Such responses to sexual harassment by labor unions have largely been shaped by the sexism rooted in the birth of the labor movement . It is the same male-centered philosophy that breeds the sexual harassment unions are intended to protect workers against. The first labor unions were over represented by male workers and as such, they discouraged women from reporting harassment by other union members, reinforcing labor unions as organizations perpetuating patriarchy . Despite the tumultuous past unions have had with sexual harassment, many are now taking a proactive approach to address the issue. UNITE HERE, in particular, has worked on addressing sexual harassment in the hospitality industry in cities across the country. Their Chicago Hands Off Pants On campaign, involved a member survey which found 58% of hotel workers and 77% of casino workers had been sexually harassed by a guest . The survey gave the union the data it needed to push for protections through both collective bargaining and eventually a city wide ordinance to protect both union and nonunion Chicago hotel workers . Moreover, in 2016, California members of the Service Employees International Union began their Ya Basta campaign and succeeded in winning statewide legislation in California requiring all janitors in the state to receive training on rape and sexual harassment prevention . Because of the growing relationship between labor unions and the cannabis industry, labor unions serve as a unique point of intervention for addressing sexual harassment in this particular workplace. Specifically,clone rack it useful to consider the social-ecological model proposed by Bronfenbrenner . His model had been referenced and adopted to fit several public interventions including violence prevention .

The individual, relationship, community and societal levels of the social ecological framework focus on personal factors, interpersonal relationships involving family, friends workplaces and neighborhoods, and broader structural factors such as health systems and polices, respectively . Each level of the model can be addressed to holistically grapple with the issue of violence, including sexual harassment in the workplace. Therefore, this particular study seeks to explore the community level as an opportunity for intervention, specifically through the growing relationship between the cannabis industry and labor unions.The study population was drawn from a source sample of cannabis employees represented by UFCW in Los Angeles County. Currently, UFCW Local 770 represents 20 dispensaries in Los Angeles County and approximately 400 workers to-date. In order to participate in the study, participants had to be at least 21 years of age, employed in a dispensary shop in Los Angeles County and had to be a current member of Local 770. As of August 2018, California is home to at least 358 retail stores with varying levels of permission including Temporary Approval or License to sell Medicinal or Adult-Use cannabis and cannabis products . Of the 358 legal stores located in the state, 188 are located in Los Angeles County , confirming Los Angeles area as an optimal location to investigate sexual harassment in the retail sector of the cannabis industry. This was a mixed methods study using a cross sectional design to analyze survey data from a convenience sample of dispensary workers represented by UFCW Local 770 in Los Angeles County. Surveys were used as part of this study to assess the demographics of the workforce, prevalence, frequency, types of sexual harassment occurring in the workplace, and workers’ knowledge on workplace policies regarding sexual harassment. In-depth interviews were conducted to better understand organizational risk factors of sexual harassment that are specific to cannabis dispensaries. Recruitment flyers with a description of the study were posted in the 20 Los Angeles dispensaries represented by Local 770.

The recruitment flyer included an online link and a quick response code to the survey for workers to easily access the survey through a mobile device. The recruitment flyer included an invitation to participate in an additional interview. The majority of recruitment occurred at shops as I explained the study to workers, and they volunteered to participate. Interview participants were recruited through snowball sampling. Each interview participant was provided a consent form, was explained the objectives of the study, and agreed to have their interview digitally recorded. Participants were assured that no identifiable information would be collected and any identifiable information that may have been mentioned throughout the interview was redacted from the transcript.Each interview took between 40 and 60 minutes and occurred at a location of the interviewee’s choosing . Each participant was compensated for their time with a 15-dollar gift card. In total, seven participants were interviewed as a part of this study. The interview protocol included providing participants with a copy of the interview questions so that they could follow along with the line of questions. In addition to interview questions, participants were provided with a formal definition of workplace violence and sexual harassment in order to establish a common understanding of the terms I would be using throughout the interview. Interviewees were first asked to explain their regular workday and their work responsibilities as well as their employment tenure in cannabis. They were then asked about sexual harassment in their workplace, if they had experienced, witnessed or heard about instances through word-of-mouth. They were also asked to explain the protocols of their workplace aimed at preventing harassment and protecting workers who seek redress. In an effort to be mindful of discussing the sensitive topic of sexual harassment, the questions focused more on factors leading up to harassment and the actions taken by workers, co-workers and their management as a result of such events. Finally, workers were also asked to describe interventions they would like to see implemented in their workplace to address harassment.

Throughout data collection, questions were added, removed, and re-written based on feedback from participants. Survey participants completed the survey on their mobile device or self-administered it in-person with a tablet. While visiting stores for recruitment, interested workers could take the survey on tablets provided by Local 770. Once the study was complete, all data collected, except for the city of employment variable, were made available to UFCW with additional information and recommendations on how they may better support members of their cannabis division. The final sample for the survey was 117 participants. All data were collected between October 2019 – February 2020.The survey instrument consisted of 37 questions designed to take no more than seven minutes to complete. The first section included questions on workplace city, duration of employment in the current workplace and total employment experience in the cannabis industry. The second section asked about sexual harassment policies in the workplace followed by questions regarding experiences with sexual harassment , perpetuators, and responses . The fourth section asked workers to select, from a list of examples, interventions they believed could be implanted to prevent or better handle instances of harassment in their workplace. The fifth and concluding section included basic demographic questions .In order to capture sexual harassment in the workplace, I adopted nine questions from Fitzgerald and colleague’s Sexual Experiences Questionnaire – Department of Defense Items . The SEQ-Dod was originally developed to measure sexual harassment of in the Armed Forces but has also frequently been adopted to capture harassment in the workplace . The original 24-item measures four dimensions of sexual harassment: sexist hostility, sexual hostility, unwanted sexual attention and sexual coercion. As defined by Fitzgerald , sexist hostility is defined as gender-based harassment, sexual hostility is also a form of gender-based harassment but involves behaviors that are more overtly sexual in nature, unwanted sexual attention is characterized by attention and behaviors that are unprompted and unreciprocated by the receiver, and lastly sexual coercion is more explicitly extortionist in nature and considered quid pro quo harassment. The tool has also since been shortened to capture the original four dimensions of sexual harassment in 16 questions . For this study,4×8 tray grow only the first three dimensions of sexual harassment were measured. The version of survey specifically adopted for this study was from the National Park Service’s Work Environment Survey as their study addressed shortcomings of the original tools including issues of recall bias and a lack of examples which constitute sexual harassment . All survey participants responded to 12 different questions of sexual harassment behaviors . Participants indicated how often they experienced the behavior in the last 12 months using a 6-point scale ranging from never , “once” , “once or less a month” , “2-3 times a month” , “once a week or more” , “to once or more times a day” . Because the 12 questions were adopted from the original instrument, a reliability analysis conducted to determine if the 12 items were interrelated produced a Cronbach’s alpha of 0.88, indicating good internal consistency. The Cronbach alphas for the three sub-scales of sexist hostility, sexual hostility, unwanted sexual attention were 0.88, 0.84, and 0.36, respectively.

Consistent with the scoring used in the WES report , responses to these questions were summed and averaged across items to determine a sexual harassment score per sub-scale. Sexual harassment scores ranged from one to five; the greater the sexual harassment score, the greater frequency of harassment experienced by a worker. Prevalence of specific dimensions or behaviors of harassment among respondents was calculated by the percentage of respondents who reported experiencing an example of harassment at least once. The greater the percentage, the more employees experienced either a specific dimension or behaviors of sexual harassment. Respondents who answered experiencing sexual harassment at least once in their workplace were asked to report who had sexually harassment them. Options included a fellow employee , subordinate, manager, customer and other or an unknown individual. The five possible options were coded as separate dichotomous variables and scored as one if selected and zero if left blank. Participants were also asked if they had ever experienced sexual harassment by someone of the same gender. Responses options were “yes” , “no” , and “unsure” . To gauge awareness of sexual harassment policies, a question was adopted from UFCW’s questionnaire on sexual harassment in the workplace. The survey provided by UFCW was originally designed for their grocery worker division and was intended to capture worker’s knowledge of workplace policies protecting them against sexual harassment as well as the impact of sexual harassment on employee work schedules. Respondents were asked if they were aware of their retail company’s policies regarding sexual harassment. They either responded “yes” or “no” . If they responded “yes,” they were further asked to select specific policies that applied to their workplace from a list of six examples . All examples were coded as dichotomous variables. If selected, the item was coded as a one, and if left blank, the item was coded as zero. A response score ranging from one to six was calculated by summing the responses to all six items. A larger score equates to workers having a greater awareness of policies in the workplace to protect them against sexual harassment. The Cronbach alpha for these six items was 0.83, indicating good internal consistency. As part the translational nature of the project, respondents were asked if they wanted more support in their workplace to protect workers against sexual harassment. This item as also adopted from UFCW’s questionnaire. If respondents selected “yes” , they were asked to indicate which actions, from a list of seven examples, they would like to see taken in their workplace to address the issue of sexual harassment . Respondents were asked to choose all options they felt were applicable to their workplace. All seven options were converted into a dichotomized variable and scored as one if the option was selected and zero if left unselected. An intervention score was created by the summing responses to seven options. This data will be especially beneficial to informing the union how they may better service their members.Covariates included in my analysis include the age of workers, their race/ethnicity, gender identity, sexual orientation and employment status. Age was treated as a continuous variable in analysis. Respondents provided their race and ethnicity by selecting from the six provided categories, all the options they felt best applied to their identity. A seventh option for “other” was provided if they wanted to write in their response. Responses for race/ethnicity were first categorized into seven levels: White, Black, Asian, Latina/o, Asian, Native American, Hawaiian or Pacific Islander and other.

The center’s ongoing research includes a multi faceted project to assess specific aspects of Northern California’s cannabis farms, including the number and size of non-compliant cultivation sites; the environmental effects of non-compliant sites ; and the challenges to regulatory compliance that cannabis cultivators encounter. According to a grant proposal associated with the research, the project is motivated by an urgent need to understand the environmental threats posed by non-compliant farms and the reasons that some farms successfully navigate state regulations while others fail. The researchers are combining high-resolution satellite images with local and state permitting data to identify permitted and non-permitted cultivation sites. In parallel, the researchers are combining permit specifications with water use models to estimate the effects on stream flows of non-permitted versus permit ted cultivation. Additionally, they are determining which factors associated with cannabis cultivation are most closely linked to compliance — whether parcels are large or small, old or new — and, through writ ten grower surveys and in-person interviews, they are seeking to understand what stands in the way of cultivator compliance. Ultimately, the work will yield a policy report outlining ways in which state and local governments can decrease the harm of non-compliant cannabis cultivation while increasing rates of compliance. The research is supported by a grant from the Campbell Foundation, provided through the Resource Legacy Fund. In another example of CRC research focused on cannabis and the environment, last year Butsic, Jennifer Carah and additional co-authors published the results of their work on “agricultural frontiers” . These are places where,plants rack due to increased profit potential for agricultural activity, land is newly cultivated — frequently resulting in environmental impacts such as forest fragmentation and threats to sensitive species.

Such transformations, the authors write, occur when economic circumstances are altered by some new mechanism — such as, in the case of cannabis, a new legal status. The researchers, documenting the emergence of such a frontier, studied cannabis cultivation sites in Humboldt and Mendocino counties from 2012 to 2016. Using satellite imagery to develop a database of cultivation sites, the research ers correlated site characteristics such as remoteness and erosion potential with the spread of agricultural frontiers. They report that, over the study period, cannabis cultivation sites in the study area nearly doubled in number, with total acreage under cultivation likewise nearly doubling, and that a significant portion of the new cultivation occurred in areas such as sensitive watersheds. They found, for example, that nearly 90% of the areas newly developed for cannabis cultivation had been covered in natural vegetation as late as 2006. The researchers argue that agricultural frontiers can develop “almost anywhere institutions fail to prevent” them — and note that, for 18 years after medicinal cannabis use became legal in California with the 1996 Compassionate Use Act, the state devoted no funds to regulating cannabis cultivation and production. In this issue of California Agriculture, Grantham and four co-authors from the North Coast Regional Water Quality Control Board present the results of their research into cannabis cultivators’ patterns of water use in several Northern California countries. For the research that resulted in “Watering the Emerald Triangle: Irrigation sources used by cannabis cultivators in Northern California” , Grantham and his colleagues analyzed reports submitted to the board by cannabis cultivators. The researchers determined how many cultivators sourced their water from wells, surface water diversions, spring diversions and other sources; how water sourcing behavior changed over the course of a year; and how water use patterns varied ac cording to whether growers operated within the state’s legal cannabis market.

