Results and discussion Directory UMM :Data Elmu:jurnal:A:Agricultural & Forest Meterology:Vol100.Issue4.Febr2000:

266 H.S.J. Hill et al. Agricultural and Forest Meteorology 100 2000 261–272 modeled. The value of the forecast information is dif- ference between the expected net returns with and without SOI-based forecasts V i p = π h,i p − π i p 4 where, V i p is the expected value additional net re- turns of the forecasting system for site i given price p . Obtaining the value of the forecast in this man- ner is consistent with previous value of information studies Hilton, 1981; Mjelde et al., 1997; Hill et al., 1998. For the information system to have value, the SOI-based forecasts must alter the optimal input com- bination relative to the prior knowledge case for at least one of the phases. That is, for either SOI-based method to have value, z k,i cannot equal x i for all k. Changes in input usage effect returns through wheat yield changes caused by adjustments in applied nitro- gen level and planting date and costs through changes in applied nitrogen level. Changes in returns and costs are reflected in the value of the forecasts. This process is repeated for each of the five price levels.

3. Results and discussion

The expected values of the 3P, 5P, and perfect fore- casts for spring and winter wheat producers vary by site and price Tables 3 and 4. In some locations, the SOI-based information is of no greater value than cli- matological information. As expected, at all sites the value of perfect information is greater than the value of either the 3P or 5P forecast method. Although changes in inputs usage caused by the use of ENSO-based fore- casts are not presented because of space considera- tions, the value of the forecasts additional net returns results from the changes in input use. 3.1. Winter wheat sites Use of either 3P method in Illinois provides no value to the producers Table 3. At the Ohio site, the CPC’s classification has no value and the −0.60.6 method has little value. For the remaining four sites, additional net returns range from US 0.55ha in Texas to US 4.77ha in Oklahoma at the lowest wheat price. At the highest wheat price, additional net returns range from US 0.04ha in Kansas to US 7.67ha in Okla- homa. Differences between sites are also noted in the Table 3 Value to US hard winter wheat producers USha of SOI-based climate forecasts Price a Illinois Kansas Ohio Oklahoma Texas Washington 3-phase Climate Prediction Center classification 1 0.00 2.03 0.00 2.75 1.01 1.25 2 0.00 1.09 0.00 3.58 1.22 1.55 3 0.00 0.69 0.00 3.84 1.06 1.48 4 0.00 0.55 0.00 4.17 0.91 1.26 5 0.00 0.48 0.00 4.42 0.97 1.35 3-phase −0.60.6 method 1 0.00 0.99 0.30 4.77 0.55 0.78 2 0.00 0.47 0.37 6.21 1.09 0.07 3 0.00 0.22 0.41 6.67 0.96 0.31 4 0.00 0.12 0.43 7.24 0.83 0.18 5 0.00 0.04 0.51 7.67 0.90 0.13 5-phase method 1 0.00 1.65 0.00 6.57 4.47 3.53 2 0.00 0.82 0.00 8.40 4.39 2.59 3 0.00 0.52 0.00 8.99 3.94 1.98 4 0.00 0.25 0.00 9.71 3.88 1.70 5 0.00 0.05 0.00 10.27 4.05 1.58 Perfect forecasts 1 8.96 23.19 12.51 35.11 19.33 25.79 2 10.98 24.55 15.32 43.25 21.80 29.22 3 11.92 24.81 16.59 45.90 22.14 30.61 4 12.51 25.20 17.38 49.20 22.57 31.03 5 14.64 25.54 20.23 51.72 22.90 31.48 a Because of differences in regional prices, the analysis uses dif- ferent prices for each region. For Kansas, Oklahoma, and Texas the five prices are US 91.97, 117.60, 125.84, 135.99, and 143.73kg. For Illinois and Ohio the prices are US 89.43, 110.49, 120.05, 125.95, and 147.03kg and for Washington the prices are US 97.30, 117.60, 132.50, 140.30, and 148.00kg. pattern of values over the range of prices. In Kansas, forecast value decreases as the wheat price increases, whereas in Oklahoma an opposite pattern is noted. For Texas and Washington, the forecast value increases as wheat price increases at the lower prices, whereas at the higher prices the forecast value declines as price increases. This result is consistent with previous find- ings on the determinants