Evapotranspiration measurements on an eroded cropland using an automated and mobile chamber system: gap filling strategies and impact of soil type and topsoil modification

In light of the ongoing global change in climatic conditions and a related trend to increases in extreme hydrological events, it is increasingly crucial to assess ecosystem resilience and - in agricultural systems - to ensure sustainable management and food security.  A comprehensive understanding of ecosystem water cycle budgets and spatio-temporal dynamics is indispensable. Evapotranspiration (ET) plays a pivotal role returning up to 90 % of ingoing precipitation back to the atmosphere. Here, we studied impacts of soil types and management on an agroecosystem's water budgets and agronomic water use efficiencies (WUE_agro). To do so, a plot experiment with winter rye (September 17, 2020 to June 30, 2021) was conducted at an eroded cropland which is located in the hilly and dry ground moraine landscape of the Uckermark region in NE Germany. Along the experimental plot (110 m x 16 m), a gantry crane mounted mobile and automated two chamber system (FluxCrane as part of the AgroFlux platform within the CarboZALF-D research site) was used for the first time to continuously measure water fluxes and determine evapotranspiration. Three soil types representing the soil erosion gradient related to the hummocky ground moraine landscape (extremely eroded: Calcaric Regosol, strongly eroded: Nudiargic Luvisol, non-eroded: Calcic Luvisol) and additional soil manipulation (topsoil removal and subsoil admixture) were investigated (randomized block design, 3 replicates per treatment). Five different gap-filling approaches were used and compared in light of their potential for reliable water budgets over the entire crop growth period as well as reproduce short-term (day-to-day, diurnal) water flux dynamics. The best calibration performance was achieved with approaches based on machine learning, such as support vector machine (SVM) and artificial neural networks (with Bayesian regularization; ANN_BR), while especially SVM yielded in most reliable predictions of measured ET during validation.

We found significant differences in dry biomass (DM) and minor in evapotranspiration between soil types, resulting in different water use efficiencies (WUE_agro). The Calcaric Regosol (extremely eroded) shows a maximum of around 37% lower evapotranspiration and a maximum of around 52% lower water use efficiency (WUE_agro) compared to the non-eroded Calcic Luvisol.  The key period contributing to ~ 70% of overall ET of the entire growth period was from April until June (harvest), however differences in the overall ET budget (ET_sum) between soil types and manipulation resulted predominantly from small differences between the treatments over the entire growth period.