Laserfiche WebLink
MEMO 89.2 <br />three times the amount of parcels added confirming the initial overestimation found in the accuracy <br />assessment. <br />Table 5. Analysis of Irrigated Parcels Based on Water User Information <br /> Number of <br />Parcels Number of Parcels <br />(Percenta e) Parcels Area <br />(Acres) Parcel Area <br />(Percenta e) <br />Final Irrigated Parcels 27,389 100.00% 1,017,130 100.00% <br />Added Irrigated Parcels <br />(Error of Omission) 1,369 5.00% 45,195 4.44% <br />Eliminated Irrigated <br />Parcels 1,685 6.15% 70,822 6.96% <br />(Error of Commission) <br />Total Parcel Changes <br />(Omission and 3,054 11.15% 116,017 11.41% <br />Commission) <br />2.7.2 Irrigated Crop Types Accuracy Assessment <br />As with the irrigated lands, the accuracy assessment of irrigated crop types used reference data collected <br />mostly from FSA (Section 2.5). After all reference data was acquired, fifty percent was selected for <br />testing in the accuracy assessment procedure using a randomization algorithm. These data were separated <br />and kept independent from the training data used in the classification. To ensure that a representative <br />sample was collected, reference data acquired from FSA was compared to 2001 Colorado Agricultural <br />Statistics (Table ~. From Table 6, with the exception of `Grass for Pasture and Hay' it is evident that all <br />major crop types in the reference sample have very similar proportions as compared to the Colorado <br />Agricultural Statistics. The reason for this is that the `Grass for Pasture and Hay' class was difficult to <br />find in FSA records because most FSA programs involve more valuable crops. On the other hand, as <br />mentioned in Section 2.4, RTi requested FSA personnel to perform database queries of crop types that <br />were difficult to find on a strictly random basis. Consequently, crop types, such as `Dry Beans' and <br />`Sugar Beets' had higher proportion in the reference data than in the Colorado Agricultural Statistics. <br />This was necessary to provide sufficient information on these crop types for the MLC classifier. <br />Table 6. Comparison Between the Proportion of Crop Types in the Colorado Agricultural Statistics 2001 and <br />the Test Sample Collected From FSA <br /> <br />Crop Type Proportion of Crops in Colorado <br />Agricultural Statistics for the SPDSS <br />Study Area Proportion of Crops in <br />Reference Test Sample <br />Collected from FSA <br />Small Grains 8.3% 10.5% <br />Dry Beans 3.6% 5.1% <br />Corn 43.8% 44.8% <br />Sugar Beets 3.3% 7.4% <br />Alfalfa 26.9% 28.5% <br />Grass for Pasture or <br />Ha <br />14.1% <br />3.7% <br />Total 100.0% 100.0% <br />Figure 17 is a diagram of the irrigated crops accuracy assessment procedure. Similar to the accuracy <br />assessment of irrigated lands, a quantitative evaluation of the classification was performed using the <br />reference data selected for testing. This evaluation consisted of an error matrix analysis and derived <br />accuracy assessment parameters, the overall accuracy, the producer's accuracy, the user's accuracy, and <br />the individual and overall Kappa coefficients (Jensen 1996, Congalton and Green 1999, Lillesand and <br />Page 25 of 45 ~R~versfde FecAnotogy, fnc. <br />4'JaYer Resources Errgi~ecr:np an~i CansuFlrnp <br />