Wildfire spread is affected by various factors including weather, fuel, topography, and human intervention. Previous studies have focused on wildfire probability modeling for purposes of wildfire management, with less attention paid to potential wildfire behavior characteristics such as wildfire speed and intensity. Remote sensing technology has excellent advantages in deriving the characteristics and fuel variables. This study aimed to model these characteristics for wildfire hazard assessment in the Yunnan Province of China. The random forest (RF) model and the extreme gradient boosting (XGBoost) model, were selected to establish the potential wildfire behavior characteristics (PWBC) models based on explanatory variables. The results verified that elevation, fuel moisture content, and infrastructure variables played a more significant role in the models. The RF-based models performed better than the XGBoost-based ones with higher overall accuracy (≥0.83) and kappa coefficient (≥0.79), indicating the effectiveness of predicting potential wildfire behavior characteristics to assess wildfire hazards.

Accurate estimation of potential fire intensity (PFI) can improve wildfire management. The PFI can be simulated by fire spread models, but with immeasurable uncertainties. There are also some difficulties in estimating PFI with multi-source drivers, since the fire spread is limited by fire suppression. This study aimed to estimate the probability density of PFI over southwestern China, using time-series fuel and weather data as well as topographic data. The Quantile Regression and Bi-directional Long Short-Term Memory were selected to establish the prediction model of PFI. The results showed that the QR-BiLSTM performed best at the 90% confidence level. The modal PFI values extracted from the estimated probability density were closer to the observed values. This study suggests the potential of probability density estimation of PFI with artificial intelligence, for which improves wildfire risk assessment.

Dead fuel moisture content (DFMC) was usually involved and being an important part in predicting ignition potential, fireline intensity, flame length, and rate of spread. Previous studies focused on model development and paid little attention to large-scale DFMC forecasting using these models, especially process-based models. In this study, we forecast spatial 1-h (fuel with a diameter less than 0.635 cm) and 10-h (fuel with a diameter between 0.635 cm and 2.54 cm) DFMC in 16 days with meteorological variables interpolated from Global Forecast System (GFS). First, we interpolated meteorological variables including air temperature (Tair), relative humidity (RH), wind speed (Ws) and precipitation (P) to 2 km from GFS data (0.25°) for each site with DFMC measurement in Liangshan Yi Autonomous Prefecture. Then, we forecasted DFMC hourly for 96 sites with three process-based models (Simard, Nelson and fuel stick moisture model (FSMM)). Our results show that the FSMM forecasted more accurate DFMC values (1-h: R2=0.54, RMSE=6.58%, MAE=5.63%; 10-h: R2=0.73, RMSE=3.7%, MAE=2.7%) compared to the Nelson and Simard model. Our results suggest that accurate DFMC forecasts from GFS data based on our methods can be used for fire risk assessment and fire behavior prediction.

Rice false smut (RFS) is a late fungal disease mainly occurring on rice panicle in recent years. This research was based on the unmanned aerial vehicle (UAV) hyperspectral remote sensing data. On the basis of genetic algorithm combined with partial least squares to select the feature bands, the correlation coefficient method and Instability Index between Classes method were used to further select the feature bands, which further eliminated 27.78% of the feature bands when the model monitoring accuracy was improved overall. The prediction accuracy of Gradient Boosting Decision Tree model and Random Forest model was the best, which were 85.62% and 84.10% respectively, and the monitoring accuracy was improved by 2.22% and 2.4% compared with that before optimization. Then, based on the UAV hyperspectral data and the characteristic bands, the sensitive band ranges of rice false smut monitoring were determined, which were 698nm-750nm and 974nm-984nm.

