UC ANR internal program coordination. Additional information gathered from the questionnaire included a more in-depth description of UC ANR’s internal programs and activities. Thirteen of the 17 respondents indicated that their counties have an active Master Gardener Program, and 10 indicated that their master gardeners work with school or after-school garden programs or Farm to School programs. This internal program coordination was cited as an important factor for implementing successful school and after-school garden programs and Farm to School programs. These results suggest that the multidisciplinary and highly collaborative UC Cooperative Extension network has the potential to provide an important framework for successful school gardens, after-school gardens and Farm to School programs. Highlights from UCCE-evaluated programs are provided below.A unique UCCE program in Contra Costa County brings many young school aged children, especially those in grades 1 and 2, to an edible garden at the county fairgrounds. The site is also home to an agriculture museum. Approximately 1,700 students, 67 teachers and many parents visited the site during the 2010-2011 school year. Process evaluation, which documents and evaluates the development of a program from its inception, has demonstrated positive attitudes toward the program among teachers,stackable planters and results support the concept that teachers value the emphasis on local agriculture in the education process. However, these evaluations lack control groups of children who did not visit the edible garden, making it difficult to draw authoritative conclusions about the program’s success.
UCCE Contra Costa and Nevada counties collaborated to initiate the UC Sustainable Community Project, a federally funded Children, Youth and Families at Risk Sustainable Community Project that will begin participant enrollment in February 2012. A key element of the project is place-based learning, including at least one field trip to a farm. Both counties are partnering with master gardeners, and all intervention sites have gardens. The program will use the 4-H Teens as Teachers model to deliver the majority of the education to the younger participants . The short term goals of the program include improvement in youth knowledge about nutrition, gardening, agriculture, cooking and health; improvement in the ability to act on this knowledge; and improvement in physical fitness. The program leaders expect to provide participants with the skills to grow and cook their own food to support their personal health goals. As this is a nationally funded project, evaluation tools have already been developed, and a research team at Arizona State University will analyze pre- and post-intervention data. An exciting aspect of this project is that it supports the recent Institute of Medicine call for innovative techniques, integrating gardening and Farm to School programs with new technologies. For example, teens will use iPad 2 applications to identify and map safe routes to school and will share their findings by teaching children about walking and biking paths in their communities. Several education lessons will be delivered using accredited applications, and all data analysis will be collected with “clicker” technology, which uses wireless student response pads that allow instructors to instantly assess how well students understand the material presented.
In San Bernardino County, a team consisting of UCCE staff and academic personnel from the Fielding Graduate University Department of Psychology used a multidisciplinary approach to evaluate the impact of school gardens on nutrition knowledge and psychological parameters including attention and mood. Students in first- and second-grade classrooms were assessed pre- and post-intervention for nutrition knowledge using the Eating Healthy from Farm to Fork: Promoting School Wellness assessment tool. Teachers were trained to deliver this curriculum in its entirety and to use the 4-H gardening curriculum See Them Sprout. In addition, students spent 30 minutes in the garden each Friday. At the end of the 14-week semester, the post-test results showed a statistically significant increase in fresh fruit and vegetable knowledge. A unique aspect of this project was the attention given to the psychological impact of the school garden. Children worked in the garden for only one semester, allowing investigators to use a cross-over design to compare gardening and non-gardening children both within and between groups. Assessments of mood and attention were conducted before and after the 30-minute garden session and before and after the matched control non-gardening activity sessions each Friday over two semesters. The following semester, this procedure was repeated with the group assignments reversed. Assessments of self-efficacy and well-being were conducted with individual students, using longer measures at the beginning and end of each semester.
Results of this study are pending analysis. While randomized controlled interventions are needed, studies using an observational pre- and post-test design can still be highly informative, especially with respect to process evaluation. UCCE in Stanislaus and Merced counties has taken a leadership role in implementing Farm to School programs that reach over 3,000 children per year. Taste tests, teacher evaluations and teacher interviews were conducted to determine taste preferences and nutrition-related behavior changes in children participating in Farm to School programs.This is likely the result of prior exposure to school garden and Farm to School programs, as these have been operational for several years. Given these high baseline preferences, no improvements in children’s taste preferences were observed. While the finding that children participating in Farm to School programs prefer fruits and vegetables is encouraging, the information we gain is limited, reinforcing the need for randomized control studies. Without controls, it is impossible to conclude that the program being evaluated actually resulted in the measured outcomes. With controls, however, researchers can sort out any outcomes that might have happened by chance or simply as a result of other factors in the environment. Similarly, with randomization, researchers can ascertain whether outcomes were the result of one study site being more determined to make changes. The Shaping Healthy Choices Program uses a randomized controlled design to determine the outcomes of a multi-component nutrition education program on student health–related outcomes. Findings will help ascertain the impact of a coordinated comprehensive nutrition education program on students’ dietary behavior and health status.While UCCE has implemented and partially evaluated Farm to School and garden-enhanced nutrition education programs, it is important to integrate these strengths into a research and education program that incorporates the constructs of the socio-ecological model. Consistent with this, the ANR Healthy Families and Communities strategic plan addresses childhood obesity prevention with a multidisciplinary approach that involves a statewide network of researchers and educators creating, developing and applying knowledge in agricultural, natural and human resources. Funded by the ANR Competitive Grants Program, the research and extension project A Multi-Component, School-Based Approach to Supporting Regional Agriculture, Promoting Healthy Behaviors, and Reducing Childhood Obesity builds upon the multidisciplinary, comprehensive approach to investigate dietary and lifestyle habits with the greatest potential for sustainable childhood obesity prevention. This 4-year study will use the socio-ecological model to implement and measure the effectiveness of an integrated, school-based,stackable flower pots multi-component intervention. The long term goal of the Shaping Healthy Choices Program is to prevent childhood obesity by improving students’ diets and increasing physical activity. A collaborative research team will work with four schools in two counties to develop a system wide, sustainable program to achieve the following objectives: increase availability, consumption and enjoyment of fruits and vegetables, improve dietary patterns and increase physical activity consistent with the 2010 U.S. Dietary Guidelines for Americans, improve science-processing skills to sustain patterns learned and adopted through participating in the program, promote positive changes in the school environment to support dietary and exercise patterns and student health and facilitate the development of an infrastructure to sustain the program beyond the funding period.
