A novel red?edge spectral index for retrieving the leaf chlorophyll content

The leaf chlorophyll content (Chlleaf) is a key indicator of the physiological condition vegetation and integral for harvesting solar radiation required to drive photosynthesis (Evans, 1989; Vernon & Seely, 1966). Retrievals Chlleaf are crucial providing important information on plant stress diseases, modelling productivity serving as proxy photosynthetic capacity within terrestrial biosphere models (Croft et al., 2017; Luo 2019). provision accurate spatially temporally continuous data at user-relevant spatial resolution very ecological science. Remote sensing provides practical approach obtaining across large swaths. red-edge wavelength reflectance (680 ~ 750 nm), which sharply increases from red band absorption maxima near-infrared (NIR) shoulder, most sensitive experiences less saturation in presence high contents Chen, 2018). An increasing number satellite sensors have sampled this spectral region since 2000s (e.g. Environmental Satellite (ENVISAT) Medium Resolution Imaging Spectrometer (MERIS), Sentinel-2 Multispectral Instrument (MSI), Sentinel-3 Ocean Land Colour (OLCI), RapidEye, WorldView-2 Gaofen-6). These provide opportunities estimate different temporal scales. Two methods been widely used remote data: physically based radiative transfer (RTM) empirical index (VI)-based approaches. RTM using inversion look-up tables (LUTs) 2020; Zarco-Tejada 2019) machine learning (Verrelst 2012) allows us model physical mechanisms underpinning light interaction with leaves or canopies. Given an parameterization, scattering contributions scenes components can be simulated then retrieve Chlleaf. Based PROSPECT+SAIL (PROSAIL) (Jacquemoud 2009) 4-Scale-PROSPECT (Chen Leblanc, 1997; Jacquemoud Baret, 1990), 300 m-resolution global product, was generated LUT ENVISAT-MERIS 2020). However, generally red-edge, NIR wavelengths both canopy parameters, such area (LAI) soil background optical properties. Different combinations Chlleaf, LAI other leaf/canopy/soil variables, well solar-observation geometry, produce same reflectance, ill-posed prime issue model-based limits its accuracy, especially when applied globally (Combal 2003). Empirical VI-based represent efficient accessible tools structural biochemical traits. Many VIs developed 2014), constructed exhibit better performance. achieve accuracy estimating scale (Chlcanopy). (CIre) found strongly correlated Chlcanopy maize soybean (Gitelson 2005). inverted (IRECI) exhibits strongest performance experiments performed situ (Frampton 2013). Furthermore, MERIS (MTCI) produced official level 2 product (Dash Curran, 2004). Nevertheless, retrieval compounded by coupling scales, LAI, angle distribution (LAD), type 2014; Demarez Gastellu-Etchegorry, 2000; Viña 2011). A VI that only but resistant structure thus retrieving For extraction concentration vegetation, primary goal decouple effect parameters process. Several efforts reduce reported, LAI. One combine responses parameters. ratio transformed (TCARI, chlorophyll) optimized soil-adjusted (OSAVI, LAI), TCARI/OSAVI, achieved highest among multiple potato (Clevers 2017). modified (MCARI) OSAVI (MCARI/OSAVI) winter wheat (Wu 2008). Another matrix-based combination remove than pair varied types, growth stages angles (Xu Therefore, matrix approaches still limited specific types regional areas, their application large-scale generation. objectives research follows: (1) develop sensitivity (CSI) highly (2) CSI-based estimation method validate it ground measurements compare existing (3) generate China-wide CSI MSI data. measured studies were collected validation. comprised 308 nine sites four types: cropland (CRP), deciduous broadleaf forest (DBF), evergreen needleleaf (ENF) grassland (GRA). Information each experiment reported Table 1. Canopy spectra validation analysis calculate field spectrometer sensor. through laboratory (Lab) spectrometer-based (Spec; Uddling 2007). Detailed descriptions some shown (Supporting 1). images (https://scihub.copernicus.eu/). European Space Agency Earth observation mission consists two satellites (Sentinel-2A Sentinel-2B) revisit frequency 5 days. onboard has 13 bands, including three bands (RE1: central = 705 nm; RE2: 740 RE3: 783 nm). 