An analysis of motorcycle injury and vehicle damage severity using ordered probit models

Problem: Motorcycles constitute about 19% of all motorized vehicles in Singapore and are generally overrepresented in traffic accidents, accounting for 40% of total fatalities. Method: In this paper, an ordered probit model is used to examine factors that affect the injury severity of motorcycle accidents and the severity of damage to the vehicle for those crashes. Nine years of motorcycle accident data were obtained for Singapore through police reports. These data included categorical assessments of the severity of accidents based on three levels. Damage severity to the vehicle was also assessed and categorized into four levels. Categorical data of this type are best analyzed using ordered probit models because they require no assumptions regarding the ordinality of the dependent variable, which in this case is the severity score. Various models are examined to determine what factors are related to increased injury and damage severity of motorcycle accidents. Results: Factors found to lead to increases in the probability of severe injuries include the motorcyclist having non-Singaporean nationality, increased engine capacity, headlight not turned on during daytime, collisions with pedestrians and stationary objects, driving during early morning hours, having a pillion passenger, and when the motorcyclist is determined to be at fault for the accident. Factors leading to increased probability of vehicle damage include some similar factors but also show some differences, such as less damage associated with pedestrian collisions and with female drivers. In addition, it was also found that both injury severity and vehicle damage severity levels are decreasing over time.     

Vegetation stress detection through chlorophyll a + b estimation and fluorescence effects on hyperspectral imagery

Physical principles applied to remote sensing data are key to successfully quantifying vegetation physiological condition from the study of the light interaction with the canopy under observation. We used the fluorescence-reflectance-transmittance (FRT) and PROSPECT leaf models to simulate reflectance as a function of leaf biochemical and fluorescence variables. A series of laboratory measurements of spectral reflectance at leaf and canopy levels and a modeling study were conducted, demonstrating that effects of chlorophyll fluorescence (CF) can be detected by remote sensing. The coupled FRT and PROSPECT model enabled CF and chlorophyll a + b (Ca + b) content to be estimated by inversion. Laboratory measurements of leaf reflectance (r) and transmittance (t) from leaves with constant Ca + b allowed the study of CF effects on specific fluorescence-sensitive indices calculated in the Photosystem I (PS-I) and Photosystem II (PS-II) optical region, such as the curvature index [CUR; (R675.R690)/R2(683)]. Dark-adapted and steady-state fluorescence measurements, such as the ratio of variable to maximal fluorescence (Fv/Fm), steady state maximal fluorescence (F'm), steady state fluorescence (Ft), and the effective quantum yield (delta F/F'm) are accurately estimated by inverting the FRT-PROSPECT model. A double peak in the derivative reflectance (DR) was related to increased CF and Ca + b concentration. These results were consistent with imagery collected with a compact airborne spectrographic imager (CASI) sensor from sites of sugar maple (Acer saccharum Marshall) of high and low stress conditions, showing a double peak on canopy derivative reflectance in the red-edge spectral region. We developed a derivative chlorophyll index (DCI; calculated as D705/D722), a function of the combined effects of CF and Ca + b content, and used it to detect vegetation stress.