Utilization of Hyperspectral Remote Sensing Imagery for Improving Burnt Area Mapping Accuracy

Wildfires pose a direct threat when occurring close to populated areas. Additionally, their significant carbon and climate feedbacks represent an indirect threat on a global, long-term scale. Monitoring and analyzing wildfires is therefore a crucial task to increase the understanding of interconnections between fire and ecosystems, in order to improve wildfire management activities. This study investigates the suitability of 232 different red/near-infrared band combinations based on hyperspectral imagery of the DESIS sensor with regard to burnt area detection accuracy. It is shown that the selection of wavelengths greatly influences the detection quality, and that especially the utilization of lower near-infrared wavelengths increases the yielded accuracy. For burnt area analysis based on the Normalized Difference Vegetation Index (NDVI), the optimal wavelength range has been found to be 660–670 nm and 810–835 nm for the red band and near-infrared band, respectively.

Learning-based classification of multispectral images for deterioration mapping of historic structures

The conservation of historic structures requires detailed knowledge of their state of preservation. Documentation of deterioration makes it possible to identify risk factors and interpret weathering mechanisms. It is usually performed using non-destructive methods such as mapping of surface features. The automated mapping of deterioration is a direction not often explored, especially when the investigated architectural surfaces present a multitude of deterioration forms and consist of heterogeneous materials, which significantly complicates the generation of thematic decay maps. This work combines reflectance imaging and supervised segmentation, based on machine learning methods, to automatically segment deterioration patterns on multispectral image composites, using a weathered historic fortification as a case study. Several spectral band combinations and image classification techniques (regression, decision tree, and ensemble learning algorithmic implementations) are evaluated to propose an accurate approach. The automated thematic mapping facilitates the spatial and semantic description of the deterioration patterns. Furthermore, the utilization of low-cost photographic equipment and easily operable digital image processing software adds to the practicality and agility of the presented methodology.

Winter Wheat Take-All Disease Index Estimation Model Based on Hyperspectral Data

Wheat take-all, caused by two variants of the fungus Gaeumannomyces gramnis (Sacc.) Arx & D. Olivier, was common in spring wheat areas in northwest and north China and occurred in winter wheat areas in north China. The yield of common disease areas was reduced by more than 20% and the yield of severe cases was reduced by more than 50%. Large-scale rapid and accurate estimation of the incidence of wheat take-all plays an important role in guiding field control and agricultural yield estimation. In this study, a portable ground spectrometer was used to collect the spectral reflectance in the 350–1050 nm band range of wheat canopy after take-all infection in the wheat grain filling stage and combined with the ground disease survey data.Then a winter wheat take-all disease index estimation model was proposed based on the spectral band division interval and selected band combination. According to the normalized difference spectral index (NDSI) and the determinative coefficient of the disease index formed by any two band combinations, the spectral index band combinations corresponding to the spectral index with high correlation in each region were screened by dividing spectral intervals. Partial least-squares regression was used to establish a binary and ternary disease index calibration model. The results showed that the model based on spectral indices of ternary variables had the highest coefficient of determination. Finally, the optimal regression model of wheat take-all disease condition index composed of NDSI(R590,R598), NDSI(R534,R742) and NDSI(R810,R834) was established: Y = 134.577 − 70.301 NDSI(R590,R598) − 223.533 NDSI(R534,R742) + 51.584 NDSI(R810,R834) (R2 = 0.743, RMSEP = 0.094, df = 15), which was the most suitable model for winter wheat take-all estimation. The construction of this model can provide new method and technical support for future evaluation and monitoring of wheat take-all disease on the field.

Assessment of Empirical Algorithms for Shallow Water Bathymetry Using Multi-Spectral Imagery of Pearl River Delta Coast, China

Pearl River Delta (PRD), as one of the most densely populated regions in the world, is facing both natural changes (e.g., sea level rise) and human-induced changes (e.g., dredging for navigation and land reclamation). Bathymetric information is thus important for the protection and management of the estuarine environment, but little effort has been made to comprehensively evaluate the performance of different methods and datasets. In this study, two linear regression models—the linear band model and the log-transformed band ratio model, and two non-linear regression models—the support vector regression model and the random forest regression model—were applied to Landsat 8 (L8) and Sentinel-2 (S2) imagery for bathymetry mapping in 2019 and 2020. Results suggested that a priori area clustering based on spectral features using the K-means algorithm improved estimation accuracy. The random forest regression model performed best, and the three-band combinations outperformed two-band combinations in all models. When the non-linear models were applied with three-band combination (red, green, blue) to L8 and S2 imagery, the Root Mean Square Error (Mean Absolute Error) decreased by 23.10% (35.53%), and the coefficient of determination (Kling-Gupta efficiency) increased by 0.08 (0.09) on average, compared to those using the linear regression models. Despite the differences in spatial resolution and band wavelength, L8 and S2 performed similarly in bathymetry estimation. This study quantified the relative performance of different models and may shed light on the potential combination of multiple data sources for more timely and accurate bathymetry mapping.

