Within-field crop leaf area index simulation using a hybrid PROSAIL-SVR approach: evaluating Sentinel-2 and PlanetScope potential

Consumer-grade cameras are being increasingly used for remote sensing applications in recent years. However, the performance of this type of cameras has not been systematically tested and well documented in the literature. The objective of this research was to evaluate the performance of original and resolution-reduced images taken from two consumer-grade cameras, a RGB camera and a modified near-infrared (NIR) camera, for crop identification and leaf area index (LAI) estimation. Airborne RGB and NIR images taken over a 6.5-square-km cropping area were mosaicked and aligned to create a four-band mosaic with a spatial resolution of 0.4 m. The spatial resolution of the mosaic was then reduced to 1, 2, 4, 10, 15 and 30 m for comparison. Six supervised classifiers were applied to the RGB images and the four-band images for crop identification, and 10 vegetation indices (VIs) derived from the images were related to ground-measured LAI. Accuracy assessment showed that maximum likelihood applied to the 0.4-m images achieved an overall accuracy of 83.3% for the RGB image and 90.4% for the four-band image. Regression analysis showed that the 10 VIs explained 58.7% to 83.1% of the variability in LAI. Moreover, spatial resolutions at 0.4, 1, 2 and 4 m achieved better classification results for both crop identification and LAI prediction than the coarser spatial resolutions at 10, 15 and 30 m. The results from this study indicate that imagery from consumer-grade cameras can be a useful data source for crop identification and canopy cover estimation.

Leaf area index (LAI) is a key driver of forest productivity and evapotranspiration; however, it is a difficult and labor-intensive variable to measure, making its measurement impractical for large-scale and long-term studies of tropical forest structure and function. In contrast, satellite estimates of LAI have shown promise for large-scale and long-term studies, but their performance has been equivocal and the biases are not well known. We measured total, overstory, and understory LAI of an Amazon-savanna transitional forest (ASTF) over 3 years and a seasonal flooded forest (SFF) during 4 years using a light extinction method and two remote sensing methods (LAI MODIS product and the Landsat-METRIC method), with the objectives of (1) evaluating the performance of the remote sensing methods, and (2) understanding how total, overstory and understory LAI interact with micrometeorological variables. Total, overstory and understory LAI differed between both sites, with ASTF having higher LAI values than SFF, but neither site exhibited year-to-year variation in LAI despite large differences in meteorological variables. LAI values at the two sites have different patterns of correlation with micrometeorological variables. ASTF exhibited smaller seasonal variations in LAI than SFF. In contrast, SFF exhibited small changes in total LAI; however, dry season declines in overstory LAI were counteracted by understory increases in LAI. MODIS LAI correlated weakly to total LAI for SFF but not for ASTF, while METRIC LAI had no correlation to total LAI. However, MODIS LAI correlated strongly with overstory LAI for both sites, but had no correlation with understory LAI. Furthermore, LAI estimates based on canopy light extinction were correlated positively with seasonal variations in rainfall and soil water content and negatively with vapor pressure deficit and solar radiation; however, in some cases satellite-derived estimates of LAI exhibited no correlation with climate variables (METRIC LAI or MODIS LAI for ASTF). These data indicate that the satellite-derived estimates of LAI are insensitive to the understory variations in LAI that occur in many seasonal tropical forests and the micrometeorological variables that control seasonal variations in leaf phenology. While more ground-based measurements are needed to adequately quantify the performance of these satellite-based LAI products, our data indicate that their output must be interpreted with caution in seasonal tropical forests.

