Enhanced Automated Canopy Characterization from Hyperspectral Data by a Novel Two Step Radiative Transfer Model Inversion Approach

Abstract

Automated, image based methods for the retrieval of vegetation biophysical and biochemical variables are often hampered by the lack of a priori knowledge about land cover and phenology, which makes the retrieval a highly underdetermined problem. This study addresses this problem by presenting a novel approach, called CRASh, for the concurrent retrieval of leaf area index, leaf chlorophyll content, leaf water content and leaf dry matter content from high resolution solar reflective earth observation data. CRASh, which is based on the inversion of the combined PROSPECT+SAILh radiative transfer model (RTM), explores the benefits of combining semi-empirical and physically based approaches. The approach exploits novel ways to address the underdetermined problem in the context of an automated retrieval from mono-temporal high resolution data. To regularize the inverse problem in the variable domain, RTM inversion is coupled with an OPEN ACCESSRemote Sens. 2009, 1 1140 automated land cover classification. Model inversion is based on a two step lookup table (LUT) approach: First, a range of possible solutions is selected from a previously calculated LUT based on the analogy between measured and simulated reflectance. The final solution is determined from this subset by minimizing the difference between the variables used to simulate the spectra contained in the reduced LUT and a first guess of the solution. This first guess of the variables is derived from predictive semi-empirical relationships between classical vegetation indices and the single variables. Additional spectral regularization is obtained by the use of hyperspectral data. Results show that estimates obtained with CRASh are significantly more accurate than those obtained with a tested conventional RTM inversion and semi-empirical approach. Accuracies obtained in this study are comparable to the results obtained by various authors for better constrained inversions that assume more a priori information. The completely automated and image-based nature of the approach facilitates its use in operational chains for upcoming high resolution airborne and spaceborne imaging spectrometers.

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation – Part 2: Validation

“…(i) The year 2006 SIAM decision tree presented in Baraldi et al (2006). (ii) The static decision tree for Spectral Classification of surface reflectance signatures (SPECL) proposed by Dorigo et al (2009), see Table 4 in the Part 1 of this paper, and implemented by the Atmospheric/Topographic Correction for Satellite Imagery (ATCOR) commercial software product (Richter & Schläpfer, 2012a, 2012b). (iii) The static decision tree for Single-Date Classification (SDC), proposed by Simonetti et al (2015).…”

“…On the one hand, no effective and efficient understanding (mapping) of a sub-symbolic EO image into a symbolic SCM is possible if DNs (pixels) are affected by low radiometric quality (GEO-CEOS, 2010). On the other hand, no effective and efficient Cal of digital numbers (DNs) into SURF values corrected for atmospheric, topographic and adjacency effects is possible without an SCM, available a priori in addition to sensory data to enforce a statistic stratification (layered) principle, synonym of class-conditional data analytics (Baraldi, 2017; Baraldi et al, 2010b; Baraldi & Humber, 2015; Baraldi, Humber, & Boschetti, 2013; Bishop & Colby, 2002; Bishop, Shroder, & Colby, 2003; DLR & VEGA, 2011; Dorigo, Richter, Baret, Bamler, & Wagner, 2009; Lück & Van Niekerk, 2016; Riano, Chuvieco, Salas, & Aguado, 2003, Richter & Schläpfer, 2012a, 2012b; Vermote & Saleous, 2007). Well known in statistics, the principle of statistic stratification guarantees that “stratification will always achieve greater precision provided that the strata have been chosen so that members of the same stratum are as similar as possible in respect of the characteristic of interest” (Hunt & Tyrrell, 2012).…”

“…EO value-adding products and services delivered by existing hybrid EO-IUSs whose input is a MS image class-conditioned (masked) by static color names encompass a large variety of low-level EO image enhancement tasks, ranging from MS image compositing to atmospheric and topographic correction of top-of-atmosphere reflectance (TOARF) into SURF values (Ackerman et al, 1998; Baraldi et al, 2010c; Baraldi & Humber, 2015; Baraldi et al, 2013; Despini, Teggi, & Baraldi, 2014; DLR and VEGA, 2011; Dorigo et al, 2009; Lück & Van Niekerk, 2016; Luo, Trishchenko, & Khlopenkov, 2008; Richter & Schläpfer, 2012a, 2012b; Vermote & Saleous, 2007) and high-level EO image understanding applications, including EO image time-series classification, ranging from cloud/cloud-shadow detection to burned area recognition (Arvor, Madiela, & Corpetti, 2016; Baraldi, 2015; Baraldi et al, 2010a, 2010b; Boschetti, Roy, Justice, & Humber, 2015; DLR & VEGA, 2011; Lück & Van Niekerk, 2016; Muirhead & Malkawi, 1989; Simonetti, Simonetti, Szantoi, Lupi, & Eva, 2015; GeoTerraImage, 2015). To the best of these authors’ knowledge, none of these prior knowledge-based MS reflectance space partitioners presented in the RS literature has ever been submitted to a GEO-CEOS stage 4 Val process (GEO-CEOS WGCV, 2015), in compliance with the GEO-CEOS QA4EO Val requirements (GEO-CEOS, 2010).…”

