Partitioning net carbon dioxide fluxes into photosynthesis and respiration using neural networks

Machine learning in photosynthesis: Prospects on sustainable crop development

Improving photosynthesis is a promising avenue to increase food security. Studying photosynthetic traits with the aim to improve efficiency has been one of many strategies to increase crop yield but analyzing large data sets presents an ongoing challenge. Machine learning (ML) represents a ubiquitous tool that can provide a more elaborate data analysis. Here we review the application of ML in various domains of photosynthetic research, as well as in photosynthetic pigment studies. We highlight how correlating hyperspectral data with photosynthetic parameters to improve crop yield could be achieved through various ML algorithms. We also propose strategies to employ ML in promoting photosynthetic pigment research for furthering crop yield.

Mapping Chinese annual gross primary productivity with eddy covariance measurements and machine learning

Annual gross primary productivity (AGPP) is the basis for grain production and terrestrial carbon sequestration. Mapping regional AGPP from site measurements provides methodological support for analysing AGPP spatiotemporal variations thereby ensures regional food security and mitigates climate change. Based on 641 site-year eddy covariance measuring AGPP from China, we built an AGPP mapping scheme based on its formation and selected the optimal mapping way, which was conducted through analysing the predicting performances of divergent mapping tools, variable combinations, and mapping approaches in predicting observed AGPP variations. The reasonability of the selected optimal scheme was confirmed by assessing the consistency between its generating AGPP and previous products in spatiotemporal variations and total amount. Random forest regression tree explained 85 % of observed AGPP variations, outperforming other machine learning algorithms and classical statistical methods. Variable combinations containing climate, soil, and biological factors showed superior performance to other variable combinations. Mapping AGPP through predicting AGPP per leaf area (PAGPP) explained 86 % of AGPP variations, which was superior to other approaches. The optimal scheme was thus using a random forest regression tree, combining climate, soil, and biological variables, and predicting PAGPP. The optimal scheme generating AGPP of Chinese terrestrial ecosystems decreased from southeast to northwest, which was highly consistent with previous products. The interannual trend and interannual variation of our generating AGPP showed a decreasing trend from east to west and from southeast to northwest, respectively, which was consistent with data-oriented products. The mean total amount of generated AGPP was 7.03 ± 0.45 PgC yr^-1 falling into the range of previous works. Considering the consistency between the generated AGPP and previous products, our optimal mapping way was suitable for mapping AGPP from site measurements. Our results provided a methodological support for mapping regional AGPP and other fluxes.

Improved estimation of global gross primary productivity during 1981-2020 using the optimized P model

Accurate estimation of terrestrial gross primary productivity (GPP) is essential for quantifying the net carbon exchange between the atmosphere and biosphere. Light use efficiency (LUE) models are widely used to estimate GPP at different spatial scales. However, difficulties in proper determination of maximum LUE (LUE_max) and downregulation of LUE_max into actual LUE result in uncertainties in GPP estimated by LUE models. The recently developed P model, as a LUE-like model, captures the deep mechanism of photosynthesis and simplifies parameterization. Site level studies have proved the outperformance of P model over LUE models. However, the global application of the P model is still lacking. Thus, the effectiveness of 5 water stress factors integrated into the P model was compared. The optimal P model was used to generate a new long-term (1981-2020) global monthly GPP dataset at a spatial resolution of 0.1° × 0.1°, called PGPP. Validation at globally distributed 109 FLUXNET sites indicated that PGPP is better than three widely-used GPP products. R² between PGPP and observed GPP equals to 0.75, the corresponding root mean squared error (RMSE) and mean absolute error (MAE) equal to 1.77 g C m^-2 d^-1 and 1.28 g C m^-2 d^-1. During the period from 1981 to 2020, PGPP significantly increased in 69.02% of global vegetated regions (p < 0.05). Overall, PGPP provides a new GPP product choice for global ecology studies and the comparison of various water stress factors provides a new idea for the improvement of GPP model in the future.

Intercomparison of methods to estimate gross primary production based on CO2 and COS flux measurements

Separating the components of ecosystem-scale carbon exchange is crucial in order to develop better models and future predictions of the terrestrial carbon cycle. However, there are several uncertainties and unknowns related to current photosynthesis estimates. In this study, we evaluate four different methods for estimating photosynthesis at a boreal forest at the ecosystem scale, of which two are based on carbon dioxide (CO₂) flux measurements and two on carbonyl sulfide (COS) flux measurements. The CO₂-based methods use traditional flux partitioning and artificial neural networks to separate the net CO₂ flux into respiration and photosynthesis. The COS-based methods make use of a unique 5-year COS flux data set and involve two different approaches to determine the leaf-scale relative uptake ratio of COS and CO₂ (LRU), of which one (LRU_CAP) was developed in this study. LRU_CAP was based on a previously tested stomatal optimization theory (CAP), while LRU_PAR was based on an empirical relation to measured radiation. For the measurement period 2013-2017, the artificial neural network method gave a GPP estimate very close to that of traditional flux partitioning at all timescales. On average, the COS-based methods gave higher GPP estimates than the CO₂-based estimates on daily (23% and 7% higher, using LRU_PAR and LRU_CAP, respectively) and monthly scales (20% and 3% higher), as well as a higher cumulative sum over 3 months in all years (on average 25% and 3% higher). LRU_CAP was higher than LRU estimated from chamber measurements at high radiation, leading to underestimation of midday GPP relative to other GPP methods. In general, however, use of LRU_CAP gave closer agreement with CO₂-based estimates of GPP than use of LRUPAR. When extended to other sites, LRU_CAP may be more robust than LRU_PAR because it is based on a physiological model whose parameters can be estimated from simple measurements or obtained from the literature. In contrast, the empirical radiation relation in LRU_PAR may be more site-specific. However, this requires further testing at other measurement sites.