The researchers determined that cannabis growers rely on well water to a greater degree than is generally supposed — and that their reliance on well water may increase as more growers join the legal market because of well water’s less restrictive permit ting requirements. In separate research, Michael Polson — a post doctoral researcher in UC Berkeley’s Department of Environmental Science, Policy, and Management — has investigated the environmental dimensions of cannabis from an anthropological perspective. In a paper published earlier this year, Polson shows how cannabis has been identified as an environmental problem that requires public intervention . On the basis of participant observation and more than 70 interviews with subjects across the cannabis spectrum — from park rangers to environmentalists to “criminalized people” — Polson demonstrates how cannabis production has been defined as pollution — “dovetail[ing] with [cannabis] prohibition’s history of marking people and substances as socially polluting.” Polson argues, as he highlights the legacy of cannabis prohibition in environmental debates, that policy making is at its most innovative when it includes a broad range of cultivators and when stigmas are explicitly addressed. Research into the environmental aspects of cannabis is also underway at UC Davis, where Mourad Ga briel is a research associate member in UCD’s School of Veterinary Medicine. In 2018, Gabriel and co-authors, including Robert Poppenga — a professor of molecular bio-sciences at the California Animal Health and Food Safety Lab at UC Davis — published the results of their research on the effects of rodenticides on owls in northwestern California forests . The researchers, working on privately owned timberland in Humboldt and Del Norte counties, investigated the prevalence of anticoagulant rodenticides in areas char acterized by illegal cannabis cultivation. Anticoagulant rodenticides, used by some cannabis cultivators to control pests, are known to affect non-target species in urban areas and recently have been shown to affect carnivores in California’s remote forest areas as well. Gabriel and his co authors undertook to deter mine whether the northern spotted owl, a threatened species, is exposed to anti coagulant rodenticides in the study area — and also to determine if barred owls, a common species, can be used as a surrogate to determine exposure levels in northern spotted owls.

The researchers analyzed liver samples from 84 barred owls and 10 northern spotted owls. Within the study area, 70% of northern spotted owls and 40% of barred owls tested positive for anticoagulant rodenticides. The researchers hypothesize that cannabis cultivation in the area is the main source point for the presence of dangerous rodenticides. They also determined that barred owls are a suitable surrogate for determining rodenticide levels in the threatened northern spotted owl. Gabriel, in his capacities as a UC researcher and as executive director of the Integral Ecology Research Center, a nonprofit organization based in Humboldt County, is currently carrying out reclamation projects at illegal cannabis cultivation sites in California and Oregon. In a project conducted this May in the Shasta-Trinity National Forest, a team representing 11 governmental and nongovernmental entities worked at 16 cultivation sites within eight large cultivation complexes, removing 6,000 pounds of trash, which included rodenticides and more than 5 miles of irrigation lines. Mourad estimates that removal of the irrigation lines restored more than 500,000 gallons of water — daily — into affected watersheds. Agencies including the California Department of Fish and Wildlife and the Law Enforcement and Investigations arm of the U.S. Forest Service have provided grant funding for 170 such projects, 112 of which have already been completed. In an entirely different vein, UC Davis–based cannabis research has been conducted since 2016 at the UC Agricultural Issues Center , a UC Agriculture and Natural Resources statewide program operating since 1985. The center’s broad mission is to provide research based information on the economic dimensions of emerging issues in agriculture. Cannabis, then, is right in the center’s wheelhouse. Dan Sumner, the center’s director,plant growing trays reports that AIC began pursuing cannabis-related work after the 2015 passage of a set of laws known collectively as the Medical Marijuana Regulation and Safety Act. This legislation laid the groundwork for state regulation of medicinal cannabis and ultimately of the recreational cannabis industry. The lead agency in regulating commercial cannabis licenses for distributors and retailers, among other business types, is the Bureau of Cannabis Control — for which, between 2016 and 2018, the AIC prepared a Standardized Regulatory Impact Analysis . In the process, the AIC advised the BCC on the economic dimensions of various regulatory scenarios — and the bureau used the center’s analysis to inform the final cannabis regulations that it issued on Jan. 16 of this year. According to Sumner, a principal insight that the AIC furnished to the BCC was that, since illegal cannabis continues to be attractive to retail buyers because it is cheaper than cannabis from regulated retailers, “much of the cannabis sold in California [after legalization] would remain in the illegal segment.” Moreover, regulations that generate benefits for consumers at lower costs will help sustain the legal marketplace. In this issue of California Agriculture, three AIC researchers — Pablo Valdes-Donoso, a postdoctoral scholar; Robin S. Goldstein, principal economic counselor; and Sumner — present their research on California’s rather stringent system for testing cannabis that enters the legal market . All cannabis sold legally in the state is tested for more than 100 contaminants. Of those contaminants, 66 are pesticides — and tolerance for 21 of those pesticides is set at zero. In many cases, allowable levels of cannabis contaminants are lower than those established for food sold in the state. The researchers, drawing on data provided by testing laboratories and manufacturers of testing equipment, estimated how much it costs to test a pound of cannabis under California’s regulatory regime, as well as the cost of collecting samples. They concluded that the need to destroy batches of cannabis that fail testing accounts for a large share of testing costs. The research ers argue that, though the availability of certifiably safe and legal cannabis products may prompt some customers to join the regulated market, other customers will remain in the cheaper illegal market.

They speculate that, over time, increased availability of data about cannabis testing and sales will allow for greater certainty about the effect of the testing regime on cannabis prices and demand for legal cannabis. Meanwhile, UC Davis is establishing a dedicated center for research into psychoactive cannabis and industrial hemp — the Cannabis Research Initiative. According to Cindy Kiel, executive associate vice chancellor for research administration at UC Davis, the initiative will draw on the comprehensive strengths of UC Davis faculty in areas ranging from agricultural and environmental impacts to legal, economic and policy outcomes to human and animal health. In particular, the initiative will benefit from UC Davis’s strong emphasis on agricultural issues such as soils, water, genomics and plant science and from faculty interest in two-way interactions such as those between cannabis and the environment. Funding is envisioned to flow from the UC Davis budget, from research funds established in Proposition 64 and from outside sources such as industry partners. The initiative will be headed by co-directors representing the agricultural and medical sides of cannabis research. In May, UC Davis faculty members including Chemistry Professor Mark Mascal, along with col leagues from the University of Reading in the United Kingdom, published an article demonstrating that a synthetic analogue of cannabidiol is as effective as CBD in controlling seizures in rats — and that it provides several benefits in comparison to CBD. The synthetic analogue is cheaper than herbal CBD, cannot be converted into psychoactive tetrahydro cannabinol and is not restricted by the Drug Enforcement Administration’s “scheduling” apparatus. Meanwhile, the UC Davis–affiliated Western Center for Agricultural Health and Safety is studying issues such as workplace safety for cannabis workers, who face risks that include unhealthy pesticide exposure. For students, UC Davis has offered cannabis courses including the graduate-level Cannabis sativa: The Plant and Its Impact on People — and, for undergraduates, Physiology of Cannabis.A brand-new entrant into UC cannabis research is the UC Nicotine and Cannabis Policy Center at UC Merced , established just last year to study tobacco- and cannabis-related issues in public health and public policy, especially in the San Joaquin Valley. The center, partnering with local public health departments and organizations such as the American Heart Association, aims to produce tobacco and cannabis research that places special emphasis on the San Joaquin Valley’s diverse population of teens and young adults and in forms policy decisions that affect the region. The center’s flagship research initiative is a long term, survey-based effort to understand issues surrounding cannabis, tobacco and e-cigarettes.

Particularly notable is FAAH’s ability to hydrolyse bioactive fatty amides, which do not bind to any of the known cannabinoid receptors: these include the satiety factor oleoylethanolamide and the anti inflammatory/ analgesic mediator palmitoylethanolamide. FAAHtightly controls brain concentrations of these compounds, but the functional significance of this regulation is unknown. FAAH is widely distributed in the rat brain, where it is expressed at high concentrations in cell bodies and dendrites of principal neurons. In the hippocampus, neocortex and cerebellum, FAAH-positive cell bodies are juxtaposed to axon terminals that contain CB1 receptors, indicating not only that FAAH participates in the inactivation of neurally generated anandamide, but also that this process occurs post synaptically. This idea can now be tested in FAAH-deficient mice or using selective FAAH inhibitors with long-lasting systemic actions.The pig brain contains two chromatographically distinct 2-AG-hydrolysing activities, one of which is probably due to the enzyme monoacylglycerol lipase . The rat brain isoform of this cytosolic serine hydrolase has been characterized both molecularly and morphologically. It has a broad distribution in the central nervous system , which partially overlaps with that of FAAH; however, whereas FAAH is predominantly found in postsynaptic structures, MGL might be mostly associated with nerve endings. In the hippocampal CA1 field, MGL-positive axon terminals surround cell bodies of pyramidal neurons containing FAAH. This localization could reflect a functional role of presynaptic MGL in terminating RETROGRADE SIGNALLING events mediated by 2-AG .CB1 is considered to be the most abundant G-protein coupled receptor in the mammalian brain, and its presence in the neocortex, hippocampus, basal ganglia, cerebellum and brainstem accounts for most of the behavioural actions of cannabinoid drugs. The four symptoms that are often used to define cannabinoid intoxication in the rodent __ hypothermia, rigid immobility,weed drying room analgesia and decreased motor activity __ are strikingly absent in mice in which the cb1 gene has been deleted by targeted recombination.

Aside from its unusually high concentrations in the brain, CB1 is a standard Gi/o-coupled receptor and can initiate signalling events typical of this class of transducing proteins. These include closure of Ca2+ channels, opening of K+ channels, inhibition of adenylyl cyclase activity and stimulation of kinases that phosphorylate tyrosine, serine and threonine residues in proteins. Each of these mechanisms seems to have distinct functions in translating CB1 -receptor occupation into biological responses. Cannabinoid agonists inhibit N- and P/Q-type voltage-activated Ca2+ channels. This effect, which has been suggested to result from a direct interaction of Gi/o-protein β−γ subunits with the channels, might underlie CB1 -mediated depression of transmitter release at GABA synapses in the CA1 field of the hippocampus and at glutamatergic synapses in the dorsal striatum. Importantly,however, endocannabinoid-mediated suppression of GABA release in hippocampal slices seems primarily to involve N-type Ca2+ channels. Cannabinoid regulation of voltage-gated K+ currents is also implicated in presynaptic inhibition at GABA and glutamate synapses. The latter include PARALLEL FIBRE–Purkinje cell synapses in the cerebellum, as well as synapses in the nucleus accumbens and lateral amygdala. The sensitivity of these responses to PERTUSSIS TOXIN implies that they are mediated by Gi/o proteins, but it is still unclear whether transduction is direct or indirect . Inhibition of cAMP formation does not seem to be involved. On the other hand, cAMP can contribute to the regulation of neuronal gene expression by CB1 . This process, which is necessary to produce lasting changes in synaptic strength, depends on the recruitment of complex networks of intracellular protein kinases87. Two components of these networks, extracellular signal-regulated kinase and focal adhesion kinase , become activated when hippocampal slices are treated with cannabinoid agonists. This activation is mimicked by inhibitors of cAMP-dependent kinase and is lost when the slices are exposed to cell-permeant cAMP analogues, implying that it might result from a decrease in intracellular cAMP concentrations.

The involvement of ERK and FAK in synaptic plasticity indicates that these protein kinases could participate in the changes in gene expression and the persistent neural adaptations that accompany cannabinoid administration90.In the rodent and human cortices, CB1 receptors are primarily found on axon terminals of cholecystokinin-8 -positive GABA interneurons. This expression pattern dominates the neocortex, hippocampal formation and amygdala, where nerve terminals that form excitatory synapses are ostensibly devoid of CB1 immunoreactivity. However, there is evidence that excitatory terminals in these regions do contain the receptor; for example, cannabinoid agonists reduce glutamatergic transmission in the amygdala of normalmice, but fail to do so in CB1 -deficient mutants. In addition, low concentrations of CB1 messenger RNA have been found in many neurons of the cortex that do not contain GABA. CB1 receptors are also expressed at very high levels throughout the basal ganglia. In the striatum they are localized to three distinct neuronal elements: glutamatergic terminals originating in the cortex, local-circuit GABA interneurons and axon terminals of GABA projection neurons. Medium spiny neurons project to striatal outflow nuclei, where CB1 receptors are especially abundant; for example, in the globus pallidus they outnumber dopamine D1 receptors by a factor of 45 . In the cerebellum, CB1 is present on excitatory terminals of climbing and parallel fibres as well as on GABA interneurons. Smaller numbers of CB1 receptors are also found in the thalamus , hypothalamus , midbrain , medulla and spinal cord. Last, CB1 is expressed in peripheral sensory neurons, where it is localized in cells that express N52, a protein marker of mechanosensitive Aβ fibres.A few cannabinoid effects persist in CB1 -null mice, implying that this receptormight not act alone in mediating brain cannabinoid signalling. Although cannabinoid agonists lose their ability to inhibit GABA and glutamate transmission in some brain regions of adult CB1 -knockout mice, they can still reduce excitatory transmission in the hippocampal CA1 field of these animals. This discrepancy is reinforced by the finding that GABA and glutamate synapses in CA1 respond in different ways to cannabinoid drugs. For example, cannabinoid depression of excitatory currents is blocked by CAPSAZEPINE, whereas depression of inhibitory currents is not.

