Hyperspectral Field Measurements Research Articles

Comparative analysis of four semi-analytical models for estimating chlorophyll-a concentration in case-2 waters using field hyperspectral reflectance

ABSTRACT Hyperspectral remote sensing can capture the complicated and variable characteristics of inland waters; thus, it is suited for the water quality assessment of Case-2 waters, and it has the potential to attain high estimation accuracy. In the present study, four improved models adapted from published approaches (three-band index, ΔΦ, BNDBI and TCARI) were investigated to estimate chlorophyll-a (chl-a) for the case of Dianshan Lake, China. Calibration and validation were provided from in situ measured chl-a and field hyperspectral measurements. The improved three-band (ITB) model, ΔΦ model, and BNDBI model yielded satisfactory results and enabled the estimation of chl-a for inland Case-2 waters with coefficients of determination (R 2) reaching 0.75, 0.76, and 0.86, respectively. In particular, the TCARI/OSAVI model presented the highest accuracy (R 2 = 0.94) compared to the other models. All of the results provide strong evidence that the hyperspectral models presented in this paper are promising and applicable to estimate chl-a in eutrophic inland Case-2 waters.

Evaluation of Chlorophyll-a Estimation Approaches Using Iterative Stepwise Elimination Partial Least Squares (ISE-PLS) Regression and Several Traditional Algorithms from Field Hyperspectral Measurements in the Seto Inland Sea, Japan.

Harmful algal blooms (HABs) occur frequently in the Seto Inland Sea, bringing significant economic and environmental losses for the area, which is well known as one of the world’s most productive fisheries. Our objective was to develop a quantitative model using in situ hyperspectral measurements in the Seto Inland Sea to estimate chlorophyll a (Chl-a) concentration, which is a significant parameter for detecting HABs. We obtained spectra and Chl-a data at six stations from 12 ship-based surveys between December 2015 and September 2017. In this study, we used an iterative stepwise elimination partial least squares (ISE-PLS) regression method along with several empirical and semi-analytical methods such as ocean chlorophyll, three-band model, and two-band model algorithms to retrieve Chl-a. Our results showed that ISE-PLS using both the water-leaving reflectance (RL) and the first derivative reflectance (FDR) had a better predictive ability with higher coefficient of determination (R2), lower root mean squared error (RMSE), and higher residual predictive deviation (RPD) values (R2 = 0.77, RMSE = 1.47 and RPD = 2.1 for RL; R2 = 0.78, RMSE = 1.45 and RPD = 2.13 for FDR). However, in this study the ocean chlorophyll (OC) algorithms had poor predictive ability and the three-band and two-band model algorithms did not perform well in areas with lower Chl-a concentrations. These results support ISE-PLS as a potential coastal water quality assessment method using hyperspectral measurements.

Canopy Hyperspectral Sensing of Paddy Fields at the Booting Stage and PLS Regression can Assess Grain Yield

Canopy hyperspectral (HS) sensing is a promising tool for estimating rice (Oryza sativa L.) yield. However, the timing of HS measurements is crucial for assessing grain yield prior to harvest because rice growth stages strongly influence the sensitivity to different wavelengths and the evaluation performance. To clarify the optimum growth stage for HS sensing-based yield assessments, the grain yield of paddy fields during the reproductive phase to the ripening phase was evaluated from field HS data in conjunction with iterative stepwise elimination partial least squares (ISE-PLS) regression. The field experiments involved three different transplanting dates (12 July, 26 July, and 9 August) in 2017 for six cultivars with three replicates (n = 3 × 6 × 3 = 54). Field HS measurements were performed on 2 October 2017, during the panicle initiation, booting, and ripening growth stages. The predictive accuracy of ISE-PLS was compared with that of the standard full-spectrum PLS (FS-PLS) via coefficient of determination (R2) values and root mean squared errors of cross-validation (RMSECV), and the robustness was evaluated by the residual predictive deviation (RPD). Compared with the FS-PLS models, the ISE-PLS models exhibited higher R2 values and lower RMSECV values for all data sets. Overall, the highest R2 values and the lowest RMSECV values were obtained from the ISE-PLS model at the booting stage (R2 = 0.873, RMSECV = 22.903); the RPD was >2.4. Selected HS wavebands in the ISE-PLS model were identified in the red-edge (710–740 nm) and near-infrared (830 nm) regions. Overall, these results suggest that the booting stage might be the best time for in-season rice grain assessment and that rice yield could be evaluated accurately from the HS sensing data via the ISE-PLS model.