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation - Part 1: Theory

“…Therefore, it is very difficult to solve and requires a priori knowledge in addition to data to become better posed for numerical solution (Cherkassky & Mulier, 1998). On the one hand, no effective and efficient Cal of digital numbers (DNs) into SURF values corrected for atmospheric, topographic and adjacency effects is possible without an SCM, available a priori in addition to data to enforce a statistical stratification principle (Hunt & Tyrrell, 2012), synonym of layered (class-conditional) data analytics (Baraldi, 2017; Baraldi et al, 2010b; Baraldi & Humber, 2015; Baraldi, Humber, & Boschetti, 2013; Bishop & Colby, 2002; Bishop, Shroder, & Colby, 2003; DLR & VEGA, 2011; Dorigo, Richter, Baret, Bamler, & Wagner, 2009; Lück & van Niekerk, 2016; Riano et al, 2003; Richter & Schläpfer, 2012a, 2012b; Vermote & Saleous, 2007). On the other hand, no effective and efficient understanding (mapping) of a sub-symbolic EO image into a symbolic SCM is possible if DNs (pixels) are affected by low radiometric quality (GEO-CEOS, 2010).…”

“…In the RS discipline, there is a long history of prior knowledge-based MS reflectance space partitioners for static color naming, alternative to SIAM’s, developed but never validated by space agencies, public organizations and private companies for use in hybrid EO-IUSs in operating mode, see Figure 11. Examples of hybrid EO image pre-processing applications in the quantitative/sub-symbolic domain of information-as-thing , where a numeric input variable is statistically class-conditioned (masked) by a static color naming first stage to generate as output another numeric variable considered more informative than the input one,  are large-scale MS image compositing (Ackerman et al 1998; Lück & van Niekerk, 2016; Luo, Trishchenko, & Khlopenkov, 2008), MS image atmospheric correction and topographic correction (Baraldi, 2017; Baraldi et al, 2010b; Baraldi & Humber, 2015; Baraldi et al, 2013; Bishop & Colby, 2002; Bishop et al, 2003; DLR & VEGA, 2011; Dorigo et al, 2009; Lück & van Niekerk, 2016; Riano et al, 2003; Richter & Schläpfer, 2012a, 2012b; Vermote & Saleous, 2007), see Figure 12, MS image adjacency effect correction (DLR & VEGA, 2011) and radiometric quality assessment of pan-sharpened MS imagery (Baraldi, 2017; Despini, Teggi, & Baraldi, 2014). Examples of hybrid EO image classification applications in the qualitative/equivocal/categorical domain of information-as-data-interpretation and statistically class-conditioned by a static color naming first stage are cloud and cloud-shadow quality layer detection (Baraldi, 2015, 2017; Baraldil., DLR & VEGA, 2011; Lück & van Niekerk, 2016), single-date LC classification (DLR & VEGA, 2011; GeoTerraImage, 2015; Lück & van Niekerk, 2016; Muirhead & Malkawi, 1989; Simonetti et al 2015a), multi-temporal post-classification LC change (LCC)/no-change detection (Baraldi, 2017; Baraldi et al, 2016; Simonetti et al, 2015a; Tiede, Baraldi, Sudmanns, Belgiu, & Lang, 2016), multi-temporal vegetation gradient detection and quantization into fuzzy sets (Arvor, Madiela, & Corpetti, 2016), multi-temporal burned area detection (Boschetti, Roy, Justice, & Humber, 2015), and prior knowledge-based LC mask refinement (cleaning) of supervised data samples employed as input to supervised data learning EO-IUSs (Baraldi et al, 2010a, 2010b).…”