Skip to main content
ARS Home » Research » Publications at this Location » Publication #225477

Title: Three Analysis Examples for Time Series Data

item Meek, David
item Laird, David
item Dinnes, Dana
item Jaynes, Dan
item Cambardella, Cynthia
item Hatfield, Jerry
item Karlen, Douglas

Submitted to: American Society of Agronomy Abstracts
Publication Type: Abstract Only
Publication Acceptance Date: 10/9/2008
Publication Date: 10/9/2008
Citation: Meek, D.W., Laird, D.A., Dinnes, D.L., Jaynes, D.B., Cambardella, C.A., Colvin, T.S., Hatfield, J.L., Karlen, D.L. 2008. Three Analysis Examples for Time Series Data [CD-ROM]. In: ASA-CSSA-SSSA Annual Meeting Abstracts, Oct. 5-9, 2008, Houston, TX.

Interpretive Summary:

Technical Abstract: With improvements in instrumentation and the automation of data collection, plot level repeated measures and time series data are increasingly available to monitor and assess selected variables throughout the duration of an experiment or project. Records and metadata on variables of interest alone or desired covariates commonly result from such measures. Regression and time series methods can often be used to provide insightful analysis particular to each experiment and variable of interest. Three such examples are provided. The first is from a paired-in-time watershed experiment. Weekly observations of nitrate-nitrogen concentration, [NO3-N], from the sub-basin tile drain outlets are paired in time over a three year period for two similar sub-basins in an agricultural watershed. The nitrogen fertilizer in the treatment sub-basin was applied at a rate recommended by a best management practice (BMP), while that for the control was with the normal management strategy. A three-phase model of the paired difference was developed that both analytically and graphically yielded insightful results: there was initially a period of similarity, followed by a period of change, then it leveled out to a period with the NO3-N concentrations for the BMP treatment having systematically lower values. The next two examples come from a soil biochar amendment study conducted in 48 repacked soil columns placed in a climate controlled chamber. The design-structure was completely randomized with 6 replications for each of 8 treatments. The treatment-structure was one-way. A no manure or a manure addition treatment was considered in the absence or presence of biochar amendments (0, 5, 10, 20 g-biochar kg-soil-1). Weekly irrigation was applied to each column followed by measuring weekly total NO3-N loads in the leachate for 45 consecutive weeks. The NO3-N load was modeled for each column then treatment differences were examined. The no manure treatments were modeled with an isotherm curve form. The manure treatments were modeled with a time distribution form superimposed on the isotherm model beginning at the date of manure application. The second example is based on examining the magnitude and variability of the time distribution model parameters. The addition of manure greatly increased the magnitude of NO3-N load but with the presence of biochar this magnitude was reduced. The third example is based on examining the rate of change of the isotherm model; the level and time of maximum NO3-N load are first estimated and then compared. The maximum loads were reduced by adding biochar to the soil and associated times of occurrence were delayed.