such as senSCOPE, we further developed simplified RTMs based on a two-leaf approach to maintain computational efficiency within QUINCY. These submodels can improve the representation of senescent material in nutrient cycling, thereby improving our understanding of ecosystem processes such as biomass production and litter decomposition. We evaluated the outputs, including gross primary productivity, aPAR, and albedo, against the standard QUINCY model over green and senescent material leaf area fractions using the goodness of fit measures (root mean square error, mean error, and mean absolute error).

By integrating the two-leaf-based advanced RTM and computationally efficient submodels within QUINCY, we achieved a more accurate and cost-effective representation of senescent material in grasslands, respectively.