USING CALIBRATED RGB IMAGERY FROM LOW-COST UAVS FOR GRASSLAND MONITORING: CASE STUDY AT THE RENGEN GRASSLAND EXPERIMENT (RGE), GERMANY

Abstract. Monitoring the spectral response of intensively managed grassland throughout the growing season allows optimizing fertilizer inputs by monitoring plant growth. For example, site-specific fertilizer application as part of precision agriculture (PA) management requires information within short time. But, this requires field-based measurements with hyper- or multispectral sensors, which may not be feasible on a day to day farming practice. Exploiting the information of RGB images from consumer grade cameras mounted on unmanned aerial vehicles (UAV) can offer cost-efficient as well as near-real time analysis of grasslands with high temporal and spatial resolution. The potential of RGB imagery-based vegetation indices (VI) from consumer grade cameras mounted on UAVs has been explored recently in several. However, for multitemporal analyses it is desirable to calibrate the digital numbers (DN) of RGB-images to physical units. In this study, we explored the comparability of the RGBVI from a consumer grade camera mounted on a low-cost UAV to well established vegetation indices from hyperspectral field measurements for applications in grassland. The study was conducted in 2014 on the Rengen Grassland Experiment (RGE) in Germany. Image DN values were calibrated into reflectance by using the Empirical Line Method (Smith & Milton 1999). Depending on sampling date and VI the correlation between the UAV-based RGBVI and VIs such as the NDVI resulted in varying R2 values from no correlation to up to 0.9. These results indicate, that calibrated RGB-based VIs have the potential to support or substitute hyperspectral field measurements to facilitate management decisions on grasslands.

ASD field hyperspectral measurements for discrimination of the ferruginous rocks and the iron ore types at El Gedida-Ghorabi area, Bahariya Oasis, Western Desert, Egypt

The ironstone succession at El Gedida-Ghorabi-Naser area of El Bahariya depression is subdivided into lagoonal manganiferous mud and fossiliferous ironstone consisting mainly of hematite and goethite-hydrogoethite. The application of the ASD field spectroradiometer measurements (spectral range) in the ASTER data led to the interpretation of the presence of ferruginous units as quartzitic sandstone, gluconitic sandy clay, and pink marly limestone. The existing iron ore minerals in the iron ore localities were also classified into high Mn hematite, low Mn hematite, goethite, hydrogoethite as well as low- and high-grade Hematite and Barite. Quartz, feldspars, rutile, and clay minerals (e.g., kaolinite and illite) are mainly associated with the iron ore. Accessory minerals of manganese, e.g., psilomelane and pyrolusite, were also present. The Barite mineral is recorded as a common mineral association with the iron ore deposits at El Gedida and Ghorabi localities. The stratigraphical units investigated in the study area include the oldest gravely clayey sandstones of the Bahariya Formation overlain by the fossiliferous and oolitic limestones of the El-Hamra, Qazzun, and Naqb Formations. Quartztic sandstones and clayey sandstones of the Radwan Formation and youngest Quaternary sediments of sandy-clayey materials were often found as intermittent cover and overburden in unconformity surfaces over the iron ore bands.

