Abstract Thermal infrared‐based remote sensing of surface energy fluxes has traditionally relied on high spatial resolution satellite data with revisit frequencies on the order of weeks. In this study, we evaluate a biophysics‐based analytical surface energy balance model for predicting latent energy (LE) and sensible heat (H) fluxes using proximal sensing observations. The Surface Temperature Initiated Closure (STIC1.2) model has been extensively validated across a wide range of spatial and temporal scales using various satellite‐derived thermal infrared data sets. Here we extend this validation by applying STIC at sub‐hourly temporal resolution over multiple growing seasons for four distinct agricultural systems. We further develop and evaluate novel STIC variants that incorporate machine learning (ML) techniques to eliminate the need for surface energy balance observations, specifically net radiation and soil heat flux, thereby enhancing model applicability in data‐sparse settings. The integration of a ML component to estimate surface available energy is shown to have strong predictive performance for both LE (R2 = 0.81–0.94) and H (R2 = 0.46–0.72) across all agricultural systems examined here, demonstrating the potential of hybrid biophysical‐machine learning approaches for surface energy balance modeling with minimal data requirements. This study concludes with a novel application of explainable machine learning (exML) to diagnose sources of model error. This exML framework attributes residual prediction errors to both model input variables and environmental drivers not explicitly included in the simulation experiments. This approach provides a new pathway for improving model design and integrating previously overlooked yet influential variables into future model iterations.

Proximal remote sensing has the potential to provide critical information on vegetation biophysical factors that can predict land-atmosphere exchange of water and energy. Latent energy (LE) flux is traditionally estimated using process-based models which rely on vegetation parameters that change during the growing season. Data-driven models have the potential to address these issues by offering flexible predictor selection and more efficient utilization of the information in predictor sets. These models require careful choice of predictors to avoid redundancy and allow robust cross-validation. In this study we present a systematic and comprehensive evaluation of machine learning (ML) models to assess the capability of meteorological and proximal sensing data for predicting LE at a half-hourly temporal resolution across multiple growing seasons for an agricultural system. The results presented here demonstrate that a model using four environmental predictors in combination with two proximal sensing variables can capture 88 % of the variability in LE. ML models using only three predictors (one meteorological and two proximal remote sensing) captured 81 % of LE variability, offering the best trade-off between performance and complexity. An ML model utilizing only two predictors, one proximal remote sensing variable and downwelling radiation, captured 77 % of LE variability. These results demonstrate the power of proximal remote sensing and meteorological observations to estimate land-atmosphere water vapor exchange, providing a solution where more direct methods such as eddy covariance are not available and for evaluations of agronomic management and genotypic variations.

Common bean (Phaseolus vulgaris L.) is the world's most important legume crop and a vital staple food for millions of people in Latin America and Africa. Given the increasing trend in bean consumption and its importance for nutrition and food security in these regions, there is an urgent need to enhance common bean seeds' nutritional value through breeding. This requires rapidly assessing large and diverse germplasm collections to uncover key nutritional traits in the available genetic diversity. To address this challenge, Near-Infrared Spectroscopy (NIRS) offers a large-scale, cost-effective and non-destructive approach for accurately predicting nutrient content in intact common bean seeds. This study describes the development of predictive models based on NIRS to predict nitrogen (N), iron (Fe) and zinc (Zn) content, using whole common bean seeds from a germplasm core collection held at the International Center for Tropical Agriculture. Spectra were captured for 1754 accessions (wild and domesticated), and reference values for N, Fe, and Zn content were measured with conventional destructive methods in a panel of 401 accessions. Prediction models of N content achieved a concordance correlation coefficient (CCC) of 0.84, while for Fe and Zn, CCC was 0.4. NIRS quantification detected higher N content in wild accessions than in domesticated accessions. These results demonstrate that NIRS can effectively estimate the N content of common bean seeds in a non-destructive manner, while providing valuable nutritional information to enhance access to large genebank collections for bean breeding.

