Deep Residual Involution Network for Hyperspectral Image Classification

Convolutional neural networks (CNNs) have achieved great results in hyperspectral image (HSI) classification in recent years. However, convolution kernels are reused among different spatial locations, known as spatial-agnostic or weight-sharing kernels. Furthermore, the preference of spatial compactness in convolution (typically, 3×3 kernel size) constrains the receptive field and the ability to capture long-range spatial interactions. To mitigate the above two issues, in this article, we combine a novel operation called involution with residual learning and develop a new deep residual involution network (DRIN) for HSI classification. The proposed DRIN could model long-range spatial interactions well by adopting enlarged involution kernels and realize feature learning in a fairly lightweight manner. Moreover, the vast and dynamic involution kernels are distinct over different spatial positions, which could prioritize the informative visual patterns in the spatial domain according to the spectral information of the target pixel. The proposed DRIN achieves better classification results when compared with both traditional machine learning-based and convolution-based methods on four HSI datasets. Especially in comparison with the convolutional baseline model, i.e., deep residual network (DRN), our involution-powered DRIN model increases the overall classification accuracy by 0.5%, 1.3%, 0.4%, and 2.3% on the University of Pavia, the University of Houston, the Salinas Valley, and the recently released HyRANK HSI benchmark datasets, respectively, demonstrating the potential of involution for HSI classification.

Risk factors for SARS-CoV-2 infection among farmworkers in Monterey County, California

ABSTRACTImportanceEssential workers in agriculture and food production have been severely affected by the ongoing COVID-19 pandemic.ObjectiveTo identify risk factors associated with SARS-CoV-2 shedding and antibody response in farmworkers in California.DesignThis cross-sectional study collected survey data and determined current SARS-CoV-2 shedding and seropositivity among 1,107 farmworkers in California’s Salinas Valley from 16 July to 30 November 2020.SettingFarmworkers receiving transcription-mediated amplification (TMA) tests for SARS-CoV-2 infection at federally qualified community clinics and community sites were invited to participate in our study.ParticipantsIndividuals were eligible if they were not pregnant, ≥18 years old, had conducted farm work since the pandemic started, and were proficient in English or Spanish.ExposuresSociodemographic, household, community, and workplace characteristics.Main Outcome(s) and Measure(s)Current (as indicated by TMA positivity) and historical (as indicated by IgG seropositivity) SARS-CoV-2 infection.ResultsMost farmworkers enrolled in the study were born in Mexico, had primary school or lower levels of educational attainment, and were overweight or obese. Current SARS-CoV-2 shedding was associated in multivariable analyses with attained only primary or lower educational levels (RR=1.32; 95% CI: 0.99-1.76), speaking an indigenous language at home (RR=1.30; 0.97-1.73), working in the fields (RR=1.60; 1.03-2.50), and exposure to known or suspected COVID-19 case at home (RR=2.98; 2.06-4.32) or in the workplace (RR=1.59; 1.18-2.14). Antibody detection was associated with residential exposures including living in crowded housing (RR=1.23; 0.98-1.53), with children (RR=1.40; 1.1-1.76) or unrelated roommates (RR=1.40; 1.19-1.64), and with a known or suspected COVID-19 case (RR=1.59; 1.13-2.24). Those who were obese (RR=1.65; 1.01-2.70) or diabetic (RR=1.31; 0.98-1.75) were also more likely to be seropositive. Farmworkers who lived in rural areas other than Greenfield (RR=0.58; 0.47-0.71), worked indoors (RR=0.68; 0.61-0.77), or whose employer provided them with information on how to protect themselves at work (RR=0.59; 0.40-0.86) had lower risk of prior infection.Conclusions and RelevanceOur findings suggest both residential and workplace exposures are contributing to SARS-CoV-2 infection among farmworkers in California. Urgent distribution of COVID-19 vaccines is warranted given this population’s increased risk of infection and the essential nature of their work.

Prevalence and clinical profile of SARS-CoV-2 infection among farmworkers in Monterey County, California: June-November, 2020

ABSTRACTAs essential personnel, United States farmworkers have continued working in-person throughout the COVID-19 pandemic. We undertook prospective surveillance of SARS-CoV-2 infection and antibody prevalence among farmworkers in California’s Salinas Valley from 15 June to 30 November, 2020. Over this period, we observed 22.1% (1514/6864) positivity for current SARS-CoV-2 by nucleic acid detection among farmworkers tested at federally-qualified migrant and community health clinics, as compared to 17.2% (1255/7305) among other adults from the same communities (risk ratio, 1.29; 95% confidence interval, 1.20-1.37). In a nested study enrolling 1,115 farmworkers, prevalence of current infection was 27.7% among farmworkers reporting ≥1 potential COVID-19 symptom, and 7.2% among farmworkers without symptoms (adjusted odds ratio 4.17; 2.86-6.09). Prevalence of anti-SARS-CoV-2 IgG antibodies increased from 10.5% (6.0-18.4%) between 16 July-31 August to 21.2% (16.6-27.4%) between 1-30 November. The high observed prevalence of infection among farmworkers underscores the need for vaccination and other preventive interventions.