Aims Leaf area index(LAI) is a commonly used parameter for quantifying canopy structure and can be quickly measured by indirect optical methods in a forest stand,but few studies have evaluated the accuracy of optical methods to estimate seasonal variations of LAI in a mixed conifer-broadleaved forest. The aims of this study are to(1) develop a practical field method for directly measuring seasonal variations in LAI for mixed conifer-broadleaved forest;(2) evaluate the accuracy of optical methods(digital hemispherical photography(DHP) and LAI-2000 plant canopy analyzer) for measuring the seasonality of LAI; and(3) determine how much the accuracy of estimating the seasonality of LAI can be improved by using optical methods after correcting for influencing factors(e.g.,woody materials and clumping effects within a canopy).Methods The seasonal variations of LAI in a mixed broadleaved-Korean pine(Pinus koraiensis) forest were estimated from litterfall and used to evaluate optical LAI(effective LAI,Le) measurements using the DHP and the LAI-2000 plant canopy analyzer. We corrected a systematic error due to incorrect automatic photographic exposure for DHP measurements. In addition to optical Le,we also measured the seasonality of other major factors influencing the determination of LAI,including woody-to-total area ratio(α),clumping index(?E) and needle-to-shoot area ratio(γE). Important findings The LAI from different methods all showed a unimodal form,and peaked in early August. Effective LAIs from the optical methods underestimated LAI throughout the growing seasons(from May to November). Le from DHP underestimated LAI by an average of 55%(ranging from 50% to 59%) and from LAI-2000plant canopy analyzer by an average of 27%(ranging from 19% to 35%). The accuracy of Le from DHP after correcting for the automatic exposure,α,?E and γE was greatly improved,but the LAI was underestimated by 6%–15%(with mean value of 9%) from May to November. In contrast,the accuracy of Le from LAI-2000 plant canopy analyzer after correcting for the α,?E and γE was also greatly improved,the difference between corrected Le from LAI-2000 plant canopy analyzer and observed LAI was less than 9%. The results from our study demonstrate that seasonal variations in LAI in mixed conifer-broadleaved forests can be optically measured with high accuracy(85% for DHP and 91% for LAI-2000 plant canopy analyzer),as long as corrections are made for the influences of woody materials and foliage clumping on the measurement.

Temporal variation of leaf area index (LAI) in two young Norway spruce stands with different densities was monitored during eight consecutive growing seasons (1998–2005). We focused on: (1) LAI dynamics and above-ground mass production of both spruce stands and their comparison, (2) leaf area duration (LADU), crop production index (CPI) and leaf area efficiency (LAE) evaluation, and (3) thinning impact on the above-mentioned parameters. Also, we tried to deduce the most effective LAI value for the Norway spruce forest investigated. The LAI values of both spruce stands showed a typical seasonal course. To describe the LAI dynamics of the stand, we recommend taking LAI measurements within short time intervals at the time of budding and needle expansion growth (i.e., in early spring) and close to the LAI peak, when the twig growth has been completed. The reason was that after reaching the seasonal maximum, no significant differences between subsequently obtained values were found in the following 2 months. Therefore, we recommend this period for the estimation of seasonally representative LAI values, enabling the comparison of various spruce stands. The maximum hemi-surface LAI value reached 12.4. Based on our results, the most effective LAI values for maximum above-ground biomass production were within the range of 10–11. We found an LAI over these values to be less effective for additional production of above-ground biomass. In forest practice, thinning intensity is mostly described by percentage of stocking reduction. We want to show that not only thinning intensity, but also the type of thinning is important information. The type of thinning significantly affected the stand above-ground biomass increment, canopy openness, stand LAI and LAI efficiency. The stimulating effect of high-type thinning was observed; the LAE as well as the CPI increased. Low-type thinning had no such effects on LAE increments compared to the high-type thinning with similar intensity.

