BEGIN:VCALENDAR VERSION:2.0 PRODID:-//pretalx//talks.staging.osgeo.org//foss4g-it-2023//speaker//8NEVRQ BEGIN:VTIMEZONE TZID:GMT BEGIN:STANDARD DTSTART:20001029T030000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10 TZNAME:GMT TZOFFSETFROM:+0100 TZOFFSETTO:+0000 END:STANDARD BEGIN:DAYLIGHT DTSTART:20000326T020000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3 TZNAME:BST TZOFFSETFROM:+0000 TZOFFSETTO:+0100 END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT UID:pretalx-foss4g-it-2023-7N8VLH@talks.staging.osgeo.org DTSTART;TZID=GMT:20230613T164500 DTEND;TZID=GMT:20230613T170000 DESCRIPTION:Crop traits monitoring is a fundamental step for controlling cr op productivity in the context of precision agriculture and field phenotyp ing. At present\, the usage of hyperspectral data in machine learning regr ession algorithms (MLRAs) has attracted increasing attention to alleviate the challenges associated with traditional crop trait measurements. Howeve r\, the performance assessment of such hyperspectral-based MLRA models for crop trait retrievals with respect to the well-known natural variations i n either structural or biochemical crop properties remains largely elusive . As such\, this experiment was set up to assess whether full-range hypers pectral data\, acquired by a handheld spectrometer (Spectral Evolution\; 3 50 – 2500 nm)\, as inputs to partial least squares regression (PLSR) and random forest (RF) models are capable of modeling different wheat crop tr aits at the canopy level. The examined crop traits were leaf area index (L AI)\, canopy water content (CWC)\, canopy chlorophyll content (CCC)\, and canopy nitrogen content (CNC). This approach allowed us\, as an overarchin g objective\, to compare the performance of the two aforementioned MLRA mo dels while also focusing on the physical interpretation of the modelling r esults for each particular crop trait. \nOverall\, PLSR provided remarkabl y higher accuracy\, tested with a cross-validation strategy\, as compared to RF for all the crop traits. More precisely\, PLSR denoted R2 (resp. nRM SE%) values of 0.72 (11.97)\, 0.77 (10.89)\, 0.70 (14.61)\, and 0.74 (14.3 8) for LAI\, CWC\, CCC\, and CNC\, respectively. All PLSR models indicated robust prediction capability with RPD values greater than 1.4\, and among st them\, CWC was found to have excellent prediction performance with an R PD higher than 2. However\, RF yielded less predictive models with R2 (res p. nRMSE%) values of 0.59 (14.59)\, 0.42 (17.42)\, 0.50 (18.86)\, and 0.42 (21.41) for LAI\, CWC\, CCC\, and CNC\, respectively. RF models for LAI a nd CCC showed good prediction capabilities (RPD > 1.4)\, whilst RF models of neither CWC nor CNC were reliable (RPD < 1.4). \nIn general\, RF band i mportance and PLSR regression coefficient results revealed physically- mea ningful and consistent patterns for each specific crop trait. Specific wav elengths at SWIR (1716-1745 nm) and NIR (1057-1120 nm)\, Green\, and the R ed-Edge bands respectively showed the highest importance for LAI retrieval . Water absorption regions around 910 nm and 1200 nm as well as the Red-Ed ge and Visible parts were of higher importance for the retrieval of CWC. T he best-performing bands were situated in Red-Edge and Green spectral chan nels for CCC retrieval. SWIR spectral regions between 1600-1800 nm and 210 0-2300 nm appeared to be important (in particular with respect to the othe r traits) alongside the Red-Edge part of the spectrum to retrieve CNC. \ nWe demonstrated that full-range hyperspectral data in combination with ML RA algorithms can provide accurate estimates of wheat crop traits at the c anopy level. The success of utilizing hyperspectral data in MLRA algorithm s was further highlighted by the physically-meaningful modelling performan ces in accordance with the subtle structural and biochemical crop properti es. Our results suggest that such spectroscopic hyperspectral-based MLRA a pproaches could be a powerful tool to accurately monitor crop status throu ghout the cropping season to improve high-throughput phenotyping activitie s and to further aid precision agricultural practices. DTSTAMP:20260516T153754Z LOCATION:Sala Videoconferenza @ PoliBa SUMMARY:Investigating PLSR and RF for retrieving wheat crop traits in a fie ld phenotyping experiment using full-range hyperspectral data: performance assessment and modelling interpretation - Ramin Heidarian Dehkordi\, Mirc o Boschetti\, Gabriele Candiani\, Federico Carotenuto\, Carla Cesaraccio\, Andrea Genangeli\, Beniamino Gioli\, Donato Cillis\, Marina_Ranghetti URL:https://talks.staging.osgeo.org/foss4g-it-2023/talk/7N8VLH/ END:VEVENT BEGIN:VEVENT UID:pretalx-foss4g-it-2023-8HU3SP@talks.staging.osgeo.org DTSTART;TZID=GMT:20230613T171500 DTEND;TZID=GMT:20230613T173000 DESCRIPTION:Non-photosynthetic vegetation (NPV) detection and quantificatio n represent a key variable in remote sensing of conservative agriculture\, and\, more recently\, in carbon farming due to its important role in wate r\, nutrient and carbon cycling. For this reason\, both mapping and charac terization of NPV represent a relevant topic in the exploitation of Earth Observation (EO) data for agriculture monitoring.