Using an airborne electromagnetic method to map saltwater intrusion in the northern Salinas Valley, California

Saltwater intrusion can pose a serious threat to groundwater quality in coastal regions. Estimating the extent of saltwater intrusion is vital for groundwater managers to plan appropriate mitigation strategies. The airborne electromagnetic (AEM) method is commonly used to evaluate groundwater resources, but it is challenging to apply in coastal environments because the low resistivity of saltwater-saturated aquifers attenuates the electromagnetic signal quickly and the relationship between electrical resistivity and pore water salinity is complex. However, if successful, the AEM method can supply information to address questions of critical importance in coastal regions. We investigated the extent of, and controls on, saltwater intrusion using the AEM method in the northern Salinas Valley, CA, USA. We collected 635 line-km of AEM data in the study area, the inversion results of which produced estimates of the electrical resistivity of the subsurface, reaching depths of between 50 and approximately 200 m below the ground surface. We have developed a relationship between the AEM electrical resistivity model and groundwater salinity, calibrated from borehole geophysical and water quality measurements, which allowed us to generate images revealing the distribution of saltwater and fresher groundwater in the study area. This fresher groundwater (defined as “a source of drinking water”) was successfully mapped out in the unconfined aquifer (the Dune Sand Aquifer) and the uppermost confined aquifer (the 180-Foot Aquifer) in the study area, illustrating a groundwater recharge process that helps mitigate saltwater intrusion in the 180-Foot Aquifer. Deep, low-resistivity bodies also were mapped, indicating regions where saltwater likely is migrating vertically from the 180-Foot Aquifer into the lower confined aquifer (the 400-Foot Aquifer). The findings from this case study demonstrate the value of acquiring AEM data for investigating the distribution of salinity in coastal aquifers impacted by saltwater intrusion.

Comparison of CNN Algorithms on Hyperspectral Image Classification in Agricultural Lands

Several versions of convolutional neural network (CNN) were developed to classify hyperspectral images (HSIs) of agricultural lands, including 1D-CNN with pixelwise spectral data, 1D-CNN with selected bands, 1D-CNN with spectral-spatial features and 2D-CNN with principal components. The HSI data of a crop agriculture in Salinas Valley and a mixed vegetation agriculture in Indian Pines were used to compare the performance of these CNN algorithms. The highest overall accuracy on these two cases are 99.8% and 98.1%, respectively, achieved by applying 1D-CNN with augmented input vectors, which contain both spectral and spatial features embedded in the HSI data.

Pre-Harvest Survival and Post-Harvest Chlorine Tolerance of Enterohemorrhagic Escherichia coli on Lettuce

In the field, foodborne pathogens such as enterohemorrhagic Escherichia coli (EHEC) are capable of surviving on produce over time, yet little is known about how these pathogens adapt to this environment. To assess the impact of pre-harvest environmental conditions on EHEC survival, we quantified survival on romaine lettuce under two relative humidity (75% and 45%) and seasonal conditions (March and June). Greenhouse-grown lettuce was spray-inoculated with EHEC and placed in a growth chamber, mimicking conditions typical for June and March in Salinas Valley, California. Bacteria were enumerated on days 0, 1, 3, and 5 post-inoculation. Overall, we found that the effect of relative humidity on EHEC survival depended on the seasonal conditions. Under June seasonal conditions, higher relative humidity led to lower survival, and lower relative humidity led to greater survival, five days post-inoculation. Under March seasonal conditions, the impact of relative humidity on EHEC survival was minimal over the five days. The bacteria were also tested for their ability to survive a chlorine decontamination wash. Inoculated lettuce was incubated under the June 75% relative humidity conditions and then washed with a 50 ppm sodium hypochlorite solution (40 ppm free chlorine). When incubated under June seasonal conditions for three to five days, EHEC strains showed increased tolerance to chlorine (adj. p < 0.05) compared to chlorine tolerance upon inoculation onto lettuce. This indicated that longer incubation on lettuce led to greater EHEC survival upon exposure to chlorine. Subsequent transcriptome analysis identified the upregulation of osmotic and oxidative stress response genes by EHEC after three and five days of incubation on pre-harvest lettuce. Assessing the physiological changes in EHEC that occur during association with pre-harvest lettuce is important for understanding how changing tolerance to post-harvest control measures may occur.