These results make a persuasive case for the existence of a hippocampal cannabinoid-sensitive site that is distinct from CB1 , but other evidence appears to contradict them; for example, in newborn CB1 -null mice, cannabinoid agonists affect neither GABA nor glutamate transmission. Although this difference could be due to the developmental stage of the preparation used __ adult versus one-week-old mice __ more studies are needed to establish whether the CB3 site is molecularly distinct from CB1 . A novel cannabinoid site has also been identified in the vascular endothelium, but seems to be different from CB3 because it is not antagonized by capsazepine or activated by the CB1 /CB2 agonist Win-55212-2 .Outside the brain, the endocannabinoids are produced on demand and act on cells located near their site of synthesis. For example, they are formed by circulating leukocytes and platelets, and induce vascular relaxation by interacting with cannabinoidreceptors on the surface of neighbouring endothelial and smooth muscle cells. Similar PARACRINE actions are thought to occur in the CNS, where the endocannabinoids might mediate a localized signalling mechanism through which principal neurons modify the strength of incoming synaptic inputs.When a pyramidal neuron in the CA1 field of the hippocampus is depolarized, the inhibitory GABA inputs received by that cell are transiently suppressed. This phenomenon, called depolarization-induced suppression of inhibition , is initiated postsynaptically by voltage-dependent influx of Ca2+ into the soma and dendrites of the neuron, but is expressed presynaptically through inhibition of transmitter release from axon terminals of GABAinterneuron. This indicates that a chemical messenger generated during depolarization of the pyramidal cell must travel backwards across the synapse to induce DSI . There is evidence that this retrograde signalling process involves an endocannabinoid substance, possibly 2-AG. First, CB1 agonists mimic DSI, whereas CB1 antagonists block it. Second, DSI is absent in CB1 – deficient mice81,109. Third, the GABA interneurons that are implicated in DSI express high levels of CB1 receptors, which are localized to their axon terminals. Fourth, neural activity and Ca2+ entry stimulate the hippocampal synthesis of 2-AG, but have no effect on anandamide concentrations. Nevertheless, we still don’t know whether the endocannabinoid actually crosses back to the presynaptic nerve ending or is produced there by the action of another, unidentified retrograde signal . The fact that DSI is induced in vitro by levels of neural activity that could also be encountered in vivo indicates that this process might have a role in normal brain function. Although this idea is still questioned, various results link DSI to the regulation of hippocampal GAMMA OSCILLATIONS. These network oscillations are coordinated by CB1 -positive GABA interneurons and are influenced by cannabinoid agonists,drying rack for weed raising the possibility that an endocannabinoid substance might modulate their expression and be involved in the organization of hippocampal cell assemblies. Another function of DSI might relate to synaptic plasticity. By weakening GABA-mediated inhibition, DSI could facilitate the induction of long-term potentiation in individual CA1 pyramidal neurons; this might contribute in turn to the formation of ‘place fields’ or to other forms of hippocampus-dependent learning. Such a cognitive-enhancing action would not contradict the well-known amnesic effects of cannabinoid drugs as the latter might result from a generalized, circuit-independent activation of CB1 receptors in the hippocampus and other brain areas. Outside the hippocampus, endocannabinoid-mediated DSI has been shown to occur at interneuron– principal cell synapses of the cerebellum and probably will soon be discovered elsewhere.

CB1 -bearing interneurons are selectively localized to a subdivision of the amygdala called the basolateral complex, a key station in the neural circuitry that processes emotions and a primary site of cannabinoid analgesia. This localization, and the fact that CB1 inactivation causes anxiety-like and aggressive responses in rodents, indicate that the endocannabinoid system might influence affective states through changes in the amygdala’s efferent activity. This idea is further supported by two findings: first, presentation of anxiogenic stimuli increases anandamide and 2-AG concentrations in the mouse amygdala; second, FAAH inhibitors exhibit marked anxiolytic-like properties in rats. Locally formed endocannabinoids could modify the amygdala’s output in two complementary ways.They could depress glutamate release from axon terminals originating in the cortex and other brain regions. In addition, by reducing GABA release from basolateral interneurons, they might disinhibit GABA cells in the adjacent intercalated nuclei and consequently decrease the activity of their postsynaptic targets, the pyramidal neurons in the central nucleus of the amygdala, which constitute the structure’s primary efferent pathway. Here, local administration of cannabinoid agonists inhibits GABA release and profoundly affects motor behaviours. Membrane depolarization and dopamine D2 -receptor activation stimulate striatal anandamide formation, indicating that this endocannabinoid might contribute to the regulation of basal ganglia function. In agreement with this hypothesis, the CB1 antagonist rimonabant enhances the stimulation of movement that is induced in rats by dopamine agonists, whereas the endocannabinoid transport inhibitor AM404 attenuates this stimulation in a CB1 -dependent manner. Anandamide might act at multiple sites in the basal ganglia, including GABA projection neurons, corticostriatal glutamatergic terminals and local-circuit interneurons. Local-circuit interneurons are particularly notable because of their functional resemblance to CB1 -positive interneurons in the hippocampus, with which they share not only a GABA-containing phenotype, but also the ability to discharge high-frequency bursts of action potentials that can inhibit firing in large assemblies of projection cells. Does locally released anandamide gain access to these interneurons? Or does it primarily act on medium spiny cells and their cortical afferents? We don’t know yet. But these unanswered questions do not diminish the significance of striatal endocannabinoid signalling, which is further highlighted by the effectiveness of cannabinoid agonists in the symptomatic treatment of LEVODOPA-INDUCED DYSKINESIAS and TOURETTE’S SYNDROME, two disorders with strong striatal underpinnings.Beside their actions in the amygdala, cannabinoid agonists can influence the central processing of pain by interacting with CB1 receptors in the periaqueductal grey, rostral ventromedial medulla128 and spinal trigeminal nucleus. At each of these sites, CB1 activation depresses GABA release through a presynaptic mechanism, without causing significant changes in somatic membrane conductances. In the trigeminal nucleus, glycinergic transmission also is inhibited. Painful stimuli elicit anandamide release in the rat periaqueductal grey, and systemic administration of CB1 antagonists produces HYPERALGESIA in rats and mice. So, noxious stimuli can engage a central analgesic circuit operated by the endocannabinoids, which, working in combination with a parallel mechanism in the periphery, could underlie the analgesic properties of cannabinoid drugs.

It is already understood that by preheating the air before it enters the cathode results in better SOFC performance because the temperature difference from inlet to outlet in the stack is much less and therefore can respond faster to load fluctuations. As is already understood, SOFC systems best serve stationary power generation applications. SolidPower recommends using its Engen-2500 system for multi-family houses, hotels, restaurants, public buildings, schools, or small companies. It can be installed in different configurations including as a stand-alone mCHP unit with water storage where it can be combined with heat and power systems that use renewable energy source or as a series of multiple modules, integrated with the heating system of the building. The particular application of the Engen-2500, discussed throughout this thesis, is for providing power to individual server racks, which typically range from a handful of servers to dozens depending on the class of server. Fuel cells are generally suitable for constant power generation applications. They perform best when fuel and oxygen supplies are steady and the load demand remains constant. Therefore, Microsoft’s vision of disconnecting large data center facilities from the electric grid and relying solely on SOFCs for providing power is definitely possible. Data center power consumption has both short and long-term variations due to workload fluctuations and servers turning on and off. The variations in load can be represented by three categories: instantaneous load changes, short term load changes, and long-term load changes. For the first case, the load of a server can change almost instantaneously reacting to a workload. This may be caused by a change in CPU utilization from 0% to 100% and can occur within milliseconds. Some of these types of rapid fluctuations in load can potentially be absorbed by the server’s power supply with its internal capacitors, however, larger instantaneous changes must be handled using an external energy storage system in cooperation with the SOFC system. Short-term load changes occur over several seconds or minutes and can be handled by the SOFC system as it ramps up or down accordingly. Yet, in extreme situations where the server has to be cold rebooted,curing cannabis the SOFC lags behind the sudden spikes in power consumption. For these extreme situations, it is always necessary to include a battery in order to keep up with the drastic load changes.

If the load changes are predictable , the SOFC system can increase its power production ahead of time to follow the load demand. Doing so may lead to reduced end-to-end efficiency of energy usage, therefore this practice is not recommended if trying to maximize production efficiency. The ideal case for applying SOFC systems in data centers is when there are long-term load changes that occur over days and weeks. These changes are typically much slower than all fuel cell’s ramp rates and therefore are no issue for the SOFC’s ability to keep up with the load demand. Under load fluctuations, it would be helpful to understand how much over provisioning of the fuel cell is necessary. The only case where SOFCs struggle is when the servers cause large load changes due to startup and shutdown processes. By staggering server power on and off events over time instead of all at once, a single server sized battery can be shared by multiple servers in a rack. Therefore, the SOFC would not need to be aggressively over provisioned. After receiving the Engen-2500 system from SolidPower and connecting it to the load bank, natural gas, and water pipelines in the testing bay facility of the NFCRC, the initial round of testing began. It was agreed upon with members of the fuel cell team at UCI, Microsoft, and SolidPower that the testing framework would begin with less severe transients in order to avoid causing damage to the stack or system components early on. The first step to characterizing and understanding the performance of any fuel cell system is to subject it to steady-state analysis. A testing matrix was established to neatly organize and document the user-constrained parameters and the corresponding values for other important parameters. For the steady-state analyses, the cathode outlet temperature and fuel utilization were the two parameters that were defined as user-constrained with the two outlet temperature values selected to be 760°C and 770°C and the three fuel utilization values selected to be 0.70, 0.72, and 0.75 as shown in Table 5. In an ideal world, a fuel cell would be able to maintain a constant voltage output for any amount of current demanded. This would appear as a flat line on the current-voltage plot at the peak voltage level, which refers to the open-circuit voltage – defined as the difference of electrical potential between the two electrodes when disconnected from a circuit . However, due to real-world physical limitations, the actual voltage output of a real fuel cell is less than the ideal thermodynamically predicted voltage due to unavoidable losses that is discussed in detail in Section 6 of this thesis.

Another very noticeable fact about the system performance is the slight voltage differences for the same current load demand. From Figures 24 and 25 the slight voltage differences observed between the two stacks in the system are thought to be caused by the slight temperature differences of the two stacks and the manufacturing variation amongst the cells. The benefits of understanding the steady-state performance by looking at the polarization curve is to provide additional information for the overall data center design such as the DC bus, power supply specifications, and DC/DC converters. For all energy conversion devices, it is always of great importance to consider and compare the operational efficiency. The fundamentals of efficiency can be broken down and represented by two important concepts, ideal and real efficiency. Since fuel cell devices produce electric and not mechanical work, thermodynamic theory suggests that the electric work available is limited by the change in Gibbs free energy, therefore, the ideal efficiency of a fuel cell is limited by. The ideal fuel cell efficiency is defined as the amount of useful energy that can be extracted from the process relative to the total energy evolved by that process. For a fuel cell, the maximum amount of energy available to do work is given by the ratio of Gibbs free energy to enthalpy. These losses are illustrated in the polarization plots and the voltage efficiency can be characterized by the ratio of the real operating voltage to the thermodynamically reversible voltage. Note that the operating voltage depends upon the current drawn, therefore the higher current demand, the lower the voltage efficiency. The fuel utilization efficiency accounts for the fact that not all of the provided fuel is used by the fuel cells. Some of the fuel may undergo side reactions that do not produce electric power or the fuel may simply flow through the fuel cell without ever reacting. The fuel utilization efficiency is therefore a ratio of the fuel used by the cells to generate electric current to the total fuel provided. With this understanding of real fuel cell efficiency and considering the initial round of steady-state tests, the steady-state efficiencies of Engen-2500 stack and system were determined. Considering the case for fuel utilization of 75% and a cathode outlet temperature of 770°C, the corresponding stack and system efficiencies are shown in Figure 26. As is evident, the electrical efficiency of just the stack alone is well above 52%, which is a remarkable electrical efficiency for a 2.5 kW electric generator. The additional stack efficiencies for all other cathode outlet temperatures and fuel utilizations listed in Table 5 are displayed in Appendix D at the end of this thesis. From previous fuel cell modeling efforts performed at the NFCRC, a spatially and temporally resolved fuel cell model was developed using the Matlab and Simulink interface environments for experimental verification. The Matlab environment was particularly chosen for modelling development efforts because of its versatility and widespread adoption in the engineering community. Simulink is a graphical block diagram environment for multi-domain simulation and model based design,cannabis dryer which is extremely useful for modeling, controlling and simulating dynamic systems. The SOFC dynamic model developed at the NFCRC and adapted for the purposes of experimental verification must inherently be accurate and sophisticated to achieve reasonable and verifiable results. The NFCRC fuel cell system model incorporates all the necessary components to analyze and assess the dynamic performance for both the SOFC and MCFC fuel cell types. Additional changes can be made to restructure the model to study other fuel cell types like PEMFCs; however, for the purposes of this thesis, the SOFC system model was utilized and adapted to accurately model the Engen-2500 experimental system.