Retrieval of Chlorophyll-a and Total Suspended Solids Using Iterative Stepwise Elimination Partial Least Squares (ISE-PLS) Regression Based on Field Hyperspectral Measurements in Irrigation Ponds in Higashihiroshima, Japan

Concentrations of chlorophyll-a (Chl-a) and total suspended solids (TSS) are significant parameters used to assess water quality. The objective of this study is to establish a quantitative model for estimating the Chl-a and the TSS concentrations in irrigation ponds in Higashihiroshima, Japan, using field hyperspectral measurements and statistical analysis. Field experiments were conducted in six ponds and spectral readings for Chl-a and TSS were obtained from six field observations in 2014. For statistical approaches, we used two spectral indices, the ratio spectral index (RSI) and the normalized difference spectral index (NDSI), and a partial least squares (PLS) regression. The predictive abilities were compared using the coefficient of determination (R2), the root mean squared error of cross validation (RMSECV) and the residual predictive deviation (RPD). Overall, iterative stepwise elimination based on PLS (ISE–PLS), using the first derivative reflectance (FDR), showed the best predictive accuracy, for both Chl-a (R2 = 0.98, RMSECV = 6.15, RPD = 7.44) and TSS (R2 = 0.97, RMSECV = 1.91, RPD = 6.64). The important wavebands for estimating Chl-a (16.97% of all wavebands) and TSS (8.38% of all wavebands) were selected by ISE–PLS from all 501 wavebands over the 400–900 nm range. These findings suggest that ISE–PLS based on field hyperspectral measurements can be used to estimate water Chl-a and TSS concentrations in irrigation ponds.

Estimation of herbage biomass and nutritive status using band depth features with partial least squares regression in Inner Mongolia grassland, China

AbstractAlthough herbage biomass and nutrient status are widely assessed from hyperspectral measurements, certain difficulties are encountered in semiarid and arid regions with low canopy cover. This study investigated the potential of band depth approaches using partial least squares (PLS) regression to estimate herbage biomass and the concentrations of nitrogen (N) and phosphorus (P) in the Inner Mongolia grassland. Field hyperspectral measurements and plant sampling were conducted in desert and typical steppes with different fertilizer levels. The PLS analyses of typical steppe, desert steppe and combined datasets were based on canopy reflectance and first derivative reflectance (FDR) at wavelengths of 400–1000 nm, with consideration of six band depth features extracted from the red absorption region (580–740 nm). The predictive accuracy of the standard full‐spectrum PLS (FS‐PLS) was compared with that of the iterative stepwise elimination PLS (ISE‐PLS) via the cross‐validated coefficient of determination (R2cv) and the ratio of prediction to standard deviation (RPD). In most of the datasets, the ISE‐PLS provided better predictive results than the FS‐PLS. The final models used band depth features to estimate herbage biomass (R2cv = 0.624–0.952, RPD = 1.506–4.539) and pasture N (R2cv = 0.437–0.888, RPD = 1.331–2.869) and reflectance and FDR to estimate pasture P (R2cv = 0.686–0.815, RPD = 1.754–2.267). The models could accurately estimate most of the grass parameters (RPD >1.5), with the exception of pasture N concentrations in the desert steppe dataset due to a range of variation that was too small. The band depth approach with ISE‐PLS improved the predictive ability of the method for estimating herbage biomass and the nutrient contents of grasses in sparse grasslands.

Developing a PCA–ANN Model for Predicting Chlorophyll a Concentration from Field Hyperspectral Measurements in Dianshan Lake, China

This paper aims at combining principle component analysis (PCA) and artificial neural network (ANN) algorithm to predict chlorophyll a concentration in Dianshan Lake, Shanghai, eastern China. Firstly, based on field hyperspectral measurements, the sensitive wavelengths were selected as the input variables to build the basic ANN model, and the estimate accuracy (R2) reached 0.85. In order to improve the accuracy and stability of the ANN model, the total nitrogen, total phosphorus, chemical oxygen demand, dissolve oxygen, and dissolved potential of hydrogen were selected as additional input variables. Consequently, the model accuracy increased to 0.9091. Further, aiming at eliminating the effect of inter-correlation of input variables, the PCA method was utilized to reduce the dimension of input variables. The result shows that the combined PCA–ANN model can reach an estimated accuracy with R2 = 0.9184 and RMSE < 5.6 mg m−3. Moreover, the stability and performance of the enhanced model was further evaluated by cross-validation of PCA–ANN model output and in situ measured datasets. The model sensitivity test through adding 10 % Gauss white noise to the input variables also proved that the enhanced PCA–ANN model has better noise tolerance ability.

