Michael Vohland

Estimating structural and biochemical parameters for grassland from spectroradiometer data by radiative transfer modelling (PROSPECT+SAIL)

As permanent grassland is a large‐scale land‐use type in Central Europe, grassland inventories are relevant for ecological and agrarian issues. The objective of this study was to assess structural and biochemical grassland parameters (LAI, chlorophyll, water and dry matter contents) from field spectroradiometer data (ASD FieldSpec II) by radiative transfer modelling (PROSPECT+SAIL). Constraints were necessary to compensate the ill‐posed nature of model inversion for accurate parameter retrieval. In this context, we found the foliage moisture content to play an important role. After coupling the equivalent water thickness and the dry matter content in a ratio of 4:1, the estimation accuracy for the LAI clearly improved. In terms of LAI, the RMSE decreased from 0.86 to 0.74, and the range of LAI values measured in the field (min = 0.10, max = 5.88) was reproduced exactly with estimates ranging from 0.05 to 5.46. The spectra reconstructed by PROSPECT+SAIL using the inverted parameter estimates coincided well with the measured spectra. Nonetheless, obtained canopy chlorophyll contents tended to be too high. When using ground measured chlorophyll data for spectra generation, simulated reflectances were clearly higher in the visible domain than the measured ones. This is partly attributed to the chlorophyll absorption coefficients of PROSPECT that may not be true for the majority of grassland plant species in reality. Neverthelesss, the obtained results prove the potential of PROSPECT+SAIL for retrieving structural and biochemical grassland parameters; results may be appropriate for assimilation in the modelling of plant growth or carbon cycle, for example.

Estimation of Some Chemical Properties of an Agricultural Soil by Spectroradiometric Measurements

Abstract The contents of nitrogen and organic carbon in an agricultural soil were analyzed using reflectance measurements ( n = 52) performed with an ASD FieldSpec-II spectroradiometer. For parameter prediction, empirical models based on partial least squares (PLS) regression were defined from the measured reflectance spectra (0.4 to 2.4 μm). Here, reliable estimates were obtained for nitrogen content, but prediction accuracy was only moderate for organic carbon. For nitrogen, the real spatial pattern of within-field variability was reproduced with high accuracy. The results indicate the potential of this method as a quick screening tool for the spatial assessment of nitrogen and organic carbon, and therefore an appropriate alternative to time- and cost-intensive chemical analysis in the laboratory.

Near-infrared spectroscopy (NIRS) predicts non-structural carbohydrate concentrations in different tissue types of a broad range of tree species

The allocation of non-structural carbohydrates (NSCs) to reserves constitutes an important physiological mechanism associated with tree growth and survival. However, procedures for measuring NSC in plant tissue are expensive and time-consuming. Near-infrared spectroscopy (NIRS) is a high-throughput technology that has the potential to infer the concentration of organic constituents for a large number of samples in a rapid and inexpensive way based on empirical calibrations with chemical analysis. The main objectives of this study were (i) to develop a general NSC concentration calibration that integrates various forms of variation such as tree species and tissue types and (ii) to identify characteristic spectral regions associated with NSC molecules. In total, 180 samples from different tree organs (root, stem, branch, leaf) belonging to 73 tree species from tropical and temperate biomes were analysed. Statistical relationships between NSC concentration and NIRS spectra were assessed using partial least squares regression (PLSR) and a variable selection procedure (competitive adaptive reweighted sampling, CARS), in order to identify key wavelengths. Parsimonious and accurate calibration models were obtained for total NSC (r2 of 0·91, RMSE of 1·34% in external validation), followed by starch (r2 = 0·85 and RMSE = 1·20%) and sugars (r2 = 0·82 and RMSE = 1·10%). Key wavelengths coincided among these models and were mainly located in the 1740–1800, 2100–2300 and 2410–2490 nm spectral regions. This study demonstrates the ability of general calibration model to infer NSC concentrations across species and tissue types in a rapid and cost-effective way. The estimation of NSC in plants using NIRS therefore serves as a tool for functional biodiversity research, in particular for the study of the growth–survival trade-off and its implications in response to changing environmental conditions, including growth limitation and mortality.