Using Moderate Resolution Imaging Spectrometer (MODIS) and Global Inventory Monitoring and Modeling Studies (GIMMS) data, leaf area index (LAI) has been compared with the simulated results by a recently developed dynamic vegetation model, Interactive Canopy Model (ICM), which includes the carbon and nitrogen cycling processes of the ecosystem. Results show that ICM has the capability of reproducing the seasonal and interannual variations of the global vegetation. However, LAI was generally overestimated in the high and low latitudes but underestimated in the middle latitudes. The underestimations in mid-latitude are always followed by the vegetation sprout for the reason that modeled growth period lag behind the observed. The significant interannual variabilities and the spatial distributions of LAI are well captured by the model. But the simulated LAI exhibits larger variabilities than the observations in most areas except for some tropical regions. The temporal and spatial evolutions of the observed LAI are well simulated in low-latitudes. The bimodal distributions in seasonal variations of the tropical evergreen broadleaf trees and crops have not been well simulated. In addition, the model gives better results in the interannual variations of the boreal shrubs, savanna and deciduous needleleaf trees than other types of vegetation. And the simulated LAI in natural vegetation cases is better than that of crop in both seasonal and interannual time scales, which suggests that it is very important to incorporate the human interferences into the dynamic vegetation model.

An accurate estimation of the leaf area index (LAI) by satellite remote sensing is essential for studying the spatial variation of ecosystem structure. The goal of this study was to estimate the spatial variation of LAI over a forested catchment in a mountainous landscape (ca. 60 km2) in central Japan. We used a simple model to estimate LAI using spectral reflectance by adapting the Monsi-Saeki light attenuation theory for satellite remote sensing. First, we applied the model to Landsat Operational Land Imager (OLI) imagery to estimate the spatial variation of LAI in spring and summer. Second, we validated the model’s performance with in situ LAI estimates at four study plots that included deciduous broadleaf, deciduous coniferous, and evergreen coniferous forest types. Pre-processing of the Landsat OLI imagery, including atmospheric correction by elevation-dependent dark object subtraction and Minnaert topographic correction, together with application of the simple model, enabled a satisfactory 30-m spatial resolution estimation of forest LAI with a maximum of 5.5 ± 0.2 for deciduous broadleaf and 5.3 ± 0.2―for evergreen coniferous forest areas. The LAI variation in May (spring) suggested an altitudinal gradient in the degree of leaf expansion, whereas the LAI variation in August (mid-summer) suggested an altitudinal gradient of yearly maximum forest foliage density. This study demonstrated the importance of an accurate estimation of fine-resolution spatial LAI variations for ecological studies in mountainous landscapes, which are characterized by complex terrain and high vegetative heterogeneity.

As the world's largest loess deposit region, the Loess Plateau's vegetation dynamics are crucial for its regional water-heat balance and ecosystem functioning. Leaf Area Index (LAI) serves as a key indicator bridging canopy architecture and plant physiological activities. Existing studies have made significant advancements in simulating LAI, yet accurate LAI simulation remains challenging. To address this challenge and gain deeper insights into the environmental controls of LAI, this study aims to accurately simulate LAI in the Loess Plateau using deep learning models and to elucidate the spatiotemporal influence of soil moisture and temperature on LAI dynamics. For this purpose, we used three deep learning models, namely Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), and Interpretable Multivariable (IMV)-LSTM, to simulate LAI in the Loess Plateau, only using soil moisture and temperature as inputs. Results indicated that our approach outperformed traditional models and effectively captured LAI variations across different vegetation types. The attention analysis revealed that soil moisture mainly influenced LAI in the arid northwest and temperature was the predominant effect in the humid southeast. Seasonally, soil moisture was crucial in spring and summer, notably in grasslands and croplands, whereas temperature dominated in autumn and winter. Notably, forests had the longest temperature-sensitive periods. As LAI increased, soil moisture became more influential, and at peak LAI, both factors exerted varying controls on different vegetation types. These findings demonstrated the strength of deep learning for simulating vegetation-climate interactions and provided insights into hydrothermal regulation mechanisms in semiarid regions.

