Agroecosystem research with big data and a modified scientific method using machine learning concepts. Issue 10 (31st October 2016)

• Record Type: Journal Article
• Authors: Moran, M. Susan; Heilman, Philip; Peters, Debra P. C.; Holifield Collins, Chandra
• Is Part Of: Ecosphere. Volume 7:Issue 10(2016)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2016-10-31
• Subjects: big data -- experimental watersheds, ranges, and forests -- Long‐term Agro‐ecosystem Research (LTAR) -- long‐term data -- network science -- regional‐scale analyses; Ecology -- Periodicals; Ecology; Periodicals; 577.05
• DOI: 10.1002/ecs2.1493
• Languages: English
• ISSNs: 2150-8925

Abstract

Abstract: Long‐term studies of agroecosystems distributed across the North American continent are providing an extraordinary understanding of regional environmental dynamics. The new Long‐Term Agro‐ecosystem Research (LTAR) network (organized in 2012) has designed an explicit cross‐site research program with multiple U.S. Department of Agriculture (USDA) experimental watersheds, ranges, and forests. Here, we report results from studies using a modified scientific method that includes learning through time that was implemented over the past five years with long‐term data from USDA experimental sites in coordination with other networks. The results offer a compelling argument for the LTAR concept of combining site‐based expertise with network‐wide coordination and collaboration to arrive at more accurate scientific conclusions than possible from individual researchers working alone. Simply put, without site‐based expertise and cross‐site communication working in parallel to provide input, feedback, and refinement to each subsequent step, similar to the way machine learning works, the interpretations and conclusions of these studies would have been incomplete, if not incorrect. Further, the up‐front time commitment to data processing and analytics above the time dedicated to place‐based studies increased the productivity of the team and the impact of the research, unlike the common perception that cross‐site research is often less efficient. In turn, this approach supported a non‐traditional system of credit for co‐authors based on citation impact of the journal selected as the publication outlet with less regard for author order. The LTAR network has embraced this modified scientific method in its shared research strategy and common experiment to address the problematic issues of mixed data quality across studies and sites, co‐author credit, research efficiency, and scientific impact on data‐intensive research. This approach can be combined with other types of collaborative and social media approaches, such as crowdsourcing, to take advantage of the wide range of expertise in the agroecological community as well as other disciplines to move science forward in the time of big data.