Next Article in Journal
A Multi-Level Cross-Attention Image Registration Method for Visible and Infrared Small Unmanned Aerial Vehicle Targets via Image Style Transfer
Next Article in Special Issue
Refined Intelligent Landslide Identification Based on Multi-Source Information Fusion
Previous Article in Journal
Research on Infrared Dim Target Detection Based on Improved YOLOv8
Previous Article in Special Issue
A Novel Attention-Based Generalized Efficient Layer Aggregation Network for Landslide Detection from Satellite Data in the Higher Himalayas, Nepal
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Advancing Volcanic Activity Monitoring: A Near-Real-Time Approach with Remote Sensing Data Fusion for Radiative Power Estimation

by
Giovanni Salvatore Di Bella
1,*,
Claudia Corradino
1,
Simona Cariello
1,2,
Federica Torrisi
1 and
Ciro Del Negro
1
1
Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Catania, Osservatorio Etneo, Piazza Roma 2, 95125 Catania, Italy
2
Department of Electrical, Electronic and Computer Engineering, University of Catania, Viale Andrea Doria 6, 95125 Catania, Italy
*
Author to whom correspondence should be addressed.
Remote Sens. 2024, 16(16), 2879; https://doi.org/10.3390/rs16162879
Submission received: 19 July 2024 / Revised: 4 August 2024 / Accepted: 5 August 2024 / Published: 7 August 2024
(This article belongs to the Special Issue Application of Remote Sensing Approaches in Geohazard Risk)

Abstract

:
The global, near-real-time monitoring of volcano thermal activity has become feasible through thermal infrared sensors on various satellite platforms, which enable accurate estimations of volcanic emissions. Specifically, these sensors facilitate reliable estimation of Volcanic Radiative Power (VRP), representing the heat radiated during volcanic activity. A critical factor influencing VRP estimates is the identification of hotspots in satellite imagery, typically based on intensity. Different satellite sensors employ unique algorithms due to their distinct characteristics. Integrating data from multiple satellite sources, each with different spatial and spectral resolutions, offers a more comprehensive analysis than using individual data sources alone. We introduce an innovative Remote Sensing Data Fusion (RSDF) algorithm, developed within a Cloud Computing environment that provides scalable, on-demand computing resources and services via the internet, to monitor VRP locally using data from various multispectral satellite sensors: the polar-orbiting Moderate Resolution Imaging Spectroradiometer (MODIS), the Sea and Land Surface Temperature Radiometer (SLSTR), and the Visible Infrared Imaging Radiometer Suite (VIIRS), along with the geostationary Spinning Enhanced Visible and InfraRed Imager (SEVIRI). We describe and demonstrate the operation of this algorithm through the analysis of recent eruptive activities at the Etna and Stromboli volcanoes. The RSDF algorithm, leveraging both spatial and intensity features, demonstrates heightened sensitivity in detecting high-temperature volcanic features, thereby improving VRP monitoring compared to conventional pre-processed products available online. The overall accuracy increased significantly, with the omission rate dropping from 75.5% to 3.7% and the false detection rate decreasing from 11.0% to 4.3%. The proposed multi-sensor approach markedly enhances the ability to monitor and analyze volcanic activity.

Graphical Abstract

1. Introduction

Technological advancements in satellite remote sensing have transformed our perception and understanding of volcanic processes. The continuous monitoring of volcanoes and the study of volcanic hazardous phenomena are of great interest and are constantly evolving [1,2]. This progress has led to the development of various sensors, both dedicated and non-dedicated, to measure volcanic thermal anomalies from space [3,4]. Among these, the Moderate Resolution Imaging Spectroradiometer (MODIS), the Sea and Land Surface Temperature Radiometer (SLSTR), and the Visible Infrared Imaging Radiometer Suite (VIIRS) onboard polar satellites, as well as the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) on geostationary satellites, are widely used for volcano monitoring. High spatial resolution satellite sensors also play a key role in monitoring the behavior of a volcano, such as the MultiSpectral Instrument (MSI) on Sentinel-2 (from ESA) and the Operational Land Imager (OLI) and the Thermal Infrared Sensor (TIRS) on Landsat-8 (from NASA & USGS) [5,6,7].
Implementing new algorithms that enhance precision and reliability in real-time observation of high-temperature volcanic features is fundamental to improving our understanding of volcanic processes, identifying renewed volcanic activity, forecasting eruptions, and assessing hazards [8,9]. Thermal infrared sensors on board satellite platforms allow for continuous monitoring of active volcanoes worldwide, providing accurate estimates of volcanic thermal emissions [10]. Several volcanic hotspot detection algorithms have been proposed, utilizing different spatial [11], temporal [12], and spectral contexts [13] of the satellite sensors. These detection algorithms can be classified into four main categories based on the information used to identify an anomalous pixel: contextual (e.g., the VAST algorithm [14]), temporal (e.g., the RAT algorithm [15]), spectral (e.g., MODVOLC [16]), and hybrid (e.g., the Okmok algorithm [17]). Further classification is possible based on the criteria used to identify hotspots. Most algorithms apply intensity thresholds to one or more bands, which can be predetermined (e.g., the NHI algorithm [18,19]), statistical (e.g., the HOTSAT algorithm [20,21]), or based on machine learning approaches [6,22].
Several volcanic monitoring satellite systems operate in near real time (NRT), such as MIROVA [23], MODVOLC, HOTSAT, HOTVOLC [24] and FASTVRP [25], utilizing traditional algorithms to automatically provide radiance information from volcanic anomalous pixels worldwide. These systems process infrared images acquired by satellite sensors to calculate relevant Volcanic Radiative Power (VRP), which measures the heat radiated by volcanic activity, analogous to the Fire Radiative Power (FRP) [26,27] measure used for fire detection. While these algorithms generally perform well in detecting intense volcanic activity [28,29], they may have limitations, particularly in detecting more subtle thermal anomalies. These limitations arise from the low spatial resolution of satellite sensors, which average incoming radiance information over a large area into a single value. Consequently, spatially-confined volcanic emissions might not significantly affect the averaged pixel value, making it appear similar to pixels covering heated slopes.
In more advanced techniques, the replication of visual system processes has been introduced to account for spatial features, such as the shapes of volcanic thermal anomalies [30]. This approach enhances the detection of volcanic anomalies by using geometrical features to overcome the intrinsic limitations of the instruments. Recent studies have demonstrated the importance of using spatial features that replicate the focus mechanism, particularly in the context of the ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) multispectral thermal infrared (TIR) dataset using statistical [31] and deep learning [32] algorithms. Regardless of the spatial resolution, this approach appears to be feasible for every satellite sensor. Typically, each satellite sensor has its own specific volcanic hotspot detection algorithm, primarily relying on intensity. To operationally utilize all available satellite data, data fusion is needed to provide VRP estimates as soon as new data becomes available.
Here, we introduce a new algorithm for multi-sensor VRP estimates, named the Remote Sensing Data Fusion (RSDF) algorithm. This algorithm employs a novel volcanic hotspot detection method that exploits both intensity and spatial features. We describe and demonstrate the operation of the RSDF algorithm by analyzing the eruptive activity at the Etna and Stromboli volcanoes from 2021 to 2023 [33,34]. Specifically, we first focus on two similar polar satellite sensors, SLSTR and MODIS, demonstrating enhanced performance compared to the SLSTR Level 2 products, distributed by Copernicus Dataspace [35], and the MODIS fire detection product, available online on the Fire Information for Resource Management System (FIRMS) [36]. We then show the potential of using this approach with the polar-orbiting VIIRS and SEVIRI sensors. Improved performance over single sensor existing algorithms is shown for the SLSTR, MODIS, VIIRS, and SEVIRI sensors.

2. Materials

2.1. Case Studies: Etna and Stromboli Volcanoes

We investigated the Etna and Stromboli volcanoes, both characterized by frequent eruptive events that produced exemplary volcanic heat sources during the analyzed period from 2021 to 2023.
Mount Etna, located on the east coast of Sicily, is one of the most active volcanoes in the world, known for its persistent activity in summit craters—consisting of degassing and explosive phenomena associated with fast-moving lava flows—and recurrent effusive eruptions from vents located on the flanks of the volcano, producing lava flows that can extend for several kilometers [37,38]. During the investigated period, Etna experienced numerous paroxysmal events, particularly in early 2021, with powerful lava fountains, ash plumes reaching several kilometers into the atmosphere, and extensive lava flows that significantly altered the landscape [39]. Notably, the series of paroxysmal events in February and March 2021 were among the most intense, resulting in widespread ash fall and disruptions to local communities and air traffic [6,40,41].
Similarly, Stromboli, situated in the Aeolian Islands, is renowned for its continuous mild explosive activity, often referred to as “Strombolian” eruptions [42,43]. From 2021 to 2023, Stromboli exhibited consistent eruptive behavior with frequent explosive bursts that ejected incandescent lava fragments, ash, and volcanic gases. The volcano’s activity included several significant episodes, such as the July 2021 eruption which featured a major explosive event, producing high-energy pyroclastic flows that reached the sea and prompted evacuation warnings for nearby residents and tourists. Additionally, its volcanic activity in late September 2023 was marked by heightened explosive activity, creating substantial volcanic heat sources and providing valuable data for our analysis [44,45,46].
The frequent and varied eruptive behaviors of these volcanoes during the study period provided an excellent opportunity to test and validate our RSDF algorithm’s ability to detect and analyze volcanic thermal activity across different sensors.