Fabian Mueller, a previous NFCRC graduate researcher, was one of the first students to work on and develop the NFCRC dynamic fuel cell modeling tools. He incorporated modeling strategies from Rivera , Xue et al. , Roberts and Gemmen , Smugeresky, Roberts et al. , and Lim et al. , using control volumes to spatially discretize system components and apply dynamic conservation equations. The preliminary modeling strategies developed by these scholars were shown to accurately capture the dynamics of fuel cell systems. However, additional details were necessary to improve the precision of the system model. Mueller outlines a host of modelling assumptions that are important to consider for a simplified analysis and incorporated them into the NFCRC fuel cell model.An essential feature of the SOFC system model is the capability of spatially resolving each component for varying degrees of resolution, dependent on the desired precision. This capability provides the user with greater accuracy and precision than relying on bulk or equivalent circuit models, especially for scaling analyses. Keep in mind that increasing the model resolution increases the precision of the localized analysis at the expense of significant computing resources and additional time to complete the simulation. Therefore, the user must be aware of the computational power of his or her computer and select a discretization resolution that maximizes the precision of the analysis while minimizing the time to reach completion. Spatial resolution yields descriptive localized analyses of the internal temperature profile and heat transfer across an individual cell and each component of the SOFC system for a given moment in time during the dynamic analysis. For the fuel cell stack, spatial resolution is achieved by taking a single cell and dividing it into a grid of smaller elements, which are referred to as nodes. Each node is broken down further into five distinct segments that comprise a single cell: the oxidant separator plate, cathode gas stream, electrolyte , anode gas stream, and fuel separator plate. Figure 27 indicates the typical structure of one SOFC cell that includes the five segments mentioned.Spatially resolving each segment of a single cell permits the localized dynamic analysis of the conservation of mass, energy, and momentum equations while also locally evaluating the temperature, species mole fractions, pressure, and other required characteristics. The dynamic analysis of one cell is then scaled to the number of cells in the stack, ultimately representing the dynamic characteristics of the entire stack component. The same discretization method just described is applied to all components of the Engen-2500 SOFC system. The transport phenomena and electrochemical reactions evaluated at each locally resolved temperature, species mole fraction, and pressure, determine the performance of each component in the SOFC system. Constructing the system model requires integration of the multiple individual components, which have been resolved following the particular component physics, chemistry and electrochemistry. The complex interactions of the integrated components are captured through simultaneous solutions from the dynamic system analysis. For higher-temperature fuel cells used in combined heat and power systems, the fuel cell stack often appears to be quite a small and insignificant part of the whole system. The extra components required depend greatly on the type of fuel cell and the fuel used. In SOFC systems, fuel and air enter the SOFC stack and electricity, exhaust gas, and hot water or steam exit the system. The difference between an SOFC stack and an SOFC system is generally referred to as the “balance-of-plant” . BOP equipment may differ for each application depending on the size of the system, the operating pressure, and the fuel used. If an SOFC stack is to be tested for commercial applications, the test is usually performed in a complete system with balance-of-plant components included and the stack integrated into the system.

Our second analysis focuses on behavior of jobs to quantify the probability that a job will have to span more racks to find resources compared to the minimum number of racks the job can oc cupy based on its requested number of nodes. In addition, this way we capture the correlation of different resource types that are assigned to the same job. In particular, we sample 210 randomly chosen timestamps from our dataset. For each timestamp, we record which jobs are executing, their resource utilization, and their size in nodes. Because KNL jobs lack memory bandwidth measure ments, we focus only on Haswell jobs. For each job’s resource utilization metric, such as memory occupancy, we measure the maximum utilization among all nodes reserved to the job through out the job’s execution. We use a job’s maximum node utilization to account for the worst-case scenario. Then, we execute 16 of the following random experiments for each randomly chosen timestamp and report the maximumprobability across experiments. In each random experiment, we look at all jobs running at the chosen timestamp and assign them to nodes. Though job placement that prioritizes resource disaggregation would likely yield better results, we leave this as future work, since such job placement policies are still emerging. Therefore, in each experiment , we allocate racks of Haswell nodes and assign them to jobs in a random fashion. For each job, the number of nodes is the same as when it was running on Cori. For each randomly chosen rack, we reserve resources to cover the job’s utilization at the randomly chosen timestamp. If the job still has resource requirements remaining, then we continue by allocating another random rack. At the end of each experiment,cannabis growing equipment we record the percentage of jobs that had to allocate resources from more racks than the minimum number of racks they could be placed in based on their size in nodes.

These are the jobs that had to span more racks because of a lack of resources. This analysis is performed for each resource type separately. Results are shown in Figure 14 . Without reducing resources, there is still some worst-case probability for a job to span more racks than the minimum to allocate memory capacity because of unfavorable random job placement. However, the average across our random experiments re mains near zero . With a 20% reduction, the same worst-case probability becomes 11%, with a 50% reduction the probability becomes 22.3%, and with 80% reduction the probability becomes 56%. For NIC and memory bandwidth, for up to a reduction of 85% the probability is near zero. For a reduction of 95%, the probability for NIC bandwidth is 12.2% and for memory bandwidth 15.5%. While these results are sensitive to our assumptions and random choices of our algorithm, they indicate that intra-rack disaggregation suffices the majority of the time except when reducing resources aggressively.To illustrate potential benefits and thus further motivate intra-rack resource disaggregation, we use per-node statistics to calculate an average resource utilization of each rack. We do this for the node-to-rack mapping in Cori but also for a large set of randomized mappings of nodes to racks, because Cori’s scheduler is agnostic to re source disaggregation. This increases the computational complexity of this analysis, so we only perform it across four days of our data set. For each mapping, we use node statistics to calcu late per-rack utilization for each resource, by average across the four-day sampling period. We then take the maximum utilization for each resource among all the node-to-rack mappings and all racks within each mapping, to capture the worst-case rack utilization for each metric. We use that maximum to derive how much we could reduce in-rack resources and still satisfy the worst-case average rack utilization in our chosen four-day period. This does not indicate that no job will have to cross a rack boundary to find resources, or that application performance will be unaffected. Instead, this analysis focuses on resource utilization statistics. Memory band width reduction is based on the per-node theoretical maximum of 136 GB/s from the eight memory modules.

The percentage we can reduce each resource and still satisfy the worst-case average rack utilization is shown in Table 2. As shown, except for memory capacity in KNL racks, the other resources can be reduced substantially. Resource disaggregation-aware job scheduling may further improve these findings.Based on Section 4, there are also opportunities for rack-level disaggregation in ML workloads and GPU-accelerated systems. We observe a strong variability of resource requirements among different neural networks and therefore application domains but also among inference and training . Inference usually requires higher CPU-to-GPU ratios than training and uses less GPU memory. In contrast, training leads to high GPU utilization in terms of computation and memory but lower CPU utilization. However, this also varies with the workload. Job schedulers can use a disaggregated system by allocating more CPUs and less GPU memory for inference and give the remaining resources to a training job, which requires lots of memory and generally less CPU time. While we observe that GPU utilization is generally high, CPU resources and network bandwidth are underutilized most of the time. Although we cannot determine peak bandwidth demands with our methodology, we see that the average bandwidth utilization over longer periods is low. As we strong-scale training to multiple nodes, the bandwidth demands increase but other resources become less utilized, such as GPU memory . A disaggregated system allows us to provision unused GPU memory for other jobs, and since bandwidth depends on the scale, we can give more bandwidth to large-scale jobs and less bandwidth to small-scale jobs.Sampling a production system has significant value, because it demonstrates what users actually execute and how resources are used in practice. At the same time, sampling a production system inevitably has practical limitations. For instance, it requires privileged access and sampling infrastructure in place. In addition, even though Cori is a top 20 system that executes a wide array of open-science HPC applications, observations are affected by the set of applications and hardware configuration of the system. Therefore, HPC systems substantially different than Cori can use our analysis as a framework to repeat a similar study. In addition, sampling typically does not capture application executable files or input data sets.

Therefore, reconstructing our analysis in a system simulator is impossible but also impractical due to the vast slowdown of a simulator for large-scale simulations compared to a real system. Similarly, sampling a production system has no method to differentiate when application demands exceed available resources as well as the resulting slow down. For this reason and our 1s sampling period, our study focuses on sustained behavior and cannot make claims for the impact of resource disaggregation to application performance.While it is hard to speculate how important HPC applications will evolve over the next decade, we have witnessed little change in HPC fundamental algorithms during the previous decade. It is those fundamentals that currently cause imbalance that motivates resource disaggregation. Another consideration is application resource demands relative to available resources in future systems. For instance, if future applications require significantly more memory than the memory available per CPU today, then this may motivate full system disaggregation of memory, especially if there is significant variability across applications. Similarly, if a subset of future applications request non-volatile memory , then this may also motivate full system disaggregation of NVM, similar to how file system storage is disaggregated today.However,cannabis drying trays future systems may have larger racks or nodes with more resources, strengthening the case for intra-rack resource disaggregation. When it comes to specialized fixed-function accel erators, a key question is how much data transfer they require and how many applications can use them. This can help determine which fixed-function accelerators should be disaggregated within racks, hierarchically, or across the system. Different resources can be disaggregated at different ranges. Ultimately, the choice should be made for each resource type for a given mix of applications, following an analysis similar to our study.Future work should explore the performance and cost trade off when allocating resources to applications whose utilization is dynamic. For instance, providing enough memory bandwidth to satisfy only the application’s average demand is more likely to degrade the application’s performance, but increases average resource utilization. Future work should also consider the impact of resource dis aggregation to application performance, which should also consider the underlying hardware to implement resource disaggregation and the software stack. Job scheduling for heterogeneous HPC systems should be aware of job resource usage and disaggregation hardware limitations. For instance, scheduling all nodes of an application in the same rack benefits locality but also increases the probability that all nodes will stress the same resource, thus hurting resource disaggregation.

By the end of the 20th century, the most pervasive world-changing technology was the internet because of how it revolutionized the daily productivity of modern society. Mark Schueler, a Ph.D. student from Southampton University illustrates the explosive, “Growth of the Internet,” from its inception until more recent years in Figure 1, showing just how rapidly new technology can pique the public’s interest when it positively influences the majority. The World Economic Forum estimates about 2.5 billion people are connected to the internet today; a third of the world’s population. It is projected that 4 billion users will be connected by 2020, more than half the global population. With so much of the world’s population currently having little or no internet connectivity, this poses the question: Can the infrastructure that society counts on to carry all this digital traffic keep up with the accelerating demand? There is a growing need to produce the most computing power per square foot at the lowest possible cost of energy and resources. More recently, the electricity used by data centers has garnered the most intense interest, partly because of the importance of these facilities to the broader economy and because the power used by the individual data centers rivals some large industrial facilities. In a report to Congress from the United States Environmental Protection Agency discussing data center electricity use leading up to 2006 , it is quickly apparent how the electricity consumption nearly tripled from year 2000 to year 2006. Had people chosen to disregard improving energy efficiency and technology within U.S. data centers, then the domestic energy conversion would have nearly doubled by 2011. Fortunately, the data center industry has remained aware of the vast energy required to operate their infrastructure and has experimented with improving data center operation as recent data would suggest. In 2008, data center electricity demand was approximately 69 billion kWh or 1.8% of the total 2008 U.S. electricity sales. It comes as no surprise with how connected everyone is to the internet that nowadays data centers are likely to be increasingly large, powerful, energy-intensive, always running and out-of-sight. Because of the sheer size and numbers of servers involved, data centers are loaded with energy inefficiencies. Considering a traditional data center connected to the electric grid, less than 35% of the energy from the fuel source that is supplied to the power plant is delivered to the data center. The most significant inefficiencies result from power plant generation losses and transmission and distribution losses. Figure 3 outlines the process of power loss through the transportation and distribution network of the electric grid starting from the fuel source and ending with the power supplied to the consumer. It is evident that the largest inefficiency stems from the power generated at the power plant level with additional losses associated with the transmission and distribution to the data center, where the data center receives roughly 30% of the total energy that could have been supplied ideally from the fuel source. Considering the operations within a data center, there are further losses associated with the infrastructure required for daily reliable operation systems. The additional power consumed by the cooling, lighting, and energy storage, means approximately less than 17.5% of the energy supplied to the power plant is ultimately delivered to the servers.Recognizing the significant energy losses at the power plant level, Microsoft made a commitment in May 2012 to make their operations carbon neutral: to achieve net zero emissions for their data centers, software development labs, offices, and employee business air travel.

