A Two-Tier Statistical Forecast Method for Agricultural and Resource Management Simulations

• Published in: Journal of applied meteorology (1988) 2008-06, Vol.47 (6), p.1573-1589
• Main Authors: Mauget, Steven A., Ko, Jonghan
• Format: Article
• Language: English
• Subjects: Agricultural and forest climatology and meteorology. Irrigation. Drainage; Agricultural and forest meteorology; Agricultural management; Agriculture; Agronomy. Soil science and plant productions; Best management practices; Biological and medical sciences; Climatic zones; Climatology, meteorology; Crop production; Earth, ocean, space; Economic forecasts; El Nino; Estimates; Exact sciences and technology; External geophysics; Forecasting models; Freshwater; Fundamental and applied biological sciences. Psychology; General agronomy. Plant production; Generalities. Techniques. Climatology. Meteorology. Climatic models of plant production; Indicators; Meteorological applications; Meteorology; Methods; Neural networks; Northwest; Planting; Precipitation; Profit forecasting; Profits; Resource management; Rivers; Sea surface temperature; Simulation; Skills; Southern Oscillation; Statistical methods; Statistical weather forecasting; Triticum aestivum; Valleys; Watershed management; Weather analysis and prediction; Weather forecasting; Winter; Winter wheat
• ISSN: 1558-8424; 0894-8763; 1558-8432; 1520-0450
• DOI: 10.1175/2007JAMC1749.1

Simple phase schemes to predict seasonal climate based on leading ENSO indicators can be used to estimate the value of forecast information in agriculture and watershed management, but may be limited in predictive skill. Here, a simple two-tier statistical method is used to hindcast seasonal precipitation over the continental United States, and the resulting skill is compared with that of ENSO phase systems based on Niño-3 sea surface temperature anomaly (SSTA) and Southern Oscillation index (SOI) persistence. The two-tier approach first predicts Niño-3 winter season SSTA, and then converts those predictions to categorical precipitation hindcasts via a simple phase translation process. The hindcasting problem used to make these comparisons is relevant to winter wheat production over the central United States. Thus, given the state of seasonal SOI and Niño-3 indicators defined before August, the goal is to predict the tercile category of the following November–March precipitation. Generally, it was found that the methods based on either predicted or persisted winter Niño-3 conditions were skillful over areas where ENSO affects U.S. winter precipitation—that is, the Southeast and the Gulf Coast, Texas, the southern and central plains, the Southwest, Northwest, and the Ohio River valley—and that the two-tier approach based on predicted Niño-3 conditions was more likely to provide the best skill. Skill based on SOI persistence was generally lower over many of those regions and was insignificant over broad parts of the central and southwest United States, but did lead the other methods over the Ohio River valley and the northwest. A more restrictive test of leading hindcast skill showed that the skill advantages of the two-tier approach over the central and western United States were not substantial, and mainly highlighted SOI persistence’s lack of skill over the central United States and leading skill over the Ohio River valley. However, two-tier hindcasts based on neural-network-predicted Niño-3 SSTA were clearly more skillful than both ENSO phase methods over areas of the Southeast. It is suggested that the relative skill advantage of the two-tier approach may be due in part to the use of arbitrary thresholds in ENSO phase systems.