2.2. Satellite Sensors

The Sea and Land Surface Temperature Radiometer (SLSTR) is on board the Sentinel-3A and 3B satellites. It is designed to monitor temperatures of Earth’s terrestrial and aquatic surfaces with a precision of under 0.2 K. The SLSTR features nine spectral channels ranging from 0.55 µm to 12.0 µm, including three in the visible and near-infrared spectrum, three in the short-wave infrared, one in the mid-infrared, and two in the thermal infrared. SLSTR data are captured in both nadir and oblique views, offering spatial resolutions of 1 km and 0.5 km for VIS/SWIR channels, across swaths of 1400 km and 750 km, respectively. The algorithm extracts information using the S7 (Medium InfraRed, MIR) and S8 (Thermal InfraRed, TIR) bands for brightness temperature, switching to Fire Channels F1 and F2 when saturation temperatures of 312 K and 500 K are exceeded. Data are provided in netCDF 4 format, covering both nadir and oblique views, with comprehensive metadata for each pixel [47].
The Moderate Resolution Imaging Spectroradiometer (MODIS) operates on NASA’s Terra and Aqua satellites, providing global coverage every 1–2 days across 36 spectral bands from 0.4 µm to 14.4 µm. It boasts a high 12-bit radiometric sensitivity with spatial resolutions of 250 m, 500 m, and 1 km, covering a 2330-km swath in a 55-degree scanning pattern from a 705 km orbit. Its optical system includes a rotating double-sided scan mirror and off-axis telescope, cooled to 83 K for infrared bands. MODIS Level 1 data, available since February 2000, are provided in a geographic projection at 1 km resolution. The radiance values are extracted from the MIR (channel 22) and TIR (channel 32) bands, switching to channel 21 when the MIR band’s saturation threshold of 328 K is exceeded. Data are encoded as 16-bit unsigned integers, requiring conversion for analysis, and are accessible via the LAADS web platform [48].
The Visible Infrared Imaging Radiometer Suite (VIIRS), deployed on the Suomi NPP and JPSS satellites, captures detailed data on Earth’s land cover, surface temperatures, and atmospheric parameters across various spectral bands. It provides global coverage, collecting data from diverse regions during each orbit. A notable feature is its ability to observe at night using low-light bands, useful for monitoring urban lighting and nocturnal fires. VIIRS L1B granules, available from the LAADS-DAAC system, include MIR (channel I4) and TIR (channel I5) brightness temperature data at 375 m resolution. These bands have saturation temperatures of 367 K and 380 K, respectively, and switch to M13 (MIR) and M15 (TIR) bands at 750 m resolution when saturated. After resampling into a UTM grid, a matrix of 67 by 67 pixels is obtained, due to the sensor’s 750-m resolution [48].
The Spinning Enhanced Visible and Infrared Imager (SEVIRI), carried on EUMETSAT’s Meteosat Second Generation satellites, provides data at 15-min intervals with spatial resolutions of 3–5 km and 1 km for the High-Resolution Visible channel. It supports weather forecasting, climate monitoring, and environmental research by assessing parameters like cloud properties, sea surface temperatures, and land surface temperature. SEVIRI data are transmitted in 12 spectral channels through high-rate transmissions, with particular focus on the IR3.9 (MIR) and IR10.8 (TIR) bands. These data are resampled into UTM coordinates, forming a 12 × 12 km grid, aiding in the monitoring of severe weather events such as storms, wildfires, and hurricanes [47].
A summary of the main characteristics of the four satellite sensors utilized is shown in Table 1.

2.3. SLSTR and MODIS Level 2 Products and Fire Radiative Power

The Sentinel-3 SLSTR Level 2 databases are available for open access via Copernicus dataspace. This platform facilitates the download of volcanic anomaly data from both the Etna and Stromboli volcanoes within the same time window. We extract Fire Radiative Power (FRP) values from the same volcanic anomalies obtained through an Active Fire Detection algorithm [27,35,49]. Originally designed to identify and describe vegetation fires occurring on land surfaces, the algorithm was also adapted to detect areas with elevated temperatures related to active volcanism and intense industrial heat sources. Fire detection employs varying fixed thresholds to identify confirmed fire pixels in near-nadir view data, adjusting for daytime and nighttime conditions. For each image, FRP values of individual hotspots are collected and summed to derive the total FRP value for each acquired image. NASA’s MODIS active fire products marked the introduction of remotely sensed fire datasets [36,50,51]. Utilizing moderate spatial resolution (~1 km), these data sets originate from advanced “fire-capable” sensors on satellite platforms [27]. By accessing MODIS Level 2 products at https://firms.modaps.eosdis.nasa.gov/, accessed on 18 January 2024, it is possible to download .csv files containing FRP values of detection alerts for a given period and geographical area [36].

3. Methods

The RSDF algorithm detects volcanic hotspots and estimates the emitted VRP for each image acquired. Information is extracted from multispectral satellite sensor images through spectral and spatial analysis, exploiting the potential of a bank of Gabor filters [31] known for their effectiveness in improving detection sensitivity and reducing the effects of clouds (Figure 1).

3.1. Spectral Analysis

The initial step involves extracting Level 1 data from each satellite sensor, including geolocation, date, time, brightness temperature, and radiance specific to the relevant spectral band. Geolocation data is linked to the target volcano’s position. Different satellites yield distinct Level 1 data based on sensor instruments. The algorithm focuses on the MIR and TIR bands, specifically targeting the MIR range of 3.973–4.128 μm and the TIR range of 10.263–11.263 μm. Band selection is influenced by the saturation temperature, which particularly affects the MIR band. The choice of these two bands is related to their frequency response; for low temperatures, the spectral radiance is similar for both bands, but for high temperatures, the MIR band shows a significant increase in spectral radiance compared to the TIR band. The next step involves performing a spectral analysis, deriving the Normalized Thermal Index (NTI) [16]. The NTI is calculated as follows, exploiting the differences between the MIR and TIR bands:
N T I = L M I R L T I R L M I R + L T I R
where L M I R and L T I R are the radiance recorded by the MIR and TIR bands, respectively. The NTI demonstrates heightened sensitivity to hot surfaces, leading to a substantial increase in L M I R readings. This enhanced sensitivity allows for the precise detection of thermal anomalies in the image, facilitating the automatic identification of active hotspot pixels.

3.2. Spatial Analysis

Following the discovery of the NTI matrix, a spatial analysis is conducted using a Spatial Standard Deviation (SSD) filter applied to each pixel in the image. The SSD involves subtracting the mean (or median) of eight neighboring pixels from the NTI values matrix to mitigate artifacts such as seas or clouds [20]. The resulting SSD is then cropped and divided into volcanic areas (VAs) and non-volcanic areas (NVAs). The former is centered on the volcano’s crater, halved from the original matrix, while the non-volcanic area is obtained by subtracting the volcanic area.

3.3. Statistical Mask

The first mask (Mask1) is defined by the maximum SSD value in the non-volcanic area (SSDNVA) serving as a threshold. It is compared to the SSD values of each pixel in the volcanic area (SSDVA), marking pixels exceeding the maximum SSD value of the non-volcanic area (maxSSDNVA) as “potential hotspots”, expressed as
S S D V A > m a x S S D N V A
The second mask (Mask2) is applied to differentiate and identify the “true hotspots” among the potential ones. The condition for Mask2 is
N T I V A > m e a n N T I V A + n s t d N T I V A
where m e a n N T I V A   a n d   s t d N T I V A are, respectively, the mean of NTI and the standard deviation of NTI of the volcanic area (VA). n serves as a parameter determining the permissible deviation of the integrated MIR pixel temperature from the mean value. For polar satellites (MODIS, VIIRS and SLSTR), a distinction between daytime and nighttime data is necessary due to the significant influence of sunlight on daytime data values. Therefore, n = 10 is chosen for daytime data, while n = 5 is selected for nighttime data for MODIS and SLSTR sensors. Regarding VIIRS, given the different spatial resolution compared to previous sensors and the size of the volcanic area, n = 8 was adopted for daytime and n = 12 for nighttime. As for geostationary satellites (SEVIRI), as they are at a low spatial resolution they are not heavily affected by solar radiation, and thus a unique n value is choosen, set to 1.

3.4. Gabor Image

We opted to employ a bank of Gabor filters to enhance the performance of the detection algorithm [31]. These filters are derived from the product of a Gaussian function and an oriented sinusoidal plane wave. Gabor filters highlight local spatial features in images by focusing on the spatial frequency content around each pixel. When used in conjunction with thermal intensities, they enable the detection of anomalies based on both temperature and spatial ‘texture’. This is especially crucial for identifying subtle anomalies characterized by low intensities, such as groups of warm pixels near volcanic crater vents.
The use of the Gabor filter on the selected image aims to extract the most significant features from MIR images. This task is achieved through the multi-step procedure described in Figure 1. Firstly, candidate features are retrieved using a bank of Gabor filters, Gaussian low-pass filter smoothing of the Gabor image, and spatial information. This results in a feature matrix with 24 Gabor features plus 2 spatial features per pixel. After normalization, i.e., the matrix is Z-score normalized to achieve zero mean and unit variance, Principal Component Analysis (PCA) is applied [31]. This step allows for the retrieving and selecting of only the most informative features, reducing the risk of overlap or redundancy.
The output is ultimately an image highlighting the most significant regions relative to the surrounding context. The choice to generate the Gabor image from the MIR band is due to the need to highlight thermal anomalies and avoid the cloud effects that influence the TIR band, which may affect the NTI. By multiplying the Gabor image with the NTI matrix obtained in the previous steps, we obtain the Gabor Weighted NTI (G-NTI).

3.5. Spatial Weighted

We generate a Spatial Gabor Weighted NTI (SG-NTI), aiming to highlight hotspots in the crater area and reduce the effect of any non-volcanic hotspots due to the uncertainty outside the volcanic zone.

3.6. Statistical Mask

In this step, similarly to Step 3, a mask is computed based on the SG-NTI matrix
M a s k 3 : S G N T I > n s t d S G N T I + m e a n S G N T I
where m e a n S G N T I   a n d   s t d S G N T I are respectively the mean of SG-NTI and the standard deviation of SG-NTI.