Fourth-generation bike sharing models may also incentivize user based redistribution by employing demand-based pricing where users receive a price reduction or credit for docking bicycles at empty docking locations. A third feature of fourth-generation systems is the seamless integration of bike sharing with public transportation and other alternative modes, such as taxis and car sharing via smart cards, which support numerous transportation modes on a single card. In 2009, the Yélo bike sharing system was launched in La Rochelle, France. This system includes a smart card, which is fully integrated with the public transportation system. This facilitates multi-modal transportation linkages and user convenience, which could lead to greater auto ownership and usage reductions, as more daily trips are supported by alternative modes. However, creating a program that coordinates various forms of transportation on a single card is challenging, as this can be costly and often requires multi-agency involvement. Another area for improvement is bicycle security, which can be supported by ongoing technological advancement, such as the design and integration of GPS units into more robust bicycle frames that further enhance existing locking mechanisms, deter bike theft, and facilitate bike recovery. However, adding GPS units is costly and can potentially increase financial losses, if bikes with built-in GPS are vandalized or stolen. Finally, to target a larger scope of bike sharing users, fourth-generation systems may be more likely to incorporate electric bicycles, which enable longer-distance trips; encourage cycling on steeper hills and slopes; and lessen physical exertion requirements, particularly when users are commuting or making work trips in business attire. Over the past century, California has built an extraordinarily complex water management system with hundreds of dams and a vast distribution network that spans the state. This system generates electricity, provides flood protection, delivers reliable water supplies to 40 million people and sup ports one of the most productive agricultural regions in the world. Yet development of the state’s water manage ment system has come at a price.

Damming waterways,microgreens grow rack diverting water from rivers and streams and altering natural flow patterns have transformed the state’s freshwater ecosystems, leading to habitat degradation, declines of freshwater species and loss of services that river ecosystems provide, including high-quality drinking water, fishing and recreational opportunities, and cultural and aesthetic values. The state aims to accommodate human water needs while maintaining sufficient stream flow for the environment. To support this mission, scientists from the U.S. Geological Survey , The Nature Conservancy and UC have developed new techniques and tools that are advancing sustainable water management in California. At the center of these new advances is the need to understand the natural ebbs and flows in the state’s rivers and streams. Natural patterns in stream flow are characterized by seasonal and annual variation in timing , magnitude , duration and frequency . California’s native freshwater species are highly adapted to these seasonally dynamic changes in stream flows. For example, salmon migration is triggered by pulses of stream flow that follow winter’s first storms, reproduction of foothill yellow-legged frogs is synchronized with the predictable spring snow melt in the Sierra Nevada, and many native fish breed on seasonally inundated flood plains, where juveniles take advantage of productive, slow-moving waters to feed and grow. When rivers are modified by dams, diversions and other activities, flows no longer behave in ways that support native species, contributing to population declines and ultimate extinction. Thus, understanding natural stream flow patterns and the role they play in supporting ecosystem health is an essential first step for developing management strategies that balance human and ecosystem needs. Unfortunately, our ability to assess alteration of natural stream flow patterns, and the ecosystem consequences, is hindered by the absence of stream flow data. California’s stream flow gauging network offers only a limited perspective on how much water is moving through our state’s rivers. In fact, it’s been estimated that 86% of California’s significant rivers and streams are poorly gauged and nearly half of the state’s historic gauges have been taken offline due to lack of funding . Of those gauges that are still in operation, most are located on rivers that are highly modified by human activities and gauge records prior to impacts are limited. These limitations can be partially overcome with modeling approaches to predict the attributes of natural stream flow expected in the absence of human influence.

The predictions can then be compared to measured stream flow at gauging locations, or they can be used to estimate natural flow conditions in ungauged streams. In 2010, Carlisle et al. developed a modeling technique to predict natural attributes of stream flow and assessed stream flow alteration at gauges throughout the United States . Soon after, UC and TNC scientists began using the approach to expand and further refine the technique for applications in California . The models have evolved over time, but all rely on stream flow monitoring data from USGS gauges located on streams with minimal influence from upstream human activities. These are referred to as reference gauges. Some reference gauge data come from historical measurements made before significant modification of flows occurred, such as the years prior to the building of a dam. The remaining data are from reference gauges located in California watersheds that remain minimally altered by human influence. Once reference gauges were identified and flow records obtained from the USGS web-based retrieval system, we used geographic information systems to characterize the watersheds above each reference gauge based on their physical attributes, such as topography, geology and soils . We also assembled monthly precipitation and temperature climate data for the past 65 years for each watershed. The watershed variables and climate data were then compiled and statistically evaluated in relation to observed flow conditions at the reference sites using a machine-learning approach that uses the power of modern computers to search for predictive relationships in large data sets. An advantage of machine-learning techniques is the ability to make predictions from multiple model iterations , which tends to increase accuracy. Once we had developed and evaluated models using observed stream flow data from reference gauges, we could predict stream flow attributes for any portion of a stream or river in California for which the climate and watershed characteristics were known . Additional technical details of the modeling approach are provided in Carlisle et al. 2016 and Zimmerman et al. 2018.In a study led by Zimmerman et al. , we applied the machine-learning technique to assess patterns of stream flow modification in California. We did this by predicting natural monthly flows at 540 streams throughout California with long-term USGS gauging stations and comparing those predictions with ob served conditions. We then assessed how observed flow conditions at the gauges deviated from predictions and recorded the frequency and degree to which flows were either higher or lower than natural expected levels, while considering the uncertainty of model predictions. We found evidence of widespread stream flow modification in California . The vast majority of sites experienced at least 1 month of modified flows over the past 20 years and many sites were modified most of the time .

When stream flows were modified, the magnitude of modification tended to be high. On average, inflated stream flows were 10 times higher than natural expected levels, whereas depleted stream flows were 20% of natural expected levels. Overall, stream flow modification in California reflects a loss of natural seasonal variability by shifting water from the wet season to the dry season and from wet areas of the state to the drier south. Stream flow inflation was most common in dry summer months and for annual minimum flows. Conversely, flow depletion was most common in winter and spring months and for annual maximum flows. Unaltered sites tended to occur in places with relatively low population density and water management infrastructure, such as the North Coast,ebb and flow flood table whereas greater magnitude and frequency of alteration was seen in rivers that feed the massive water infrastructure in the Central Valley and the poulated Central Coast and South Coast regions. A key water management goal in California is to manage river flows to support native freshwater biodiversity. By estimating natural river flows and the degree to which they are modified, our work provides a foundation for assessing “ecological flow” needs, or the river flows necessary to sustain ecological functions, species and habitats. Assessments of ecological flow needs are generally performed at stream reach to regional scales , but rarely for an area as large and geo graphically complex as California. In 2017, a technical team that includes scientists from UC, TNC, USGS, California Trout, Southern California Coastal Water Research Project and Utah State University began developing a statewide approach for assessing ecological flows. The team has identified a set of ecologically relevant stream flow attributes for California streams that reflect knowledge of specific flow requirements for key freshwater species and habitats . Our modeling technique is now being extended to predict natural expectations for these new stream flow attributes. Model predictions of the natural range of variability for these ecologically relevant stream flow attributes will provide the basis for setting initial ecological flow criteria for all streams and rivers in California by the State Water Resources Control Board and other natural resource agencies.

These ecological flow criteria will be based on unimpaired hydro logic conditions, but they can be refined in locations where management and ecological objectives require a more detailed approach. For example, refined approaches would likely be required in rivers that must be managed for species listed under the Endangered Species Act or in rivers where substantial flow and physical habitat alteration makes reference hydrology less relevant for setting ecological flow criteria, such as in the Central Valley or in populated watersheds of coastal California. Our technical team also was involved in establishing the California Environmental Flows Work group of the California Water Quality Monitoring Council . The mission of the Work group is to advance the science of ecological flows assessment and to provide guidance to natural resource management agencies charged with balancing environmental water needs with consumptive uses. The Work group is comprised of representatives from state and federal agencies, tribes, and nongovernmental organizations involved in the management of ecological flows. It serves as a forum to facilitate communication between science and policy development and to provide a common vision for the use of tools and science-based information to support decision-making in the evaluation of ecological flow needs and allocation of water for the environment. The modeling technique described above has also been used to evaluate statewide water allocations. Grantham and Viers analyzed California’s water rights database to evaluate where and to what extent water has been allocated to human uses relative to natural supplies. They calculated the maximum annual volume of water that could be legally diverted according to the face value of all appropriative water rights in the SWRCB’s water rights database. Water rights were distributed according to their location of diversion, and the permitted diversion volumes were aggregated at the watershed scale to estimate a maximum water demand for each of the state’s watersheds. These permitted water diversion volumes were compared with modeled predictions of average annual supplies to estimate the degree of appropriation of surface water resources throughout the state . The study found that appropriative water rights exceed average supplies in more than half of the state’s large river basins, including most of the major watersheds draining to the Central Valley, such as the Sacramento, Feather, Yuba, American, Mokelumne, Tuolumne, Merced and Kern rivers. In the San Joaquin River, appropriative water rights were eight times the volume of estimated natural water supplies . The volume of water rights allocations would be much higher if pre-1914 and riparian water rights had been included, but these data were not available at the time.For example, comparisons of allocations with water use suggest that in most of California only a fraction of claimed water is being used. In a well-functioning water rights system where allocations are closely tracked and verified, an excess of water rights relative to supplies is not necessarily a problem. During water shortages, holders of junior appropriative rights would be required to curtail their water use. When water is abundant, most water rights holders should be able to fully exercise their claims.

An additional issue related to product regulation of marijuana edibles is the high THC potency per package without adequate requirements that these products clearly be demarcated to explicitly communicate the actual size of an individual serving to the consumer.In Colorado and Washington, product regulations allow for each package to contain up to ten 10 mg servings of THC or 100 mg of THC. Poor product labeling in Colorado and Washington contributed to an increase in calls to poison control centers and self-reports of adult intoxication.In Colorado, marijuana-related calls to the poison control center increased from 44 in 2006 to 227 in 2015,while in Washington calls increased by 79% from 111 in 2010 to 199 in 2014.Since commercialization, calls increased by 55% from 129 in 2012 to 199 in 2014.Liquor control boards in charge of approving products prior to market release allowed fruit and candy flavored marijuana products to enter the legal markets in Washington and Colorado. Despite a rule that the Washington State Liquor and Cannabis Control Board not approve any marijuana-infused edible products “especially appealing to children” such as, but not limited to, “gummy candies, lollipops, cotton candy, or brightly colored products” did not block approval of fruit flavored sodas and candy, chocolate and peanut butter flavored cookies and brownies, and chocolate truffles,including Mirth Provisions’ Legal Sparkling Rainier Cherry Soda and Nana’s Secret Soda in Orange Cream and Peach Flavors.Colorado does not have even such nominal restrictions and similar products have entered the market, including Dixie Elixir’s Crispy Cracken and Chocolate Cherry Pretzel. Marijuana edibles may be a safer alternative for adult consumers than marijuana cigarettes because they avoid combustion. However, because edibles are being produced in a wide array of flavors and variations that often are appealing to children, it is questionable whether these products contribute to less harm. Avoiding these harms could be achieved through tight regulation,plant growing stand including low limits on potency, large warning labels, accurate labeling, standardization of dosing, and standardization of packaging to avoid accidental ingestion by children and adults.

