Modeling and Predicting Carbon and Water Fluxes Using Data-Driven Techniques in a Forest Ecosystem

Hybrid AI model for estimating the canopy photosynthesis of eggplants

Modern models for estimating canopy photosynthetic rates (A_c) can be broadly classified into two categories, namely, process-based mechanistic models and artificial intelligence (AI) models, each category having unique strengths (i.e., process-based models have generalizability to a wide range of situations, and AI models can reproduce a complex process using data without prior knowledge about the underlying mechanism). To exploit the strengths of both categories of models, a novel "hybrid" canopy photosynthesis model that combines process-based models with an AI model was proposed. In the proposed hybrid model, process-based models for single-leaf photosynthesis and image analysis first transform raw inputs (environmental data and canopy images) into the single-leaf photosynthetic rate (A_L) and effective leaf area index (L_c)), after which A_L and L_c are fed into an artificial neural network (ANN) model to predict A_c. The hybrid model successfully predicted the diurnal cycles of A_c of an eggplant canopy even with a small training dataset and successfully reproduced a typical A_c response to changes in the CO₂ concentration outside the range of the training data. The proposed hybrid AI model can provide an effective means to estimate A_c in actual crop fields, where obtaining a large amount of training data is difficult.

Predicting carbon and water vapor fluxes using machine learning and novel feature ranking algorithms

Gap-filling eddy covariance flux data using quantitative approaches has increased over the past decade. Numerous methods have been proposed previously, including look-up table approaches, parametric methods, process-based models, and machine learning. Particularly, the REddyProc package from the Max Planck Institute for Biogeochemistry and ONEFlux package from AmeriFlux have been widely used in many studies. However, there is no consensus regarding the optimal model and feature selection method that could be used for predicting different flux targets (Net Ecosystem Exchange, NEE; or Evapotranspiration -ET), due to the limited systematic comparative research based on the identical site-data. Here, we compared NEE and ET gap-filling/prediction performance of the least-square-based linear model, artificial neural network, random forest (RF), and support vector machine (SVM) using data obtained from four major row-crop and forage agroecosystems located in the subtropical or the climate-transition zones in the US. Additionally, we tested the impacts of different training-testing data partitioning settings, including a 10-fold time-series sequential (10FTS), a 10-fold cross validation (CV) routine with single data point (10FCV), daily (10FCVD), weekly (10FCVW) and monthly (10FCVM) gap length, and a 7/14-day flanking window (FW) approach; and implemented a novel Sliced Inverse Regression-based Recursive Feature Elimination algorithm (SIRRFE). We benchmarked the model performance against REddyProc and ONEFlux-produced results. Our results indicated that accurate NEE and ET prediction models could be systematically constructed using SVM/RF and only a few top informative features. The gap-filling performance of ONEFlux is generally satisfactory (R² = 0.39-0.71), but results from REddyProc could be very limited or even unreliable in many cases (R² = 0.01-0.67). Overall, SIRRFE-refined SVM models yielded excellent results for predicting NEE (R² = 0.46-0.92) and ET (R² = 0.74-0.91). Finally, the performance of various models was greatly affected by the types of ecosystem, predicting targets, and training algorithms; but was insensitive towards training-testing partitioning. Our research provided more insights into constructing novel gap-filling models and understanding the underlying drivers affecting boundary layer carbon/water fluxes on an ecosystem level.

CO2 Flux Characteristics of Different Plant Communities in a Subtropical Urban Ecosystem