of information value. Hilton showed there is no monotonic relationship between the determinants of information value in this case wheat price and the value of information. In Ohio and Oklahoma, the −0.60.6 scheme pro- vides higher value than the CPC method. For the Ohio site the increases in net returns are small US 0.30–0.50ha, whereas at the Oklahoma site net H.S.J. Hill et al. Agricultural and Forest Meteorology 100 2000 261–272 267 Table 4 Value to Canadian and US hard red spring wheat producers USha of SOI-based climate forecasts Price a Man. b Sask. 1 Sask. 2 Alta. South Dakota North Dakota Montana 3-phase Climate Prediction Center classification 1 1.99 0.41 0.95 0.37 2.68 0.59 1.46 2 2.47 0.95 2.30 0.37 2.37 0.71 1.17 3 3.84 2.13 4.72 0.55 2.24 0.93 0.94 4 4.23 4.45 5.77 0.87 2.19 1.09 0.88 5 6.15 6.23 6.92 1.21 2.23 1.36 0.82 3-phase −0.60.6 method 1 1.94 0.03 0.00 0.00 0.07 0.11 1.34 2 2.66 0.14 1.02 0.00 0.00 0.00 0.90 3 3.69 0.61 1.68 0.07 0.00 0.00 0.80 4 4.63 2.00 2.20 0.00 0.00 0.00 0.88 5 6.38 2.41 2.70 0.00 0.13 0.00 1.06 5-phase method 1 5.02 0.00 0.59 0.64 4.95 1.19 0.91 2 5.64 0.84 2.14 0.97 4.72 1.02 0.25 3 7.35 2.13 3.96 0.71 5.08 1.14 0.25 4 7.99 3.81 4.93 0.94 5.44 1.42 0.43 5 9.78 4.47 6.07 1.58 5.82 1.78 0.61 Perfect knowledge 1 16.38 12.01 9.20 13.87 29.67 22.25 26.37 2 19.38 20.19 16.76 16.99 31.75 24.41 28.73 3 24.28 24.94 22.07 19.30 33.66 26.26 30.79 4 26.47 30.81 28.09 20.90 35.13 27.69 32.36 5 32.49 34.47 32.77 23.80 36.78 29.21 34.09 a Because of differences in regional prices, the analysis uses different prices for each region. Prices for the US sites are US 111.25, 131.88, 147.71, 159.17, and 171.46kg. For Alberta sites the prices are US 56.43, 77.03, 96.81, 115.42, and 132.09kg, for Saskatchewan the prices are US 54.12, 66.48, 86.05, 109.00, and 130.71kg, and for Manitoba US 61.32, 71.50, 87.90, 96.04, and 116.69kg. b Abbreviations are Carmen, Manitoba, Aneroid, Saskatchewan, Watson, Saskatchewan, Vermillion, Alberta. returns almost double between the CPC and −0.60.6 scheme. This result is unexpected, because the CPC uses information beyond October planting time to classify the years. It was expected such additional information would provide increase value. By using additional information, however, the classification maybe ignoring conditions early in the growing sea- son that may have a greater impact on yields than later conditions. At Kansas, Texas, and Washington sites, the CPC method provides higher value. Very similar patterns are observed with the 5P method. Illinois and Ohio producers obtain no value from using the 5P method. At the four remaining sites, additional net returns range from US 1.65ha in Kansas to US 6.57ha in Oklahoma at the low- est wheat price. At the highest price, additional net returns range from US 0.05ha in Kansas to US 10.27ha in Oklahoma. The value of the forecast across price follows patterns similar to those observed for the 3P method, except for Washington. At the Washington site, the value of the forecast decreases as wheat price increases. Depending on the 3P method selected, the site, and price, the value associated with the 5P method ranges from 0.1 to 12.1 times the value of the 3P forecasts ignoring sites and prices with an expected value of zero for the 3P andor 5P methods. Differences in the value of the forecasts are partially caused by the climatic variability experienced at each site and the strength of the association between climatic variability and the Southern Oscillation. Producers at all sites obtain value from the use of perfect climate forecasts. Depending on price and site, additional net returns associated with perfect forecasts range from US 8.96–51.72ha. Producers in Illinois and Ohio gain the least. Within any site, the value of perfect forecasts increases with price increase, a 268 H.S.J. Hill et al. Agricultural and Forest Meteorology 100 2000 261–272 pattern not consistent with the SOI-based forecast methods for some sites. The value of