3.7. Calculation of VRP

The final binary mask identifying real volcanic hotspots is the product of the three previously computed masks (Figure 2). Once we identify and localize hotspot pixels (i.e., Mask = Mask1 × Mask2 × Mask3), we calculate the VRP mask using
V R P M I R = σ ϵ A p i x e l a ϵ M I R L M I R , h V R P M I R = k L M I R L M I R , b g
where A p i x e l is the pixel ground sampling area ( m 2 ) and L M I R L M I R , b g is the excess of MIR radiance [6,22]. The background is the average value of the MIR band of the hotspots surrounding the “true hotspots” detected by the output of the total mask. The constant k is specific for each sensor and depends directly on the parameter a ( k M O D I S = 1.89 × 10 7   W m 2 s r 1 μ m 1 K 4 , k S L S T R = 1.70 × 10 7   W m 2 s r 1 μ m 1 K 4 , k S E V I R I = 3.70 × 10 7   W m 2 s r 1 μ m 1 K 4 , k V I I R S 375 = 2.48 × 10 7   W m 2 s r 1 μ m 1 K 4 , and k V I I R S 750 = 1.11 × 10 7   W m 2 s r 1 μ m 1 K 4 ).
The summation of the VRP values is then computed to estimate the total VRP value of the analyzed image:
i V R P M I R ( i )

3.8. Eruptive Parameters

The time series of VRP (Volcanic Radiative Power) is necessary for characterizing three eruptive parameters: the Time-Averaged Discharge Rate (TADR [m3/s]), the volume of lava flows, and the Volcanic Radiative Energy (VRE [J]). The TADR refers to the radiant density approach [52], connecting it directly to the VRP in cubic meters per second. This relationship is established through a single best-fit coefficient known as ‘radiant density’ ( c r a d ):
T A D R = V R P c r a d
where:
c r a d = 6.45 10 25 X S i O 2 10.4
where X S i O 2 10.4 represents the concentration of silicon dioxide adjusted by a factor of 10.4. Values of the parameter c r a d used for Etna and Stromboli are shown in Table 2.
Due to the uncertainty in deriving the TADR parameter, we calculated its maximum, minimum, and mean values using three different values of c r a d   [ J/m3]. Estimating c r a d is vital for understanding lava dynamics. With a ±50% uncertainty, minimum and maximum TADR values are computed for accuracy [53,54]. The volume of the lava flow is given by [55]:
T A D R m e a n = T A D R i + T A D R i 1 2 V o l u m e i = T A D R m e a n t
The last parameter is the Volcanic Radiated Energy (VRE), which represents the total energy emitted by a volcano, including thermal infrared radiation. The VRE is measured in Joules [J]. It serves as a crucial parameter for monitoring and understanding volcanic activity, enabling the analysis of eruption intensity and lava volume through remote sensing techniques such as satellite-based sensors. The calculation of VRE involves trapezoidal integration of VRP/FRP time series [56].

3.9. Uncertainties and Limits

The algorithm for Volcanic Radiative Power (VRP) estimation faces limitations, notably the MIR radiance method’s constraint to emitter temperatures above 600 K. This restricts its accuracy for colder emitters, yet it maintains reliability for warmer ones. Challenges like atmospheric clouds, volcanic clouds, fires, and satellite viewing angles introduce noise during acquisitions, generating false hotspots and masking potential thermal anomalies. Clouds complicate satellite observations, making it challenging to discern real hotspots from false positives linked to cloud presence. Introducing the cloud coverage index ( i c l o u d ) quantifies cloud presence within each pixel, aiding in understanding the impact of clouds on the image under investigation, expressed as
i c l o u d = N c l o u d V A N V A
where N c l o u d V A   is the number of cloudy pixels inside the volcanic area and N V A   is the total number of pixels inside the volcanic area. The detection of cloudy pixels, using satellite data, is achieved by considering atmospheric and surface conditions to calculate the probability of clear sky for each pixel. A cloud coverage index of 1 denotes that the image is categorized as cloudy, while an index of 0 indicates the absence of identified cloudy pixels within the volcanic region. We consciously avoided modifying datasets through image inspections or filters to eliminate cloudy or geometrically challenging scenes. This decision aims to ensure a precise comparison between time series datasets, maintaining the integrity of the original data and allowing for a more accurate assessment of cloud impact on volcanic observations over time.
Another significant consideration is the satellite viewing geometry, meaning that the acquired images could have high zenith and azimuth angle values. In volcanic satellite monitoring, this poses several challenges: geometric distortion of images, reduced spatial resolution affecting small volcanic features, thermal measurement inaccuracies due to atmospheric interference, and compromised data quality from increased atmospheric dispersion and absorption. These factors hinder accurate identification of hotspots in deep craters and areas at the base of craters. Using the information on zenith and azimuth angles contained in the data provided by the satellite sensors, it is possible to calculate the value of the scan angle. When this value often exceeds 70/80, it indicates a certain degree of projected thermal anomaly distortion, visible as an increase in the number of pixels with the same exact value. However, the validation of this uncertainty in the algorithm’s results, as with cloudiness, has not undergone filtering to maintain its integrity.

4. Results

4.1. Etna

The RSDF algorithm identified 764 alerts ( N a l e r t ) between 2021 and 2023, recognized as satellite acquisitions with thermal anomalies in the volcanic area by the SLSTR sensor. This is nearly two times the 456 alerts identified by the Level 2 data processing. Considering the 2545 SLSTR overpasses ( N p a s s ) , there is a noticeable contrast in alert detection frequency (f% = N a l e r t / N p a s s ). The algorithm registers a frequency of 30% ( N a l e r t = 764), a marked difference compared to the 18% seen in the SLSTR product level 2 ( N a l e r t = 445).
Similar to SLSTR, the results of MODIS active fire products, provided by NASA, indicate insufficient sensitivity for continuous volcanic monitoring. The RSDF algorithm provides a N a l e r t = 858   with f = 17%, while the Level 2 product provides N a l e r t = 13   with a frequency of 0.002%, both data related to N p a s s = 5171, as achieved by the TERRA and AQUA satellites.
The temporal series of VRP from both sensors reveal a pattern that allows the identification of four volcanic eruptive phases, each representing a period of heightened volcanic activity over the analyzed timeframe.
Phase 1 spans from 15 February 2021 to 3 April 2021. Subsequent phases include phase 2, from 17 May 2021 to 28 September 2021, phase 3, from 12 May 2022 to 15 June 2022, and phase 4, from 27 November 2022 to 6 February 2023 (Figure 3).
In the period of February to May 2021, and generally throughout 2021, Mount Etna exhibited exceptionally high volcanic activity, validated by the occurrence of 57 paroxysmal events [39]. These events significantly influence VRP calculations, with SLSTR showing VRP values ranging from 400 MW to 1000 MW. Notably, in phase 1, there is a peak VRP of 2139 MW detected on 28 February 2021, at 09:40 UTC.
Regarding MODIS, it exhibits very high peaks in the early part of 2021, with the series’ maximum reaching 7684 MW on 21 May 2021, at 01:40 UTC. It is worth noting the absence of MODIS active fire products measurements. The period from April 2022 to June 2023 remains consistent in terms of magnitude. The highest peak during this timeframe for SLSTR was observed on 6 June 2022, at 09:36 UTC, with a VRP value of 1848 MW. On the other hand, MODIS recorded its peak on 30 May 2022, at 20:25 UTC, featuring a VRP value of 2054 MW.
An interesting observation is that in the months following intense volcanic activity, low-level thermal anomalies are rarely detected by Level 2 and MODIS active fire products data processing. In contrast, the implemented RSDF algorithm demonstrates enhanced accuracy and sensitivity in identifying these anomalies, enabling continuous monitoring of thermal anomalies throughout all phases of volcanic activity, including the presence of low-intensity thermal anomalies inside the volcanic crater.
Generally, higher Fire Radiative Power (FRP) values from the Sea and Land Surface Temperature Radiometer (SLSTR) are found compared to Volcanic Radiative Power (VRP). This correlation is due to the Radiative Surface Detection Function (RSDF) having lower false detection rates, resulting in lower total VRP values. This is especially true during periods of greater volcanic intensity, such as the 2021 lava fountain period on Etna. During such times, RSDF proves to be more accurate in detecting the number of hotspots, which tend to lower the VRP value.
Statistical analysis serves a dual purpose: it validates existing hypotheses regarding alert numbers and establishes thresholds capable of distinguishing between different volcanic thermal regimes. In long-term monitoring, the temporal complementarity of datasets is crucial, facilitating a point-to-point comparison of the diverse values obtained.
The distribution of alerts (Figure 4) confirms the RSDF algorithm’s heightened sensitivity in detecting hotspots with low VRP values, evident for all values below a logVRP threshold of approximately 8.5 MW for SLSTR and 7.8 MW for MODIS. The absence of alerts in the Level 2 product dataset and MODIS active fire products further emphasizes this sensitivity. However, a similar trend is observed above the threshold, as demonstrated earlier, indicating the occurrence of high-intensity thermal volcanic events.
This analysis of the alert distribution for MODIS highlights the lack of thermal observations in the MODIS active fire products. The use of a probabilistic plot, where any alteration in the curve’s slope is associated with a change in thermal regime, confirms that surpassing the previously defined thresholds causes the logVRP points to deviate from the theoretical line, manifesting extraordinary behavior that is attributed to phenomena such as lava fountains and paroxysms. This trend is also evident in the datasets obtained via processing SLSTR datasets. Consistently, this analysis holds true for the MODIS dataset processed by the RSDF algorithm, emphasizing the low reliability and sensitivity of the MODIS active fire products. Comparing RSDF performance, computed using the Etna and Stromboli study cases, with MIROVA outcomes for the same volcanoes, we find an improvement in terms of omitted detections (3.7% vs. 10%) and false detections (4.3% vs. 5%) [28]. The two metrics calculated for the two volcanoes refer to different datasets compared to this work [28]. By utilizing the calculation of average VRP values on a weekly scale (VRPw) for each sensor in each dataset, it is noted that the absence of intermediate values between two significant eruptive events confirms the previously shown results (Figure 5). This underscores the reliability and sensitivity demonstrated by the RSDF algorithm. From Figure 5, it is observed that the trends of MODIS and SLSTR processed through the RSDF algorithm are nearly coincident and overlapping, indicating considerable numerical compatibility. In the graph, the deviation of individual values of SLSTR and MODIS, processed via RSDF, is attributable to the different revisit times of the two sensors and the uncertainty conditions present for each satellite pass.