There is concern that the high potency of these products as well as their appeal to children may result in adverse health consequences.Indeed, it is likely that such youth appealing products are a major contributor to an increase in accidental childhood ingestion since legalization in Colorado402 and Washington.Prior to legalization in Colorado and Washington there were few cases involving marijuana related accidental poisonings in children. Children admitted to the emergency room for accidental marijuana ingestion increased from 0 to 14 two years following liberalization of medical marijuana laws in Colorado.Following implementation of retail marijuana laws in Colorado in 2013, an additional 14 children were admitted to the hospital for ingestion of edibles,with the first 9 occurring in the first 6 months of legalization.Washington, which modeled its product labeling and potency rules on Colorado’s,experienced a similar increase in adverse outcomes. In 2014, 45% of calls to poison control center were related to marijuana intoxication for those under age twenty – since legalization in 2012, these calls have increased from 50 in 2012 to 90 in 2014. Significantly, the highest number of calls in 2015 were regarding children under the age of five.Of the calls reported for the first nine months of 2015, 51% were in the marijuana/cannabis category, 42% were associated with infused-products, and 7% were related to marijuana oil. Youth accounted for 43% of the statewide calls during this nine-month period in 2015. National data from the United States show similar trends for accidental childhood ingestion. Between 2005-2011 there was an annual 30% increase in marijuana exposure in medical marijuana states while non-medical marijuana states showed no increase.To address the issue of accidental consumption of marijuana edibles, Colorado and Oregon enacted legislation requiring marijuana producers to place a THC warning symbol on their products . Colorado, Washington, and Oregon developed mass media campaigns aimed at preventing youth marijuana use , not general market campaigns designed to minimize overall population use as is done for tobacco. These campaigns were targeted at youth with messages on health risks of impaired memory, developmental delays, increased risk for addiction, depression, anxiety, psychosis, or other mental illnesses.

Messages related to the consequences of marijuana use include ineligibility for receiving financial aid and how marijuana-related charges may lead to school suspensions and expulsions. State health departments public awareness messages in Colorado and Washington directed to adults only cautioned adults, particularly new users, to “be safe and sensible” when using newly legal marijuana rather than discouraging use altogether. Colorado contracted with the University of Colorado to evaluate the impact of its mass media campaign on change risk perceptions and use behaviors as well as increasing knowledge of marijuana laws, health impacts of use, safe storage practices, and prevention.Adult exposure to the 2015 “Good to Know Campaign” was associated with an increased likelihood of accurately identifying retail marijuana laws compared to adults with zero exposure, with the proportion adult acute awareness increasing from 62% to 73% at follow up. There were moderate effects on knowledge of harms associated with use and perceptions of risk related to underage use , use around children , and high risk use . The survey did not question respondents whether or not the campaign impacted use behavior or thoughts on quitting, intentions to quit, or quit attempts .Taxation can both be used to raise marijuana prices artificially in order to discourage consumption,and to prevent taxpayers from subsidizing the regulatory, public education, and prevention and control program, and the marijuana education and research program and adjusted periodically for inflation. Additional tax increases could be used as a way to raise the price to reduce marijuana initiation and promote cessation. While all four US states that had legalized recreational marijuana as of October 2016 and Uruguay tax marijuana, these tax rates were not set at levels designed to cover regulatory, public health education, and medical costs associated with marijuana legalization.In Colorado,Washington,and Oregon, state legislators were directed by the ballot initiatives voters enacted to adjust the retail sales tax to make retail marijuana competitive with black market prices. Washington and Oregon ballot initiatives also include criteria for adjusting marijuana taxes to discourage use, and an additional requirement in Oregon to maximize net revenue from the marijuana tax. In Uruguay, officials of the IRRCA have determined that marijuana will be taxed at $1 per gram to compete with black market prices, despite national legislation requiring that government officials develop and fund an enforcement system and education and prevention programs.

Shortly after legal sales went into effect, state legislators in Colorado,Washington,and Oregon reduced marijuana taxes to compete with the black market. Colorado reduced the retail sales tax from 10% in 2014 to 8% in 2015, while Washington consolidated the state’s three-tier tax system to a single ad valorem excise tax of 37% at the retail sales level to reduce the marijuana industry’s federal income tax liability because consumers would pay the tax and so would technically not be considered part of the retailers’ gross income. Oregon also modified its wholesale tax in 2015 to a price-based excise tax of 17% of the retail sale, with up to an additional 3% tax levied at the local level, to increase state revenue through increased sales stimulated by lower prices.In the three states where marijuana taxes were reduced, state legislators were more concerned with short-term gains of competing with the black market and maximizing state revenue than long-term public health impact and costs associated with reduced use through higher taxes. There are no requirements for marijuana to be taxed based on a percentage of tetrahydrocannabinol content, which may in effect provide incentive for manufacturers to increase the THC content of cannabis.Indeed, US marijuana producers have been increasing product potency over the last 20 years.Between 1995 and 2014, marijuana potency increased from 5% to 12%, with a corresponding decline in cannabinol. The result was a THC/CBD ratio increase from a factor of 14 in 1995 to a factor of 80 in 2014.In jurisdictions with legal marijuana sales, edibles and cannabis concentrates, where THC concentration can be as high as 70%,plant grow table has increased in recent years. A weight-based tax, or a tax based on the unit of THC per weight or volume could be a solution to this problem. Another policy worth considering from the alcohol control literature is implementation of minimum unit pricing .Evidence from Canada show that MUP for alcohol is associated with reduced consumption and alcohol-related harms.Longitudinal estimates from British Colombia suggest that a 10% increase in MUP for a given alcohol product would result in a 16.1% drop in consumption.As is the case with most parts of the new regulatory framework for marijuana, implementation of MUP for marijuana should be considered at the same time as legalization in order to avoid potential legal battles with a professionalized marijuana industry. In 2012, Scotland was the first country to pass national legislation requiring MUP for all alcohol products.However, implementation of MUP in Scotland has been met with fierce opposition from the drinks industry, with claims of MUP as a violation of European Union trade law.The US states took varied approaches to regulating restrictions on marijuana business locations, none of which protect those most likely to regularly use marijuana In Colorado, local governments were prohibited by state law from granting a license to a business within 1,000 feet of a school defined as “public or private preschool or a public or private elementary, middle, junior high, or high school or institution of higher education”, alcohol or drug treatment facility, principal campus of college, university or seminary, or a residential child care facility.Although Washington lawmakers prohibited marijuana businesses within 1,000 feet of K-12 schools, recreational center or facility, child care center, public park, public transit center, library, or any game arcade, it allowed local governments to pass rules to reduce the distance requirement to a minimum of 100 feet from areas where children and adolescents are likely to congregate.As of September 2016, four Washington cities reduced the buffer zone for marijuana businesses to 500 feet, and one city reduced its buffer zone to 100 feet for parks, recreational/community centers, libraries, childcare centers, game arcades, and public transit centers.Oregon lawmakers did not prohibit retail store locations within 1,000 feet of colleges or universities despite the fact that many college students are under 21. Retail stores in Alaska were prohibited under the legalization initiative within 500 feet of child-sensitive areas, defined as facilities that provide services for persons under 21, a building in which religious services are regularly conducted, or a correctional facility. Colleges and universities are not explicitly included. Retail outlet density is positively associated with youth and young adult smoking,heavy alcohol consumption,and marijuana use.Despite the fact that use is higher in areas where there are more retail outlets, marijuana regulatory regimes in the four US states have failed to implement licensing systems to control retail density in ways that would protect vulnerable populations . Similar to tobacco and alcohol outlets marijuana businesses appear to be concentrated in low-income, minority communities. By 2016, Colorado marijuana businesses were more likely to be located in census tracts with higher proportions of racial/ethnic minorities , lower proportion of young people, lower median household incomes , higher crime rates, and higher concentrations of alcohol outlets .Similar findings were observed in California neighborhoods with medical marijuana dispensaries.Research on US state implementation of retail marijuana laws has focused on potential impacts of these laws on risk perceptions,use,health harms and stakeholder participation in the regulatory process.There is only a limited literature on the impact of marijuana policies on use and associated harms from the experiences in the Netherlands,Uruguay,and the United States.However, variability in US state medical marijuana laws makes it difficult to make strong generalizations, which likely explains why there is no scientific consensus on how legalization will impact risk perceptions or use patterns.There is limited evidence on the complexities of how a policy is implemented and when it is implemented having a dramatic effect on health-related outcomes. It is important to consider perceptions of risk when assessing the public health impacts of marijuana legalization laws.

Exposure to alcohol advertising is independently associated with initiating drinking, drinking dependence, and binge drinking among young adults .Middle and high school students that own alcohol branded merchandise are more likely to report ever alcohol use. Ownership of alcohol branded merchandise is positively associated with youth perceptions on peer use and peer acceptance of alcohol.Nicotine cravings are enhanced by alcohol use and alcohol cravings are enhanced by nicotine use.Indeed, cigarette companies use imagery of alcohol use in their cigarette advertisements in print media, which disproportionately impacts young adults, particularly college students. Likewise, exposure to television food commercials is an important predictor for unhealthy food choice, brand preference, and high caloric food consumption.Receptivity to television fast food marketing is associated with youth obesity, with a one point increase in marketing receptivity being associated with a 19% increased odds of being obese.Electronic commerce such as internet, mail order, text messaging, and social media sales are difficult to regulate, leading to increased youth sales, tax evasion, and illicit trade compared to traditional tobacco sales.Although tobacco companies advertise on the internet, a substantial amount of tobacco promotion occurs through social media and user-generated promotional media, and the content is predominantly positive. These messages reach both adults and adolescents.In addition, internet sales have provided new avenues for tobacco companies to market their products to youth.A 2002 study that examined cigarette advertising on the Internet in the USA found that nearly 20% of cigarette-selling websites did not include warnings that sales to minors are illegal or prohibited. Among those websites that required some form of age-verification,vertical rack more than half required that a buyer confirm legal purchase age , 15% required that buyers manually type in their date of birth, and 7% required buyers manually enter information from a driver’s license.

At least fifteen US attorneys general have conducted Internet stings and found that children as young as 9 years old were able to purchase cigarettes. For example, a New York sting operation found that 93% of websites observed had sold to children under 18 . A 2004 study found that more than 96% of minors aged 15-16 were able to find an Internet cigarette vendor and place an order in less than 25 minutes, with most completing the order in seven minutes.A study in California found that 101 websites selling tobacco failed to comply with California laws regarding age and ID verification to prevent youth sales. A detailed 22-page summary of the scientific evidence through 2011 on tobacco sales through the Internet submitted to the US Food and Drug Administration to conclude that youth access to tobacco cannot be prevented by existing rules and procedures in the US, including those by which sellers conduct age verification were ineffective at preventing youth access to tobacco products.A 2013 report by the World Health Organization shows that 96 countries banned internet tobacco advertising,141 but enforcing such bans has proven difficult. For example, while the sale of snus is illegal in all European Union countries except Sweden, online vendors in Sweden target online marketing activities toward EU citizens outside of Sweden, including sales promotions, price discounts, and gifts with purchase. A study that made test purchases in ten EU member states reported a 96% success rate . Age-verification relied on self-reports from buyer, and the majority of these sales applied Swedish taxes only, contrary to EU requirements.According to smokers in Western countries, aside from television, the most common source of health information regarding the risks of smoking comes from tobacco product packaging.Indeed, evidence from the Global Adult Tobacco Survey shows that among 12 countries surveyed between 2008 and 2010, more than 90% of men had reported seeing the health warning label on cigarette packages.Large graphic warnings and plain packaging reduce tobacco use, discourage nonsmokers from initiating, and encourage smokers to quit. Large warnings specified in the WHO Framework Convention on Tobacco Control have spread across the world as countries have implemented the FCTC, with slower adoption of effective warning labels in countries that had previously entered into voluntary agreements with the tobacco industry.

The extent to which health warning labels on tobacco packages impact risk perceptions and smoking behavior largely depends upon the size, prominence, position, and design of these messages.Warning labels that cover up to at least 50% or more of principal display areas, and not just limited to the sides of the tobacco package,are associated with increases in health knowledge and motivation to quit. Experimental studies in Canada demonstrate that increasing the warning label from 50% to 75%, 90%, or 100% increased its effectiveness among youth.Studies evaluating graphic, pictorial warning labels in Canada and Australia have shown high levels of cognitive processing and an association between cognitive processing, intentions to quit, and quit attempts.In Brazil and Thailand, countries with strong pictorial depictions on the health impacts of smoking, had the strongest impact on thinking about quitting among current smokers.Nationally representative data from Canada demonstrate that 80% of youth reported pictorial health warning messages decreased the attractiveness of smoking.Compared to small, text-only warning labels, large warning labels that include images in addition to text are more effective at communicating health risks associated with use, evoking an emotional response, provoking thoughts about quitting, increasing motivations and quit attempts among smokers.National data from Canada show that 95% of youth rated pictorial health warnings as more effective at communicating health risks than text-only versions.Large pictorial warnings have longer lasting effects on increasing risk perceptions, encouraging quitting and quit attempts among smokers,and are more likely to be seen by low-literacy adults and children.In contrast, small, text-only warning labels, such as those used for tobacco in the United States, have low impact on youth tobacco use.In addition, these warning labels do not effectively communicate health messages on the specific health risks of tobacco consumption to the public.Young people are less likely to recall seeing text-only warning labels.Among participants that report text-only warning label recall, only one-third were able to accurately recall message content.Additional requirements for effective warning labels include positioning health messages on front and back, and on the top of all principal display areas. Warning labels on tobacco packages are more effective when novel health warnings and messages are used, and the content, layout, and design of the warning label are rotated periodically to avoid “burn out” of stale messages.