Containerized plant systems using soilless substrates are increasingly vital to global food and horticultural production, offering efficient resource use and expanded crop production in areas where native soils are not suitable. However, open-air production of containerized plants introduces risks for extreme root zone temperatures (RZTs) that impair root function, reduce growth and increase crop loss. While passive mitigation strategies exist, growers often rely on irrigation to cool substrates during high-temperature events. Without reliable real-time monitoring or predictive tools, RZT management remains sub-optimal, ineffective and wasteful. To address this problem we evaluated the potential for predicting substrate temperature using accessible environmental observations and low-complexity modeling approaches. Environmental conditions and substrate temperatures were monitored from June to October 2024 at experimental sites in Tennessee and Ohio. Each site included trials comparing two ground cover types commonly found in open-air crop production, limestone gravel and black landscape fabric. Despite significant differences in cover temperature, only minor differences in substrate temperatures were observed. Ambient air temperature, downwelling shortwave radiation and vapor pressure deficit all demonstrated strong correlations with substrate temperature, with maximum correlations occurring between 2 and 5 hours prior to the substrate temperature observation. These strong lagged correlations motivated the development of relatively simple (one and two predictor) models of substrate temperature using both machine learning (shallow neural networks) and linear statistical models. Models incorporating air temperature and shortwave radiation achieved strong predictive performance for substrate temperatures (R2 = 0.84–0.93) with lead times up to five hours, demonstrating the utility of models developed using widely available weather observations.

The study used the DSSAT model to assess potential soybean yields in different regions of India and validated it under diverse agroecological conditions. The average simulated yield under irrigated conditions was 3794 kg ha−1 relative to the simulated average rainfed yield of 2446 kg ha−1, showing a 35.52% reduction in grain yield due to adverse moisture conditions under rainfed conditions. Relative to simulated yield, the average observed (actual) rainfed yield across 43 districts of India was 1025 kg ha−1, which was 2769 and 1421 kg ha−1 lower than irrigated and rainfed potential yield, respectively. A significant positive correlation was observed between simulated water non-limited yield and solar radiation (R2 = 0.55, p ≤ 0.05). The simulated rainfed grain yield (R2 = 0.66, p ≤ 0.05) had a significant, positive, and curvilinear relationship with growing season rainfall. On the other hand, the actual yield (R2 = 0.008) showed a non-significant relationship with mean crop seasonal rainfall across locations. The gap between simulated yield under irrigated and rainfed conditions is huge at locations with low seasonal rainfall and narrows with increasing rainfall. In addition, the gap between actual yield and simulated yield under rainfed conditions was larger, even in high seasonal rainfall areas. The yield gap under rainfed conditions is due to the non-adoption of improved crop management practices and could be reduced with proper interventions. This includes adapting drought-resistant varieties, conserving rainwater, changing land configuration, and adopting waterlogging-tolerant varieties using improved technology to reduce the soybean yield gap.

Wheat stripe rust (WSR), a fungal disease capable of inflicting severe crop loss, threatens most of global wheat production. Breeding for genetic resistance is the primary defense against stripe rust infection. Further development of rust-resistant wheat varieties depends on the ability to accurately and rapidly quantify rust resilience. In this study we demonstrate the ability of visible through shortwave infrared reflectance spectroscopy to effectively provide high-throughput classification of wheat stripe rust severity and identify important spectral regions for classification accuracy. Random forest models were developed using both leaf-level and canopy-level hyperspectral reflectance observations collected across a breeding population that was scored for WSR severity using 10 and 5 severity classes, respectively. The models were able to accurately diagnose scored disease severity class across these fine scoring scales between 45-52% of the time, which improved to 79-96% accuracy when allowing scores to be off-by-one. The canopy-level model demonstrated higher accuracy and distinct spectral characteristics relative to the leaf-level models, pointing to the use of this technology for field-scale monitoring. Leaf-level model performance was strong despite clear variation in scoring conducted between wheat growth stages. Two approaches to reduce predictor and model complexity, principal component dimensionality reduction and backward feature elimination, were applied here. Both approaches demonstrated that model classification skill could remain high while simplifying high-dimensional hyperspectral reflectance predictors, with parsimonious models having approximately 10 unique components or wavebands. Through the use of a high-resolution infection severity scoring methodology this study provides one of the most rigorous tests of the use of hyperspectral reflectance observations for WSR classification. We demonstrate that machine learning in combination with a few carefully-selected wavebands can be leveraged for precision remote monitoring and management of WSR to limit crop damage and to aid in the selection of resilient germplasm in breeding programs.

