Abstract:Early acoustic cues of Sitophilus activity could reduce fumigation frequency in bulk stores. We record controlled silo segments and compare shallow time–frequency features with compact convolutional models under label scarcity. Shallow models remain competitive at small sample sizes, while deep models gain margin when augmented with synthetic chirp noise.
Abstract:Deferred canal maintenance undermines water reliability but labor contributions are poorly documented. We combine focus groups with sketch maps digitized into a participatory GIS and quantify hours by canal reach, season, and household role. Findings motivate rotating maintenance rosters that align with peak transplanting periods.
Abstract:High-elevation smallholder plots often lack dense weather grids. We interpolate daily thermal time and moisture indices using geostatistical kernels and feed them into a nonparametric yield model calibrated on extension trial data. Holdout counties show reduced error versus ridge regression when station density drops below one per 400 km2.
Abstract:Indoor production shifts environmental burdens from land to electricity and materials. We couple hourly cultivation records with marginal grid emission factors and Monte Carlo uncertainty for nutrients, substrates, and packaging. Outcomes are compared to regional open-field benchmarks for lettuce, highlighting seasons when indoor systems outperform on climate indicators.
Abstract:Pastoral mobility reshapes contact networks for foot-and-mouth-like pathogens in ways mean-field models miss. We embed herd agents on a corridor graph with stochastic mixing at wells and markets. Scenario sweeps show that staggered market days reduce peak prevalence more than blanket movement bans, with trade-offs for herder income stability.
Abstract:Intercropping with grain legumes is promoted as a low-input pathway to rebuild carbon stocks, yet reported gains vary widely. We synthesize 47 field experiments using hierarchical models and explore moderators of residue management, tillage intensity, and follow-up duration. Legume effects are positive on average but sensitive to baseline SOC and measurement depth.
Abstract:Timely irrigation decisions in steep vineyards require indicators that scale across row geometry. We combine narrow-band reflectance cubes with thermal profiles and apply constrained unmixing to isolate sunlit leaf, shaded leaf, and soil endmembers. Cross-validation against stem water potential shows improved lead time over NDVI alone, with minimal added flight time.
Abstract:Conservation tillage spreads unevenly even when agronomic benefits are well documented. We specify a structural equation model linking risk perception, peer networks, credit access, and training intensity to stated adoption intent among 412 smallholders. Bootstrapped indirect effects suggest peer demonstration dominates direct extension messaging. Results inform sequencing of incentives and field-day investments.
Abstract:Mixed-mode systems promise energy savings by combining natural ventilation episodes with mechanical cooling, yet comfort outcomes vary with operator protocols. We analyzed longitudinal survey responses from twelve floors in two retrofitted towers in a humid subtropical city, spanning two cooling seasons. Logistic models related acceptability votes to outdoor enthalpy, wind speed, and floor-level damper schedules. Occupants on perimeter zones reported wider comfort bands when night flushing was consistent. Mid-core zones remained sensitive to humidity spikes when economizer logic was conservative. Facility managers prioritized predictive maintenance alerts over additional sensors. Results inform commissioning checklists for climates with frequent shoulder-season ambiguity.
Abstract:Digitization accelerates access to specimens but also raises concerns about substituted or manipulated scans. We curated a balanced set of authentic high-resolution sheet images and controlled forgeries produced by splicing label regions and altering color histograms. A compact convolutional architecture with frequency-domain augmentations achieved area-under-curve above 0.91 on holdout accessions from three institutions. False positives clustered on faded type specimens with uneven lighting. Human expert review remained necessary for borderline cases involving legitimate conservation treatment. We release feature scripts and stress that automated scores should support, not replace, curatorial judgment. Ethical use guidelines prohibit deployment against living collectors without consent frameworks.