4.2. Stromboli

The application of the RSDF algorithm to a volcano under non-ideal conditions, such as the small island of Stromboli with its proximity to a sea that introduces noise during processing, demonstrates robust sensitivity for continuous monitoring over an extensive time scale. The algorithm detected 605 alerts out of 2545 overpasses, rendering the detection of thermal activity consistent, with an alarm frequency of 24%. For the same number of overpasses, the Level 2 product presents only 25 alerts (Nalert) and a frequency (f%) of 0.01% highlighting the insufficient sensitivity of the Level 2 product for sustained monitoring.
Turning to MODIS, the RSDF algorithm registers 570 alerts (Nalert) with a frequency (f%) of 11%. The relatively low frequency and alert count, compared to the total detections ( N p a s s = 5171), is attributed to the presence of the sea, which introduces noise into the system and negatively influences the calculation of VRP. Even less precise and reliable are the MODIS active fire products, which report only 11 alerts, considering the same number of overpasses ( N p a s s = 5171) and a frequency of 0.002%.
Examining Stromboli’s volcanic activity from 2021 to 2023 reveals a dynamic, three-year period marked by a continuum of volcanic events, ranging from ordinary to extraordinary and encapsulated in three distinct phases. The first phase (13 May 2021–27 May 2021) witnesses pyroclastic flows and lava overflows. The second phase (11 May 2022–8 June 2022) encompasses a significant explosion on May 13, resulting in substantial material emission. The final phase spans from 27 November 2023 to 11 December 2023. While the VRP data from Level 2 and MODIS active fire products are slightly over a dozen units, the RSDF algorithm manages to detect peaks during the first eruptive cycle. Notably, on 19 May 2021 at 20:22 UTC, the maximum recorded VRP value is 959 MW. In the case of MODIS, the values are slightly lower, with a peak VRP of 997 MW on 19 May 2021 at 12:55 UTC, corresponding to the occurrence of a pyroclastic flow and overflow that reached the sea (Figure 6).
The utilization of Level 2 product and MODIS active fire product databases for both sensors proves inadequate for continuous monitoring of thermal anomalies in Stromboli. Consequently, leveraging the SWIR bands becomes crucial, capturing large thermal anomalies with very high VRP/FRP values. In this study, for a consistent and complementary comparison, the results of the processing are obtained exclusively using the MIR and TIR bands.
In general, the VRP values are lower than in the case of Etna, strongly influenced by the numerous paroxysmal events of 2021. Statistically analyzing the distribution of SLSTR sensor alerts reveals that below a logVRP threshold of 7.3 MW, the Level 2 product exhibits a complete absence of detections. The RSDF algorithm demonstrates an increasing trend up to an 8 MW threshold before a decline in the number of alerts corresponding to extraordinary volcanic events (Figure 7a). At this specific threshold, observed in the probabilistic plot, there is a marked change in slope, with points deviating from the theoretical distribution line.
The frequency distribution of alerts from MODIS mirrors the trend observed in the statistical analysis of SLSTR. In Figure 7b, the threshold of 8 MW divides the graph into two distinct zones. On the left, there is a zone indicative of ordinary events, while on the right, a zone emerges with points deviating from the theoretical line, signaling extraordinary events.
Extracting the weekly average VRP [48], as done in the case of Etna, a similar temporal trend is observed in Figure 8, especially for data derived from SLSTR. The continuity in detecting hotspots is also highlighted here, even during periods of mild volcanic activity.
The trends of MODIS and SLSTR, as processed by the RSDF algorithm, exhibit similar patterns in Figure 8, with MODIS showing lower numerical values. The discrepancy in SLSTR and MODIS values can be attributed to the varying revisit times of the sensors and uncertainty conditions during satellite passes.

4.3. VRE (Volcanic Radiated Energy)

The final volcanic parameter under analysis is the VRE, obtained from the time series of VRP and FRP from both MODIS and SLSTR. Table 3 presents the VRE values (in Joules), allowing for comparisons of eruptions in the different phases identified earlier for both Etna and Stromboli. An initial quantitative comparison between Etna and Stromboli highlights VRE values for Etna that are several orders of magnitude higher than those for Stromboli. This disparity arises from substantial differences in their volcanic activities.
Regarding the results of the RSDF algorithm compared to the Level 2 products and MODIS active fire products, it is important to note that during periods between two eruptive activities a constant segment is observed in the Level 2 product, whereas the RSDF algorithm shows a segment that grows linearly before reaching a peak marking the beginning of a new significant volcanic event. For Etna, it is interesting to note that SLSTR VRE values—both for the data processed by the RSDF algorithm and for the Level 2 data—as well as the MODIS data produced by the RSDF algorithm, show similar behavior in quantitative terms. The substantial difference is observed in the graph of the MODIS active fire products, with shows values differing by several orders of magnitude due to the limited amount of data available (Figure 9a,c). In the case of Stromboli, interpreting the comparison is more challenging due to the limited data recorded by the Level 2 products of both sensors. The efficiency of the RSDF algorithm enables tracking the evolution of the volcano’s state and distinguishing various significant volcanic events. Separately, in both cases, the results of VRE derived from the MODIS Level 2 product differ significantly from previous values due to lower reliability (Figure 9b,d). The higher VRE values derived from the Level 2 SLSTR products exceed those from RSDF as they are calculated from the VRP time series, highlighting this intensity difference.

5. Discussion

5.1. Performances

The temporal distribution of Etna’s VRP demonstrates the RSDF algorithm’s remarkable sensitivity in continuous volcanic monitoring over an extended time period. This sensitivity contrasts sharply with the low reliability of observations provided by SLSTR Level 2 products and MODIS active fire products, particularly in the detection of low-intensity thermal anomalies (Figure 2). The consistent results reported above show the algorithm’s superior performance in detecting thermal anomalies for the Etna and Stromboli volcanoes when compared to already-processed Level 2 data and MODIS active fire products. Two specific indices, the omissions rate and the false detection rate, were calculated to quantify the algorithm’s performance. The omission rate refers to pixels where volcanic activity is present but goes undetected, whereas the false detection rate refers to pixels indicated as true hotspots where no volcanic activity exists. A random set of images was selected from the dataset, and comparisons were conducted on an image by image basis, examining NTI, MIR, and the final mask obtained from the algorithm to identify false or omitted detected pixels. For Etna, using a sample of 200 data points from March 2021 to April 2021 taken from the SLSTR sensor, the RSDF algorithm exhibited a 2.9% omission rate and a 3.6% false detection rate, while the Level 2 product shows a 62% omission rate and a 12% false detection rate. Regarding MODIS data, analyzing a sample of 346 data points from July 2021 to August 2021, the RSDF algorithm showed a 4.5% omission rate and a 4.8% false detection rate, while the MODIS active fire products showed an 89% omission rate and a 10% false detection rate.
In the case of Stromboli, using a sample of 150 data points from June 2021 to August 2022 from the SLSTR sensor dataset, the RSDF algorithm showed a 5.4% omission rate and a 5.1% false detection rate. Conversely, the Level 2 product showed a 90% omission rate and 95% false detection rate. For MODIS data, using a sample of 150 data points from the same period, the RSDF algorithm showed a 5.7% omission rate and a 4.7% false detection rate, while the MODIS active fire products showed a 99% omission rate and a 1% false detection rate (Table 4). The Nalert and Npass parameters were calculated for the entire dataset for both Etna and Stromboli.
The anomaly detection frequency is higher for both Etna and Stromboli during daytime as compared to nighttime, primarily due to the influence of solar radiation, which is particularly significant during summer months. The RSDF algorithm demonstrates even greater improvement compared to active fire products for Stromboli. In this case, volcanic products exhibit significant numerical deficiencies, with omission rates of 80% for SLSTR and 99% for MODIS. Generally, a high false rate mainly results from false daytime alerts, often due to solar radiation. However, higher omitted rates tend to balance this out by a reduction in daytime alerts compared to nighttime alerts, given that nighttime thermal gradients are typically higher. SLSTR level 2 products exhibit a significantly higher omitted rate (62%) than false rate (12%), so the overall effect is largely driven by the higher omitted rate. Specifically, SLSTR has a lower overall percentage of Nalert compared to RSDF (18% vs. 30%), with notably higher daytime omissions, having only 5% daytime alerts compared to 13%.
A further classification can be made based on various intensity levels, categorized as low (green), moderate (yellow), high (orange), and extreme (red). Values below the green line belong to the low intensity region, values between the green and the yellow lines belong to the moderate region, values between the yellow and the orange lines are classed as high, and values above the orange line are classified as extreme. The gap between the various levels should be approximately an order of magnitude to distinguish between different volcanic events. These levels are extracted from the datasets using a statistical indicator like the percentile, enabling the comparison of a specific value with the entire matrix. In Figure 10a,b, the radiative power time series is represented in a logarithmic scale, using datasets from the MODIS and SLSTR sensors processed by the RSDF algorithm, for both Etna and Stromboli.
This classification allows for a clearer understanding of the intensity and frequency of volcanic events, facilitating better risk assessment and decision-making. The use of the RSDF algorithm thus not only enhances detection accuracy but also provides a structured approach to analyzing and categorizing volcanic activities based on their thermal intensity. This data fusion method proves effective for continuous and integrated volcano monitoring in terms of accuracy and the variety of data used, especially for active and complex volcanic regions like Etna and Stromboli.