This study explores the use of leaf-level visible-to-shortwave infrared (VSWIR) reflectance observations and partial least squares regression (PLSR) to predict foliar concentrations of macronutrients (nitrogen, phosphorus, potassium, calcium, magnesium, and sulfur), micronutrients (boron, copper, iron, manganese, zinc, molybdenum, aluminum, and sodium), and moisture content in winter wheat. A total of 360 fresh wheat leaf samples were collected from a wheat breeding population over two growing seasons. These leaf samples were used to collect VSWIR reflectance observations across a spectral range spanning 350 to 2,500 nm. These samples were then processed for nutrient composition to allow for the examination of the ability of reflectance to accurately model diverse chemical components in wheat foliage. Models for each nutrient were developed using a rigorous cross-validation methodology in conjunction with three distinct component selection methods to explore the trade-offs between model complexity and performance in the final models. We examined absolute minimum predicted residual error sum of squares (PRESS), backward iteration over PRESS, and Van der Voet’s randomized t-test as component selection methods. In addition to contrasting component selection methods for each leaf trait, the importance of spectral regions through variable importance in projection scores was also examined. In general, the backward iteration method provided strong model performance while reducing model complexity relative to the other selection methods, yielding R2 [relative percent difference (RPD), root mean squared error (RMSE)] values in the validation dataset of 0.84 (2.45, 6.91), 0.75 (1.97, 18.67), 0.78 (2.13, 16.49), 0.66 (1.71, 17.13), 0.68 (1.75, 14.51), 0.66 (1.72, 12.29), and 0.84 (2.46, 2.20) for nitrogen, calcium, magnesium, sulfur, iron, zinc, and moisture content on a wet basis, respectively. These model results demonstrate that VSWIR reflectance in combination with modern statistical modeling techniques provides a powerful high throughput method for the quantification of a wide range of foliar nutrient contents in wheat crops. This work has the potential to advance rapid, precise, and nondestructive field assessments of nutrient contents and deficiencies for precision agricultural management and to advance breeding program assessments.

Soil heat flux (SHF) is a key component of the surface energy balance and driver of soil physiochemical and biological processes. Accurate estimation of soil heat flux is challenging due to variations in soil composition, overlying vegetation density and phenology, and highly variable environmental forcings. Existing SHF process-based and data-driven estimation methods have focused on midday landscape scale estimates that correspond to satellite acquisitions, despite the high variability that SHF displays at diurnal scales and throughout the growing season. Recently developed data-driven techniques have emphasized pre-determined predictor variables, with model complexity and predictor selection not carefully evaluated, leaving a gap in our understanding of what information is required to accurately predict SHF. Here we developed and evaluated a suite of ensemble machine learning (ML) models to quantify the ability of meteorological and remote sensing data to predict SHF variability at half-hourly temporal resolution throughout a growing season, producing a comprehensive evaluation of the importance of predictor set composition for SHF estimation at high temporal resolution. We compared this suite of machine learning models to six semi-empirical models and found that the machine learning models broadly outperformed the existing models in capturing diurnal variability across the growing season for four agro-ecosystems (soybean, corn, sorghum, and miscanthus). Crop-specific ML models were able to capture over 86% of the variability in SHF using only two predictor variables, pointing to the need for careful evaluation of predictor sets to identify synergistic combinations. ML models developed using pooled data across all crops captured almost 80% of SHF variability using three predictors, demonstrating the power and generalizability of these methods independent of crop type. Shapley additive explanations (SHAP) were used to examine model interpretability, providing insights into the typically opaque ML modelling process and interaction of predictor variables. Models trained with fewer input variables tended to display more linear and interpretable feature attribution, motivating the use of interpretability as an important consideration in parsimonious model selection. These results provide a robust demonstration of the ability of ML to capture the variability in SHF at sub-hourly resolution across growing seasons spanning a wide range of phenological variation for unique agricultural systems. This study provides a comprehensive evaluation of predictor requirements for model performance, guiding future applications that will take advantage of the next generation of satellite-based observing systems, or in-situ proximal observations of vegetation status and meteorological conditions.