Spectral Index for Quantifying Leaf Area Index of Winter Wheat by Field Hyperspectral Measurements: A Case Study in Gifu Prefecture, Central Japan

Timely and nondestructive monitoring of leaf area index (LAI) using remote sensing techniques is crucial for precise and efficient management of crops. In this paper, a new spectral index (SI) for estimating LAI of winter wheat (Triticum aestivum L.) is proposed on the basis of field hyperspectral measurements. A simple index based on the empirical relationships between LAIs and SIs of all available two-waveband combinations from hyperspectral data is developed by considering the difference between reflectance values at 760 and 739 nm (DSIR760–R739 = R760 – R739). Among published and newly developed SIs, DSIR760–R739 exhibited a significant and strong linear relationship with LAI and showed outstanding performance in LAI assessments. The permissible bandwidths for broad-band DSIR760–R739 investigated using simulated reflectance were 5 nm for both 760 and 739 nm center wavelengths. The results indicate that the linear regression model based on the narrow-band and broad-band DSIR760–R739 is a simple but accurate method for timely and nondestructive monitoring of LAI.

Variability in spectral characteristics of trampled high-mountain grasslands

Abstract The goal of the paper is a presentation of field remote sensing methods for the analysis of the trampled plants of a highly protected mountain meadow ecosystem (M&amp;B UNESCO Reserve and one of the most important Polish National Parks). The research area covers a core part of the Western Tatras - the Gąsienicowa Valley and Kasprowy Wierch summit, which are among the most visited destinations of the Polish Tatras. The research method is based on field hyperspectral measurements, using the ASD FieldSpec 3 spectrometer, on the dominant plant species of alpine swards. Sampling sites were located on trampled areas (next to trails) and reference plots, with the same species, but located more than 10 m from the trail (where the probability of trampling was very low, but the same composition of analysed plants). In each case, homogenous plots with a domination of one plant species were investigated. Based on the hyperspectral measurements, spectral characteristics as well as vegetation indices were analysed with the ANOVA statistical test. This indicated a varied resistance to trampling of the studied plant species. The analysis of vegetation indices enabled the selection of those groups which are the most useful for research into mountain vegetation condition: the broadband greenness group; the narrowband greenness group, measuring chlorophyll content and cell structure; and the canopy water content group. The results of the analyses show that vegetation of the High Tatras is characterised by optimal ranges of remote sensing indices. Only plants located nearest to the trails were in a worse condition (chlorophyll and water content was lower for the reference targets). These differences are statistically significant, but the measured values indicate a good condition of vegetation along trampled trails, within the range of optimum plant characteristics.

A hyperspectral indicator system for rangeland degradation on the Tibetan Plateau: A case study towards spaceborne monitoring

The Tibetan Plateau suffers from progressive degradation caused by over-grazing due to improper livestock management, global climate change and herbivory from small mammals. Therefore, a robust indicator system for rangeland degradation has to be developed and tested. This paper investigates local patterns of degradation at two sites (Lake Namco and Mt. Kailash) in Xizang province (China) that are covered by vegetation types typical of a large portion of the plateau. The suitability of a two-indicator system is analysed using hyperspectral field measurements, and its transferability to spaceborne data is tested. The indicators are (1) land-cover fractions derived from linear spectral unmixing and (2) chlorophyll content as a proxy for nutrient and water availability calculated using hyperspectral vegetation indices and partial least squares regression. Because cattle remain near settlements overnight in the local semi-nomadic pastoral system, it can be expected that grazing intensity is highest near the settlement and declines with increasing distance. Therefore, we tested the effect of distance on both indicators using a Spearman correlation analysis. The predicted chlorophyll content and land cover fractions of the indicator system were in good agreement with field observations (correlation coefficients between 0.70 and 0.98). High correlations between distance from settlements and land-cover fractions at both study sites demonstrated that the land-cover fraction is a reliable indicator for degradation. A positive correlation between distance from settlements and photosynthetically active vegetation (PV) revealed over-grazing patterns at the first site. Furthermore, the chlorophyll indicator was proven suitable because chlorophyll concentration declined with increasing distance from settlements. This underlines the over-grazing pattern because cattle excrement was the only external source of nutrients in the ecosystem and it was positively correlated with grazing intensity. However, at the second site, we found a significant positive effect of distance on the amount of photosynthetically non-active vegetation; no effect of distance on PV and chlorophyll content was found. Therefore, no evidence of pasture degradation was detected at the second site. Regarding the potential use of satellite data for degradation monitoring, we found that (1) the land-cover indicator derived from multispectral data was more robust than using noise-filtered hyperspectral information and (2) the chlorophyll amount indicator was estimated from simulated EnMAP data with low error rates. Because the proposed two-indicator system can be derived from multi- and hyperspectral satellite data and combines site conditions and local plant cover, it provides a time-saving and robust method to measure pasture degradation across large areas, assuming that respective satellite data are available.

