Phenological benchmarking with a land surface model: an in-silico experiment for temperate forests

Plant phenology affects biotic interactions, water, carbon and nutrient cycling, ultimately influencing global climate, and making understanding and quantifying phenological dynamics central for ecological forecasting. This however varies depending on data source, smoothing methodology or date-extraction method. The selection of an appropriate phenological processing pipeline for the system of study thus requires accurate benchmarking in a controlled environment.

In this study we use a Land Surface Model (LSM) to simulate ecosystem dynamics at four temperate forest sites under five lesser studied climate perturbations. We use the generated synthetic dataset to control for environmental conditions while testing commonly used phenological detection methods (relative thresholds at 10%, 20% and 50%, highest rate of change, detection of inflection points in the curvature’s angle, detection of recovery and senescence points).

Amongst tested methods, we find general agreement in terms of response to climate alterations. Inflection and low thresholds (10%) methods provide the closest match to the model’s predicted start of season, with a maximum discrepancy of 4 days between the two. Higher thresholds (20%, 50%) are closer to the models’ predicted end of season, although with discrepancies of up to 10 days. In addition, high thresholds and methods based on the rate of change of vegetation parameters successfully detect the peak of season. We also show how some methods can pick up on effects that are ignored by temperature-only type growing seasons, thereby supporting the need for increased realism of phenological representations in land surface models.

This is, to our knowledge, the first time an LSM is used to generate an ‘in silico’ experiment. This study not only tests phenological extraction methodologies within a common framework, but demonstrates how land surface models can be used as tools for ecological testing beyond their current use as predictive tools.