5.2. Combining SEVIRI, MODIS, SLSTR, and VIIRS

The decision to use several sensors is closely tied to the desire to generate a unique and consistent VRP signal by integrating the findings from the VRP series of individual sensors. This method takes advantage of the various characteristics of the sensors in terms of spatial and temporal resolution. By integrating the results obtained from the RSDF algorithm for VRP, using data from SEVIRI, MODIS, SLSTR, and VIIRS, each with low to medium spatial resolution, it is possible to identify and confirm thermal anomalies across all four sensors. Figure 11a,b shows the time series of above average volcanic activity for Etna, from 1 February 2021 to 30 April 2021, and for Stromboli, from 27 September 2023 to 10 October 2023, respectively.
The dense columns of observable points during significant volcanic events can be attributed to SEVIRI’s high temporal resolution, which allows for the monitoring of high-intensity thermal anomalies more frequently than the other sensors.
The adaptability of the RSDF algorithm to different sensor types allows for the derivation of volcanic parameters such as the Time-Averaged Discharge Rate (TADR) and lava flow volume, leveraging the high temporal resolution (every 15 min) of a geostationary sensor like SEVIRI. The volume calculation involves dividing the TADR value by the time difference between consecutive observations and then aggregating it through a cumulative function to yield the total erupted lava volume.
Cumulative volume curves, derived from various TADR values, exhibit linear growth during the lava flow emplacement, followed by a plateau representing the cooling period between successive flows. For instance, during an effusive event on Mount Etna from 14 May 2022 to 16 June 2022, the volume of erupted lava stabilized at around 680,197 m3 (Figure 12).
To evaluate the adaptability of the RSDF algorithm through the integration of an additional sensor like SEVIRI, we examined a recent effusive eruption on Stromboli from 25 September 2023 to 10 October 2023. This event was used to compute TADR and the volume of lava flows, quantifying the size and magnitude of the volcanic events. The VRP flow analysis identified five distinct volcanic events, each linked to a corresponding lava flow. Notably, the initial event on September 27 at 14:30 UTC significantly contributed to the total erupted volume, constituting 45% of the maximum volume (112.6 m3/s out of 251.3 m3/s), the mean volume (56.3 m3/s out of 125.7 m3/s), and the minimum volume (37.3 m3/s out of 83.8 m3/s) (Figure 13). The limitations requiring greater attention are related to factors of uncertainty that negatively influence the accurate detection of thermal anomalies. Factors such as clouds and satellite viewing angles necessitate careful validation of each individual image. Upcoming developments will include comparing our system with other existing systems to evaluate different data validation methodologies.

6. Conclusions

Effective monitoring of active volcanoes requires the use of multiple sensors and accurate hotspot detection algorithms. In this study we demonstrate several key points:
Importance of SLSTR for Volcanic Monitoring: SLSTR, typically not used for volcanic activity monitoring, has proven to be an important sensor due to its ability to detect thermal anomalies effectively.
Enhanced Detection Algorithm: The proposed detection algorithm, which incorporates the Gabor filter on medium spatial resolution satellite sensors, significantly improves sensitivity and reduces false positives. This improvement is crucial for accurate and reliable volcanic monitoring.
Versatility Across Sensors: The detection algorithm’s capacity to be applied uniformly across both polar and geostationary sensors showcases its versatility. This uniform application ensures consistent monitoring regardless of the sensor’s specific characteristics.
The analysis conducted on the Etna and Stromboli volcanoes from 2021 to 2023 highlighted several improvements. Specifically, the overall accuracy increased significantly, with the omission rate dropping from 75.5% to 3.7% and the false detection rate decreasing from 11% to 4.3%. Additionally, intensity levels based on the logarithmic scale of VRP time series were calculated. These intensity levels are instrumental in pinpointing the volcano’s activity status at any given moment.
We underscore the necessity of not only relying on individual sensors but also of integrating information from multiple sensors with different characteristics. By doing so, it is possible to address each sensor’s weaknesses and leverage their strengths, depending on their spatial and temporal resolutions. This integrative approach enables the monitoring of hotspots characterized by both high and medium-low intensity, which are particularly challenging for sensors with low spatial resolution.
Finally, the implementation of a multi-sensor approach combined with an enhanced detection algorithm significantly advances the capability to monitor and analyze volcanic activity. This method provides an accurate satellite remote sensing data fusion algorithm for volcano thermal monitoring, essential for timely warnings and forecasting volcanic hazards. The fusion of polar and geostationary data is crucial for advancing volcanic monitoring and ensuring better preparedness for volcanic events. The study demonstrates that leveraging various sensors and sophisticated algorithms is key to improving volcanic monitoring and preparedness. Future developments will involve expanding the number of volcanoes investigated to enable global volcanic monitoring and exploring super resolution techniques.

Author Contributions

Conceptualization, G.S.D.B. and C.C.; methodology, G.S.D.B. and C.C.; software, G.S.D.B., S.C., F.T. and C.C.; validation, S.C. and F.T.; formal analysis, S.C. and F.T.; data curation, G.S.D.B. and C.C.; writing—original draft preparation, G.S.D.B. and C.C.; writing—review and editing, C.D.N., S.C. and F.T.; visualization, G.S.D.B.; supervision, C.D.N.; project administration, C.D.N.; funding acquisition, C.D.N. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by ATHOS Research Programme (INGV OB.FU. 0867.010), by the 2019 Strategic Project FIRST—ForecastIng eRuptive activity at Stromboli volcano: timing, eruptive style, size, intensity, and duration—of the INGV Volcanoes Department (Delibera n. 144/2020), and by the Project INGV Pianeta Dinamico VT_ORME 2023–2025 (INGV OB.FU. 1020.010).

Data Availability Statement

All SLSTR, MODIS, VIIRS, and SEVIRI data utilized in this work are available on the Google Earth Engine.

Acknowledgments

This work was developed within the framework of the Laboratory of Technologies for Volcanology (TechnoLab) at the INGV in Catania (Italy). We are grateful to the European Space Agency (ESA), the Italian Space Agency (ASI), and the National Aeronautics and Space Administration (NASA) for satellite data. Note: Google CoLaboratory™ is a trademark of Google LLC—©2018 Google LLC All rights reserved.

Conflicts of Interest

The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Abbreviations

FRPFire Radiative Power
MIRMedium Infrared
MODISModerate Resolution Imaging Spectroradiometer
NTINormal Thermal Index
NVANon-Volcanic Area
RSDFRemote Sensing Data Fusion
SEVIRISpinning Enhanced Visible And Infrared Imager
SLSTRSea And Land Surface Temperature Radiometer
SSDSpatial Standard Deviation
TIRThermal Infrared
VAVolcanic Area
VREVolcanic Radiative Energy
VRPVolcanic Radiative Power