Leaf area index (LAI) underpins terrestrial ecosystem functioning, yet our ability to predict LAI remains limited. Across Amazon forests, mean LAI, LAI seasonal dynamics and leaf traits vary with soil moisture stress. We hypothesise that LAI variation can be predicted via an optimality-based approach, using net canopy C export (NCE, photosynthesis minus the C cost of leaf growth and maintenance) as a fitness proxy. We applied a process-based terrestrial ecosystem model to seven plots across a moisture stress gradient with detailed in situ measurements, to determine nominal plant C budgets. For each plot, we then compared observations and simulations of the nominal (i.e. observed) C budget to simulations of alternative, experimental budgets. Experimental budgets were generated by forcing the model with synthetic LAI timeseries (across a range of mean LAI and LAI seasonality) and different leaf trait combinations (leaf mass per unit area, lifespan, photosynthetic capacity and respiration rate) operating along the leaf economic spectrum. Observed mean LAI and LAI seasonality across the soil moisture stress gradient maximised NCE, and were therefore consistent with optimality-based predictions. Yet, the predictive power of an optimality-based approach was limited due to the asymptotic response of simulated NCE to mean LAI and LAI seasonality. Leaf traits fundamentally shaped the C budget, determining simulated optimal LAI and total NCE. Long-lived leaves with lower maximum photosynthetic capacity maximised simulated NCE under aseasonal high mean LAI, with the reverse found for short-lived leaves and higher maximum photosynthetic capacity. The simulated leaf trait LAI trade-offs were consistent with observed distributions. We suggest that a range of LAI strategies could be equally economically viable at local level, though we note several ecological limitations to this interpretation (e.g. between-plant competition). In addition, we show how leaf trait trade-offs enable divergence in canopy strategies. Our results also allow an assessment of the usefulness of optimality-based approaches in simulating primary tropical forest functioning, evaluated against in situ data.

The leaf area index (LAI) of a forest is a key determinant of its primary productivity. Estimation of LAI via remote sensing of forest canopies offers an alternative to direct measurement, which is difficult in tropical rainforests (TRFs). Our objectives were to determine the variation of LAI and Normalized Difference Vegetation Index (NDVI) of a range of TRFs of Sri Lanka across a wide altitudinal range (117-2,132 m above sea level) and develop a predictive model to estimate LAI from NDVI and canopy architectural properties. Ten permanent sampling plots (PSPs) of one hectare each were established in forest reserves of Kanneliya (117-174 m), Sinharaja-Pitadeniya (509-618 m), Sinharaja-Enasalwatte (1,042-1,065 m), Rilagala (1,668 m), Hakgala (1,804 m), Piduruthalagala (2,080 m) and Horton Plains (2,132 m). Canopy LAI and its architectural properties (i.e., Mean Leaf Angle (MLA) and leaf angle distribution) were computed by analysis of ‘fish-eye’ images obtained from September 2019 to July 2020, using hemispherical photography and HemiView software. Satellite images for March-April, 2020 were downloaded from Landsat 8 OLI/TIRS C2L1. NDVI (NDVI=λNIR-λRED/λNIR+λRED) was calculated from ENVI software, where λNIR and λRED are reflectances of near-infrared and red wavebands. ENVI software computed the maximum NDVI (NDVIMax) and minimum NDVI (NDVIMin) values among 30 m×30 m pixels within each PSP. Mean NDVI (NDVIMean) was computed by taking the mean of NDVI values of all pixels within a PSP. Canopy LAI ranged from 1.94 (Pidurutalagala) to 3.38 (Pitadeniya). The corresponding ranges for NDVIMax, NDVIMean and NDVIMin were 0.620-0.767, 0.594-0.764 and 0.429-0.747 respectively. Canopy LAI, NDVIMax and NDVIMean showed significant (p<0.05) linear decreasing trends with increasing altitude. For every 1,000 m increase in altitude, LAI, NDVIMax and NDVIMean decreased by 0.396 (Adjusted-R2=0.407, AIC=-19.92), 0.066 (0.698,-66.72) and 0.058 (0.493,-61.27). In contrast, NDVIMin did not show a significant trend with altitude. Second-order polynomial functions showed greater explanatory power than the linear functions, in terms of adjusted-R2and AIC, in fitting the variation of LAI (Adj.-R2=0.492, AIC=-20.77) and NDVIMax with altitude (0.755,-68.15). The estimated maximum LAI and NDVIMax were at 637 and 329 m above sea level respectively. From among a range of multiple linear regression models using different combinations of NDVI, canopy architectural properties and altitude, the following two models were selected for predicting LAI, based on their adj-R2 and AIC values: LAI=1.269+[3.567.NDVIMax]–[1.598.NDVIMin]–[0.000215.Altitude] (Adj[1]R2=0.544, AIC=-21.41); LAI=-0.673+[5.887.NDVIMax]–[1.490.NDVIMin]–[0.000452.MLA] (0.489, -20.27).
 Keywords: Normalized Difference Vegetation Index, Leaf Area Index, Tropical rainforests, Altitude, Canopy architecture