While youth perceive health messages on US warning labels for tobacco products to be believable, 186 few reported that these messages were informative or relevant, and that these messages were “vague”, “stale”, and “worn-out”.Warning labels that include messaging with cessation information and a toll-free quitline number are associated with an increase in calls to the quitline,particularly among male smokers and those from low socioeconomic groups,and help to address tobacco-related health disparities. Implementation of comprehensive warning labels for tobacco packaging has been actively opposed by tobacco industry interference in the policy process.Between 1984 and 2003, countries without mandated HWL on tobacco packages transitioned to having either some form of HWL or a voluntary industry HWL passed by the tobacco companies. Countries with voluntary industry HWLs were less likely to adopt comprehensive HWLs, which were compliant with FCTC guidelines than countries with previously enacted mandated HWLs.These findings also point to the importance of implementing at the time of legalization a comprehensive set of demand reduction policies for marijuana before a large marijuana industry develops and works to weaken or defeat public health strategies to control use.Cigarette pack design is a key component to tobacco company marketing techniques.Package design establishes brand identity and promotes brand appeal,seedling grow rack particularly among youth. Tobacco companies design products that are attractive to children while being marketed toward young adult peers. A longitudinal study on youth attitudes toward cigarette brands found a ten-point increase in the proportion of teenage girls reporting a favorite cigarette brand between 2007 and 2008. The study coincided with the launch of RJ Reynold’s campaign for Camel No. 9, a brand that appears to be specifically designed to attract teenage girls, and which accounted for the majority of the increase in brand preference.116 Similar impacts on brand preference were found among young people in Mexico191 that had reported a greater exposure to tobacco marketing and advertising. Tobacco companies use package design techniques to mislead consumers into perceiving their products as less harmful or safer than other tobacco products. Tobacco product packaging with descriptors such as “natural”, “light”, “mild”, and “organic” are associated with false beliefs of the health risks of smoking,and are perceived as less harmful or healthier than tobacco products without these descriptors,likely because the tobacco companies target concerned195 and older smokers at risk of quitting. Indeed, the cigarette companies consider the color of the package as an “ingredient” of the cigarettes that can be used to manipulate users’ perception of the taste of the product in ways interchangeable with changes in the physical product itself.The effectiveness of health warnings may be enhanced through the use of standardized packaging ,a strategy used to reduce attractiveness and appeal of tobacco, to increase the prominence of health warnings,and to correct misperceptions on the health risks of smoking. Plain packaging enhances the effectiveness of health warnings by increasing their notice ability, and has been shown to make smoking less appealing and has the potential to reduce the level of false beliefs about the risks of different brands. Compared to branded packages, tobacco products in standardized packaging are associated with reduced brand awareness and identification,and reduced brand appeal,particularly among young women.

Consistent with previous research in high-income countries, plain packaging in low and middle-income countries have similar impacts on reducing tobacco product appeal.Consistent with adopting a comprehensive tobacco control approach, plain packaging may be useful even if nations have adequately funded mass media campaigns . Unlike media campaigns, packaging changes have almost universal reach and ongoing frequency of exposure. Packaging changes cost little to governments, unlike media campaigns that constantly have to justify their funding allocations against ongoing efforts by tobacco companies to defund media campaigns.As discussed in detail in the next section, plain packs with larger graphic health warning labels complement media campaign messages, amplifying their impact. There is broad scientific consensus that mass media campaigns aimed at the general population are an important element of a comprehensive program to prevent youth initiation of tobacco use and reduce its prevalence.US Surgeon General Report concludes that there is sufficient evidence to infer a causal relationship between the level of funding for anti-smoking media campaigns and reduced smoking prevalence among youth.The effectiveness of well-done anti-tobacco media campaigns is not an argument against other elements of a comprehensive tobacco control policy. Indeed, media campaigns can amplify the effects of other policies, such as plain packaging, advertising restrictions, graphic warning labels and smoke free laws, as well as the other way around, since marketing prohibitions reduce the salience of pro-smoking cues, and increase and reinforce anti-smoking norms. In particular, in Australia, introduction of pictorial health warnings on cigarette packets was supported by a televised media campaign highlighting illnesses featured in two of the warning labels .Between 2005 and 2006, the proportion of smokers aware that gangrene is caused by smoking increased by 11.2 percentage points , and awareness of the link between smoking and mouth cancer increased by 6.6 percentage points . In contrast, awareness of throat cancer decreased by 4.3 percentage points, and this illness was mentioned in the pack warnings but not the advertisements. Smokers who had prior exposure to the warnings were significantly more likely to report positive responses to the advertisements and stronger post-exposure quitting intentions. Thus, anti-smoking television advertisements and pictorial health warnings on cigarette packets reinforced each other to positively influence awareness of the health consequences of smoking and motivation to quit. Analysis of the impact of tobacco control policies and mass media campaigns on smoking prevalence in Australian adults found that stronger smoke free laws, tobacco price increases and greater exposure to mass media campaigns combined to independently explain 76% of the decrease in smoking prevalence from February 2002 to June 2011.For example, youth exposure to anti-tobacco media campaigns reduced the odds of current cigarette use by 15% and smokeless use by 30% compared to students with zero media exposure.

Regarding ketene that has been suggested to be formed by vaping or pyrolytic heating of VEA,it is not clear whether it is identifiable with our methods or is not formed at the temperatures tested here. Products like duroquinone and durohydroquinone are reported to be formed below a vaping coil temperature of 300 °C; however, we do not observe them with the preparation or detection methods used in this work. The selectivity and solubility of GC-MS extraction solvent could be a reason why products like quinones were not observed in the current study. These results underscore the fact that THC oil is a complex mixture, the complexity of which increases with thermal degradation chemistry and the addition of VEA. Further research on individual components is still needed for a better understanding of aerosol composition from vaping cannabis extracts and their mixtures with diluents.Cannabis is the most commonly used federally illicit drug by U.S. adolescents and young adults . One in five 12th grade students reported past-month use of cannabis and 7% reported daily use . In the last few years, there has been an evolving landscape of cannabis products on the market, including various combustible , blunts , vaporized , and edible products . Although smoking cannabis remains the most popular mode of consumption among AYAs , studies indicate a rise in use of non-smoked cannabis products , particularly vaporized cannabis , among AYAs. As cannabis use in this developmental period poses concerns of negative effects on brain development and mental function , preventing AYA cannabis use in all forms is of public health importance. Research shows that AYAs’ perceptions of cannabis is a major driver of use, with low perceived risks associated with initiation and continued use of cannabis . Concurrent with the rise in use, AYAs’ perceived risks of cannabis have steadily declined over the past decade . The expanding legalization of cannabis nationwide may increase acceptability and ease of access among AYAs . Despite AYAs reporting using various forms of cannabis ,grow rack the extant literature predominately focuses on their perceptions of health risks associated with combustible cannabis, or “marijuana” in general.

Lacking are studies concerning AYAs’ perceptions associated with non-combustible products and with blunts. In addition, prior research has asked about perceptions of more general outcomes rather than of specific risk or benefit outcomes . Such nuanced data on specific perceived risks and benefits regarding both health and social impacts across product types could inform the development of tailored messages and educational content for cannabis use prevention . To address these gaps, we analyzed cross-sectional data collected among 433 California AYAs during 2017–2018. According to the 2019 Youth Risk Behavior Survey, California adolescents had lower prevalence of ever and current use of cannabis compared to national estimates of adolescent cannabis use . However, as California is considered the largest legal cannabis market in the US after legalization occurred in January 2018 , cannabis-related data in this state provide important information since policies pioneered in California are often adopted by other states, and so may reflect and inform future use patterns and drivers. This study examined perceptions of not only risks but also benefits related to short-term and long-term use of different cannabis products . Given that blunts are a well-documented form of co-use of cannabis and tobacco among AYAs, and previous research showing different perceptions related to blunts and other combustible cannabis products , we examined perceptions related to blunts and other forms of combustible cannabis separately. We also compared perceptions between participants who had ever used cannabis and those who had never used cannabis. Understanding AYAs’ perceived risks and benefits across different types of cannabis products and by use status is critical to inform public health and education messaging strategies aimed at preventing and reducing use of all forms of cannabis. Means and standard deviations were computed for each perception item. We used generalized linear models to account for correlation of students’ responses clustered within school and unbalanced group sizes. Outcomes were continuous variables indicating perceived chance of having a given health or social risk if using cannabis.

For each perception outcome, we estimated a model comparing means of that outcome among four cannabis products , and another model of that outcome between ever and never cannabis users . All models were adjusted for age and sex. There were six pairwise comparisons among four cannabis products. The statistical significance across pairwise comparison tests was controlled by using the Tukey-Kramer method with p-values adjusted based on the studentized range distribution . All tests were two-tailed with a significance level of α less than 0.05. Analyses were conducted using SAS v9.4. Among the risks assessed , the most common perceived short-term and long-term risks were “get into trouble” and “become addicted,” respectively. Across the cannabis products, perceived percent chance of experiencing the short- and long-term risks was highest for blunts and combustible cannabis, followed by vaporized cannabis, and the lowest for edible cannabis. Almost all of the pairwise comparisons of perceived health risks among the cannabis products were significantly different. Comparisons between combustible cannabis and blunts were not significantly different for most of the perceived health risks. The only significant difference between the two products was that long-term use of blunts was perceived to have a higher likelihood of resulting in lung cancer and heart attack than long-term use of other combustible cannabis products. Combustible cannabis and blunts were perceived to have greater risks than vaporized cannabis for all the health outcomes, except for having a heart attack. Vaporized cannabis was perceived to have greater health risks than edibles, except for having heart attack. For perceived short-term risk of experiencing social problems if used cannabis, the only significant comparison was between blunts and edible cannabis, indicating that blunts were perceived as more likely than edible cannabis to lead to getting into trouble and having friends upset. Compared to never users, ever cannabis users perceived less risk of getting lung cancer and experiencing social problems for the four cannabis products. Likewise, perceived risk of getting oral cancer was lower among ever users for all types of cannabis products, except combustible cannabis. For edible cannabis, ever users perceived lower short-term health risks and addiction than never users. Likewise, ever users perceived lower risk of having trouble catching breath and getting lung disease for vaporized cannabis, and lower risk of becoming addicted for combustible cannabis.Among the benefits assessed , the most common perceived short-term and long term benefits from using cannabis were “feel high or buzzed” and “feel less anxious,” respectively.

For pairwise comparisons of short-term benefits, combustible cannabis was perceived as having greater benefits of all the short-term benefits than vaporized cannabis, and having greater benefits for short-term mental health outcomes than blunts and edibles. Participants thought that smoking combustible cannabis or blunts would be more likely to result in someone “looking cool” than using vaporized or edible cannabis. Participants perceived feeling high or buzzed from using edible cannabis more than from using vaporized cannabis. Regarding long-term benefits, combustible cannabis was perceived as reducing anxiety and depression better than all other cannabis products. Ever cannabis users perceived greater benefits of using cannabis on reducing mental health problems for using all types of cannabis products. Compared to never users, ever users also perceived greater benefits of having better concentration for all types of cannabis, except for vaporized cannabis. In addition, ever users perceived greater benefits of looking cool for combustible cannabis and blunts, while they perceived greater benefits of feeling high or buzzed for edible cannabis. While combustible cannabis and blunts remain the most common forms of AYA cannabis use, increasing rates of use of vaporized and edible cannabis have been observed . However,greenhouse grow tables most studies on perceptions of cannabis focus on combustible cannabis, or “marijuana” more generally. This study extends the literature by examining AYAs’ perceptions of short-term and long-term risks and benefits of cannabis use, with a focus on different types of cannabis products , and between ever and never users of cannabis. The main finding is that AYAs differently perceive risks and benefits across the four cannabis products, and ever cannabis users generally perceive lower risks and greater benefits of cannabis use than never users. Consistent with previous studies among AYAs , the most common cannabis products used in our sample were combustible cannabis and blunts. Interestingly, these products on average were perceived to have greater short- and long-term risks than vaporized and edible cannabis. However, it should be noted that ever cannabis users perceived less risks of using these products than never users. The paradox between perceived risk and actual use of combustible cannabis and blunts may be explained by another finding that AYAs also had greater perceived benefits related to these products. These findings indicate that AYAs’ use of combustible cannabis and blunts are based on a balance of both their perceived risks and perceived benefits of these products, highlighting the role of these sets of opposite perceptions in AYAs’ behavioral decision-making . Prevention and intervention efforts often focus on communicating health risks related to cannabis use rather than social risks , yet such health outcome-focused strategies may not always be effective for young populations , 2021. Indeed, among the perceptions of risks assessed in our study, the most common perceived short-term and long-term risks were social negative outcomes and addiction, respectively. Likewise, a previous qualitative study found that adolescents expressed a concern about addiction and social impacts of cannabis use . This finding suggests that, along with communicating about health risks, increasing AYAs’ awareness of social risks and the addictive nature of cannabis use may be an additional focus to deter AYA cannabis use, as has been shown with other research including both social and health risks for tobacco prevention .