Automatic extraction of crop row centerline is an important technology for agricultural automation, and it has a wide range of applications in automated operations, such as automatic agricultural navigation, automatic harvesting, automatic weeding and automatic seedling replenishment. In this study, the method of row centerline detection is proposed by combining image segmentation and the technique of feature point extraction, and it is applied to the extraction of corn missing seedling locations. Firstly, image segmentation is performed by combining the improved vegetation index ExGG and a double-threshold algorithm (the OTSU method combined with the Particle Swarm Optimization algorithm), and most of the pseudo-feature points are removed using median filtering to initially separate corn seedlings from weeds and soil. Then, the number of crop rows is obtained using the vertical projection method; the micro-region of interest(micro-ROI) is used to find the center of mass and extract the feature points. Finally, the remaining pseudo-feature points are removed by the location clustering method, and the crop row centerline is fitted using the linear regression method of least squares. This study extracts the location and number of missing seedlings of corn based on the information from the row centerline, providing technical support for the subsequent seedling replenishment operation. The experimental results show that the accuracy of the proposed method for detecting the centerline of corn seedling rows is 0.016°, which is better than the Hough transform.

Understanding the extent to which climate affects interannual wildfire variability is key to project and mitigate wildfire. However, obtaining a robust quantification of the climate-wildfire relationship at large spatial scales remains challenging. This study employed hierarchical Bayesian framework to estimate the effect of drought on forest wildfire frequency in subtropical China. We quantified the drought-wildfire relationship across the subtropical China that the probability of excess wildfire (i.e., above normal wildfire activity level) shown disproportionate growth from 2.4% to 76.7% when vapor pressure deficit (VPD) increased from -3 to 3 (z-score). The extreme wildfires only occurred when VPD exceeds 0.5 (z-score). Our results suggest that Bayesian hierarchical model performs better in quantifying the impact of drought on wildfire by taking into account spatiotemporal heterogeneity of climate-wildfire relationship.

The neutral detergent fiber (NDF) content is a key factor in the forage palatability and livestock digestibility of alpine grassland. Traditional methods of measuring NDF are time-consuming and laborious, while multispectral remote sensing often has difficulty capturing subtle changes due to its broad spectral channels. In this study, we evaluate NDF using hyperspectral data and environmental factors (EFs) during 2016–2019. The artificial neural network (ANN), support vector machine (SVM), and random forest (RF) are used to construct models. The results show that: 1) compared with the correlation analysis, the least absolute shrinkage and selection operator (LASSO) regression can more effectively select the important variables for estimating NDF and achieve significant dimension reduction; 2) estimation accuracy of the models (coefficient of determination ($R^{2}$) between 0.44 and 0.51, relative root mean square error (RMSE) between 4.31% and 4.61%) based on first derivative (FD) spectra is relatively better than the models based on original spectra (OR), log transformation (Log), and continuum removal (CR) spectra ($R^{2}$between 0.35 and 0.46, RMSE between 4.39% and 4.92%). The RF model based on FD-Log-vegetation indices (VIs)-EFs is the best model ($R^{2} = 0.62$, RMSE = 3.98%); and 3) compared with the EFs, hyperspectral feature plays a pivotal role in estimating NDF, with a total contribution of 81.66%, and red-edge and shortwave infrared (SWIR) regions are significant for estimating NDF. In general, the combination of hyperspectral data with the EFs significantly improves the accuracy of estimating NDF. This study provides important guidance for future efforts on estimating NDF of natural grassland using remote sensing and machine learning algorithms.

Although self quotient image (SQI) has been popularly used for face recognition as a retinex-based based illumination normalization method, it suffers from strong shadows in faces with abrupt intensity change. In this paper, we propose SQI-based illumination normalization for face recognition based on discrete wavelet transform (DWT). To remove shadow edges while preserving features, we combine DWT with SQI into retinex-based illumination normalization. First, we divide a face image into four sub-bands using wavelet analysis to enhance vertical, horizontal, and diagonal edges. Then, we perform histogram truncation to remove very bright and dark regions. Finally, we conduct illumination normalization from SQI and the histogram truncation result based on the retinex theory. Experimental results demonstrate that the proposed method outperforms state-of-the-art retinex-based ones in terms of face recognition accuracy.