Diagnosing and predicting evaporation through satellite-based surface energy balance (SEB) and land surface models (LSMs) is challenging due to the non-linear responses of aerodynamic (ga) and stomatal conductance (gcs) to the coalition of soil and atmospheric drought. Despite a soaring popularity in refining gcs formulation in the LSMs by introducing a link between soil-plant hydraulics and gcs, the utility of gcs has been surprisingly overlooked in SEB models due to the overriding emphasis on eliminating ga uncertainties and the lack of coordination between these two different modeling communities. Therefore, a persistent challenge is to understand the reasons for divergent evaporation estimates from different models during strong soil-atmospheric drought. Here we present a virtual reality experiment over two contrasting European forest sites to understand the apparent sensitivity of the two critical conductances and evaporative fluxes to a water-stress factor (b-factor) in conjunction with land surface temperature (soil drought proxy) and vapor pressure deficit (atmospheric drought proxy) by using a non-parametric diagnostic model (Surface Temperature Initiated Closure, STIC1.2) and a prognostic model (Community Land Model, CLM5.0). Results revealed the b-factor and different functional forms of the two conductances to be a significant predictor of divergent response of the conductances to soil and atmospheric drought, which subsequently propagated in the evaporative flux estimates between STIC1.2 and CLM5.0. This analysis reaffirms the need for consensus on theory and models that capture the sensitivity of the biophysical conductances to the complex coalition of soil and atmospheric drought for better evaporation prediction.

Stomatal conductance (gs) is a key leaf-level function controlling water, carbon, and energy exchange between vegetation and the surrounding environment. Conventionally, semi-empirical models have been used to model gs, but these models require re-parameterization as ecosystems undergo phenological changes over the growing season. In contrast, machine learning (ML) models offer a potential path to overcome this problem but are less interpretable than process-based models. This study explores ML as an approach to develop flexible and robust models of gs for a range of plant functional types (PFTs), including C3 crops, C3 grasses, shrubs, and tree species across different continents. An explainable machine-learning approach (eXML) was used here to provide novel interpretations and insights into the ML model formulations and relative predictor importance. We contrast the performance of three ML architectures: extreme gradient boosting, random forests, and neural networks. Models were developed and examined using many combinations of environmental and physiological predictors. The results demonstrated that ML models significantly outperform conventional semi-empirical models in predicting gs responses to the environment, while not requiring re-parameterization as is required in the semi-empirical paradigm. Particular focus is placed on models formulated around predictor sets that are: (a) relevant to gs estimation in modern terrestrial biophysical simulation models, and (b) composed of variables describing environmental and physiological drivers that can be remotely sensed non-invasively. “Generalized” models developed using data from all four PFTs demonstrated strong predictive performance using only three predictor variables, capturing 63–80 % of the variability in stomatal conductance across all ML architectures. Four predictor variables resulted in models capturing 79–83 % of gs variability, and models developed using all five predictor variables examined here were able to capture as much as 87 % of gs variability across all PFTs. Uncertainty in gs predictions was quantified using quantile regression. Shapley additive explanations was applied to unravel instance-based positive and negative contributions of environmental and physiological predictors to gs modeling, while illustrating that the models are consistent with the underlying ecophysiology. This work demonstrates the power of ML to introduce a new paradigm in the simulation of highly dynamic ecophysiological processes critical to environmental prediction.