Genetic algorithm‐based partial least squares regression for estimating legume content in a grass‐legume mixture using field hyperspectral measurements

AbstractThis study investigated the ability of a field hyperspectral radiometer (400–2350 nm) and genetic algorithm‐based partial least squares (GA‐PLS) regression to estimate legume content in a mixed sown pasture in Hokkaido, Japan. Canopy reflectance data and plant samples were obtained from 50 selected sites in the spring (May) and summer (July) of 2007 (n = 100). The predictive accuracy of GA‐PLS was compared with that of multiple linear regression (MLR) and of standard full‐spectrum PLS (FS‐PLS) for the spring and summer datasets. Overall, the highest coefficient of determination (R2) and the lowest root mean squared error of cross validation (RMSECV) values were obtained in the GA‐PLS models for both datasets (R2 = 0.72–0.86, RMSECV = 4.10–5.73%). Selected hyperspectral wavebands in the GA‐PLS models did not perfectly match wavelengths identified previously using MLR, but in most cases, they were within 20 nm of previously known wavelength regions.

Feasibility of Invasive Grass Detection in a Desertscrub Community Using Hyperspectral Field Measurements and Landsat TM Imagery

Invasive species’ phenologies often contrast with those of native species, representing opportunities for detection of invasive species with multi-temporal remote sensing. Detection is especially critical for ecosystem-transforming species that facilitate changes in disturbance regimes. The African C4 grass, Pennisetum ciliare, is transforming ecosystems on three continents and a number of neotropical islands by introducing a grass-fire cycle. However, previous attempts at discriminating P. ciliare in North America using multi-spectral imagery have been unsuccessful. In this paper, we integrate field measurements of hyperspectral plant species signatures and canopy cover with multi-temporal spectral analysis to identify opportunities for detection using moderate-resolution multi-spectral imagery. By applying these results to Landsat TM imagery, we show that multi-spectral discrimination of P. ciliare in heterogeneous mixed desert scrub is feasible, but only at high abundance levels that may have limited value to land managers seeking to control invasion. Much higher discriminability is possible with hyperspectral shortwave infrared imagery because of differences in non-photosynthetic vegetation in uninvaded and invaded landscapes during dormant seasons but these spectra are unavailable in multispectral sensors. Therefore, we recommend hyperspectral imagery for distinguishing invasive grass-dominated landscapes from uninvaded desert scrub.

Estimating the spatial distribution of green herbage biomass and quality by geostatistical analysis with field hyperspectral measurements