Using Variable Selection and Wavelets to Exploit the Full Potential of Visible–Near Infrared Spectra for Predicting Soil Properties

In soil spectroscopy a series of strategies exists to optimise multivariate calibrations. We explore this issue with a set of topsoil samples for which we estimated soil organic carbon (OC) and total nitrogen (N) from visible–near infrared (vis–NIR) spectra (350–2500 nm). In total, 172 samples were collected to cover the soil heterogeneity in the study area located in western Rhineland-Palatinate, Germany. There, soils with varying properties developed from very diverse parent materials, e.g., ranging from very acidic sandstone to dolomitic marl. We defined four sample sets each of a different size and heterogeneity. Each set was subdivided into a calibration and a validation set. The first strategy that we tested to improve prediction accuracies was spectral variable selection using competitive adaptive reweighted sampling (CARS) and iteratively retaining informative variables (IRIV), both in combination with partial least squares regression (PLSR). In addition, continuous wavelet transformation (CWT) with the Mexican Hat wavelet was applied to decompose the measured spectra into multiple scale components (dyadic scales 21−25) and thus to represent the high and low frequency features contained in the spectra. CARS was then applied to select wavelet coefficients from the different scales and to introduce them in the PLSR approach (CWT-CARS-PLSR). Regarding prediction power, CWT-CARS-PLSR outperformed the other approaches. For the smallest data set with 30 validation samples, prediction accuracy for OC increased from approximately quantitative with full spectrum-PLSR (r2 = 0.81, residual prediction deviation (RPD) = 2.27) to excellent when using wavelet decomposition and CARS-PLSR (r2 = 0.93, RPD = 3.60). For N, predictions improved from unsuccessful (r2 = 0.63, RPD = 1.36) to approximately quantitative (r2 = 0.84, RPD = 2.03). In case of OC, predictions were worst for the largest dataset with 57 validation samples: CWT-CARS-PLSR achieved approximately quantitative predictions (r2 = 0.82, RPD = 2.31), whereas full spectrum-PLSR provided estimates that allowed only separating between high and low values (r2 = 0.72, RPD = 1.88). Accuracy of N estimation for this dataset using CWT-CARS-PLSR was also approximately quantitative. Concerning the tested spectral variable selection techniques, both methods provided similar results in the prediction. The application of IRIV was limited due to long processing times.

The use of airborne hyperspectral data for tree species classification in a species-rich Central European forest area

Abstract The success of remote sensing approaches to assess tree species diversity in a heterogeneously mixed forest stand depends on the availability of both appropriate data and suitable classification algorithms. To separate the high number of in total ten broadleaf tree species in a small structured floodplain forest, the Leipzig Riverside Forest, we introduce a majority based classification approach for Discriminant Analysis based on Partial Least Squares (PLS-DA), which was tested against Random Forest (RF) and Support Vector Machines (SVM). The classifier performance was tested on different sets of airborne hyperspectral image data (AISA DUAL) that were acquired on single dates in August and September and also stacked to a composite product. Shadowed gaps and shadowed crown parts were eliminated via spectral mixture analysis (SMA) prior to the pixel-based classification. Training and validation sets were defined spectrally with the conditioned Latin hypercube method as a stratified random sampling procedure. In the validation, PLS-DA consistently outperformed the RF and SVM approaches on all datasets. The additional use of spectral variable selection (CARS, “competitive adaptive reweighted sampling”) combined with PLS-DA further improved classification accuracies. Up to 78.4% overall accuracy was achieved for the stacked dataset. The image recorded in August provided slightly higher accuracies than the September image, regardless of the applied classifier.

Use of A Portable Camera for Proximal Soil Sensing with Hyperspectral Image Data

In soil proximal sensing with visible and near-infrared spectroscopy, the currently available hyperspectral snapshot camera technique allows a rapid image data acquisition in a portable mode. This study describes how readings of a hyperspectral camera in the 450–950 nm region could be utilised for estimating soil parameters, which were soil organic carbon (OC), hot-water extractable-C, total nitrogen and clay content; readings were performed in the lab for raw samples without any crushing. As multivariate methods, we used PLSR with full spectra (FS) and also combined with two conceptually different methods of spectral variable selection (CARS, “competitive adaptive reweighted sampling” and IRIV, “iteratively retaining informative variables”). For the accuracy of obtained estimates, it was beneficial to use segmented images instead of image mean spectra, for which we applied a regular decomposing in sub-images all of the same size and k-means clustering. Based on FS-PLSR with image mean spectra, obtained estimates were not useful with RPD values less than 1.50 and R2 values being 0.51 in the best case. With segmented images, improvements were marked for all soil properties; RPD reached values ≥ 1.68 and R2 ≥ 0.66. For all image data and variables, IRIV-PLSR slightly outperformed CARS-PLSR.