Introducción. Los métodos convencionales de laboratorio para analizar el contenido de nutrientes del suelo, generalmente requieren mucho tiempo y son costosos. En contraparte, la espectroradioscopia visible e infrarroja ofrece una técnica rápida para caracterizar el suelo en laboratorio. Las firmas espectrales codifican información sobre las características inherentes del suelo, como la composición mineral, el contenido de nutrientes, los compuestos orgánicos y el agua. Objetivo. El objetivo principal de este proyecto fue construir bibliotecas espectrales para los suelos tropicales de Costa Rica y determinar las bandas hiperespectrales óptimas en el rango espectral visible infrarrojo cercano e infrarrojo de onda corta para caracterizar propiedades de suelo. Materiales y métodos. Las mediciones hiperespectrales se llevaron a cabo con un espectroradiómetro ASD FieldSpec 4 para generar las firmas de reflectancia espectral de más de 1300 muestras de suelo de Costa Rica, pre-procesadas en el Laboratorio de Suelo de la Universidad EARTH. Se determinó el contenido de nutrientes de cada muestra de suelo, mediante plasma acoplado inductivamente. Además, se evaluaron el pH, la acidez extraíble, la saturación de bases, la saturación de acidez, la capacidad efectiva de intercambio catiónico, el carbono, la materia orgánica y la textura del suelo. Se desarrollaron modelos de regresión de mínimos cuadrados parciales (PLSR) en MATLAB para predecir esas propiedades de suelo a partir de firmas hiperespectrales. Resultados. Este enfoque espectroradioscópico, combinado con modelos de aprendizaje automático, permitió identificar bandas de ondas óptimas específicas en zonas espectrales en las que se puede predecir cada nutriente. Se logró una estimación precisa del contenido de diferentes componentes (Ca, Mg, Fe, C, N y CICE) con un R2 superior a 0,8 y un error cuadrático medio (RMSE) inferior a 10%. Conclusión. Los análisis espectroscópicos combinados con el método Mínimo Cuadrático Parcial (PLS), pueden proporcionar una herramienta muy útil para la agricultura de precisión en los suelos tropicales de Costa Rica.

Thermal infrared (TIR) remote sensing has a wide array of applications in the environmental sciences, but such applications often require absolute temperature estimates with a high degree of accuracy. Low cost microbolometer-based imaging sensors present a possible alternative for such applications, being lightweight enough for deployment on small Unmanned Aerial Systems (UASs), and thus potentially opening up a new range of applications requiring high spatial or temporal resolution and flexible flight planning. These sensors however lack temperature stabilization of the imaging focal plane array (FPA), prohibiting the reliable retrieval of absolute temperature. Here we present a radiometric calibration methodology developed in laboratory settings using a temperature-controlled chamber and programmable blackbody, allowing for independent control of sensor and target temperatures. These laboratory data provided the basis for linear calibration equations that account for both mean and non-uniformity corrections of the FPA raw radiance counts, as a function of ambient sensor operating temperature. Multiple independent experimental trials were used to extensively validate the algorithm in the laboratory, demonstrating a retrieval error of less than 1 degree Celsius. The calibration methodology was tested under realistic field conditions during a two-day field campaign that utilized ground-based observations of land surface temperature (LST) for both a collection of ground targets with a range of reflectance / emissivity properties, and agricultural plots in Northern California. These field experiments included the deployment of the uncooled microbolometer imaging sensor on a UAS, with acquisitions made throughout a highly variable diurnal period. These UAS experiments demonstrated the effectiveness of the pre-flight calibration methodology under field conditions with excellent agreement between retrieved LST and ground-based infrared thermometers for both homogeneous tarps (R^2 = 0.95) and heterogeneous vegetation plots (R^2 = 0.69 across all crop types), with the full range of target temperatures spanning approximately 15-60 degrees Celsius throughout the campaign. The prediction error for absolute temperature estimates of field targets was found to be within 1 degree Celsius, within the range considered acceptable for many vegetation monitoring applications. We further present results of the application of these UAS-based remote measurements of LST to quantify evapotranspiration (ET) for multiple crop systems. UAS flights were conducted over wheat, soybean and maize fields throughout diurnal periods during the growing season of each crop. LST observations were integrated into the Surface Temperature Initiated Closure (STIC) biophysical evapotranspiration model to estimate ET. Validation against eddy covariance system estimates of evapotranspiration (latent energy flux) shows high predictive accuracy (R^2 > 0.95).