This study aimed to estimate green herbage biomass (GBM) and crude protein (CP) concentrations of a mixed-sown pasture in Hokkaido, Japan using ground based hyperspectral measurements and geostatistical analysis. The mixed-sown pasture consisted of a relatively flat section renovated by over-seeding a grass (Subunit 1, 2.6 ha) and a hilly aged section (Subunit 2, 5.0 ha). Hyperspectral reflectance and plant data were collected for 22 days in August 2009 from 88 plots within the two subunits. For mapping, separate spectral readings, without plant sampling, were obtained from a total of 347 plots along permanent transects in the pasture. Genetic algorithm-based wavebands selection with partial least squares (GA-PLS) regression analyses was performed to predict GBM and CP concentrations using both reflectance and first derivative reflectance (FDR) datasets. Then, geostatistical analysis with semivariograms was conducted to determine sampling interval of GBM and CP concentration in Subunits 1 and 2. In the GA-PLS analysis, the most accurate results were obtained by calibration of GBM using FDR (cross-validated coefficient of determination, ; cross-validated root mean square error, RMSECV = 42) and of the CP concentration using raw reflectance (, RMSECV = 2.02). Geostatistical analysis with semivariograms showed that, at the landscape scale, the GBM patch sizes in Subunits 1 and 2 were 31 and 67 m, respectively, and those of the CP concentration were 37 and 54 m. These values indicate that the spatial distribution patterns of these pasture parameters were more heterogeneous in Subunit 1 than in Subunit 2. The analysis result also indicates that the sampling interval for GBM and CP concentration should be <15 m. This work shows the ability to estimate the spatial distribution of GBM and CP concentration for the implementation of site-specific grazing management.

Testing genetic algorithm as a tool to select relevant wavebands from field hyperspectral data for estimating pasture mass and quality in a mixed sown pasture using partial least squares regression

AbstractIt is now accepted that partial least squares (PLS) regression with waveband selection might improve their predictive accuracy in multivariate calibration of models to describe pasture mass and quality. Recently, genetic algorithm (GA) has been shown to be a suitable method for selecting wavebands in the laboratory calibrations. This study aimed to investigate the performance of genetic algorithms PLS (GA‐PLS) regression analyses for estimating green forage biomass (GBM), neutral detergent fiber (NDF), acid detergent fiber (ADF), acid detergent lignin (ADL) and crude protein (CP) concentrations of herbage and herbage mass in CP (CPm) from field hyperspectral data at canopy scale. The predictive ability of GA‐PLS was compared with that of iterative stepwise elimination PLS (ISE‐PLS) and standard full‐spectrum PLS (FS‐PLS), using first derivative reflectance (FDR) spectra data. Canopy reflectance measurements and plant sampling were obtained from 50 selected sites in each of two seasons; spring (May) and summer (July) 2007. For all parameters, cross‐validated coefficients of determination (R2) increased and root mean square error values decreased, respectively, with GA wavebands selection. The number of wavebands selected in the GA‐PLS model ranged between 63 (4.0% of all 1563 available wavebands) and 167 (10.7%), suggesting that over 89% of wavebands were redundant. Between ISE‐PLS and GA‐PLS models, higher R2 values and lower root mean squared errors of prediction were obtained from GA‐PLS prediction for all parameters except CP concentration. Particularly, GA based waveband selection greatly improved ADF (R2 = 0.68–0.77) and ADL (R2 = 0.47–0.59) predictions. These results suggest that pasture quality and GBM can be predicted from field hyperspectral measurements using a GA‐PLS model, and that the GA‐PLS has the advantage of tuning the optimum bands for PLS regression, giving better predictive ability.

Evaluation of Oil-Palm Fungal Disease Infestation with Canopy Hyperspectral Reflectance Data

Fungal disease detection in perennial crops is a major issue in estate management and production. However, nowadays such diagnostics are long and difficult when only made from visual symptom observation, and very expensive and damaging when based on root or stem tissue chemical analysis. As an alternative, we propose in this study to evaluate the potential of hyperspectral reflectance data to help detecting the disease efficiently without destruction of tissues. This study focuses on the calibration of a statistical model of discrimination between several stages of Ganoderma attack on oil palm trees, based on field hyperspectral measurements at tree scale. Field protocol and measurements are first described. Then, combinations of pre-processing, partial least square regression and linear discriminant analysis are tested on about hundred samples to prove the efficiency of canopy reflectance in providing information about the plant sanitary status. A robust algorithm is thus derived, allowing classifying oil-palm in a 4-level typology, based on disease severity from healthy to critically sick stages, with a global performance close to 94%. Moreover, this model discriminates sick from healthy trees with a confidence level of almost 98%. Applications and further improvements of this experiment are finally discussed.