References

  1. Corradino, C.; Malaguti, A.B.; Ramsey, M.S.; Del Negro, C. Quantitative Assessment of Volcanic Thermal Activity from Space Using an Isolation Forest Machine Learning Algorithm. Remote Sens. 2024, 16, 2001. [Google Scholar] [CrossRef]
  2. Torrisi, F.; Corradino, C.; Cariello, S.; Del Negro, C. Enhancing detection of volcanic ash clouds from space with convolutional neural networks. J. Volcanol. Geotherm. Res. 2024, 448, 108046. [Google Scholar] [CrossRef]
  3. Coppola, D.; Cardone, D.; Laiolo, M.; Aveni, S.; Campus, A.; Massimetti, F. Global radiant flux from active volcanoes: The 2000–2019 MIROVA database. Front. Earth Sci. 2023, 11, 1240107. [Google Scholar] [CrossRef]
  4. Williams, D.B.; Ramsey, M.S. Infraredspectroscopy of volcanoes: From laboratoryto orbital scale. Front. Earth Sci. 2024, 12, 1308103. [Google Scholar] [CrossRef]
  5. Mia, M.B.; Fujimitsu, Y.; Nishijima, J. Thermal activity monitoring of an active volcano using Landsat 8/OLI-TIRS sensor images: A case study at the Aso volcanic area in southwest Japan. Geosciences 2017, 7, 118. [Google Scholar] [CrossRef]
  6. Corradino, C.; Amato, E.; Torrisi, F.; Del Negro, C. Data-Driven Random Forest Models for Detecting Volcanic Hot Spots in Sentinel-2 MSI Images. Remote Sens. 2022, 14, 4370. [Google Scholar] [CrossRef]
  7. Marchese, F.; Genzano, N.; Nolde, M.; Falconieri, A.; Pergola, N.; Plank, S. Mapping and characterizing the Kīlauea (Hawaiʻi) lava lake through Sentinel-2 MSI and Landsat-8 OLI observations of December 2020–February 2021. Environ. Model. Softw. 2022, 148, 105273. [Google Scholar] [CrossRef]
  8. Del Negro, C.; Cappello, A.; Neri, M.; Bilotta, G.; Herault, A.; Ganci, G. Lava flow hazards at Mount Etna: Constraints imposed by eruptive history and numerical simulations. Sci. Rep. 2013, 3, 3493. [Google Scholar] [CrossRef]
  9. Torrisi, F.; Amato, E.; Corradino, C.; Mangiagli, S.; Del Negro, C. Characterization of Volcanic Cloud Components Using Machine Learning Techniques and SEVIRI Infrared Images. Sensors 2022, 22, 7712. [Google Scholar] [CrossRef]
  10. Ramsey, M.S.; Harris, A.J.L. Volcanology 2020: How will thermal remote sensing of volcanic surface activity evolve over the next decade? J. Volcanol. Geotherm. Res. 2013, 249, 217–233. [Google Scholar] [CrossRef]
  11. Steffke, A.M.; Harris, A.J.L. A review of algorithms for detecting volcanic hot spots in satellite infrared data. Bull. Volcanol. 2011, 73, 1109–1137. [Google Scholar] [CrossRef]
  12. Tramutoli, V.; Cuomo, V.; Filizzola, C.; Pergola, N.; Pietrapertosa, C. Assessing the potential of thermal infrared satellite surveys for monitoring seismically active areas: The case of Kocaeli (İzmit) earthquake, August 17, 1999. Remote Sens. Environ. 2005, 96, 409–426. [Google Scholar] [CrossRef]
  13. Wright, R.; Flynn, L.; Garbeil, H.; Harris, A.; Pilger, E. MODVOLC: Near-real-time thermal monitoring of global volcanism. J. Volcanol. Geotherm Res. 2004, 135, 29–49. [Google Scholar] [CrossRef]
  14. Higgins, J.; Harris, A. VAST: A program to locate and analyse volcanic thermal anomalies automatically from remotely sensed data. Comput. Geosci. 1997, 23, 627–645. [Google Scholar] [CrossRef]
  15. Tramutoli, V. Robust AVHRR techniques (RAT) for environmental monitoring: Theory and applications. Proc. SPIE Earth Surf. Remote Sens. II 1998, 3496, 101–113. [Google Scholar] [CrossRef]
  16. Wright, R.; Flynn, L.; Garbeil, H.; Harris, A.; Pilge, E. Automated volcanic eruption detection using MODIS. Remote Sens. Environ. 2002, 82, 135–155. [Google Scholar] [CrossRef]
  17. Dehn, J.; Dean, K.; Engle, K. Thermal Monitoring of North Pacific Volcanoes from Space. Geology 2000, 28, 755–758. [Google Scholar] [CrossRef]
  18. Genzano, N.; Pergola, N.; Marchese, F. A Google Earth Engine Tool to Investigate, Map and Monitor Volcanic Thermal Anomalies at Global Scale by Means of Mid-High Spatial Resolution Satellite Data. Remote Sens. 2020, 12, 3232. [Google Scholar] [CrossRef]
  19. Falconieri, A.; Genzano, N.; Mazzeo, G.; Pergola, N.; Marchese, F. First Implementation of a Normalized Hotspot Index on Himawari-8 and GOES-R Data for the Active Volcanoes Monitoring: Results and Future Developments. Remote Sens. 2022, 14, 5481. [Google Scholar] [CrossRef]
  20. Ganci, G.; Vicari, A.; Del Negro, C. The HOTSAT Volcano Monitoring System Based on a Combined Use of SEVIRI and MODIS Multispectral Data. Ann. Geophys. 2011, 54, 5338. [Google Scholar] [CrossRef]
  21. Ganci, G.; Cappello, A.; Bilotta, G.; Del Negro, C. How the variety of satellite remote sensing data over volcanoes can assist hazard monitoring efforts: The 2011 eruption of Nabro volcano. Remote Sens. Environ. 2020, 236, 111426. [Google Scholar] [CrossRef]
  22. Corradino, C.; Ganci, G.; Bilotta, G.; Cappello, A.; Del Negro, C.; Fortuna, L. Smart Decision Support Systems for Volcanic Applications. Energies 2019, 12, 1216. [Google Scholar] [CrossRef]
  23. Coppola, D.; Laiolo, M.; Cigolini, C.; Massimetti, F.; Delle Donne, D.; Ripepe, M.; Arias, H.; Barsotti, S.; Parra, C.B.; Centeno, R.G.; et al. Thermal remote sensing for global volcano monitoring experiences from the Mirova system. Front. Earth Sci. 2020, 7, 362. [Google Scholar] [CrossRef]
  24. Gouhier, M.; Guéhenneux, Y.; Labazuy, P.; Cacault, P.; Decriem, J.; Rivet, S. HOTVOLC: A web-based monitoring system for volcanic hot spots. Geol. Soc. Lond. Spec. Publ. 2016, 426, 223–241. [Google Scholar] [CrossRef]
  25. Torrisi, F.; Amato, E.; Corradino, C.; Del Negro, C. The FastVRP automatic platform for the thermal monitoring of volcanicactivity using VIIRS and SLSTR sensors: FastFRP to monitor volcanic radiative power. Ann. Geophys. 2023, 65, 1. [Google Scholar]
  26. Wooster, M. Fire radiative energy for quantitative study of biomass burning derivation from the bird experimental satellite and comparison to MODIS fire products. Remote Sens Environ. 2003, 86, 83–107. [Google Scholar] [CrossRef]
  27. Xu, W.; Wooster, M.J. Sentinel-3 SLSTR active fire (AF) detection and FRP daytime product-Algorithm description and globalintercomparison to MODIS, VIIRS and landsat AF data. Sci. Remote Sens. 2023, 7, 100087. [Google Scholar] [CrossRef]
  28. Coppola, D.; Laiolo, M.; Cigolini, C.; Delle Donne, D.; Ripepe, M. Enhanced volcanic hot-spot detection using MODIS IR data: Results from the MIROVA system. Geol. Soc. Spec. Publ. 2016, 426, 181–205. [Google Scholar] [CrossRef]
  29. Campus, A.; Aveni, S.; Laiolo, M.; Coppola, D. Thermal unrest at La Fossa (Vulcano Island, Italy): The 2021–2023 VIIRS 375 m MIROVA-processed dataset. Bull. Volcanol. 2024, 86, 25. [Google Scholar] [CrossRef]
  30. Cariello, S.; Corradino, C.; Torrisi, F.; Del Negro, C. Cascading Machine Learning to Monitor Volcanic Thermal Activity Using Orbital Infrared Data: From Detection to Quantitative Evaluation. Remote Sens. 2024, 16, 171. [Google Scholar] [CrossRef]
  31. Ramsey, M.; Corradino, C.; Thompson, J.; Leggett, T.N. Statistical retrieval of volcanic activity in long time series orbital data: Implications for forecasting future activity. Remote Sens. Environ. 2023, 295, 113704. [Google Scholar] [CrossRef]
  32. Corradino, C.; Ramsey, M.S.; Pailot-Bonnetat, S.; Harris, A.J.L.; Del Negro, C. Detection of Subtle Thermal Anomalies: Deep Learning Applied to the ASTER Global Volcano Dataset. IEEE Trans. Geosci. Remote Sens. 2023, 61, 5000715. [Google Scholar] [CrossRef]
  33. De Beni, E.; Proietti, C.; Scollo, S.; Cantarero, M.; Mereu, L.; Romeo, F.; Pioli, L.; Sciotto, M.; Alparone, S. A Hidden Eruption: The 21 May 2023 Paroxysm of the Etna Volcano (Italy). Remote Sens. 2024, 16, 1555. [Google Scholar] [CrossRef]
  34. Calvari, S.; Nunnari, G. Statistical Insights on the Eruptive Activity at Stromboli Volcano (Italy) Recorded from 1879 to 2023. Remote Sens. 2023, 15, 4822. [Google Scholar] [CrossRef]
  35. Wooster, M.J.; Xu, W.; Nightingale, T. Sentinel-3 SLSTR Active Fire Detection and FRP Product: Pre-Launch Algorithm Development and Performance Evaluation Using MODIS and ASTER Datasets. Remote Sens. Environ. 2012, 120, 236–254. [Google Scholar] [CrossRef]
  36. Justice, C.O.; Giglio, L.; Korontzi, S.; Owens, J.; Morisette, J.T.; Roy, D.; Descloitres, J.; Alleaume, S.; Petitcolin, F.; Kaufman, Y. The MODIS Fire Products. Remote Sens. Environ. 2002, 83, 244–262. [Google Scholar] [CrossRef]
  37. Del Negro, C.; Cappello, A.; Ganci, G. Quantifying lava flow hazards in response to effusive eruption. Geol. Soc. Am. Bull. 2016, 128, 752–763. [Google Scholar] [CrossRef]
  38. Del Negro, C.; Cappello, A.; Bilotta, G.; Ganci, G.; Hérault, A.; Zago, V. Living at the Edge of an Active Volcano: Risk from Lava Flows on Mt Etna. GSA Bull. 2020, 132, 1615–1625. [Google Scholar] [CrossRef]
  39. Calvari, S.; Nunnari, G. Comparison between automated and manual detection of lava fountains from fixed monitoring thermalcameras at Etna volcano, Italy. Remote Sens. 2022, 14, 2392. [Google Scholar] [CrossRef]
  40. Marchese, F.; Filizzola, C.; Lacava, T.; Falconieri, A.; Faruolo, M.; Genzano, N.; Mazzeo, G.; Pietrapertosa, C.; Pergola, N.; Tramutoli, V.; et al. Mt. Etna paroxysms of February–April 2021 monitored and quantified through a multi-platform satellite observing system. Remote Sens. 2021, 13, 3074. [Google Scholar] [CrossRef]