Abstract. The climate regime shift during the 1980s had a substantial impact on the terrestrial ecosystems and vegetation at different scales. However, the mechanisms driving vegetation changes, before and after the shift, remain unclear. In this study, we used a biophysical dynamic vegetation model to estimate large-scale trends in terms of carbon fixation, vegetation growth, and expansion during the period 1958–2007, and to attribute these changes to environmental drivers including elevated atmospheric CO2 concentration (hereafter eCO2), global warming, and climate variability (hereafter CV). Simulated leaf area index (LAI) and gross primary production (GPP) were evaluated against observation-based data. Significant spatial correlations are found (correlations > 0.87), along with regionally varying temporal correlations of 0.34–0.80 for LAI and 0.45–0.83 for GPP. More than 40 % of the global land area shows significant positive (increase) or negative (decrease) trends in LAI and GPP during 1958–2007. Regions over the globe show different characteristics in terms of ecosystem trends before and after the 1980s. While 11.7 % and 19.3 % of land have had consistently positive LAI and GPP trends, respectively, since 1958, 17.1 % and 20.1 % of land saw LAI and GPP trends, respectively, reverse during the 1980s. Vegetation fraction cover (FRAC) trends, representing vegetation expansion and/or shrinking, are found at the edges of semi-arid areas and polar areas. Environmental drivers affect the change in ecosystem trend over different regions. Overall, eCO2 consistently contributes to positive LAI and GPP trends in the tropics. Global warming mostly affects LAI, with positive effects in high latitudes and negative effects in subtropical semi-arid areas. CV is found to dominate the variability of FRAC, LAI, and GPP in the semi-humid and semi-arid areas. The eCO2 and global warming effects increased after the 1980s, while the CV effect reversed during the 1980s. In addition, plant competition is shown to have played an important role in determining which driver dominated the regional trends. This paper presents new insight into ecosystem variability and changes in the varying climate since the 1950s.

Implementation of a dynamic specific leaf area (SLA) into a land surface model (LSM) incorporated crop-growth model

Improvement of spatially and temporally continuous crop leaf area index by integration of CERES-Maize model and MODIS data

Abstract. Leaf seasonality impacts a variety of important biological, chemical, and physical Earth system processes, which makes it essential to represent leaf phenology in ecosystem and climate models. However, we are still lacking a general, robust parametrisation of phenology at global scales. In this study, we use a simple process-based model, which describes phenology as a strategy for carbon optimality, to test the effects of the common simplification in global modelling studies that plant species within the same plant functional type (PFT) have the same parameter values, implying they are assumed to have the same species traits. In a previous study this model was shown to predict spatial and temporal dynamics of leaf area index (LAI) well across the entire global land surface provided local grid cell parameters were used, and is able to explain 96 % of the spatial variation in average LAI and 87 % of the variation in amplitude. In contrast, we find here that a PFT level parametrisation is unable to capture the spatial variability in seasonal cycles, explaining on average only 28 % of the spatial variation in mean leaf area index and 12 % of the variation in seasonal amplitude. However, we also show that allowing only two parameters, light compensation point and leaf age, to be spatially variable dramatically improves the model predictions, increasing the model's capability of explaining spatial variations in leaf seasonality to 70 and 57 % of the variation in LAI average and amplitude, respectively. This highlights the importance of identifying the spatial scale of variation of plant traits and the necessity to critically analyse the use of the plant functional type assumption in Earth system models.