10 m visible 20 bands. Images downloaded over Reusel, Borden Huailai times nearest surface inversion. 2019 2020 processed Google Engine derive China. resampled 30 neighbour time needed downloading publishing storage resource. Before 2015, ENVISAT main instrument observing chlorophyll-sensitive between 2002 2012 m. In paper, full-resolution (FR) 300-m, 7-day used. FR one (central 708 series preprocessing steps, radiometric, geometric atmospheric corrections. compared ground-measured before 2012. Global Cover Fine Classification System (GLC_FC30-2020) (Liu 2020) define 30-m GLC_FCS30-2020 land cover plants China classified into five study: CRP, (BF), (NF), shrubland (SHR) GRA. regression relationship Satellite-derived response function. PROSAIL utilized simulate CRP GRA, whose canopies considered turbid media homogeneous horizontal layers. DBF, ENF SHR canopies, used, accounts groups, crown shape clumping. Parameters listed 2. also carry out evaluation uncertainties brought carotenoids. changes studied. Normalized calculated divided max depicted Figure 1a,b. Derivatives normalized quantitatively analyse (Figure 1c,d). variables indices (Table 3) present study. 4 describes model-simulated type-specific equations fitting (coefficient determination (R2) root mean square error (RMSE)) band. validated Section 3.2, 3.4. y 73.46 x ? 3.19 (0.59, 13.66) 121.57 15.28 (0.55, 17.60) 92.64 + 8.12 (0.36, 21.08) 81.31 10.74 (0.61, 12.63) 21.01 4.66 (0.43, 15.97) 32.71 7.54 (0.32, 21.64) 22.82 26.11 (0.18, 23.84) 22.68 0.63 (0.48, 14.57) 4.89 27.75 (0.56, 16.87) 10.17 16.14 (0.66, 13.73) 10.99 22.74 (0.54, 15.82) 5.29 15.25 (0.62, 12.42) 16.06 16.53 (0.39, 16.59) 14.43 36.13 (0.17, 24.00) 25.62 39.39 (0.12, 24.66) 12.22 16.04 15.21) 33.07 17.78 (0.44, 15.90) 33.76 35.40 (0.20, 23.42) 58.95 39.14 (0.15, 24.19) 25.74 16.52 14.53) 4.97 10.75 (0.73, 11.16) 5.73 13.87 (0.86, 8.86) 7.83 11.81 (0.80, 10.30) 4.14 12.41 (0.76, 9.83) 163.69 90.80 (0.69, 11.75) 265.68 161.05 14.54) 212.51 – 110.92 18.97) 183.09 109.94 (0.72, 10.69) 117.25 52.09 9.59) 216.36 123.42 (0.82, 11.08) 221.86 120.85 (0.79, 12.03) 131.00 65.90 (0.77, 9.64) 9.98/(0.12 x) (0.51, 12.04) 9.66/(0.04 8.90) 7.57/(0.02 (0.83, 9.55) 7.66/(0.06 9.96) 122.39 56.84 (0.81, 9.33) 226.79 133.20 237.76 135.66 11,47) 135.85 70.26 76.92 2.00 (0.64, 8.28) 99.31 9.78 (0.93, 6.04) 121.99 15.97 6.18) 89.18 0.03 (0.99, 6.61) variations examined PROSAIL-simulated 5). CV_Chlleaf values MTCI, CIre (>50%), indicating chlorophyll. CV_LAI MTCI (>20%) higher those CSI, indicates they /CV_LAI suggests ability (3.43) Datt99 (3.22) larger TCARI/OSAVI (3.08), Macc01 (2.88), (2.63) (<2), suggesting more capable values. SPs Datt99, MND, 40 ?g cm?2 suggest severe problems SP problem significantly mitigated. 5, shows linear R2 (0.99) 11 indices. As 3, wet (psoil 0) dry 1), exception well-performing 3.1.1 fluctuate approximately %, these perform removing moisture content. 3b individual LAD types. sparse (1), (2)), all six 7% patterns. increased, Macc01, changed little (3), b(4)). became dense (4)), fluctuated 19.62% 9.38% LADs respectively. 4a, decreases 20% Car increased 16 cm?2. 3.1.2 do not substantially Car. 4b, MND N. decrease 5% maximum N, followed 20%. show 40% value N 1 3. strong similar result, displays weaker canopy-scale stronger leaf-scale effects will discussed 4. (psoil) geometry (solar zenith angle, SZA view VZA) induce difference CSI-estimated change slightly influences estimated Chlleaf: differences ?1.80 +2.44 (Car) (N) variance +5.57 ?3.74 +4.48 ?2.70 exerts greatest results. ?70% +100% reference causes increase 7 6, lower RMSE relative (rRMSE) (RMSE 9.39 cm?2, 0.49) 13.00 0.19) 13.76 0.23) next accuracies. illustrates results modelled 9.51 rRMSE 21.98% 0.40), 10.49 24.24%, 0.29) 10.68 24.68%, 0.27) accuracy. > 6b,d). 7.04 25.84% 0.7) ?15.00 ?55.06%). Existing tend overestimate ENF, biases 11.08 bias 8.79 cm?2) 12.49 7.02 currently (9.52 (?1.22 VIs. lowest (11.01 cm?2), (?0.29 cm?2). accuracies 14.07 0.02) 14.31 0.03) after underestimation significant (bias ?13.65 ?10.09 analysed 127 Xiaotangshan Nebraska (Figures 8). 8a, remains relatively low stable 6.13–10.19 TCARI/OSAVI-based estimations 12.82 (LAI 3.15) 6.30 4.95). VIs, increases. 