  41. Guerrieri, L.; Corradini, S.; Theys, N.; Stelitano, D.; Merucci, L. Volcanic Clouds Characterization of the 2020–2022 Sequence of Mt. Etna Lava Fountains Using MSG-SEVIRI and Products’ Cross-Comparison. Remote Sens. 2023, 15, 2055. [Google Scholar] [CrossRef]
  42. Métrich, N.; Bertagnini, A.; Pistolesi, M. Paroxysms at Stromboli Volcano (Italy): Source, Genesis and Dynamics. Front. Earth Sci. 2021, 9, 593339. [Google Scholar] [CrossRef]
  43. Corradino, C.; Amato, E.; Torrisi, F.; Calvari, S.; Del Negro, C. Classifying Major Explosions and Paroxysms at Stromboli Volcano (Italy) from Space. Remote Sens. 2021, 13, 4080. [Google Scholar] [CrossRef]
  44. Plank, S.; Marchese, F.; Filizzola, C.; Pergola, N.; Neri, M.; Nolde, M.; Martinis, S. The July/August 2019 lava flows at the Sciara Del Fuoco, Stromboli—Analysis from multi-sensor infrared satellite imagery. Remote Sens. 2019, 11, 2879. [Google Scholar] [CrossRef]
  45. Giudicepietro, F.; Calvari, S.; De Cesare, W.; Di Lieto, B.; Di Traglia, F.; Esposito, A.M.; Orazi, M.; Romano, P.; Tramelli, A.; Nolesini, T.; et al. Seismic and thermal precursors of crater collapses and overflows at Stromboli volcano. Sci. Rep. 2023, 13, 11115. [Google Scholar] [CrossRef]
  46. Calvari, S.; Di Traglia, F.; Ganci, G.; Bruno, V.; Ciancitto, F.; Di Lieto, B.; Gambino, S.; Garcia, A.; Giudicepietro, F.; Inguaggiato, S.; et al. Multi-parametric study of an eruptive phase comprising unrest, major explosions, crater failure, pyroclastic density currents and lava flows: Stromboli volcano, 1 December 2020–30 June 2021. Front. Earth Sci. 2022, 10, 899635. [Google Scholar] [CrossRef]
  47. EUMETSAT. Available online: https://data.eumetsat.int/ (accessed on 30 November 2023).
  48. LAADS-DAAC. Available online: https://ladsweb.modaps.eosdis.nasa.gov/ (accessed on 18 January 2024).
  49. COPERNICUS. Available online: https://sentinels.copernicus.eu/documents/247904/0/SLSTR_Level-2_Fire_ATBD.pdf/ (accessed on 1 November 2023).
  50. Giglio, L.; Descloitres, J.; Justice, C.O.; Kaufman, Y. An enhanced contextual fire detection algorithm for MODIS. Remote Sens. Environ. 2003, 87, 273–282. [Google Scholar] [CrossRef]
  51. Giglio, L.; Schroeder, W.; Justice, C.O. The Collection 6 MODIS active fire detection algorithm and fire products. Remote Sens. Environ. 2016, 178, 31–41. [Google Scholar] [CrossRef]
  52. Coppola, D.; Laiolo, M.; Piscopo, D.; Cigolini, C. Rheological control on the radiant density of active lava flows and domes. J. Volcan. Geotherm. Res. 2013, 249, 39–48. [Google Scholar] [CrossRef]
  53. Ganci, G.; Vicari, A.; Cappello, A.; Del Negro, C. An emergent strategy for volcano hazard assessment: From thermal satellite monitoring to lava flow modeling. Remote Sens. Environ. 2012, 119, 197–207. [Google Scholar] [CrossRef]
  54. Plank, S.; Massimetti, F.; Soldati, A.; Hess, K.U.; Nolde, M.; Martinis, S.; Dingwell, D.B. Estimates of lava discharge rate of 2018 Kīlauea Volcano, Hawaiʻi eruption using multi-sensor satellite and laboratory measurements. Int. J. Remote Sens. 2021, 42, 1492–1511. [Google Scholar] [CrossRef]
  55. Harris, A.J.L.; Steffke, A.; Calvari, S.; Spampinato, L. Thirty years of satellite-derived lava discharge rates at Etna: Implications for steady volumetric output. J. Geophys. Res.–Solid Earth 2011, 116, B8. [Google Scholar] [CrossRef]
  56. Campus, A.; Laiolo, M.; Massimetti, F.; Coppola, D. The Transition from MODIS to VIIRS for Global Volcano Thermal Monitoring. Sensors 2022, 22, 1713. [Google Scholar] [CrossRef] [PubMed]
Figure 1. (1) Derivation of the Normalized Thermal Index (NTI) obtained by combining the radiance of the MIR and the radiance of the TIR. (2) Application of the Spatial Standard Deviation (SSD) filter to each pixel in the image. (3) Definition of two statistical masks, Mask1 and Mask2, to identify “potential” and “true” hotspots, applied on the SSD and NTI of the volcanic area (VA). (4) Application of Gabor filter to extract the significant features of the image, resulting in a matrix called Gabor Weighted NTI (G-NTI). (5) Highlighting hotspots in the crater area and defining the Spatial Gabor Weighted NTI (SG-NTI). (6) Application of a statistical mask to the previously extracted matrix. (7) Calculation of the final VRP.
Figure 1. (1) Derivation of the Normalized Thermal Index (NTI) obtained by combining the radiance of the MIR and the radiance of the TIR. (2) Application of the Spatial Standard Deviation (SSD) filter to each pixel in the image. (3) Definition of two statistical masks, Mask1 and Mask2, to identify “potential” and “true” hotspots, applied on the SSD and NTI of the volcanic area (VA). (4) Application of Gabor filter to extract the significant features of the image, resulting in a matrix called Gabor Weighted NTI (G-NTI). (5) Highlighting hotspots in the crater area and defining the Spatial Gabor Weighted NTI (SG-NTI). (6) Application of a statistical mask to the previously extracted matrix. (7) Calculation of the final VRP.
Remotesensing 16 02879 g001
Figure 2. Workflow image of the RSDF algorithm. Study cases: (a) Etna on 2 December 2023 at 01:10 UTC, MODIS sensor; (b) Etna on 15 January 2023 at 20:46 UTC, SLSTR sensor; (c) Stromboli on 3 October 2023 at 13:10 UTC, MODIS sensor; (d) Stromboli on 23 October 2023 at 09:08 UTC, SLSTR sensor.
Figure 2. Workflow image of the RSDF algorithm. Study cases: (a) Etna on 2 December 2023 at 01:10 UTC, MODIS sensor; (b) Etna on 15 January 2023 at 20:46 UTC, SLSTR sensor; (c) Stromboli on 3 October 2023 at 13:10 UTC, MODIS sensor; (d) Stromboli on 23 October 2023 at 09:08 UTC, SLSTR sensor.
Remotesensing 16 02879 g002
Figure 3. Time series of the Etna volcano. The panels show VRP calculated respectively from the RSDF Algorithm SLSTR (blue triangles) SLSTR Level 2 (red triangles), the RSDF Algorithm MODIS (blue triangles), MODIS Level 2 (red triangles). (a,c) shows data from January 2021 to April 2022, (b,d) from April 2022 to June 2023.
Figure 3. Time series of the Etna volcano. The panels show VRP calculated respectively from the RSDF Algorithm SLSTR (blue triangles) SLSTR Level 2 (red triangles), the RSDF Algorithm MODIS (blue triangles), MODIS Level 2 (red triangles). (a,c) shows data from January 2021 to April 2022, (b,d) from April 2022 to June 2023.
Remotesensing 16 02879 g003
Figure 4. Histograms (a,c) and probability plots (b,d) for Etna datasets. (a,c) Histograms display data distribution related to VRP (and FRP) in logarithmic scale; (a) blue bars represent the distribution of SLSTR-–RSDF algorithm processed data, and red bars represent SLSTR Level 2 product data; (c) blue bars represent the distribution of MODIS–RSDF algorithm processed data, and red bars represent MODIS active fire products. (b,d) Probability plots for normal distribution of RSDF algorithm processed data (blue), and Level 2 product data (red). The dashed grey lines represent the reference lines of the theoretical distributions, and the black dashed line in (b) corresponds to the slope change associated with the transition between regimes of background and high thermal activity.
Figure 4. Histograms (a,c) and probability plots (b,d) for Etna datasets. (a,c) Histograms display data distribution related to VRP (and FRP) in logarithmic scale; (a) blue bars represent the distribution of SLSTR-–RSDF algorithm processed data, and red bars represent SLSTR Level 2 product data; (c) blue bars represent the distribution of MODIS–RSDF algorithm processed data, and red bars represent MODIS active fire products. (b,d) Probability plots for normal distribution of RSDF algorithm processed data (blue), and Level 2 product data (red). The dashed grey lines represent the reference lines of the theoretical distributions, and the black dashed line in (b) corresponds to the slope change associated with the transition between regimes of background and high thermal activity.
Remotesensing 16 02879 g004
Figure 5. Stacked time series of VRPw (weekly mean) retrieved for the SLSTR–RSDF algorithm processed data (blue), SLSTR Level 2 product data (red), MODIS–RSDF algorithm processed data (green), and SLSTR Level 2 product data (black) at the Etna volcano, displayed on a logarithmic scale.
Figure 5. Stacked time series of VRPw (weekly mean) retrieved for the SLSTR–RSDF algorithm processed data (blue), SLSTR Level 2 product data (red), MODIS–RSDF algorithm processed data (green), and SLSTR Level 2 product data (black) at the Etna volcano, displayed on a logarithmic scale.
Remotesensing 16 02879 g005
Figure 6. VRP time series of the Stromboli volcano. The panels show VRP calculated respectively from the RSDF Algorithm SLSTR (blue triangles) SLSTR Level 2 (red triangles), the RSDF Algorithm MODIS (blue triangles), MODIS Level 2 (red triangles). (a,c) shows data from January 2021 to April 2022, (b,d) from April 2022 to June 2023.
Figure 6. VRP time series of the Stromboli volcano. The panels show VRP calculated respectively from the RSDF Algorithm SLSTR (blue triangles) SLSTR Level 2 (red triangles), the RSDF Algorithm MODIS (blue triangles), MODIS Level 2 (red triangles). (a,c) shows data from January 2021 to April 2022, (b,d) from April 2022 to June 2023.
Remotesensing 16 02879 g006