To document student outcomes and environmental changes resulting from this multi-component, multidisciplinary approach to obesity prevention, a randomized, controlled, double-blind intervention will be implemented for one academic year through collaboration among faculty and staff from UC Davis, UC ANR, the Agricultural Sustainability Institute at UC Davis and the UC Davis Betty Irene Moore School of Nursing. The factors contributing to obesity are numerous and interrelated. Meeting the complex challenges of obesity prevention will require extensive and diverse collaboration with shared responsibility and common goals. The study will explore and document the effectiveness of an interdisciplinary team in developing comprehensive nutrition and lifestyle education programs that can be delivered throughout the state. In the future, these teams will include UC faculty; UCCE nutrition and youth development specialists and advisors, and Agricultural Sustainability Institute staff; food and agriculture industry representatives; public school educators, administrators, after school providers and families; community members; health practitioners; farmers; and state/county agency nutrition, food science, agriculture and health-care representatives — all developing coordinated programs that can be delivered throughout the state. Introduced in 2004 at the USENIX Symposium on Operating Systems Design and Implementation, the model is inspired from the functional programming construct ”map”. As such, MapReduce consists of the application of uniform functions ”map” and ”reduce” to a set of input elements sub-divided into multiple chunks to be processed by machines, part of a distributed computing system. The appeal of the model stems from the fact that it absolves the programmer from the burden of input management, parallelism, and synchronization constraints. MapReduce programs are written as single node programs by the user, and are subsequently parallelized by the framework. The details of input distribution, synchronization of necessary data structures, as well as handling machine failures, are all abstracted away from the user, and hidden in the paradigm itself. The MapReduce model in its most popular form, Hadoop, uses the Hadoop Distributed File System to serve as the Input/Output manager and fault-tolerance support system for the framework. The use of Hadoop, and of the HDFS is however not directly compatible with HPC environments such as NERSC, the NY state Grid, the Open Science Grid, and TeraGrid. This is so because Hadoop implicitly assumes dedicated resources with non-shared disks attached. At NERSC’s Magellan cluster, the system administrator has isolated a part of the large cluster for use with Hadoop. This isolation not only limits the resources available to MapReduce programs, but also produces performance penalties with MapReduce programs, as HDFS on top of an underlying file system introduces performance hampering layers of indirection. In contrast, applications using the rest of the cluster interact directly with GPFS. Furthermore, even though Hadoop has successfully worked at large scale for a myriad of applications , it is not suitable for scientific applications that rely on a POSIX compliant file systems in the grid/cloud setting. The HDFS is not POSIX compliant. In this paper, we investigate the use of Global Parallel File System, Network File System for large scale data support in a MapReduce context. For this purpose we picked three application groups, two of them, UrlRank and ”Distributed Grep”, from the Hadoop repository, and a third of our own: XML parsing of arrays of double, tested here under induced node failures, for fault tolerance testing purposes. The first two being provided with the Hadoop application package, have been shown to provide scalable performance with Hadoop. Even as we limit our evaluation to NFS and GPFS, our proposed design is compatible with a wide set of parallel and shared-block file systems, such as LUSTRE, pNFS, GFS2, and Oracle Cluster FS. We present the diverse design implications for a successful MapReduce framework in HPC contexts, and show the performance data collected from the evaluation of this approach to MapReduce along side Apache Hadoop at the National Energy Research Scientific Computing Center Magellan cluster.Hadoop uses the HDFS, inspired from the GFS for various background tasks such as input management, distribution, locality, output collection, performance but also fault tolerance. As a data manager, the HDFS is tasked with dividing the input among participating nodes in the cluster, keeping a myriad of accounting tallies, including chunk size, location, and duplication counts. The HDFS insures input distribution and rally in providing the user with an interface whose role is to provide bits of given data files to cluster nodes. Among its chief advantages, the Hadoop Distributed File System provides input locality by enabling nodes hosting input shards to apply their processing on such chunks, rather than on remotely stored data. This design provides significant performance benefits as the computation is brought to the data, rather than the data to the computation. In line with its data management role, the HDFS collects output data processed by nodes, and ”shuffles” them for the reducer to operate on them.