8b under conditions. large. tends maintaining almost impervious 8–28 August presented 9a 4). 9b seasonal phenologies sites. 9b, phenologies, spring remaining summer. autumn, EBF minimum exceed winter. reaches 70 60 RTM-based retrieved Supporting S2–S4). blue band; thus, Car, differs Car/Chlleaf 9) changes. summer, ranges 0.15 0.3, 4.27 absolute (AE) autumn spring, caused increases: AE 7.59 (autumn) 6.29 (spring). winter, 0.4 3.0, 6.58 (Car/Chlleaf 0.4) 1.63 3.0). slight, tendency overestimation becomes obvious probability overestimation. blue-band RE1-band previous study (Sola 2018) become denser, slightly. According 2018), overestimations uncorrected Level-1C (L1C) product. (RE) 90% measurements. RE L1C lower, ranging ?12.5% 16.56%. substantial, 59.72% 295.59% corrected Level-2A (L2A) SEN2COR atmosphere correction processor (Main-Knorn 2017), considerably >90% 18.7%–20.3%, although L2A reduced ?11.48% ?3.96%, make underestimate coverage Precise essential methods. affected Figures 3 Uncertainties arising estimates CSI. dependency psoil ensures conditions 8a). decreased beginning end growing season due interference avoided algorithm Due insensitivity background, improve samples scenarios complicated DBF least 7.96 showing cross-type ability. much structure, biochemistry characterized factors largely display similarly contrast, obtained dramatically. example, performs high. achieves species 7). Thus, potential expand applicability local areas future research, products. Because correlations light, drought (Khayatnezhad, 2012; Park Matsumoto, would potentially yield valuable concerning biotic abiotic stresses. It convenient understand ecosystem applications. correlation careful selection ratios construction utilizes band, many Additionally, crucially incorporates confirmed recent (Jin Wang, 2019; O'Reilly Werdell, Multiplying NDVIre ? / improves resistance saturation. Problems indices, successfully 7d–g). carotenoids conditions, errors applying images. vary 9), summer slight independent factor affecting (Lewandowska Jarvis, 1977; Thomas Gausman, 1977). decisive determining 7g 8 low, illustrate increase. may result proportion reflectance. challenge weak available leaves. Further canopy/soil estimation. 4) indicate parameter them far finding, limited. definition, introduced , calculator NDVIre, doubles generating aerosols If pixel contaminated clouds correction, overestimates underestimated (3.96% - 11.48%), 10). 6) prove CSI-derived ?4.954 (>45 further another influential represents complexity internal structure. pigments inside leaf. capturing information. definition enlarges RE1 Recent phenological (Boren trained functional periods mixed widespread vegetative (Yu understorey plays role (Nilsson Wardle, CSI–Chlleaf pixels deserves researchers, measure varied, dataset located mid-high latitudes. influenced dataset, covering wider geographic helpful evaluate new index, proposed research. wavelengths, designed strengthen positive negative By multiplying derived showed overall 23.83%, best CRP; DBF; 9.52 ENF; 11.01 GRA). (larger contents, overestimated 10-day method. continental Future should focus training cover/vegetation regions beneficial Hu Zhang, Jing Li Qinhuo Liu conceived ideas methodology. Liangyun Liu, H. Croft, Jan. Clevers Chenpeng Gu Xiaohan Zhaoxing Zhao, Yadong Dong Wentao Yu Shangrong Lin led writing manuscript. Alfredo Huete Yelu Zeng reviewed edited All authors contributed critically drafts gave final approval publication. This work supported National Key Research Development Program (2019YFE0126700), Natural Science Foundation (No. 41871265). We appreciate Center Advanced Management Technologies (CALMIT), University Nebraska–Lincoln, datasets. no conflict interest declare. peer review history article https://publons.com/publon/10.1111/2041-210X.13994. Data Bank (DOI: https://doi.org/10.11922/sciencedb.j00001.00265, https://www.scidb.cn/en/detail?dataSetId=846695127865884672#) (Li 2021). Code figures https://doi.org/10.5281/zenodo.7088530 (Zhang, 2022). S1 Please note: publisher responsible functionality any supporting supplied authors. Any queries (other missing content) directed corresponding author article.