Leaf area index (LAI) is a key input in models describing biosphere processes and has widely been used in monitoring crop growth and in yield estimation. In this study, a hybrid inversion method is developed to estimate LAI values of winter oilseed rape during growth using high spatial resolution optical satellite data covering a test site located in southeast China. Based on PROSAIL (coupling of PROSPECT and SAIL) simulation datasets, nine vegetation indices (VIs) were analyzed to identify the optimal independent variables for estimating LAI values. The optimal VIs were selected using curve fitting methods and the random forest algorithm. Hybrid inversion models were then built to determine the relationships between optimal simulated VIs and LAI values (generated by the PROSAIL model) using modeling methods, including curve fitting, k-nearest neighbor (kNN), and random forest regression (RFR). Finally, the mapping and estimation of winter oilseed rape LAI using reflectance obtained from Pleiades-1A, WorldView-3, SPOT-6, and WorldView-2 were implemented using the inversion method and the LAI estimation accuracy was validated using ground-measured datasets acquired during the 2014–2015 growing season. Our study indicates that based on the estimation results derived from different datasets, RFR is the optimal modeling algorithm amidst curve fitting and kNN with R2 > 0.954 and RMSE <0.218. Using the optimal VIs, the remote sensing-based mapping of winter oilseed rape LAI yielded an accuracy of R2 = 0.520 and RMSE = 0.923 (RRMSE = 93.7%). These results have demonstrated the potential operational applicability of the hybrid method proposed in this study for the mapping and retrieval of winter oilseed rape LAI values at field scales using multi-source and high spatial resolution optical remote sensing datasets. Details provided by this high resolution mapping cannot be easily discerned at coarser mapping scales and over larger spatial extents that usually employ lower resolution satellite images. Our study therefore has significant implications for field crop monitoring at local scales, providing relevant data for agronomic practices and precision agriculture.

The water cloud model (WCM) can be inverted to estimate leaf area index (LAI) using the intensity of backscatter from synthetic aperture radar (SAR) sensors. Published studies have demonstrated that the WCM can accurately estimate LAI if the model is effectively calibrated. However, calibration of this model requires access to field measures of LAI as well as soil moisture. In contrast, machine learning (ML) algorithms can be trained to estimate LAI from satellite data, even if field moisture measures are not available. In this study, a support vector machine (SVM) was trained to estimate the LAI for corn, soybeans, rice, and wheat crops. These results were compared to LAI estimates from the WCM. To complete this comparison, in situ and satellite data were collected from seven Joint Experiment for Crop Assessment and Monitoring (JECAM) sites located in Argentina, Canada, Germany, India, Poland, Ukraine and the United States of America (U.S.A.). The models used C-Band backscatter intensity for two polarizations (like-polarization (VV) and cross-polarization (VH)) acquired by the RADARSAT-2 and Sentinel-1 SAR satellites. Both the WCM and SVM models performed well in estimating the LAI of corn. For the SVM, the correlation (R) between estimated LAI for corn and LAI measured in situ was reported as 0.93, with a root mean square error (RMSE) of 0.64 m2m−2 and mean absolute error (MAE) of 0.51 m2m−2. The WCM produced an R-value of 0.89, with only slightly higher errors (RMSE of 0.75 m2m−2 and MAE of 0.61 m2m−2) when estimating corn LAI. For rice, only the SVM model was tested, given the lack of soil moisture measures for this crop. In this case, both high correlations and low errors were observed in estimating the LAI of rice using SVM (R of 0.96, RMSE of 0.41 m2m−2 and MAE of 0.30 m2m−2). However, the results demonstrated that when the calibration points were limited (in this case for soybeans), the WCM outperformed the SVM model. This study demonstrates the importance of testing different modeling approaches over diverse agro-ecosystems to increase confidence in model performance.