Figure 7. Histograms (a,c) and probability plots (b,d) for Stromboli datasets. (a,c) Histograms display data distribution related to VRP (and FRP) in logarithmic scale; (a) blue bars represent the distribution of SLSTR–RSDF algorithm processed data, and red bars represent SLSTR Level 2 product data; (c) blue bars represent the distribution of MODIS–RSDF algorithm processed data, and red bars represent MODIS active fire products. (b,d) Probability plots for normal distribution of RSDF algorithm processed data (blue), and Level 2 product data and MODIS active fire products (red). The dashed grey lines represent the reference lines of the theoretical distributions, and the black dashed line in (b) corresponds to the slope change associated with the transition between regimes of background and high thermal activity for Stromboli.
Figure 7. Histograms (a,c) and probability plots (b,d) for Stromboli datasets. (a,c) Histograms display data distribution related to VRP (and FRP) in logarithmic scale; (a) blue bars represent the distribution of SLSTR–RSDF algorithm processed data, and red bars represent SLSTR Level 2 product data; (c) blue bars represent the distribution of MODIS–RSDF algorithm processed data, and red bars represent MODIS active fire products. (b,d) Probability plots for normal distribution of RSDF algorithm processed data (blue), and Level 2 product data and MODIS active fire products (red). The dashed grey lines represent the reference lines of the theoretical distributions, and the black dashed line in (b) corresponds to the slope change associated with the transition between regimes of background and high thermal activity for Stromboli.
Remotesensing 16 02879 g007
Figure 8. Stacked time series of VRPw (weekly mean) retrieved for SLSTR–RSDF algorithm processed data (blue), SLSTR Level 2 product data (red), MODIS–RSDF algorithm processed data (green), and MODIS active fire products (black) at the Etna volcano, displayed on a logarithmic scale.
Figure 8. Stacked time series of VRPw (weekly mean) retrieved for SLSTR–RSDF algorithm processed data (blue), SLSTR Level 2 product data (red), MODIS–RSDF algorithm processed data (green), and MODIS active fire products (black) at the Etna volcano, displayed on a logarithmic scale.
Remotesensing 16 02879 g008
Figure 9. Cumulative Volcanic Radiative Energy (VRE) calculated from VRP (and FRP) using the trapezoidal rule for integration. The blue line represents VRESLSTR, the red dashed line FREMODIS, the green dashed line VREMODIS, and the black dashed line FREMODIS. Panels (a,c) show data for Etna; panels (b,d) show data for Stromboli.
Figure 9. Cumulative Volcanic Radiative Energy (VRE) calculated from VRP (and FRP) using the trapezoidal rule for integration. The blue line represents VRESLSTR, the red dashed line FREMODIS, the green dashed line VREMODIS, and the black dashed line FREMODIS. Panels (a,c) show data for Etna; panels (b,d) show data for Stromboli.
Remotesensing 16 02879 g009
Figure 10. Radiative power time series from SLSTR– and MODIS–RSDF algorithm data with intensity limits categorized as low, moderate, high, and extreme. (a) Etna, (b) Stromboli.
Figure 10. Radiative power time series from SLSTR– and MODIS–RSDF algorithm data with intensity limits categorized as low, moderate, high, and extreme. (a) Etna, (b) Stromboli.
Remotesensing 16 02879 g010
Figure 11. Temporal trend of VRP values derived from the RSDF algorithm for SEVIRI, SLSTR, MODIS, and VIIRS over two periods at Mt. Etna: (a) 1 February 2021–30 April 2021, and (b) 27 September 2023–10 October 2023.
Figure 11. Temporal trend of VRP values derived from the RSDF algorithm for SEVIRI, SLSTR, MODIS, and VIIRS over two periods at Mt. Etna: (a) 1 February 2021–30 April 2021, and (b) 27 September 2023–10 October 2023.
Remotesensing 16 02879 g011
Figure 12. TADR and lava flow volume flux during the effusive event at Etna from 14 May 2022 to 16 June 2022. TADR_max, TADR_mean, and TADR_min are represented by blue, red, and green points, respectively. The total volume_max, volume_mean, and volume_min are represented by blue, red, and green lines, respectively.
Figure 12. TADR and lava flow volume flux during the effusive event at Etna from 14 May 2022 to 16 June 2022. TADR_max, TADR_mean, and TADR_min are represented by blue, red, and green points, respectively. The total volume_max, volume_mean, and volume_min are represented by blue, red, and green lines, respectively.
Remotesensing 16 02879 g012
Figure 13. TADR and lava flow volume flux during the effusive event at Stromboli from 27 September 2023 to 10 October 2023. TADR_max, TADR_mean, and TADR_min are represented by blue, red, and green points, respectively. The total volume_max, volume_mean, and volume_min are represented by blue, red, and green lines, respectively.
Figure 13. TADR and lava flow volume flux during the effusive event at Stromboli from 27 September 2023 to 10 October 2023. TADR_max, TADR_mean, and TADR_min are represented by blue, red, and green points, respectively. The total volume_max, volume_mean, and volume_min are represented by blue, red, and green lines, respectively.
Remotesensing 16 02879 g013
Table 1. Characteristics of the SLSTR, MODIS, VIIRS, and SEVIRI sensors.
Table 1. Characteristics of the SLSTR, MODIS, VIIRS, and SEVIRI sensors.
SLSTR
(Sentinel-3)
MODIS
(AQUA/TERRA)
VIIRS
(S-NPP/N20)
SEVIRI
(MSG)
Orbit altitude (km)81470582438.500
Type of SatellitesPolarPolarPolarGeostationary
Equator crossing time10:00 LT10:30 LT/13:30 LT13:30 LT/12:40 LT
Spatial Resolution (km)110.75–0.3753
Temporal Resolution1–2 days1–2 days12 h10–15 min
Spectral coverage of thermal bands (μm)3.700–12.0003660–14.3853.550–12.4883.480–13.400
ID MIR BandS7/F122/21I04/M-13IR-039
Spectral Range (μm)3.700/3.7003.940–4.001/3.929–3.9893.973–4.1283.480–4.360
Tmax312 K/500 K331 K/500 K367 K/634 K335 K
ID TIR BandS8/F131I05/M-15I-108
Spectral Range (μm)10.850/12.00010.780–11.28010.263–11.26311.000–13.400
Tmax323 K/400 K400 K380 K/343 K
Table 2. Values of the parameter crad for Etna and Stromboli.
Table 2. Values of the parameter crad for Etna and Stromboli.
EtnaStromboli
c r a d 2.16 × 10 8 1.65 × 10 8
c r a d m a x 1.08 × 10 8 8.25 × 10 8
c r a d m i n 3.24 × 10 8 2.47 × 10 8
Table 3. Table displaying cumulative total VRE for each Etna phase and Stromboli phase.
Table 3. Table displaying cumulative total VRE for each Etna phase and Stromboli phase.
RSDF Algorithm
SLSTR
SLSTR Level 2
Product
RSDF Algorithm
MODIS
MODIS Level 2
Product
Etna
15 February 2021–3 April 2021 1.03   ×   10 10 2.05   ×   10 10 1.21   ×   10 10 8.45   ×   10 6
17 May 2021–28 September 2021 1.21   ×   10 10 1.54   ×   10 10 1.58   ×   10 10 8.29   ×   10 7
12 May 2022–15 June 2022 1.56   ×   10 10 2.32   ×   10 10 1.08   ×   10 10 1.28   ×   10 7
27 November 2022–6 February 2023 1.58   ×   10 10 1.48   ×   10 10 9.46   ×   10 9 0
2021–2023 5.94   ×   10 10 8.42   ×   10 10 5.81   ×   10 10 6.86   ×   10 7
Stromboli
13 May 2021–27 May 2021 1.73   ×   10 9 2.11   ×   10 9 9.02   ×   10 8 0
11 May 2022–8 June 2022 2.93   ×   10 8 0 2.03   ×   10 8 0
27 November 2022–11 December 2022 1.01   ×   10 9 0 4.94   ×   10 8 0
2021–2023 5.12   ×   10 9 2.25   ×   10 9 3.01   ×   10 9 2.59   ×   10 8
Table 4. Comparison between datasets of SLSTR and MODIS from 2021 to 2023 for selected volcanoes.
Table 4. Comparison between datasets of SLSTR and MODIS from 2021 to 2023 for selected volcanoes.
False
Rate
Omitted RateNpassNalert (f%)Mean VRP (MW)f%
Daytime
f%
Nightime
Etna
SLSTR RSDF3.6%2.9%2545764 (30%) 6.83   ×   10 7 13%17%
SLSTR Level 212%62%2545445 (18%) 5.16   ×   10 7 5%12%
MODIS RSDF4.8%4.5%5171858 (17%) 5.44   ×   10 7 10%7%
MODIS Level 210%89%517117 (0.002%) 3.06   ×   10 7 0.001%0.001%
Stromboli
SLSTR RSDF5.1%5.4%2545605 (24%) 2.16   ×   10 7 17%7%
SLSTR Level 22%80%254525 (0.01%) 2.36   ×   10 7 0.007%0.001%
MODIS RSDF4.7%5.7%5171570 (11%) 6.55   ×   10 6 7%4%
MODIS Level 21%99%517118 (0.002%) 6.21   ×   10 6 0.002%0.0%
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Di Bella, G.S.; Corradino, C.; Cariello, S.; Torrisi, F.; Del Negro, C. Advancing Volcanic Activity Monitoring: A Near-Real-Time Approach with Remote Sensing Data Fusion for Radiative Power Estimation. Remote Sens. 2024, 16, 2879. https://doi.org/10.3390/rs16162879

AMA Style

Di Bella GS, Corradino C, Cariello S, Torrisi F, Del Negro C. Advancing Volcanic Activity Monitoring: A Near-Real-Time Approach with Remote Sensing Data Fusion for Radiative Power Estimation. Remote Sensing. 2024; 16(16):2879. https://doi.org/10.3390/rs16162879

Chicago/Turabian Style

Di Bella, Giovanni Salvatore, Claudia Corradino, Simona Cariello, Federica Torrisi, and Ciro Del Negro. 2024. "Advancing Volcanic Activity Monitoring: A Near-Real-Time Approach with Remote Sensing Data Fusion for Radiative Power Estimation" Remote Sensing 16, no. 16: 2879. https://doi.org/10.3390/rs16162879

APA Style

Di Bella, G. S., Corradino, C., Cariello, S., Torrisi, F., & Del Negro, C. (2024). Advancing Volcanic Activity Monitoring: A Near-Real-Time Approach with Remote Sensing Data Fusion for Radiative Power Estimation. Remote Sensing, 16(16), 2879. https://doi.org/10.3390/rs16162879

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop