Application of artificial neural networks and logistic regression to the prediction of forest fire danger in Galicia using MODIS data

2012 ◽  
Vol 21 (8) ◽  
pp. 1025 ◽  
Author(s):  
Mar Bisquert ◽  
Eduardo Caselles ◽  
Juan Manuel Sánchez ◽  
Vicente Caselles
Keyword(s):  
Remote Sensing ◽  
Neural Networks ◽  
Logistic Regression ◽  
Artificial Neural Networks ◽  
Surface Temperature ◽  
Forest Fire ◽  
Land Surface Temperature ◽  
Land Surface ◽  
Fire Danger ◽  
Forest Fire Danger

Fire danger models are a very useful tool for the prevention and extinction of forest fires. Some inputs of these models, such as vegetation status and temperature, can be obtained from remote sensing images, which offer higher spatial and temporal resolution than direct ground measures. In this paper, we focus on the Galicia region (north-west of Spain), and MODIS (Moderate Resolution Imaging Spectroradiometer) images are used to monitor vegetation status and to obtain land surface temperature as essential inputs in forest fire danger models. In this work, we tested the potential of artificial neural networks and logistic regression to estimate forest fire danger from remote sensing and fire history data. Remote sensing inputs used were the land surface temperature and the Enhanced Vegetation Index. A classification into three levels of fire danger was established. Fire danger maps based on this classification will facilitate fire prevention and extinction tasks.

scholarly journals Relating Spatiotemporal Patterns of Forest Fires Burned Area and Duration to Diurnal Land Surface Temperature Anomalies

2018 ◽  
Vol 10 (11) ◽  
pp. 1777 ◽  
Author(s):  
Carmine Maffei ◽  
Silvia Alfieri ◽  
Massimo Menenti
Keyword(s):  
Remote Sensing ◽  
Time Series ◽  
Surface Temperature ◽  
Land Surface Temperature ◽  
Forest Fires ◽  
Land Surface ◽  
Fire Danger ◽  
Parametric Models ◽  
Burned Area ◽  
Informative Content

Forest fires are a major source of ecosystem disturbance. Vegetation reacts to meteorological factors contributing to fire danger by reducing stomatal conductance, thus leading to an increase of canopy temperature. The latter can be detected by remote sensing measurements in the thermal infrared as a deviation of observed land surface temperature (LST) from climatological values, that is as an LST anomaly. A relationship is thus expected between LST anomalies and forest fires burned area and duration. These two characteristics are indeed controlled by a large variety of both static and dynamic factors related to topography, land cover, climate, weather (including those affecting LST) and anthropic activity. To investigate the predicting capability of remote sensing measurements, rather than constructing a comprehensive model, it would be relevant to determine whether anomalies of LST affect the probability distributions of burned area and fire duration. This research approached the outlined knowledge gap through the analysis of a dataset of forest fires in Campania (Italy) covering years 2003–2011 against estimates of LST anomaly. An LST climatology was first computed from time series of daily Aqua-MODIS LST data (product MYD11A1, collection 6) over the longest available sequence of complete annual datasets (2003–2017), through the Harmonic Analysis of Time Series (HANTS) algorithm. HANTS was also used to create individual annual models of LST data, to minimize the effect of varying observation geometry and cloud contamination on LST estimates while retaining its seasonal variation. LST anomalies where thus quantified as the difference between LST annual models and LST climatology. Fire data were intersected with LST anomaly maps to associate each fire with the LST anomaly value observed at its position on the day previous to the event. Further to this step, the closest probability distribution function describing burned area and fire duration were identified against a selection of parametric models through the maximization of the Anderson-Darling goodness-of-fit. Parameters of the identified distributions conditional to LST anomaly where then determined along their confidence intervals. Results show that in the study area log-transformed burned area is described by a normal distribution, whereas log-transformed fire duration is closer to a generalized extreme value (GEV) distribution. The parameters of these distributions conditional to LST anomaly show clear trends with increasing LST anomaly; significance of this observation was verified through a likelihood ratio test. This confirmed that LST anomaly is a covariate of both burned area and fire duration. As a consequence, it was observed that conditional probabilities of extreme events appear to increase with increasing positive deviations of LST from its climatology values. This confirms the stated hypothesis that LST anomalies affect forest fires burned area and duration and highlights the informative content of time series of LST with respect to fire danger.

scholarly journals Introducing a New Remote Sensing-Based Model for Forecasting Forest Fire Danger Conditions at a Four-Day Scale