The Use of Laboratory Spectroscopy and Optical Remote Sensing for Estimating Soil Properties

The success of precision agriculture requires accurate methods for monitoring the state and health of crops. An additional key issue is the availability of accurate and efficient techniques for in-situ determination of soil properties . Reflectance spectroscopy , a technique which can be applied in the laboratory, in the field and from remote observation systems has attracted the attention of scientists in a variety of disciplines. In soil science, this technology as it relates to precision farming is rapidly developing and has triggered new research initiatives. Although a number of studies are available where soil properties have been derived from reflectance spectra the approach involves substantial scaling problems when transferring methods from laboratory spectroscopy to optical sensor systems onboard satellites and aircrafts. The analysis of reflectance images also requires dealing with data having limited signal-to-noise level, being distorted by atmospheric effects and largely affected by bidirectional effects in reflectance distribution. Starting with a short review of the state-of-the-art we present the potential use of reflectance spectroscopy for retrieving useful soil parameters based on several case studies. These studies serve to illustrate the existing limitations for retrieving soil properties over large heterogeneous areas.

Validating MODIS and Sentinel-2 NDVI Products at a Temperate Deciduous Forest Site Using Two Independent Ground-Based Sensors

Quantifying the accuracy of remote sensing products is a timely endeavor given the rapid increase in Earth observation missions. A validation site for Sentinel-2 products was hence established in central Germany. Automatic multispectral and hyperspectral sensor systems were installed in parallel with an existing eddy covariance flux tower, providing spectral information of the vegetation present at high temporal resolution. Normalized Difference Vegetation Index (NDVI) values from ground-based hyperspectral and multispectral sensors were compared with NDVI products derived from Sentinel-2A and Moderate-resolution Imaging Spectroradiometer (MODIS). The influence of different spatial and temporal resolutions was assessed. High correlations and similar phenological patterns between in situ and satellite-based NDVI time series demonstrated the reliability of satellite-based phenological metrics. Sentinel-2-derived metrics showed better agreement with in situ measurements than MODIS-derived metrics. Dynamic filtering with the best index slope extraction algorithm was nevertheless beneficial for Sentinel-2 NDVI time series despite the availability of quality information from the atmospheric correction procedure.

Quantification of Soil Variables in a Heterogeneous Soil Region With VIS–NIR–SWIR Data Using Different Statistical Sampling and Modeling Strategies

Estimation accuracies obtained for soil properties from spectroradiometer data markedly depend on the individual sample set. The choice of the statistical method to sample a calibration set and the extension of the multivariate modeling approach with bagging and/or spectral variable selection may optimize predictions. We studied this with a set of 172 arable topsoils from a region near Trier (Germany) that covered-as often typical for medium to large-scale applications of soil spectroscopy-a wide range of different soil situations. Yet, differences concerning target variables-organic carbon (OC), nitrogen (N), microbial biomass (Cmic) and thermostable carbon (Cinert)-were small. Based on a split of calibration and validation data with the Kennard-Stone algorithm, we found only moderate improvements towards partial least squares regression (PLSR) when combining PLSR with bagging and, for spectral variable selection, with “competitive adaptive reweighted sampling” (CARS). R2 improved for OC (from 0.75 to 0.79), N (from 0.72 to 0.77) and Cinert (from 0.66 to 0.68) in the validation. Additionally, we used individual calibration sets for each validation sample. In this “local” approach, we clustered calibration samples in the spectral feature space and selected individually the most similar sample from each cluster. Combining bagging-CARS-PLSR with this local approach improved R2 markedly to 0.76 for Cinert, and slightly to 0.82 for OC and to 0.76 (previously 0.73) for Cmic. Effects of the local approach were twofold, as it removed improper samples from the calibration and balanced skewness in the data distribution.

Quantification of Soil Properties with Hyperspectral Data: Selecting Spectral Variables with Different Methods to Improve Accuracies and Analyze Prediction Mechanisms

We explored the potentials of both non-imaging laboratory and airborne imaging spectroscopy to assess arable soil quality indicators. We focused on microbial biomass-C (MBC) and hot water-extractable C (HWEC), complemented by organic carbon (OC) and nitrogen (N) as well-studied spectrally active parameters. The aggregation of different spectral variable selection strategies was used to analyze benefits for reachable estimation accuracies and to explore spectral predictive mechanisms for MBC and HWEC. With selected variables, quantification accuracies improved markedly for MBC (laboratory: RPD = 2.32 instead of 1.33 with full spectra; airborne: 2.35 instead of 1.80) and OC (laboratory: RPD = 3.08 instead of 2.36; airborne: 2.20 instead of 1.94). Patterns of selected variables indicated similarities between HWEC and OC, but significant differences between all other soil variables. This agreed to our results of indirect approaches in which both (i) wet-chemical data of OC and N and (ii) spectra fitted to measured OC and N values were used to estimate MBC and HWEC. Compared to these approaches, we found marked benefits of laboratory and airborne data for a direct spectral quantification of MBC (but not for HWEC). This suggests specificity of spectra for MBC, usable for the determination of this important soil parameter.