Application of deep learning with stratified K-fold for vegetation species discrimation in a protected mountainous region using Sentinel-2 image

Adagbasa, Efosa Gbenga; Adelabu, Samuel; Okello, Tom W.

doi:10.1080/10106049.2019.1704070

Cited by 40 publications

(41 citation statements)

References 74 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For this purpose, aerial hyperspectral data, such as AVIRIS (Airborne Visible/Infra-Red Imaging Spectrometer; 224 spectral bands), DAIS 7915 (Digital Airborne Imaging Spectrometer, 79 bands), AISA Dual (Airborne Imaging Spectrometer; 494 bands), or APEX (288 bands) data were employed [7,15,19]. These attributes allow to classify heterogeneous mountain vegetation at the community level and obtain high accuracy (74%-84% OA), despite its complicated structure; however, multispectral Sentinel-2 data also has the advantage of spectral resolution due to existence of SWIR, NIR, and red-edge bands, which was confirmed by a study of mountain vegetation [10] and other studies [28,30,32]. In our case, SWIR (11 and 12) and NIR (8a) bands were the most important in the classification of the entire ABCD dataset, which confirms the 20 first variables ( Figure A2 in Appendix A).…”

Section: Mountain Vegetation Classificationmentioning

confidence: 95%

Multi-Temporal Sentinel-2 Data in Classification of Mountain Vegetation

Wakulińska

Marcinkowska-Ochtyra

2020

Remote Sensing

View full text Add to dashboard Cite

The electromagnetic spectrum registered via satellite remote sensing methods became a popular data source that can enrich traditional methods of vegetation monitoring. The European Space Agency Sentinel-2 mission, thanks to its spatial (10–20 m) and spectral resolution (12 spectral bands registered in visible-, near-, and mid-infrared spectrum) and primarily its short revisit time (5 days), helps to provide reliable and accurate material for the identification of mountain vegetation. Using the support vector machines (SVM) algorithm and reference data (botanical map of non-forest vegetation, field survey data, and high spatial resolution images) it was possible to classify eight vegetation types of Giant Mountains: bogs and fens, deciduous shrub vegetation, forests, grasslands, heathlands, subalpine tall forbs, subalpine dwarf pine scrubs, and rock and scree vegetation. Additional variables such as principal component analysis (PCA) bands and selected vegetation indices were included in the best classified dataset. The results of the iterative classification, repeated 100 times, were assessed as approximately 80% median overall accuracy (OA) based on multi-temporal datasets composed of images acquired through the vegetation growing season (from late spring to early autumn 2018), better than using a single-date scene (70%–72% OA). Additional variables did not significantly improve the results, showing the importance of spectral and temporal information themselves. Our study confirms the possibility of fully available data for the identification of mountain vegetation for management purposes and protection within national parks.

show abstract

Section: Mountain Vegetation Classificationmentioning

confidence: 95%

Multi-Temporal Sentinel-2 Data in Classification of Mountain Vegetation

Wakulińska

Marcinkowska-Ochtyra

2020

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…The spectral library was used to generate 100 random sample locations for training and cross-validation. The locations were randomly split into a training set (70%) to train the classifiers (31,35) and a test set (30%) for testing purposes (31,36).…”

Section: Methodsmentioning

confidence: 99%

“…The GEE code editor random forest machine learning classifier with ten trees was used to process the image collections to classify the images into induvial species classes (31,37). Ancillary data such as Normalized Difference Vegetation Index (NDVI) were mapped into the image collection on GEE before classification was done to improve classification accuracy.…”

Section: Methodsmentioning

confidence: 99%

“…However, these studies focus briefly on a short period, usually between one to five years, because, before now, only high-resolution hyperspectral images could give accurate vegetation species discrimination at individual levels (9,19,(27)(28)(29)(30). Nevertheless, recent studies have shown that free low-resolution satellite images like Sentinel-2 MSI and Landsat 8 OLI can be used to accurately map and monitor grassland vegetation species (31)(32)(33). Vegetation species succession and diversity monitoring can be done over an extended period using these low-resolution imageries in combination with machine learning.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Mapping Vegetation Species Succession in a Mountainous Grassland ecosystem using Landsat and Sentinel-2 data