2019 ◽  
Vol 11 (18) ◽  
pp. 2101 ◽  
Author(s):  
M. Ahmed ◽  
Quazi Hassan ◽  
Masoud Abdollahi ◽  
Anil Gupta
Keyword(s):  
Remote Sensing ◽  
Forest Fire ◽  
Forest Fires ◽  
Land Surface ◽  
Vegetation Index ◽  
Significant Risk ◽  
Fire Danger ◽  
Area Of Interest ◽  
Forest Fire Danger ◽  
Very High

Forest fires are natural disasters that create a significant risk to the communities living in the vicinity of forested landscape. To minimize the risk of forest fires for the resilience of such urban communities and forested ecosystems, we proposed a new remote sensing-based medium-term (i.e., four-day) forest fire danger forecasting system (FFDFS) based on an existing framework, and applied the system over the forested regions in the northern Alberta, Canada. Hence, we first employed moderate resolution imaging spectroradiometer (MODIS)-derived daily land surface temperature (Ts) and surface reflectance products along with the annual land cover to generate three four-day composite for Ts, normalized difference vegetation index (NDVI), and normalized difference water index (NDWI) at 500 m spatial resolution for the next four days over the forest-dominant regions. Upon generating these four-day composites, we calculated the variable-specific mean values to determine variable-specific fire danger maps with two danger classes (i.e., high and low). Then, by assuming the cloud-contaminated pixels as the low fire danger areas, we combined these three danger maps to generate a four-day fire danger map with four danger classes (i.e., low, moderate, high, and very high) over our study area of interest, which was further enhanced by incorporation of a human-caused static fire danger map. Finally, the four-day scale fire danger maps were evaluated using observed/ground-based forest fire occurrences during the 2015–2017 fire seasons. The results revealed that our proposed system was able to detect about 75% of the fire events in the top two danger classes (i.e., high and very high). The system was also able to predict the 2016 Horse River wildfire, the worst fire event in Albertian and Canadian history, with about 67% agreement. The higher accuracy outputs from our proposed model indicated that it could be implemented in the operational management, which would be very useful for lessening the adverse impact of such fire events.

Estimating Land Surface Temperature from AMSR-E and AMSR2 Data with Convolutional Neural Network

2021 ◽  
Author(s):  
Shaofei Wang ◽  
Ji Zhou ◽  
Xiaodong Zhang ◽  
Zichun Jin
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
Neural Networks ◽  
Surface Temperature ◽  
Convolutional Neural Network ◽  
Root Mean Square ◽  
Air Temperature ◽  
Land Surface Temperature ◽  
Land Surface ◽  
Mean Square

<p>Land surface temperature (LST) is a key factor in earth–atmosphere interactions and an important indicator for monitoring environmental changes and energy balance on Earth's surface. Thermal infrared (TIR) remote sensing can only obtain valid observations under clear-sky conditions, which results in the discontinuities of the LST time series. In contrast, passive microwave (PMW) remote sensing can help estimate the LST under cloudy conditions and the LST generated by PMW observations is an important input parameter for generating medium-resolution (e.g., 1km) all-weather LST. Neural networks, especially the latest deep learning, have exhibited good ability in estimating surface parameters from satellite remote sensing. However, thorough examinations of neural networks in the estimation of LST from satellite PMW observations are still lacking. In this study, we examined the performances of the traditional neural network (NN), deep belief network (DBN), and convolutional neural network (CNN) in estimating LST from the Advanced Microwave Scanning Radiometer for EOS (AMSR-E) and Advanced Microwave Scanning Radiometer 2 (AMSR2) data over the Chinese landmass. The examination results show that CNN is better than NN and DBN by 0.1–0.4 K. Different combinations of input parameters were examined to get the best combinations for the daytime and nighttime conditions. The best combinations are the brightness temperatures (BTs), NDVI, air temperature, and day of the year (DOY) for the daytime and BTs and air temperature for the nighttime. Compared with the MODIS LST, the CNN LST estimates yielded root-mean-square differences (RMSDs) of 2.19–3.58 K for the daytime and 1.43–2.14 K for the nighttime for diverse land cover types for AMSR-E. Validation based on the in-situ LST demonstrates that the CNN LST yielded root-mean-square errors of 2.10–5.34 K and the error analysis confirms that the main reason for the errors is the scale mismatching between the ground stations and the MW pixels. This study helps better the understanding of the use of neural networks for estimating LST from satellite MW observations.</p>

Sign in / Sign up

Export Citation Format

Share Document