In addition, our study did not identify whether our participants had experienced any health or social risks related to cannabis. Future research should elucidate actual consequences of specific perceived risks on cannabis use among AYAs. In addition to considering perceived social risks in research and prevention efforts, the perceived benefit construct is also part of many health behavior models , yet rarely included in studies on cannabis use perceptions. We found the most common perceived long-term benefit among AYAs was “feel less anxious.” Beyond recreational purposes, AYAs report using cannabis as self medication to cope with their anxiety and other mental health issues . Our finding that AYAs perceived cannabis use as beneficial to their mental health is particularly important given that recent national data report a concerning trend in poor mental health in this age group . Furthermore, recent reviews indicate that AYA cannabis use is associated with poorer outcomes among those with mood and anxiety disorders and increased risk for developing major depression and suicidality . Our finding highlights a need for correcting AYAs’ misperception regarding benefits of cannabis use and educating AYAs on mental health risks related to use of cannabis. In addition, including screening for mental health problems into routine clinical care and integrating resilience training in prevention programs may offer AYAs better ways to cope with their mental health issues , which in turn may prevent their cannabis use. We also found that AYAs perceive use of edible cannabis as the least harmful to their health compared to other types of cannabis use. Legalization of recreational cannabis use and permission of home cultivation have been associated with a higher likelihood of AYAs using edible cannabis . This changing policy landscape coupled with low perceived risks of edibles call for more attention on preventing use of this product among AYAs. Collectively, our study has implications for public health efforts to prevent cannabis use and its negative health effects on AYAs. Mass media campaigns and educational programs should address both perceived risks and benefits of cannabis use and consider all types of cannabis products. Effective communication strategies may be those that increase perceptions of both health and social risks and correct misperceptions of mental health benefits related to cannabis use. As AYAs’ perceptions differ by cannabis product type, messages should be tailored to specific cannabis products, especially non-combustible products, rather than focus on cannabis use generally. For example, contents on vaporized cannabis could highlight risk of vaping-related lung injuries, while those on edible cannabis could highlight risk for over-consumption and intoxication. This study has several limitations.

A study of Bicing bike sharing user activity in Barcelona found that the average difference in elevation between origin and destination stations for e-bike sharing trips was +6.21 meters, compared to -3.11 meters for conventional bike sharing trips . Although bike sharing offers the opportunity to expand cycling mode share, the evidence from traditional bike sharing ridership suggests that bike sharing users are not socio-demographically representative of the broader population in areas they operate. Existing studies of station-based bike sharing in North America have shown that bike sharing use is strongly correlated with certain user characteristics such as: gender, age, and race. Station-based bike sharing users tend to be younger and upper-to-middle income, with higher levels of educataional attainment than the general population . Station citing has been found to reflect the socio-demographic inbalances in bike sharing ridership, with one study of 42 U.S. bike sharing system reporting that the 60 percent of census tracts with greatest economic hardship contained less than 25 percent of bike sharing stations . Moreover, bike sharing station activity increases in locations with higher percentages of white residents and decreases in relation to older populations . A growing emphasis on transportation equity, particularly with respect to emerging mobiliy services, has motivated many agencies to incorporate equity-focused provisions in their shared micromobility programs . Common approaches to promote equity across station-based bike sharing systems have included offering discounted annual memberships to low income riders, citing stations based on equity reasons,commercial greenhouse supplies providing payment plan options and assistance in obtaining bank accounts, credit, and/or debit cards in order to lower access barriers to bike sharing .

Many cities have required that shared micromobility operators provide such options as a condition for obtaining an operating permit. However, additional barriers to shared micromobility use remain unaddressed. Shaheen et al. introduced the STEPS to Transportation Equity framework to evaluate transportation equity by recognizing the opportunities and limitations of Spatial, Temporal, Economic, Physiological, and Social elements . The STEPS framework can be used to evaluate whether a shared mobility system provides equitable transportation services by identifying specific barriers and opportunities within each category. In particular, spatial factors such as steep terrain and low population density may constrain bike sharing use in certain cities with these characteristics. Temporal factors, which pertain to travel time considerations of travel, may be an issue in cities where shared micromobility demand is unbalanced during peak hours, generating concerns about the reliability of available vehicles. Economic factors include both direct costs and indirect costs that may create hardship for particular groups of travelers. Physiological factors may have posed a serious limitation to bike sharing use that is reflected in the age distribution of riders, though there may be an opportunity to expand shared micromobility use for older and less physically active individuals through electric bike sharing and scooter sharing. Finally, social factors encompass social, cultural, safety, and language barriers that may inhibit an individual’s use of a particular service. Our study consists of three major analytical components: 1) a comparative analysis of bike sharing travel behavior, 2) a discrete choice analysis using a destination choice model, and 3) a geospatial suitability analysis based on the STEPS framework using the DCA coefficients. To inform our analysis, we employed two datasets from February 2018 of Ford GoBike and JUMP, composed of 77,841 docked, conventional pedal bike sharing trips and 24,270 dockless e-bike sharing trips that occurred in San Francisco. We note that February 2018 in San Francisco was slightly warmer than average and relatively dry, with 10 mm of precipitation compared to an average of 112 mm .

The high temperature and low precipitation may have resulted in greater observed ridership than would be expected during this time of year . The trip-level data include trip duration and start and end times. The origin and destination of a trip are docking stations for GoBike and census blocks in which the trip started and ended for JUMP. The age and membership status of GoBike users are also included for each trip. The datasets do not include further information regarding user identification, user characteristics, or the trajectories taken for each trip. Our analysis is thus constrained to the revealed preferences of unidentified, unlinked bike sharing users. Rather than perform a traditional discrete choice model in which individuals’ preferences for specific alternatives among a finite set of choices are modeled, we implemented a destination choice model . We modeled the decision to travel to a particular destination given that a trip originating in a particular location is made using a particular bike sharing service. We supplemented the trip-level data with: tract-level population, job count, employment rate, age, income, and gender distributions from the U.S. Census . From Open Street Map, we used the locations of bike lanes and public bike racks to determine the density of these facilities in each census tract in San Francisco . Finally, we queried the Google Directions and Elevations Application Programming Interfaces for estimates of travel distance, duration, and elevation gain along suggested bike routes for each bike sharing trip . Queries to the Google Directions API used the latitude and longitude of specified trip OD pairs to generate a suggested route that provide a path, estimated travel time, and distance for each query. These paths were then used to query the Google Elevations API for elevation samples at 100 meter intervals, which were used to estimate the total elevation gain of each trip. All unique OD pairs in the activity dataset were used in this querying process, as well as OD pairs for all alternative trips used in the DCA. Alternative GoBike trips included all possible OD pairs starting and ending at a GoBike station in San Francisco, while alternative JUMP trips were generated as the set of all actual origins of JUMP trips paired with the centroid of every census tract in San Francisco. We applied the results of the destination choice model and the STEPS framework in a suitability analysis, which is a geographic information system -based method for determining the ability of a system to meet a user’s needs . In our analysis, we examined the geospatial distribution of bike sharing suitability in San Francisco.

In the following sections, we detail the steps taken to process data, specify a destination choice model, and apply the model and the STEPS framework in a suitability analysis. In this study, observed bike sharing trip destinations are modeled as choices among a discrete set of alternative destinations. Although techniques exist to estimate continuous models , neither the GoBike nor JUMP datasets entail location data on a continuous scale. The GoBike OD locations are constrained to the discrete locations where GoBike stations exist, while the JUMP OD locations are classified by the census block in which the trip started or ended for the purpose of privacy protection. With such discrete spatial data, we took the approach of aggregating trip OD pairs to the census tract level for two reasons to: 1) avoid high correlation between very close OD pairs and 2) simplify the model analysis. Aggregating the data by census tracts also allows for the inclusion of additional attributes to the model such as: demographics, employment rate, job density, and population density,cannabis dry rack all of which can be measured at the census tract level. With aggregation of the data to the census tract level, we note a major limitation in the computability of a model with as many alternatives as there are census tracts in the coverage areas of the two SF bike sharing systems. Forty-six census tracts are serviced by the Ford GoBike system, and 192 census tracts are serviced by JUMP. Discrete choice models generally include Alternative Specific Constants that aim to capture the biases toward each alternative that is not explicitly explained by the other model attributes. To avoid overfitting and aid in the interpretability of our model, we reduced the number of ASCs by clustering the census tracts based on their attributes. We included just one ASC in the model for each of the k clusters, making reasonable assumptions that clustered alternatives have similar unexplained bias. Several techniques can be applied to solve this unsupervised clustering problem. We considered three commonly used techniques for clustering: 1) DBSCAN, 2) Gaussian Mixture Models , and 3) k-means . We decided to work with k-means as it offers two desirable properties: 1) clusters tend to have similar sizes, and 2) clusters are grouped around a centroid. The last property suited our objective of having an average ASC for the entire cluster. K-means is a distance-based algorithm that requires preprocessing of the data to avoid biases due to differences in scale. First, we apply standard normal scaling on every census-level attribute available in our data sets. As our final objective is to determine the relative likelihood of trips destined for a location, we performed a Cross Correlation Analysis between the attributes of a tract and the number of trips that end in the tract. This process produces a projection of the set of attributes so that the clustering analysis favors attributes with a strong correlation to ridership . Figure 2 presents the resulting clusters with an intuitive interpretation of each, based on our a-priori understanding of the neighborhoods they represent. For both systems, we computed elevation gain by summing all increases in elevation observed in the 100 meter intervals sampled.

A complete list of attributes included in the final model are found in Table 1. This model excludes some parameters that were found to be insignificant to the destination choices of bike sharing users. Among them, unemployment measures such as the unemployment ratio or employment to population ratio were not significant when accounting for the log number of jobs. We chose not to include trip cost and membership considerations, as they differed considerably across the GoBike and JUMP systems. GoBike members pay annual membership fees, resulting in variable per-trip costs for each member depending on the frequency with which they use the service. We also did not have information on which of the short-term pass options were used by nonmembers. Though start time has a tremendous impact on destination choice, this choice maker attribute can only be incorporated in the model by interacting it with other relevant features. We chose not to add this refinement for model simplicity. Finally, the distribution of race or ethnicity at trip destinations were found to be highly correlated with economic attributes of destinations thus were not included in the final model. For the JUMP system, we considered every tract in San Francisco County as an alternative, while for GoBike we constrained the choice set to the tracts that contain at least a GoBike station to account for trip feasibility given the service area at the time the data were collected. The sample sizes for each model amounted to 70,779 trips, with 45 alternatives for GoBike and 24,034 trips with 192 alternatives for JUMP. Including all trips and alternatives, our datasets exceeded our computational power to fit the models using the PyLogit Python package. We employed an ensemble method that combines several “weak learners” to divide the workload. In this case, a weak learner is a MNL model trained on a sample of choice experiments. For the GoBike model, each weak learner was trained on 500 choice experiments using all 45 alternatives. However, for JUMP considering all alternatives would result in keeping too few choice experiments. So, we chose to have an approach similar to those employed in stated preference surveys by restricting the number of alternatives for each choice experiment. To fit the JUMP model, we randomly sampled 110 alternatives to use for each weak learner with 500 choice experiments. We chose to use the bootstrap as it measures the variance in the parameters, indicating which parameters are not relevant in the model and can be removed. On the other hand, cross validation is more focused on assessing predictive power . Since we are more concerned with narrowing attributes to those that are most influential in destination choice rather than producing a model that predicts exactly where a bike sharing user will travel to, we considered the bootstrap a more appropriate method for this analysis. Estimating identical models separately on the two datasets required that we keep attributes that happened to be significant for one system but not for the other.