\nStudies on NPV mapping by EO data benefit from the availability of hyperspectral data due to the high spectral resolution particularly at wavelengths from 1.6 to 2.3m\, where the spectral features of carbon-based constituents of plants are di stinctive. The launch of new generation hyperspectral satellites\, as PRIS MA (PRecursore IperSpettrale della Missione Applicativa) and\, more recent ly\, EnMAP (Environmental Mapping and Analysis Program) offers research op portunities in the field\, which before was mainly investigated by proxima l and aerial sensing.\nEarly studies already proved the potential of PRISM A in NPV due to the prominence of the cellulose-lignin key absorption feat ure at 2.1m. More recent studies on PRISMA make use of machine learning regression algorithm (MLRA) trained on the basis of radiative transfer mo del simulations\, or on the basis of Exponential Gaussian Optimization (EG O) of specific absorption features on sensed data.\nThis second approach\, proposed in this study\, is aimed at the determination of Crop Residue Co ver (CRC) using PRISMA hyperspectral imagery by a two-step approach making use of: i) firstly\, an Exponential Gaussian Optimization to model pre-se lected absorption features\, also reducing the spectral dimension\; ii) se condly\, a Random Forest paradigm\, performing non-linear regression to fi nally predict and map CRC.\nThis study exploits for the training phase an extensive and well documented spectral library\, namely “Reflectance spe ctra of agricultural field conditions supporting remote sensing evaluation of non-photosynthetic vegetation cover” made available online by USGS ( https://doi.org/10.5066/P9XK3867). It consists of 916 in situ surface refl ectance spectra collected using a proximal full range spectroradiometer (3 50 to 2500 nm). Spectra are annotated with the corresponding fractions of NPV\, Soil and (if any) Green Vegetation\, as estimated by point sampling digital photograph of the radiometer field-of-view. \nThis spectral librar y was resampled to PRISMA spectral resolution\, prior to the Gaussian Expo nential Optimization (EGO) on 4 spectral intervals of interest\, already t ested in previous studies\, and corresponding to absorption bands of: cell ulose-lignin\, plant pigments\, vegetation water content and clays.\nThe E GO algorithm optimizes continuum-removed spectra by 4 parameters - absorpt ion band depth\, center\, width and asymmetry – and since this is perfor med for each spectral interval\, it results in 16 parameters. This is a re duced space as compared to the one of the input spectra (around 230 bands) . This parameter space was used to train a Random Forest to model the regr ession between Crop Residue Cover percentage and EGO parameters\, achievin g a determination coefficient around 0.8 (RPD ˜2.1\; MSE ˜ 0.02) on the test set. \nThe RF model was firstly validated against an independent spec tral library of around 100 spectra\, collected during a proximal sensing s urvey with a portable full range spectroradiometer\, conducted in a large farm test site (3800ha) located in Jolanda di Savoia (Italy). Also in this case\, spectra are annotated with Crop Residue Cover percentages\, and re sampled to PRISMA spectral resolution. The model performance on this datas et is in agreement with the test on the USGS spectral library. \nFinally\, the regression model was applied to a PRISMA image \, acquired on the Jol anda di Savoia farm (June 21st 2021)\, for CRC mapping. The resulting map was validated against field observations: the CRC map show values and patt erns in good agreement with ground data confirming encouraging prediction capabilities of the model \nIn conclusion\, the proposed classification ap proach\, trained on a spectral library is predictive\, as proved on an ind ependent spectral data set and on the PRISMA image. Further work will enco mpass testing the robustness of the model by collecting field ground data of Crop Residue Cover at the PRISMA scale\; monitoring CRC dynamics on PRI SMA time series\; and\, the use of Radiative Transfer Model simulations to enlarge the training set\, accounting also for different factors controll ing reflectance (e.g. soil moisture). DTSTAMP:20260516T153754Z LOCATION:Sala Videoconferenza @ PoliBa SUMMARY:Spectroscopic Determination of Crop Residue Cover using Exponential -Gaussian Optimization of absorption features and Random Forest - Ramin He idarian Dehkordi\, Monica Pepe\, katayoun Fakherifard URL:https://talks.staging.osgeo.org/foss4g-it-2023/talk/8HU3SP/ END:VEVENT END:VCALENDAR