SOI-based forecasts is not uniform across regions. Development and use of SOI-based forecasts will potentially benefit some producers more than others. Income distributional questions these re- sults raise, however, are beyond the scope of this study. These distributional issues need to be addressed in a larger societal context. The value of the 3P and 5P methods within a region are not equal except for Illi- nois producers. With the exceptions of the Ohio site and the Kansas site at the lowest two prices, the 5P method yields a higher value than either 3P method. In Washington, Kansas, and Texas the 3P-CPC method captures 5, 9, and 5 of the value of perfect forecasts at the lowest price, whereas the −0.60.6 method captures 3–4. At the highest price, the 3P-CPC and −0.60.6 methods capture less than 4 of the value of perfect forecasts in Kansas, Texas, and Washington. In Oklahoma, the 3P-CPC method captures approximately 8, whereas the −0.60.6 method captures 14 of the value of perfect knowl- edge at both the highest and lowest price. In Illinois, no method captures any portion of the value of per- fect forecasts. The −0.60.6 method in Ohio captures approximately 2 of the value of perfect forecasts, whereas the CPC method captures none of the value. Except for the Midwest sites and Kansas, the 5P method captures a higher percentage of the value of perfect forecasts at all prices. Kansas producers obtain more value from the 3P-CPC method at low prices and the 5P method at higher prices. At the lowest price, the percentage of the value of perfect forecasts captured by the 5P method for Texas, Oklahoma, Washington, and Kansas are 23, 19, 14, and 7, whereas at the highest price, these percentages are 18, 19, 5, and 0. The comparatively high percentage of the value of perfect forecasts captured by the SOI-based methods suggests skill is present in the current SOI-based forecasts. Economic results mirror meteorological relationships between SOI events and weather pat- terns. The strongest relationships between the SOI and climate variability have been found in the South and Northwest Ropelewski and Halpert, 1986, 1987, 1989; Kiladis and Diaz, 1989. Previous meteoro- logical findings corroborate the results of this study which shows greater value to ENSO-based forecasts in Oklahoma, Texas, and Washington. Further, in Oklahoma, SOI-based forecasts may capture a large percentage of the perfect forecast’s value because Oklahoma lies near the boundary between two Pa- cific North American PNA atmospheric pressure patterns. The PNA is an important teleconnection pat- tern in the US Wallace and Gutzler, 1981. Further, it appears there are interactions between the PNA and SOI Nemanishen, 1998. Weaker signals have been found between ENSO and climate variability for the Midwest and Great Plains Ropelewski and Halpert, 1986, 1987, 1989. Weather, soil types, and wheat type also effect the value of the SOI-based forecasts. Illinois and Ohio are areas with fertile soils and more advantageous rainfall patterns than other wheat growing areas. This growing environment in combi- nation with the regions’ weak SOI signal appears to reduce the value of SOI-based and perfect knowledge forecasts. 3.2. Spring wheat results For the spring wheat sites, the relationship between 3P, 5P, and perfect forecasts is different Table 4 than for winter wheat. The 3P-CPC method provides more valuable forecasts than the −0.60.6 method at all sites, except the Manitoba site where the two meth- ods have nearly identical expected values. At three sites, Alberta, South Dakota, and North Dakota, the − 0.60.6 method has no value. For the sites showing positive value from using the 3P-CPC method at the lowest wheat price, additional net returns range from US 0.41ha for the Aneroid, Saskatchewan site to US 2.68ha in South Dakota. At the wheat highest price, additional net returns range from US 0.82ha in Montana to US 6.92ha in Watson, Saskatchewan. Further, at the Montana and South