Laserfiche WebLink
MEMO 89.2 <br />November 4. The composite synthesizes the information contained in the individual NDVI images into a <br />single image. For example, parcels with high NDVI values on May 12, and low values in the remaining <br />two dates, such as small grains, appear in blue tones in the NDVI composite. Conversely, parcels with <br />high NDVI values in August, and low NDVI values in the other two dates, such as corn, appear in green <br />tones. This principle can be related to the mean temporal NDVI trend displayed in Figure 4, where mean <br />NDVI values for selected crop types are displayed over time. Figure 4 and Figure 5 are two different <br />ways to illustrate the development of crop patterns throughout the agricultural season and the relevance of <br />the NDVI enhancement and multitemporal analysis in the classification of crop types. <br />2.4 Reference Data Acquisition <br />As described in SPDSS Memorandum 89.1, RTi collected ground truth data from primary and secondary <br />sources. The primary sources were field observations conducted by RTi personnel of permanent cover <br />types such as rangeland, wetlands, deciduous trees, urban areas, etc., as well as semi-permanent cover <br />types like irrigated meadows and hay parcels, dry land agriculture, etc. Secondary sources included <br />reference data obtained from the Farm Service Agency (FSA), Colorado State University, and private <br />parties such as vegetable farms. <br />In preparation for reference data collection activities, RTi determined the definitive list of crop types, in <br />coordination with the Consumptive Use contractor. The following criteria was used in this determination: <br />(1) crops and acreage reported by the Colorado Agricultural Statistics, (2) spectral characteristics of the <br />crops obtained from expert knowledge and research literature, and (3) input from Consumptive Use <br />contractor on categories of crops with similar water requirements. A definitive list of crop types is shown <br />in Section 2.5.2 -Classification of Irrigated Crops. <br />Ground truth data was collected following standard statistical sampling procedures while taking practical <br />considerations into account. As recommended in the literature (Jensen 1996, Congalton and Green 1999), <br />these data included a minimum of 50 parcels for each major crop type. Special attention was given to <br />crop types that were difficult to find on a strictly random basis, such as vegetables and dry beans. In <br />addition to this, other reference data collected from FSA included dry land crops used in NDVI <br />calibrations applied in the preliminary irrigated lands classification, described below in Section 2.5.1. <br />Table 2 shows a list of reference data collected from FSA. <br />Page 7 of 45 ~RFversfde TecAnotogy, fnc. <br />4'JaYCr Resources Errgi~ecr:np an~i CansaFlrnp <br />