Cross-biome synthesis of source versus sink limits to tree growth

Uncertainties surrounding tree carbon allocation to growth are a major limitation to projections of forest carbon sequestration and response to climate change. The prevalence and extent to which carbon assimilation (source) or cambial activity (sink) mediate wood production are fundamentally important and remain elusive. We quantified source-sink relations across biomes by combining eddy-covariance gross primary production with extensive on-site and regional tree ring observations. We found widespread temporal decoupling between carbon assimilation and tree growth, underpinned by contrasting climatic sensitivities of these two processes. Substantial differences in assimilation-growth decoupling between angiosperms and gymnosperms were determined, as well as stronger decoupling with canopy closure, aridity, and decreasing temperatures. Our results reveal pervasive sink control over tree growth that is likely to be increasingly prominent under global climate change.

A novel approach to partitioning evapotranspiration into evaporation and transpiration in flooded ecosystems

Reliable partitioning of micrometeorologically measured evapotranspiration (ET) into evaporation (E) and transpiration (T) would greatly enhance our understanding of the water cycle and its response to climate change related shifts in local-to-regional climate conditions and rising global levels of vapor pressure deficit (VPD). While some methods on ET partitioning have been developed, their underlying assumptions make them difficult to apply more generally, especially in sites with large contributions of E. Here, we report a novel ET partitioning method using artificial neural networks (ANNs) in combination with a range of environmental input variables to predict daytime E from nighttime ET measurements. The study uses eddy covariance data from four restored wetlands in the Sacramento-San Joaquin Delta, California, USA, as well as leaf-level T data for validation. The four wetlands vary in their vegetation make-up and structure, representing a range of ET conditions. The ANNs were built with increasing complexity by adding the input variable that resulted in the next highest average value of model testing R² across all sites. The order of variable inclusion (and importance) was: VPD > gap-filled sensible heat flux (H_gf) > air temperature (T_air ) > friction velocity (u_∗ ) > other variables. The model using VPD, H_gf, T_air , and u_∗ showed the best performance during validation with independent data and had a mean testing R²  value of 0.853 (averaged across all sites, range from 0.728 to 0.910). In comparison to other methods, our ANN method generated T/ET partitioning results which were more consistent with CO₂ exchange data especially for more heterogeneous sites with large E contributions. Our method improves the understanding of T/ET partitioning. While it may be particularly suited to flooded ecosystems, it can also improve T/ET partitioning in other systems, increasing our knowledge of the global water cycle and ecosystem functioning.

Estimating Gross Primary Productivity (GPP) over Rice–Wheat-Rotation Croplands by Using the Random Forest Model and Eddy Covariance Measurements: Upscaling and Comparison with the MODIS Product

Despite advances in remote sensing–based gross primary productivity (GPP) modeling, the calibration of the Moderate Resolution Imaging Spectroradiometer (MODIS) GPP product (GPPMOD) is less well understood over rice–wheat-rotation cropland. To improve the performance of GPPMOD, a random forest (RF) machine learning model was constructed and employed over the rice–wheat double-cropping fields of eastern China. The RF-derived GPP (GPPRF) agreed well with the eddy covariance (EC)-derived GPP (GPPEC), with a coefficient of determination of 0.99 and a root-mean-square error of 0.42 g C m−2 d−1. Therefore, it was deemed reliable to upscale GPPEC to regional scales through the RF model. The upscaled cumulative seasonal GPPRF was higher for rice (924 g C m−2) than that for wheat (532 g C m−2). By comparing GPPMOD and GPPEC, we found that GPPMOD performed well during the crop rotation periods but underestimated GPP during the rice/wheat active growth seasons. Furthermore, GPPMOD was calibrated by GPPRF, and the error range of GPPMOD (GPPRF minus GPPMOD) was found to be 2.5–3.25 g C m−2 d−1 for rice and 0.75–1.25 g C m−2 d−1 for wheat. Our findings suggest that RF-based GPP products have the potential to be applied in accurately evaluating MODIS-based agroecosystem carbon cycles at regional or even global scales.

Coupling Photosynthetic Measurements with Biometric Data to Estimate Gross Primary Productivity (GPP) in Mediterranean Pine Forests of Different Post-Fire Age

Quantification of forest Gross Primary Productivity (GPP) is important for understanding ecosystem function and designing appropriate carbon mitigation strategies. Coupling forest biometric data with canopy photosynthesis models can provide a means to simulate GPP across different stand ages. In this study we developed a simple framework to integrate biometric and leaf gas-exchange measurements, and to estimate GPP across four Mediterranean pine forests of different post-fire age. We used three different methods to estimate the Leaf Area Index (LAI) of the stands, and monthly gas exchange data to calibrate the photosynthetic light response of the leaves. Upscaling of carbon sequestration at the canopy level was made by implementing a Big Leaf and a Sun/Shade model, using both average and variant (monthly) photosynthetic capacity values. The Big Leaf model simulations systematically underestimated GPP compared to the Sun/Shade model simulations. Our simulations suggest an increasing GPP with age up to a stand maturity stage. The shape of the GPP trend with stand age was not affected by the method used to parameterise the model. At the scale of our study, variability in stand and canopy structure among the study sites seems to be the key determinant of GPP.