Using IAP (plasma analyzer) and ISL (Langmuir probe) experiments onboard DEMETER (Detection of Electromagnetic Emissions Transmitted from Earthquake Regions) satellite and GPS (Global Positioning System) measurements, we have statistically analyzed the variations of the electron and ion densities to search for disturbances in the vicinity of four large earthquakes prior to events. The indices D_st and K_p were used to distinguish pre-earthquake anomalies from the other anomalies related to the geomagnetic activities. For each studied case, a very good agreement was found between the different parameters estimated by DEMETER and GPS data in the detection of pre-seismic anomalies. Our statistics results show that the anomalous deviations prior to earthquakes have different sign from case to case, and that their amplitude depends on the magnitude of the earthquake. It has also been found that the electron density measured by the ISL experiment at night detects anomalous variations significantly before the earthquakes. The appearance of positive and negative anomalies in both of DEMETER and TEC (Total Electron Content) data during 1 to 5 days before all studied earthquakes during quiet geomagnetic conditions indicates that these anomalous behaviors are highly regarded as seismo-ionospheric precursors.

Wildfires are major natural disasters negatively affecting human safety, natural ecosystems, and wildlife. Timely and accurate estimation of wildfire burn areas is particularly important for post-fire management and decision making. In this regard, Remote Sensing (RS) images are great resources due to their wide coverage, high spatial and temporal resolution, and low cost. In this study, Australian areas affected by wildfire were estimated using Sentinel-2 imagery and Moderate Resolution Imaging Spectroradiometer (MODIS) products within the Google Earth Engine (GEE) cloud computing platform. To this end, a framework based on change analysis was implemented in two main phases: (1) producing the binary map of burned areas (i.e., burned vs. unburned); (2) estimating burned areas of different Land Use/Land Cover (LULC) types. The first phase was implemented in five main steps: (i) preprocessing, (ii) spectral and spatial feature extraction for pre-fire and post-fire analyses; (iii) prediction of burned areas based on a change detection by differencing the pre-fire and post-fire datasets; (iv) feature selection; and (v) binary mapping of burned areas based on the selected features by the classifiers. The second phase was defining the types of LULC classes over the burned areas using the global MODIS land cover product (MCD12Q1). Based on the test datasets, the proposed framework showed high potential in detecting burned areas with an overall accuracy (OA) and kappa coefficient (KC) of 91.02% and 0.82, respectively. It was also observed that the greatest burned area among different LULC classes was related to evergreen needle leaf forests with burning rate of over 25 (%). Finally, the results of this study were in good agreement with the Landsat burned products.

Thermal anomaly is known as a significant precursor of strong earthquakes, therefore Land Surface Temperature (LST) time series have been analyzed in this study to locate relevant anomalous variations prior to the Bam (26 December 2003), Zarand (22 February 2005) and Borujerd (31 March 2006) earthquakes. The duration of the three datasets which are comprised of MODIS LST images is 44, 28 and 46 days for the Bam, Zarand and Borujerd earthquakes, respectively. In order to exclude variations of LST from temperature seasonal effects, Air Temperature (AT) data derived from the meteorological stations close to the earthquakes epicenters have been taken into account. The detection of thermal anomalies has been assessed using interquartile, wavelet transform and Kalman filter methods, each presenting its own independent property in anomaly detection. The interquartile method has been used to construct the higher and lower bounds in LST data to detect disturbed states outside the bounds which might be associated with impending earthquakes. The wavelet transform method has been used to locate local maxima within each time series of LST data for identifying earthquake anomalies by a predefined threshold. Also, the prediction property of the Kalman filter has been used in the detection process of prominent LST anomalies. The results concerning the methodology indicate that the interquartile method is capable of detecting the highest intensity anomaly values, the wavelet transform is sensitive to sudden changes, and the Kalman filter method significantly detects the highest unpredictable variations of LST. The three methods detected anomalous occurrences during 1 to 20 days prior to the earthquakes showing close agreement in results found between the different applied methods on LST data in the detection of pre-seismic anomalies. The proposed method for anomaly detection was also applied on regions irrelevant to earthquakes for which no anomaly was detected, indicating that the anomalous behaviors can be related to impending earthquakes. The proposed method receives its credibility from the overall capabilities of the three integrated methods.

InSAR processing is vastly used for land deformation monitoring from the space. Machine learning methods are known as strong tools for data modeling as well as predicting. In this study, we are going to model and predict the future behavior of land subsidence by InSAR processing and leveraging deep learning methods over the lands in the vicinity of Lake Urmia (located in the northwest of Iran). Accordingly, Sentinel-1 data over 56 months from November 2014 to June 2019 and small baseline subsets (SBAS) InSAR methods were utilized. Several regions with a high rate of subsidence were identified (maximum monthly subsidence of 13.3 mm). Furthermore, environmental factors affecting subsidence were considered. Therefore, parameters such as rainfall, groundwater, and lake area variations were measured using TRMM, GRACE, and MODIS satellite data, respectively. In order to determine and assess the relation between land deformations and environmental variations, several machine learning methods were implemented. The environmental parameters were used as the input of models and ground deformations as the target to be predicted. Eventually, ground deformations were estimated using multi-layer perceptron (MLP), convolutional neural network (CNN), and long short-term memory (LSTM) networks, in which each network had strengths and weaknesses on different occasions. Thus, by blending the forecast of the three models, a weighted ensemble was constructed, which outperformed the single models and reached the root mean square error (RMSE), mean absolute error (MAE), and standard deviation (SD) of 8.2 mm, 6.4 mm, and ±5.2 mm, respectively. The result indicated that although each single model had proper accuracy, an ensemble model can improve land deformation anticipation using the strength of networks in various conditions.

On 11 March 2011 at 14:46:23 LT, the 4th largest earthquake ever recorded with a magnitude of 9.0 occurred near the northeast coast of Honshu in Japan (38.322^° N, 142.369^° E, Focal depth 29.0 km). In order to acknowledge the capabilities of Total Electron Content (TEC) ionospheric precursor, in this study four methods including mean, median, wavelet transform, and Kalman filter have been applied to detect anomalous TEC variations concerning the Tohoku earthquake. The duration of the TEC time series dataset is 49 days at a time resolution of 2 h. All four methods detected a considerable number of anomalous occurrences during 1 to 10 days prior to the earthquake in a period of high geomagnetic activities. In this study, geomagnetic indices (i.e. D_st, K_p, A_p and F10.7) were used to distinguish pre-earthquake anomalies from the other anomalies related to the geomagnetic and solar activities. A good agreement in results was found between the different applied anomaly detection methods on TEC data.