Water, Energy, and Carbon with Artificial Neural Networks (WECANN): A statistically-based estimate of global surface turbulent fluxes and gross primary productivity using solar-induced fluorescence

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.

A global 0.05° dataset for gross primary production of sunlit and shaded vegetation canopies from 1992 to 2020

Distinguishing gross primary production of sunlit and shaded leaves (GPP_sun and GPP_shade) is crucial for improving our understanding of the underlying mechanisms regulating long-term GPP variations. Here we produce a global 0.05°, 8-day dataset for GPP, GPP_shade and GPP_sun over 1992–2020 using an updated two-leaf light use efficiency model (TL-LUE), which is driven by the GLOBMAP leaf area index, CRUJRA meteorology, and ESA-CCI land cover. Our products estimate the mean annual totals of global GPP, GPP_sun, and GPP_shade over 1992–2020 at 125.0 ± 3.8 (mean ± std) Pg C a⁻¹, 50.5 ± 1.2 Pg C a⁻¹, and 74.5 ± 2.6 Pg C a⁻¹, respectively, in which EBF (evergreen broadleaf forest) and CRO (crops) contribute more than half of the totals. They show clear increasing trends over time, in which the trend of GPP (also GPP_sun and GPP_shade) for CRO is distinctively greatest, and that for DBF (deciduous broadleaf forest) is relatively large and GPP_shade overwhelmingly outweighs GPP_sun. This new dataset advances our in-depth understanding of large-scale carbon cycle processes and dynamics.

Measurement(s)	gross primary production of sunlit and shaded vegetation canopies
Technology Type(s)	the revised TL-LUE model
Sample Characteristic - Environment	carbon cycling
Sample Characteristic - Location	terrestrial ecosystem

Ensemble Machine Learning Outperforms Empirical Equations for the Ground Heat Flux Estimation with Remote Sensing Data

Estimating evapotranspiration at the field scale is a major component of sustainable water management. Due to the difficulty to assess some major unknowns of the water cycle at that scale, including irrigation amounts, evapotranspiration is often computed as the residual of the instantaneous surface energy budget. One of the Surface Energy Balance components with the largest uncertainties in their quantification over bare soils and sparse vegetation areas is the ground heat flux (G). Over the last decades, the estimation of G with remote sensing (RS) data has been mainly achieved with empirical equations, on the basis of the G and net radiation (Rn) ratio, G/Rn. The G/Rn empirical equations generally require vegetation data (Type I empirical equations), in combination with surface temperature (Ts) and albedo (Type II empirical equations). In this article, we aim to evaluate the estimation of G with RS data. Here, we compared eight G/Rn empirical equations against two types of machine learning (ML) methods: an ensemble ML type, the Random Forest (RF), and the Neural Networks (NN). The comparison of each method was evaluated using a wide range of climate and land cover datasets, including data from Eddy-Covariance towers that extend along the mid-latitude areas that encompass the European and African continents. Our results have shown evidence that the driver of G in bare soils and sparse vegetation areas (Fraction of Vegetation, Fv ≤ 0.25) is Ts, instead of vegetation greenness indexes. On the other hand, the accuracy in the estimation of G with Rn, Ts or Fv decreases in densely vegetated areas (Fv ≥ 0.50). There are no significant differences between the most accurate Type I and II empirical equations. For bare soils and sparse vegetation areas the empirical equation which combines the Leaf Area Index (LAI) and Ts (E7) estimates G best. In densely vegetated areas, an exponential empirical equation based on Fv (E4), shows the best performance. However, ML better estimates G than the empirical equations, independently of the Fv ranges. An RF model with Rn, LAI and Ts as predictor variables shows the best accuracy and performance metrics, outperforming the NN model.

Estimation of Global Cropland Gross Primary Production from Satellite Observations by Integrating Water Availability Variable in Light-Use-Efficiency Model

Satellite-based models have been widely used to estimate gross primary production (GPP) of terrestrial ecosystems. Although they have many advantages for mapping spatiotemporal variations of regional or global GPP, the performance in agroecosystems is relatively poor. In this study, a light-use-efficiency model for cropland GPP estimation, named EF-LUE, driven by remote sensing data, was developed by integrating evaporative fraction (EF) as limiting factor accounting for soil water availability. Model parameters were optimized first using CO2 flux measurements by eddy covariance system from flux tower sites, and the optimized parameters were further spatially extrapolated according to climate zones for global cropland GPP estimation in 2001–2019. The major forcing datasets include the fraction of absorbed photosynthetically active radiation (FAPAR) data from the Copernicus Global Land Service System (CGLS) GEOV2 dataset, EF from the ETMonitor model, and meteorological forcing variables from ERA5 data. The EF-LUE model was first evaluated at flux tower site-level, and the results suggested that the proposed EF-LUE model and the LUE model without using water availability limiting factor, both driven by flux tower meteorology data, explained 82% and 74% of the temporal variations of GPP across crop sites, respectively. The overall KGE increased from 0.73 to 0.83, NSE increased from 0.73 to 0.81, and RMSE decreased from 2.87 to 2.39 g C m−2 d−1 in the estimated GPP after integrating EF in the LUE model. These improvements may be largely attributed to parameters optimized for different climatic zones and incorporating water availability limiting factor expressed by EF into the light-use-efficiency model. At global scale, the verification by GPP measurements from cropland flux tower sites showed that GPP estimated by the EF-LUE model driven by ERA5 reanalysis meteorological data and EF from ETMonitor had overall the highest R2, KGE, and NSE and the smallest RMSE over the four existing GPP datasets (MOD17 GPP, revised EC-LUE GPP, GOSIF GPP and PML-V2 GPP). The global GPP from the EF-LUE model could capture the significant negative GPP anomalies during drought or heat-wave events, indicating its ability to express the impacts of the water stress on cropland GPP.

Intercomparison of global terrestrial carbon fluxes estimated by MODIS and Earth system models

Earth system models (ESMs) have been widely used to simulate global terrestrial carbon fluxes, including gross primary production (GPP) and net primary production (NPP). Assessment of such GPP and NPP products can be valuable for understanding the efficacy of certain ESMs in simulating the global carbon cycle and future climate impacts. In this work, we studied the model performance of 22 ESMs participating in the fifth and sixth phases of the Coupled Model Intercomparison Project (CMIP5 and CMIP6) by comparing historical GPP and NPP simulations with satellite data from MODIS and further evaluating potential model improvement from CMIP5 to CMIP6. In CMIP6, the average global total GPP and NPP estimated by the 22 ESMs are 16% and 13% higher than MODIS data, respectively. The multi-model ensembles (MME) of the 22 ESMs can fairly reproduce the spatial distribution, zonal distribution and seasonal variations of both GPP and NPP from MODIS. They perform much better in simulating GPP and NPP for grasslands, wetlands, croplands and other biomes than forests. However, there are noticeable differences among individual ESM simulations in terms of overall fluxes, temporal and spatial flux distributions, and fluxes by biome and region. The MME consistently outperforms all individual models in nearly every respect. Even though several ESMs have been improved in CMIP6 relative to CMIP5, there is still much work to be done to improve individual ESM and overall CMIP performance. Future work needs to focus on more comprehensive model mechanisms and parametrizations, higher resolution and more reasonable coupling of land surface schemes and atmospheric/oceanic schemes.

Canopy Solar-Induced Chlorophyll Fluorescence and Its Link to Transpiration in a Temperate Evergreen Needleleaf Forest during the Fall Transition

Northern hemisphere evergreen needleleaf forest (ENF) contributes a significant fraction of global water exchange but regional transpiration (T) observation in ENF ecosystems is still challenging. Traditional remote sensing techniques and terrestrial biosphere models reproduce the transpiration seasonality with difficulty, and with large uncertainties. Solar-induced chlorophyll fluorescence (SIF) emission from vegetation correlates to photosynthesis at multiple spatial and temporal scales. However, how SIF links to transpiration of evergreen forest during seasonal transition is unclear. Here, we explored the relationship between canopy SIF and T retrieved from ground observation towers in ENF. We also examined the role of meteorological and soil factors on the relationship between SIF and T. A slow decrease of SIF and T with a fast reduction in photosynthetically active radiation (PAR), air temperature, vapor pressure deficit (VPD), soil temperature and soil water content (SWC) were found in the ENF during the fall transition. The correlation between SIF and T at hourly and daily scales varied significantly among different months (Pearson correlation coefficient = 0.29–0.68, p < 0.01). SIF and T were significantly linearly correlated at hourly (R2 = 0.53, p < 0.001) and daily (R2 = 0.67, p < 0.001) timescales in the October. Air temperature and PAR were the major moderating factors for the relationship between SIF and T in the fall transition. Soil water content (SWC) influenced the SIF-T relationship at an hourly scale. Soil temperature and VPD’s effect on the SIF-T relationship was evident at a daily scale. This study can help extend the possibility of constraining ecosystem T by SIF at an unprecedented spatiotemporal resolution during season transitions.

Interannual and Seasonal Drivers of Carbon Cycle Variability Represented by the Community Earth System Model (CESM2)

Earth system models are intended to make long-term projections, but they can be evaluated at interannual and seasonal time scales. Although the Community Earth System Model (CESM2) showed improvements in a number of terrestrial carbon cycle benchmarks, relative to its predecessor, our analysis suggests that the interannual variability (IAV) in net terrestrial carbon fluxes did not show similar improvements. The model simulated low IAV of net ecosystem production (NEP), resulting in a weaker than observed sensitivity of the carbon cycle to climate variability. Low IAV in net fluxes likely resulted from low variability in gross primary productivity (GPP)-especially in the tropics-and a high covariation between GPP and ecosystem respiration. Although lower than observed, the IAV of NEP had significant climate sensitivities, with positive NEP anomalies associated with warmer and drier conditions in high latitudes, and with wetter and cooler conditions in mid and low latitudes. We identified two dominant modes of seasonal variability in carbon cycle flux anomalies in our fully coupled CESM2 simulations that are characterized by seasonal amplification and redistribution of ecosystem fluxes. Seasonal amplification of net and gross carbon fluxes showed climate sensitivities mirroring those of annual fluxes. Seasonal redistribution of carbon fluxes is initiated by springtime temperature anomalies, but subsequently negative feedbacks in soil moisture during the summer and fall result in net annual carbon losses from land. These modes of variability are also seen in satellite proxies of GPP, suggesting that CESM2 appropriately represents regional sensitivities of photosynthesis to climate variability on seasonal time scales.

Accurate Simulation of Both Sensitivity and Variability for Amazonian Photosynthesis: Is It Too Much to Ask?

Estimates of Amazon rainforest gross primary productivity (GPP) differ by a factor of 2 across a suite of three statistical and 18 process models. This wide spread contributes uncertainty to predictions of future climate. We compare the mean and variance of GPP from these models to that of GPP at six eddy covariance (EC) towers. Only one model's mean GPP across all sites falls within a 99% confidence interval for EC GPP, and only one model matches EC variance. The strength of model response to climate drivers is related to model ability to match the seasonal pattern of the EC GPP. Models with stronger seasonal swings in GPP have stronger responses to rain, light, and temperature than does EC GPP. The model to data comparison illustrates a trade-off inherent to deterministic models between accurate simulation of a mean (average) and accurate responsiveness to drivers. The trade-off exists because all deterministic models simplify processes and lack at least some consequential driver or interaction. If a model's sensitivities to included drivers and their interactions are accurate, then deterministically predicted outcomes have less variability than is realistic. If a GPP model has stronger responses to climate drivers than found in data, model predictions may match the observed variance and seasonal pattern but are likely to overpredict GPP response to climate change. High or realistic variability of model estimates relative to reference data indicate that the model is hypersensitive to one or more drivers.

Artificial Neural Network Model of Soil Heat Flux over Multiple Land Covers in South America

Soil heat flux (G) is an important component for the closure of the surface energy balance (SEB) and the estimation of evapotranspiration (ET) by remote sensing algorithms. Over the last decades, efforts have been focused on parameterizing empirical models for G prediction, based on biophysical parameters estimated by remote sensing. However, due to the existing models’ empirical nature and the restricted conditions in which they were developed, using these models in large-scale applications may lead to significant errors. Thus, the objective of this study was to assess the ability of the artificial neural network (ANN) to predict mid-morning G using extensive remote sensing and meteorological reanalysis data over a broad range of climates and land covers in South America. Surface temperature (Ts), albedo (α), and enhanced vegetation index (EVI), obtained from a moderate resolution imaging spectroradiometer (MODIS), and net radiation (Rn) from the global land data assimilation system 2.1 (GLDAS 2.1) product, were used as inputs. The ANN’s predictions were validated against measurements obtained by 23 flux towers over multiple land cover types in South America, and their performance was compared to that of existing and commonly used models. The Jackson et al. (1987) and Bastiaanssen (1995) G prediction models were calibrated using the flux tower data for quadratic errors minimization. The ANN outperformed existing models, with mean absolute error (MAE) reductions of 43% and 36%, respectively. Additionally, the inclusion of land cover information as an input in the ANN reduced MAE by 22%. This study indicates that the ANN’s structure is more suited for large-scale G prediction than existing models, which can potentially refine SEB fluxes and ET estimates in South America.

Actual Evapotranspiration Estimates in Arid Cold Regions Using Machine Learning Algorithms with In Situ and Remote Sensing Data

Actual evapotranspiration (ETa) estimations in arid regions are challenging because this process is highly dynamic over time and space. Nevertheless, several studies have shown good results when implementing empirical regression formulae that, despite their simplicity, are comparable in accuracy to more complex models. Although many types of regression formulae to estimate ETa exist, there is no consensus on what variables must be included in the analysis. In this research, we used machine learning algorithms—through implementation of empirical linear regression formulae—to find the main variables that control daily and monthly ETa in arid cold regions, where there is a lack of available ETa data. Meteorological data alone and then combined with remote sensing vegetation indices (VIs) were used as input in ETa estimations. In situ ETa and meteorological data were obtained from ten sites in Chile, Australia, and the United States. Our results indicate that the available energy is the main meteorological variable that controls ETa in the assessed sites, despite the fact that these regions are typically described as water-limited environments. The VI that better represents the in situ ETa is the Normalized Difference Water Index, which represents water availability in plants and soils. The best performance of the regression equations in the validation sites was obtained for monthly estimates with the incorporation of VIs (R2 = 0.82), whereas the worst performance of these equations was obtained for monthly ETa estimates when only meteorological data were considered. Incorporation of remote-sensing information results in better ETa estimates compared to when only meteorological data are considered.

Modeling Transpiration with Sun-Induced Chlorophyll Fluorescence Observations via Carbon-Water Coupling Methods

Successfully applied in the carbon research area, sun-induced chlorophyll fluorescence (SIF) has raised the interest of researchers from the water research domain. However, current works focused on the empirical relationship between SIF and plant transpiration (T), while the mechanistic linkage between them has not been fully explored. Two mechanism methods were developed to estimate T via SIF, namely the water-use efficiency (WUE) method and conductance method based on the carbon–water coupling framework. The T estimated by these two methods was compared with T partitioned from eddy covariance instrument measured evapotranspiration at four different sites. Both methods showed good performance at the hourly (R2 = 0.57 for the WUE method and 0.67 for the conductance method) and daily scales (R2 = 0.67 for the WUE method and 0.78 for the conductance method). The developed mechanism methods provide theoretical support and have a great potential basis for deriving ecosystem T by satellite SIF observations.

Highly sampled measurements in a controlled atmosphere at the Biosphere 2 Landscape Evolution Observatory

Land-atmosphere interactions at different temporal and spatial scales are important for our understanding of the Earth system and its modeling. The Landscape Evolution Observatory (LEO) at Biosphere 2, managed by the University of Arizona, hosts three nearly identical artificial bare-soil hillslopes with dimensions of 11 × 30 m2 (1 m depth) in a controlled and highly monitored environment within three large greenhouses. These facilities provide a unique opportunity to explore these interactions. The dataset presented here is a subset of the measurements in each LEO's hillslopes, from 1 July 2015 to 30 June 2019 every 15 minutes, consisting of temperature, water content and heat flux of the soil (at 5 cm depth) for 12 co-located points; temperature, relative humidity and wind speed above ground at 5 locations and 5 different heights ranging from 0.25 m to 9-10 m; 3D wind at 1 location; the four components of radiation at 2 locations; spatially aggregated precipitation rates, total subsurface discharge, and relative water storage; and the measurements from a weather station outside the greenhouses.

Accounting for Training Data Error in Machine Learning Applied to Earth Observations

Remote sensing, or Earth Observation (EO), is increasingly used to understand Earth system dynamics and create continuous and categorical maps of biophysical properties and land cover, especially based on recent advances in machine learning (ML). ML models typically require large, spatially explicit training datasets to make accurate predictions. Training data (TD) are typically generated by digitizing polygons on high spatial-resolution imagery, by collecting in situ data, or by using pre-existing datasets. TD are often assumed to accurately represent the truth, but in practice almost always have error, stemming from (1) sample design, and (2) sample collection errors. The latter is particularly relevant for image-interpreted TD, an increasingly commonly used method due to its practicality and the increasing training sample size requirements of modern ML algorithms. TD errors can cause substantial errors in the maps created using ML algorithms, which may impact map use and interpretation. Despite these potential errors and their real-world consequences for map-based decisions, TD error is often not accounted for or reported in EO research. Here we review the current practices for collecting and handling TD. We identify the sources of TD error, and illustrate their impacts using several case studies representing different EO applications (infrastructure mapping, global surface flux estimates, and agricultural monitoring), and provide guidelines for minimizing and accounting for TD errors. To harmonize terminology, we distinguish TD from three other classes of data that should be used to create and assess ML models: training reference data, used to assess the quality of TD during data generation; validation data, used to iteratively improve models; and map reference data, used only for final accuracy assessment. We focus primarily on TD, but our advice is generally applicable to all four classes, and we ground our review in established best practices for map accuracy assessment literature. EO researchers should start by determining the tolerable levels of map error and appropriate error metrics. Next, TD error should be minimized during sample design by choosing a representative spatio-temporal collection strategy, by using spatially and temporally relevant imagery and ancillary data sources during TD creation, and by selecting a set of legend definitions supported by the data. Furthermore, TD error can be minimized during the collection of individual samples by using consensus-based collection strategies, by directly comparing interpreted training observations against expert-generated training reference data to derive TD error metrics, and by providing image interpreters with thorough application-specific training. We strongly advise that TD error is incorporated in model outputs, either directly in bias and variance estimates or, at a minimum, by documenting the sources and implications of error. TD should be fully documented and made available via an open TD repository, allowing others to replicate and assess its use. To guide researchers in this process, we propose three tiers of TD error accounting standards. Finally, we advise researchers to clearly communicate the magnitude and impacts of TD error on map outputs, with specific consideration given to the likely map audience.

Remote sensing of solar-induced chlorophyll fluorescence (SIF) in vegetation: 50 years of progress

Remote sensing of solar-induced chlorophyll fluorescence (SIF) is a rapidly advancing front in terrestrial vegetation science, with emerging capability in space-based methodologies and diverse application prospects. Although remote sensing of SIF - especially from space - is seen as a contemporary new specialty for terrestrial plants, it is founded upon a multi-decadal history of research, applications, and sensor developments in active and passive sensing of chlorophyll fluorescence. Current technical capabilities allow SIF to be measured across a range of biological, spatial, and temporal scales. As an optical signal, SIF may be assessed remotely using highly-resolved spectral sensors and state-of-the-art algorithms to distinguish the emission from reflected and/or scattered ambient light. Because the red to far-red SIF emission is detectable non-invasively, it may be sampled repeatedly to acquire spatio-temporally explicit information about photosynthetic light responses and steady-state behaviour in vegetation. Progress in this field is accelerating with innovative sensor developments, retrieval methods, and modelling advances. This review distills the historical and current developments spanning the last several decades. It highlights SIF heritage and complementarity within the broader field of fluorescence science, the maturation of physiological and radiative transfer modelling, SIF signal retrieval strategies, techniques for field and airborne sensing, advances in satellite-based systems, and applications of these capabilities in evaluation of photosynthesis and stress effects. Progress, challenges, and future directions are considered for this unique avenue of remote sensing.

NCA-LDAS: Overview and Analysis of Hydrologic Trends for the National Climate Assessment

Terrestrial hydrologic trends over the conterminous United States are estimated for 1980-2015 using the National Climate Assessment Land Data Assimilation System (NCA-LDAS) reanalysis. NCA-LDAS employs the uncoupled Noah version 3.3 land surface model at 0.125°× 1258° forced with NLDAS-2 meteorology, rescaled Climate Prediction Center precipitation, and assimilated satellite-based soil moisture, snow depth, and irrigation products. Mean annual trends are reported using the nonparametric Mann-Kendall test at p < 0.1 significance. Results illustrate the interrelationship between regional gradients in forcing trends and trends in other land energy and water stores and fluxes. Mean precipitation trends range from +3 to +9 mm yr-1 in the upper Great Plains and Northeast to -1 to -9 mm yr-1 in the West and South, net radiation flux trends range from 10.05 to 10.20 W m-2 yr-1 in the East to -0.05 to -0.20 W m-2 yr-1 in the West, and U.S.-wide temperature trends average about +0.03 K yr-1. Trends in soil moisture, snow cover, latent and sensible heat fluxes, and runoff are consistent with forcings, contributing to increasing evaporative fraction trends from west to east. Evaluation of NCA-LDAS trends compared to independent data indicates mixed results. The RMSE of U.S.-wide trends in number of snow cover days improved from 3.13 to 2.89 days yr-1 while trend detection increased 11%. Trends in latent heat flux were hardly affected, with RMSE decreasing only from 0.17 to 0.16 W m-2 yr-1, while trend detection increased 2%. NCA-LDAS runoff trends degraded significantly from 2.6 to 16.1 mm yr-1 while trend detection was unaffected. Analysis also indicated that NCA-LDAS exhibits relatively more skill in low precipitation station density areas, suggesting there are limits to the effectiveness of satellite data assimilation in densely gauged regions. Overall, NCA-LDAS demonstrates capability for quantifying physically consistent, U.S. hydrologic climate trends over the satellite era.

Advances in the Remote Sensing of Terrestrial Evaporation

Characterizing the terrestrial carbon, water and energy cycles depends strongly on a capacity to accurately reproduce the spatial and temporal dynamics of land surface evaporation. For this, and many other reasons, monitoring terrestrial evaporation across multiple space and time scales has been an area of focused research for many decades. Much of this activity has been supported by developments in satellite remote sensing, which have been leveraged to deliver new process insights, model development and methodological improvements. In this Special Issue, published contributions explored a range of research topics directed towards the enhanced estimation of terrestrial evaporation. Here we summarize these cutting-edge efforts and provide an overview of some of the state-of-the-art approaches for retrieving this key variable. Some perspectives on outstanding challenges, issues, and opportunities are also presented.

Exploring the Potential of Satellite Solar-Induced Fluorescence to Constrain Global Transpiration Estimates

The opening and closing of plant stomata regulates the global water, carbon and energy cycles. Biophysical feedbacks on climate are highly dependent on transpiration, which is mediated by vegetation phenology and plant responses to stress conditions. Here, we explore the potential of satellite observations of solar-induced chlorophyll fluorescence (SIF)—normalized by photosynthetically-active radiation (PAR)—to diagnose the ratio of transpiration to potential evaporation (‘transpiration efficiency’, τ). This potential is validated at 25 eddy-covariance sites from seven biomes worldwide. The skill of the state-of-the-art land surface models (LSMs) from the eartH2Observe project to estimate τ is also contrasted against eddy-covariance data. Despite its relatively coarse (0.5°) resolution, SIF/PAR estimates, based on data from the Global Ozone Monitoring Experiment 2 (GOME-2) and the Clouds and Earth’s Radiant Energy System (CERES), correlate to the in situ τ significantly (average inter-site correlation of 0.59), with higher correlations during growing seasons (0.64) compared to decaying periods (0.53). In addition, the skill to diagnose the variability of in situ τ demonstrated by all LSMs is on average lower, indicating the potential of SIF data to constrain the formulations of transpiration in global models via, e.g., data assimilation. Overall, SIF/PAR estimates successfully capture the effect of phenological changes and environmental stress on natural ecosystem transpiration, adequately reflecting the timing of this variability without complex parameterizations.

Satellite and In Situ Observations for Advancing Global Earth Surface Modelling: A Review

In this paper, we review the use of satellite-based remote sensing in combination with in situ data to inform Earth surface modelling. This involves verification and optimization methods that can handle both random and systematic errors and result in effective model improvement for both surface monitoring and prediction applications. The reasons for diverse remote sensing data and products include (i) their complementary areal and temporal coverage, (ii) their diverse and covariant information content, and (iii) their ability to complement in situ observations, which are often sparse and only locally representative. To improve our understanding of the complex behavior of the Earth system at the surface and sub-surface, we need large volumes of data from high-resolution modelling and remote sensing, since the Earth surface exhibits a high degree of heterogeneity and discontinuities in space and time. The spatial and temporal variability of the biosphere, hydrosphere, cryosphere and anthroposphere calls for an increased use of Earth observation (EO) data attaining volumes previously considered prohibitive. We review data availability and discuss recent examples where satellite remote sensing is used to infer observable surface quantities directly or indirectly, with particular emphasis on key parameters necessary for weather and climate prediction. Coordinated high-resolution remote-sensing and modelling/assimilation capabilities for the Earth surface are required to support an international application-focused effort.

Seasonal patterns of canopy photosynthesis captured by remotely sensed sun-induced fluorescence and vegetation indexes in mid-to-high latitude forests: A cross-platform comparison

Characterized by the noticeable seasonal patterns of canopy photosynthesis, mid-to-high latitude forests are sensitive to climate change and crucial for understanding the global carbon cycle. To monitor the seasonal cycle of the canopy photosynthesis from space, several remotely sensed indexes, such as normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and leaf area index (LAI) have been implemented within the past decades. Recently, satellite-derived sun-induced fluorescence (SIF) has shown great potential of providing retrievals that are more related to photosynthesis process. However, the potentials of different canopy measurements have not been thoroughly assessed in the context of recent advances of new satellites and proposals of improved indexes. At 15 forested sites, we present a cross-platform intercomparison of one emerging remote sensing based index of phenology index (PI) and two SIF datasets against the conventional indexes such as NDVI, EVI, and LAI to capture the seasonal cycles of canopy photosynthesis. NDVI, EVI, LAI, and PI were calculated from Moderate Resolution Imaging Spectroradiometer (MODIS) measurements, while SIF were evaluated from Global Ozone Monitoring Experiment-2 (GOME-2) and Orbiting Carbon Observatory-2 (OCO-2) observations. Results indicated that GOME-2 SIF was highly correlated with gross primary production (GPP) and absorbed photosynthetically active radiation during the growing seasons. The SIF-GPP relationship can generally be considered linear at the 16-day scale. Key phenological metrics such as start of the seasons and end of the seasons captured by SIF from GOME-2 and OCO-2 matched closely with photosynthesis phenology as inferred by GPP. However, the applications of OCO-2 SIF for phenological studies may be limited only for a small range of sites (at site-level) due to a limited spatial sampling. Among the MODIS estimations, PI and NDVI provided most reliable predictions of start of growing seasons, while no indexes accurately captured the end of growing seasons.

Reconstructed Solar-Induced Fluorescence: A Machine Learning Vegetation Product Based on MODIS Surface Reflectance to Reproduce GOME-2 Solar-Induced Fluorescence

Solar-induced fluorescence (SIF) observations from space have resulted in major advancements in estimating gross primary productivity (GPP). However, current SIF observations remain spatially coarse, infrequent, and noisy. Here we develop a machine learning approach using surface reflectances from Moderate Resolution Imaging Spectroradiometer (MODIS) channels to reproduce SIF normalized by clear sky surface irradiance from the Global Ozone Monitoring Experiment-2 (GOME-2). The resulting product is a proxy for ecosystem photosynthetically active radiation absorbed by chlorophyll (fAPARCh). Multiplying this new product with a MODIS estimate of photosynthetically active radiation provides a new MODIS-only reconstruction of SIF called Reconstructed SIF (RSIF). RSIF exhibits much higher seasonal and interannual correlation than the original SIF when compared with eddy covariance estimates of GPP and two reference global GPP products, especially in dry and cold regions. RSIF also reproduces intense productivity regions such as the U.S. Corn Belt contrary to typical vegetation indices and similarly to SIF.

Estimating surface soil moisture from SMAP observations using a Neural Network technique

A Neural Network (NN) algorithm was developed to estimate global surface soil moisture for April 2015 to March 2017 with a 2-3 day repeat frequency using passive microwave observations from the Soil Moisture Active Passive (SMAP) satellite, surface soil temperatures from the NASA Goddard Earth Observing System Model version 5 (GEOS-5) land modeling system, and Moderate Resolution Imaging Spectroradiometer-based vegetation water content. The NN was trained on GEOS-5 soil moisture target data, making the NN estimates consistent with the GEOS-5 climatology, such that they may ultimately be assimilated into this model without further bias correction. Evaluated against in situ soil moisture measurements, the average unbiased root mean square error (ubRMSE), correlation and anomaly correlation of the NN retrievals were 0.037 m³m^-3, 0.70 and 0.66, respectively, against SMAP core validation site measurements and 0.026 m³m^-3, 0.58 and 0.48, respectively, against International Soil Moisture Network (ISMN) measurements. At the core validation sites, the NN retrievals have a significantly higher skill than the GEOS-5 model estimates and a slightly lower correlation skill than the SMAP Level-2 Passive (L2P) product. The feasibility of the NN method was reflected by a lower ubRMSE compared to the L2P retrievals as well as a higher skill when ancillary parameters in physically-based retrievals were uncertain. Against ISMN measurements, the skill of the two retrieval products was more comparable. A triple collocation analysis against Advanced Microwave Scanning Radiometer 2 (AMSR2) and Advanced Scatterometer (ASCAT) soil moisture retrievals showed that the NN and L2P retrieval errors have a similar spatial distribution, but the NN retrieval errors are generally lower in densely vegetated regions and transition zones.