SATELLITE-DERIVED BATHYMETRY USING CONVOLUTIONAL NEURAL NETWORKS AND MULTISPECTRAL SENTINEL-2 IMAGES

Satellite-Derived Bathymetry (SDB) has been used in many applications related to coastal management. SDB can efficiently fill data gaps obtained from traditional measurements with echo sounding. However, it still requires numerous training data, which is not available in many areas. Furthermore, the accuracy problem still arises considering the linear model could not address the non-relationship between reflectance and depth due to bottom variations and noise. Convolutional Neural Networks (CNN) offers the ability to capture the connection between neighbouring pixels and the non-linear relationship. These CNN characteristics make it compelling to be used for shallow water depth extraction. We investigate the accuracy of different architectures using different window sizes and band combinations. We use Sentinel-2 Level 2A images to provide reflectance values, and Lidar and Multi Beam Echo Sounder (MBES) datasets are used as depth references to train and test the model. A set of Sentinel-2 and in-situ depth subimage pairs are extracted to perform CNN training. The model is compared to the linear transform and applied to two other study areas. Resulting accuracy ranges from 1.3 m to 1.94 m, and the coefficient of determination reaches 0.94. The SDB model generated using a window size of 9x9 indicates compatibility with the reference depths, especially at areas deeper than 15 m. The addition of both short wave infrared bands to the four visible bands in training improves the overall accuracy of SDB. The implementation of the pre-trained model to other study areas provides similar results depending on the water conditions.

Paradigm of Geological Mapping of the Adıyaman Fault Zone of Eastern Turkey Using Landsat 8 Remotely Sensed Data Coupled with PCA, ICA, and MNFA Techniques

A principal and independent component analysis (PCA and ICA) and a minimum noise fraction analysis (MNFA) were applied in this study to Landsat 8 Operational Land Imager (OLI) images along the Adıyaman fault zone in Eastern Turkey. These analyses indicated that the lithologic units, fault patterns, and the morphological and structural features can be mapped highly accurately by using spectral-matching techniques in regions where rocks are well exposed. An inspection of all possible band combinations indicated that the PCA 134 and 231 and the ICA 132 band combinations give the best false color composite images for identifying the lithological units and contacts. The findings of the MNFA band combinations show that the MNFA 521 band combination also is robust for discriminating the lithological units, particularly Quaternary clastic units (colluvium/alluvium). MNFA band 1 alone provides the best image for tracing the tectonic and structural elements in the study area. The new up-to-date lithologic map of the Adıyaman fault zone we produced upon the interpretation of the processed OLI images reveals several river channels that are offset and beheaded by the Adıyaman fault, which verifies its Quaternary activity. This study demonstrated that, when used with the OLI data, the PCA, ICA, and MNFA are very powerful for lithological and structural mapping in actively deforming tectonic zones and hence can be applied to other regions elsewhere in the world where the climate is arid to semiarid, and the vegetation cover is scarce.

Analyzing the Effects of Hyperspectral ZhuHai-1 Band Combinations on LAI Estimation Based on the PROSAIL Model

Leaf area index (LAI) is a key biophysical variable to characterize vegetation canopy. Accurate and quantitative LAI estimation is significant for monitoring vegetation growth status. ZhuHai-1 (ZH-1), which is a commercial remote sensing micro-nano satellite, provides a possibility for quantitative detection of vegetation with high spatial and spectral resolution. However, the band characteristics of ZH-1 are closely related to the accuracy of vegetation monitoring. In this study, a simulation dataset containing 32 bands of ZH-1 was generated by using the PROSAIL model, which was used to analyze the performance of 32 bands for LAI estimation by using the hybrid inversion method. Meanwhile, the effect of different band combinations on LAI estimation was discussed based on sensitivity analysis and the correlation between bands. Then, the optimal band combination from ZH-1 hyperspectral satellite data for LAI estimation was obtained. LAI estimation was performed based on the selected optimal band combination of ZH-1 satellite images in Xiantao city, Hubei province, and compared with the Sentinel-2 normalized difference vegetation index (NDVI) values and LAI product. The results demonstrated that the obtained LAI map based on the optimal band combination of ZH-1 was generally consistent with the overall distribution of Sentinel-2 NDVI and the LAI product, but had a moderate correlation with Sentinel-2 LAI (R = 0.60), which may not favorably indicate the validity of indirect validation. However, the method of this study on the analysis of hyperspectral data bands has application potential to provide a reference for selecting appropriate bands of hyperspectral satellite data to estimate LAI and improve the application of hyperspectral data such as ZH-1 in vegetation monitoring.

Mapping Forest Cover Changes using Sentinel-2A Imagery in the Municipality of Zubin Potok, Republic of Kosovo

This research aimed to investigate the changes in forest cover, utilizing Sentinel-2A imagery data. Annual results of deforestation, non-forest, and forest area in the Municipality of Zubin Potok (Kosovo) between 2016 and 2017 were presented and analyzed by applying the image difference change detection method on a Normalized Difference Vegetation Index (NDVI) product derived for both years. The study reveals that forest coverage in this municipality has changed due to human activity (harvested and burnt forests). The footprint of changes was evidenced by using Sentinel 2A band combinations and very high resolution (VHR) images available in Google Earth (GE). From the overall forest-covered area of 24,873.61 hectares, the detected changes during the annual period are as follows: 24,423.57 ha or 98.19 % is mapped as forest, 113.75 hectares or 0.46 % as non-forest, and 336.77 or 1.35 % of the area forest is mapped as deforestation. These results can be used to identify human-made deforestation and to develop monitoring forest plans for the coming years.