Feedbacks between atmospheric processes like precipitation and land surface fluxes including evapotranspiration are difficult to observe, but critical for understanding the role of the land surface in the Earth System. To quantify global surface-atmosphere feedbacks we use results of a process network (PN) applied to 251 eddy covariance sites from the LaThuile database to train a neural network across the global terrestrial surface. There is a strong land–atmosphere coupling between latent (LE) and sensible heat flux (H) and precipitation (P) during summer months in temperate regions, and between H and P during winter, whereas tropical rainforests show little coupling seasonality. Savanna, shrubland, and other semi-arid ecosystems exhibit strong responses in their coupling behavior based on water availability. Feedback couplings from surface fluxes to P peaks at aridity (P/potential evapotranspiration ETp) values near unity, whereas coupling with respect to clouds, inferred from reduced global radiation, increases as P/ETp approaches zero. Spatial patterns in feedback coupling strength are related to climatic zone and biome type. Information flow statistics highlight hotspots of (1) persistent land–atmosphere coupling in sub-Saharan Africa, (2) boreal summer coupling in the central and southwestern US, Brazil, and the Congo basin and (3) in the southern Andes, South Africa and Australia during austral summer. Our data-driven approach to quantifying land atmosphere coupling strength that leverages the global FLUXNET database and information flow statistics provides a basis for verification of feedback interactions in general circulation models and for predicting locations where land cover change will feedback to climate or weather. “Big data” methods reveal robust hotspots of land–atmosphere coupling. Sparse observations and inadequate analytical tools have hindered our understanding of land–atmosphere feedbacks, including the exchange of energy, water, and CO2. Emergent methods such as machine learning, however, offer new opportunities, as Tobias Gerken from Montana State University, USA, and colleagues, demonstrate. A “big data” approach is adopted to characterise the spatial and temporal variability of land–atmosphere coupling without a priori assumptions: information flows are computed from 251 FLUXNET sites which are subsequently used to train a neural network. Distinct regional differences in the magnitude of land–atmosphere feedbacks are found, related to climatic zone and biome type; coupling in semi-arid ecosystems, for example, are strongly related to seasonal water availability. Complementing model studies with such empirical approaches may assist in quantifying climate change impacts on ecosystem services.

The Midwestern US is dominated by corn (Zea mays L.) and soybean (Glycine max [L.] Merr.) production, and the carbon dynamics of this region are dominated by these production systems. An accurate regional estimate of gross primary production (GPP) is imperative and requires upscaling approaches. The aim of this study was to upscale corn and soybean GPP (referred to as GPPcalc) in four counties in Central Iowa in the 2016 growing season (DOY 145–269). Eight eddy-covariance (EC) stations recorded carbon dioxide fluxes of corn (n = 4) and soybean (n = 4), and net ecosystem production (NEP) was partitioned into GPP and ecosystem respiration (RE). Additional field-measured NDVI was used to calculate radiation use efficiency (RUEmax). GPPcalc was calculated using 16 MODIS satellite images, ground-based RUEmax and meteorological data, and improved land use maps. Seasonal NEP, GPP, and RE ( x ¯ ± SE) were 678 ± 63, 1483 ± 100, and −805 ± 40 g C m−2 for corn, and 263 ± 40, 811 ± 53, and −548 ± 14 g C m−2 for soybean, respectively. Field-measured NDVI aligned well with MODIS fPAR (R2 = 0.99), and the calculated RUEmax was 3.24 and 1.90 g C MJ−1 for corn and soybean, respectively. The GPPcalc vs. EC-derived GPP had a RMSE of 2.24 and 2.81 g C m−2 d−1, for corn and soybean, respectively, which is an improvement to the GPPMODIS product (2.44 and 3.30 g C m−2 d−1, respectively). Corn yield, calculated from GPPcalc (12.82 ± 0.65 Mg ha−1), corresponded well to official yield data (13.09 ± 0.09 Mg ha−1), while soybean yield was overestimated (6.73 ± 0.27 vs. 4.03 ± 0.04 Mg ha−1). The approach presented has the potential to increase the accuracy of regional corn and soybean GPP and grain yield estimates by integrating field-based flux estimates with remote sensing reflectance observations and high-resolution land use maps.