Adagbasa

Mukwada

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Vegetation species succession and composition are significant factors determining the rate of ecosystem biodiversity recovery after being disturbed and subsequently vital for sustainable and effective natural resource management and biodiversity. The succession and composition of grasslands ecosystems worldwide have significantly been affected by accelerated environmental changes due to natural and anthropogenic activities. Therefore, understanding spatial data on the succession of grassland vegetation species and communities through mapping and monitoring is essential to gain knowledge on the ecosystem and other ecosystem services. This study used a random forest machine learning classifier on the Google Earth Engine platform to classify grass vegetation species with Landsat 7 ETM+ and ASTER multispectral imager (MI) data resampled with the current Sentinel-2 MSI data to map and estimate the changes in vegetation species succession. The results indicate that ASTER IM has the least accuracy of 72%, Landsat 7 ETM+ 84%, and Sentinel-2 had the highest of 87%. The result also shows that other species had replaced four dominant grass species totaling about 49 km 2 throughout the study.

show abstract

“…Intelligent methods that have their accuracy values (considering their respective deviations) close to that found by the Dummy were considered not capable of classifying the inputs. The Stratified KFold (n_splits = 4 and shuffle = True) and Cross Validate methods (Forman and Scholz, 2010;Adagbasa et al, 2019) were used to minimize the effect of chance on the distribution of data used for training and testing.…”

Section: Training and Testingmentioning

confidence: 99%

Fruit Herbivory Alters Plant Electrome: Evidence for Fruit-Shoot Long-Distance Electrical Signaling in Tomato Plants

Reissig

Oliveira

et al. 2021

Front. Sustain. Food Syst.

View full text Add to dashboard Cite

The electrical activity of tomato plants subjected to fruit herbivory was investigated. The study aimed to test the hypothesis that tomato fruits transmit long-distance electrical signals to the shoot when subjected to herbivory. For such, time series classification by machine learning techniques and analyses related to the oxidative response were employed. Tomato plants (cv. “Micro-Tom”) were placed into a Faraday's cage and an electrode pair was inserted in the fruit's peduncle. Helicoverpa armigera caterpillars were placed on the fruit (either green and ripe) for 24 h. The time series were recorded before and after the fruit's exposure of the caterpillars. The plant material for chemical analyses was collected 24 and 48 h after the end of the acquisition of electrophysiological data. The time series were analyzed by the following techniques: Fast Fourier Transform (FFT), Wavelet Transform, Power Spectral Density (PSD), and Approximate Entropy. The following features from FFT, PSD, and Wavelet Transform were used for PCA (Principal Component Analysis): average, maximum and minimum value, variance, skewness, and kurtosis. Additionally, these features were used in Machine Learning (ML) analyses for looking for classifiable patterns between tomato plants before and after fruit herbivory. Also, we compared the electrome before and after herbivory in the green and ripe fruits. To evaluate an oxidative response in different organs, hydrogen peroxide, superoxide anion, catalase, ascorbate peroxidase, guaiacol peroxidase, and superoxide dismutase activity were evaluated in fruit and leaves. The results show with 90% of accuracy that the electrome registered in the fruit's peduncle before herbivory is different from the electrome during predation on the fruits. Interestingly, there was also a sharp difference in the electrome of the green and ripe fruits' peduncles before, but not during, the herbivory, which demonstrates that the signals generated by the herbivory stand over the others. Biochemical analysis showed that herbivory in the fruit triggered an oxidative response in other parts of the plant. Here, we demonstrate that the fruit perceives biotic stimuli and transmits electrical signals to the shoot of tomato plants. This study raises new possibilities for studies involving electrical signals in signaling and systemic response, as well as for the applicability of ML to classify electrophysiological data and its use in early diagnosis.

show abstract

Application of deep learning with stratified K-fold for vegetation species discrimation in a protected mountainous region using Sentinel-2 image

Cited by 40 publications

References 74 publications

Multi-Temporal Sentinel-2 Data in Classification of Mountain Vegetation

Multi-Temporal Sentinel-2 Data in Classification of Mountain Vegetation

Mapping Vegetation Species Succession in a Mountainous Grassland ecosystem using Landsat and Sentinel-2 data

Fruit Herbivory Alters Plant Electrome: Evidence for Fruit-Shoot Long-Distance Electrical Signaling in Tomato Plants

Contact Info

Product

Resources

About