Dakota sites, the value of the forecast decreases as the producer’s expected price increases. At the remaining sites the converse is true. All of the sites placed value on the 5P method at all prices except the Aneroid, Saskatchewan site at the lowest wheat price. Additional net returns range from US 0.00–5.02ha at the lowest price. At the highest price level, additional net returns range be- tween US 0.61 and 9.78ha. For Manitoba and the two Saskatchewan sites, the value of the forecast in- creases as price increases. At the other sites, there is no H.S.J. Hill et al. Agricultural and Forest Meteorology 100 2000 261–272 269 clearly discernible pattern between the forecast value and the expected price level. The value of perfect forecasts increased consistently as wheat price increases. At the lowest price, addi- tional net returns associated with perfect forecast range between US 9.20 and 29.67ha. At the highest price, additional net returns range between US 23.80 and 36.78ha. The value of SOI-based climate forecasts for spring wheat production is not uniform across regions, again raising income distribution issues. All spring wheat producing sites benefit from using the 3P or 5P fore- cast methods, although for some sites, method, and prices, the benefit is small. With three exceptions, the two Saskatchewan sites and the Montana site, the 5P method provides greater value to spring wheat pro- ducers than the 3P method. With the 3P-CPC method, the sites with positive forecast value capture at the lowest price between 3 and 12 of the value of the perfect forecast. At the highest price, the 3P-CPC method captures be- tween 2 and 19 of the value of perfect forecasts. The −0.60.6 method captures between 0.2 and 12 of the value of a perfect forecast at the lowest price, whereas between 0.3 and 20 is captured at the high- est price. At the lowest price, the 5P method captures 3.4–22 of the value of perfect forecasts, whereas at the highest price, the 5P forecasts capture 1.8–30 of the perfect forecast information. For spring wheat, the 5P method ranges between 0.03 and 70 times more valuable than the 3P methods as before ignoring sites that have zero expected value. 3.3. Value comparison by phase The question remains, ‘does knowledge of every phase or does knowledge of only certain phases pro- vide value to the producer?’ Oklahoma and Washing- ton sites are used to address this question for win- ter wheat, while Manitoba and North Dakota sites are used as spring wheat examples. These sites provide different answers to this question. The reader is cau- tioned that the results are site specific and generaliza- tions are difficult. Differences in expected net returns using the 5P and 3P-CPC forecast methods for the Ok- lahoma and Washington sites are presented in Fig. 2 by phase for price level two. As noted earlier, these differences in expected net returns arise from changes in input usage. At the Oklahoma site, changes in in- put use and consequently changes in expected net re- turns occur in the other and cold phase when using the 3P-CPC method Fig. 2, although the majority of the value arises during the cold phase. Using the 5P method, input usage changes over the prior knowl- edge strategies in Phases 2 and 5. Phase 2 is similar to the cold event in the 3P-CPC, whereas Phase 5 cor- responds somewhat to the other event in the 3P-CPC method. In contrast, decision makers in Washington alter input use in all phases when using either the 3P-CPC or 5P forecasts. Phase 4 additional net returns is small, however. In the 3P-CPC method, the cold and warm phases provide the Manitoba and North Dakota producers with valuable information. Manitoba producers gain more from the cold phase, whereas North Dakota pro- ducers obtain more value from information concern- ing the warm phase. For the 5P method, Manitoba has additional net returns in Phases 1, 2, 3, and 4, whereas North Dakota producers’ net returns increase in Phases 1, 3, and 4.

4. Conclusions