Assessing and Modeling Ecosystem Carbon Exchange and Water Vapor Flux of a Pasture Ecosystem in the Temperate Climate-Transition Zone

Abstract

The rising frequency of extreme weather events and global warming are greatly challenging pastoral ecosystem productivity, particularly in the temperate climate-transition regions. While this could cause greater gross primary production (GPP) mainly contributed by the warm-season vegetation, the consequences for the dynamics of net ecosystem exchange (NEE) and hydrological responses (e.g., evapotranspiration, ET) on an ecosystem level are poorly known. Here, we investigated the evolution of plant phenology, nutritive value, energy balance, and carbon/water budgets of a cool-season dominated pastoral ecosystem in the temperate zone; integrating both eddy covariance (EC) flux measurement and simulation modeling-based uncertainty analysis. Throughout the two-year duration (2017–2018) of this study, the entire pasture ecosystem remained a strong carbon sink (NEE = −1.23 and −1.95 kg C m−2, respectively) with 74% and 62% of available energy loss explained by EC fluxes, respectively. The cumulative ET was 735.8 and 796.8 mm, respectively; and the overall ecosystem water use efficiency (EWUE) were calculated as 6.5 g C kg−1 water across both growing seasons. The above-ground biomass yield agreed with the cumulative GPP and was inversely correlated with grass nutritive value. The uncertainty analysis indicated that accurate EC flux gap-filling models could be constructed using support vector machine trained time-series models (NEE, R2 = 0.77, RMSE = 11.8; ET, R2 = 0.90, RMSE = 73.8). The performance benchmarking tests indicated that REddyProc-based gap-filling performance was very limiting and highly variable (NEE, R2 = 0.21–0.64; ET, R2 = 0.79–0.87), particularly for estimating NEE. Overall, the warm-season vegetation encroachment greatly filled the production gap of cool-season grasses, leading to greater cumulative NEE and EWUE on a system level, compared with those from many other reported field-crop or grassland studies using EC approaches. The complex and dynamic nature of grassland ecosystems greatly challenged the conventional REddyProc-based EC flux gap-filling performance. However, accurate machine learning models could be constructed for error/uncertainty control purposes and, thus, should be encouraged in future studies.

Machine Learning and Regression Analysis Reveal Different Patterns of Influence on Net Ecosystem Exchange at Two Conifer Woodland Sites

“…• Adaptive boosting-ADA [59,60] • Decision tree-DT [61,62] • K-nearest neighbor-KNN [63,64] • Multi-layer perceptron-MLP (artificial neural network) [65][66][67] • Random forest-RF [30,[68][69][70] • Support-vector regressor-SVR [71][72][73] • Extreme gradient boosting-XGB [74][75][76] These ML algorithms apply distinctive methodologies, making it useful to compare their results when applied to complex datasets. They can be categorized as regression-based (SVR), single-tree (DT), ensemble-tree (ADA, RF, XGB), data-matching (KNN) and neural-network (MLP) algorithms.…”