Because of the possibility of getting the right answers for the wrong reasons, the predictive performance of a complex systems model is not by itself a reliable indicator of hypothesis quality for the purposes of scientific learning about processes. The predictive performance of a structurally adequate model should be an emergent property of its functional performance. In this context, any Pareto trade-off between measures of predictive performance versus functional performance indicates process-level error in the model; this trade-off, if it exists, indicates that the model's predictions are right for the wrong functional reasons. This paper demonstrates a novel concept based on information theory that is capable of attributing observed errors to specific processes. To demonstrate that the concept and method hold true for models and observations of real systems, we employ a minimal single-parameter-variation sensitivity analysis using a sophisticated ecohydrology model, MLCan, for a well-monitored field site (Bondville IL Ameriflux Soybean). We identify both functional and predictive error in MLCan, and also evidence of the hypothesized tradeoffs between the two. This trade-off indicates structural error within MLCan. For example, the sensible heat flux process can be calibrated to achieve good predictive performance at the cost of poor functional performance. In contrast, we find little structural error for processes driven by solar radiation, which appear "right for the right reasons." This method could be applied broadly to pinpoint process error and structural error in a wide range of system models, beyond the ecohydrological scope demonstrated here.

Thermal infrared sensing of evapotranspiration (E) through surface energy balance (SEB) models is challenging due to uncertainties in determining the aerodynamic conductance (g(A)) and due to inequalities between radiometric (T-R) and aerodynamic temperatures (T-0). We evaluated a novel analytical model, the Surface Temperature Initiated Closure (STIC1.2), that physically integrates T-R observations into a combined Penman-Monteith Shuttleworth-Wallace (PM-SW) framework for directly estimating E, and overcoming the uncertainties associated with T0 and gA determination. An evaluation of STIC1.2 against high temporal frequency SEB flux measurements across an aridity gradient in Australia revealed a systematic error of 10-52% in E from mesic to arid ecosystem, and low systematic error in sensible heat fluxes (H) (12-25%) in all ecosystems. Uncertainty in TR versus moisture availability relationship, stationarity assumption in surface emissivity, and SEB closure corrections in E were predominantly responsible for systematic E errors in arid and semi-arid ecosystems. A discrete correlation (r) of the model errors with observed soil moisture variance (r = 0.33-0.43), evaporative index (r = 0.77-0.90), and climatological dryness (r = 0.60-0.77) explained a strong association between ecohydrological extremes and T-R in determining the error structure of STIC1.2 predicted fluxes. Being independent of any leaf-scale biophysical parameterization, the model might be an important value addition in working group (WG2) of the Australian Energy and Water Exchange (OzEWEX) research initiative which focuses on observations to evaluate and compare biophysical models of energy and water cycle components. Plain Language Summary Evapotranspiration modeling and mapping in arid and semi-arid ecosystems are uncertain due to empirical approximation of surface and atmospheric conductances. Here we demonstrate the performance of a fully analytical model which is independent of any leaf-scale empirical parameterization of the conductances and can be potentially used for continental scale mapping of ecosystem water use as well as water stress using thermal remote sensing satellite data.d

The major goal of the PhotoSpec program was to develop a set of robust ground-based spectrometers that meet the measurement requirements to retrieve solar-induced chlorophyll fluorescence by exploiting solar Fraunhofer lines.