Shanghai, China, is a city that is relatively representative of various cities in China due to its geographical location and socio-economic dynamics. The role of urban vegetation in the carbon cycle of urban developments in these types of cities is now being studied. We focus on identifying which urban plant community types have a greater influence on CO2 flux in cities, thus providing a scientific basis for low-carbon urban greening. Based on the eddy covariance (EC) observation system, ART Footprint Tool, plant inventory, and ecological community classification, we show that the CO2 flux characteristics of different plant communities vary temporally. The carbon sink duration during summer was the longest (up to 10 h) and the carbon sink duration was the shortest during winter (7.5 h). In addition, we discovered that the CO2 flux contribution rates of different plant community types are distinct. The annual average CO2 contribution rates of the Cinnamomum camphora-Trachycarpus fortunei community and the Metasequoia glyptostroboides+Sabina chinensis community are 11.88% and 0.93%, respectively. The CO2 flux contribution rate of the same plant community differs according to season. The CO2 contribution rate of the Cinnamomum camphora-Trachycarpus fortunei community exhibits local maxima during winter and summer, with a maximum difference of 11.16%. In contrast, the Metasequoia glyptostroboides+Sabina chinensis community has a CO2 contribution rate of 0.35% during the same period. In general, summer is the season with the lowest CO2 flux contribution rate of plant communities, and winter is the season with the highest CO2 flux contribution rate. However, the Cinnamomum camphora+Salix babylonica community and the Cinnamomum camphora+Sabina chinensis community present the opposite pattern. Finally, the diurnal variation characteristics of CO2 flux in different communities have the same trend, but the peak values differ significantly. Overall, daily CO2 flux peak value of the Metasequoia glyptostroboides community and the Cinnamomum camphora-Trachycarpus fortunei community indicate that these two plant communities exhibit a strong capacity for CO2 absorption in the study area. According to these research results, urban greening efforts in subtropical climates can increase the green space covered by the Cinnamomum camphora-Trachycarpus fortunei and the Metasequoia glyptostroboides community types when urban greening, so as to appropriately reduce the CO2 emitted into the atmosphere.

Modeling and estimating aboveground biomass of Dacrydium pierrei in China using machine learning with climate change

Accurate estimations of the aboveground biomass (AGB) of rare and endangered species are particularly important for protecting forest ecosystems and endangered species and for providing useful information to analyze the influence of past and future climate change on forest AGB. We investigated the feasibility of using three developed and two widely used models, including a generalized regression neural network (GRNN), a group method of data handling (GMDH), an adaptive neuro-fuzzy inference system (ANFIS), an artificial neural network (ANN) and a support vector machine (SVM), to estimate the AGB of Dacrydium pierrei (D. pierrei) in natural forests of China. The results showed that these models could explain the changes in the AGB of the D. pierrei using a limited amount of meteorological data. The GRNN and ANN models are superior to the other models for estimating the AGB of D. pierrei. The GMDH model consistently produced comparatively poor estimates of the AGB. Three climate scenarios, including the representative concentration pathway (RCP) 2.6, RCP 4.5, and RCP 8.5, were compared with the climate situation of 2013-2017. Under these scenarios, the AGB of D. pierrei females with the same diameter at breast height (DBH) would increase by 13.0 ± 31.4% (mean ± standard deviation), 16.6 ± 30.7%, and 18.5 ± 30.9% during 2041-2060 and 15.6 ± 32.1%, 21.2 ± 33.2%, and 24.8 ± 32.7% during 2061-2080; the AGB of males would increase by 16.3 ± 32.3%, 21.7 ± 32.5%, and 22.9 ± 32.6% during 2041-2060 and 22.3 ± 30.8%, 27.2 ± 31.8%, and 30.1 ± 34.4% during 2061-2080. The R² values of all models range from 0.82 to 0.95. In conclusion, this study suggests that these advanced models are recommended to estimate the AGB of forests, and the AGB of forests would increase in 2041-2080 under future climate scenarios.

Estimating forest carbon fluxes using four different data-driven techniques based on long-term eddy covariance measurements: Model comparison and evaluation

With the recent availability of large amounts of data from the global flux towers across different terrestrial ecosystems based on the eddy covariance technique, the use of data-driven techniques has been viable. In this study, two advanced techniques, namely adaptive neuro-fuzzy inference system (ANFIS) and extreme learning machine (ELM), were developed and investigated for their viability in estimating daily carbon fluxes at the ecosystem level. All the data used in this study were based upon the long-term chronosequence observations derived from the flux towers in eight forest ecosystems. Both ANFIS and ELM methods were further compared with the most widely used artificial neural network (ANN) and support vector machine (SVM) methods. Moreover, we also focused on probing into the effects of internal parameters on their corresponding approaches. Our estimates showed that most variation in each carbon flux could be effectively explained by the developed models at almost all the sites. Moreover, the forecasting accuracy of each method was strongly dependent upon their respective internal algorithms. The best training function for ANN model can be acquired through the trial and error procedure. The SVM model with the radial basis kernel function performed considerably better than the SVM models with the polynomial and sigmoid kernel functions. The hybrid ELM models achieved similar predictive accuracy for the three fluxes and were consistently superior to the original ELM models with different transfer functions. In most instances, both the subtractive clustering and fuzzy c-means algorithms for the ANFIS models outperformed the most popular grid partitioning algorithm. It was demonstrated that the newly proposed ELM and ANFIS models were able to produce comparable estimates to the ANN and SVM models for forecasting terrestrial carbon fluxes.