Laserfiche WebLink
MEMO 93 <br />4.2 Crop Type Classification <br />The crop type classification was also hierarchical for the 1987 and 1976 time period, using unsupervised <br />classifications, multi-temporal NDVI signature analysis and spectral signature analysis, GIS data (i.e. <br />parcel data), as well as other ancillary data in a combined MLC classification procedure (described in <br />Section 4.2.1.1 of this memo). In addition, ancillary data was used to refine the classification of <br />questionable crop signatures. The irrigated crop classification for 1987 and 1976 consisted of three steps: <br />(1) irrigated lands masking, (2) NDVI spectral signature analysis and maximum likelihood classification, <br />and (3) post-classification refinement (Figure ~. Crop types for 1950s irrigated parcels were adapted <br />from final crop types from 1976 as described in Section 4.2.1.3. <br />As in SPDSS Technical Memorandum 89.2, Section 2.5.2.1, a list of crop types to be classified for all <br />time periods was determined by (1) availability of 2001 crop types for comparison, (2) crops and acreage <br />reported by the CASS, (3) spectral characteristics of the crops, and (4) input from the CU contractor on <br />categories of crops with similar water requirements. Based on this information, the following crops or <br />crop groups were classified: Small Grains, Dry Beans, Corn and Sorghum, Sugar Beets, Alfalfa, and <br />Grass for Pasture and Hay. The crop types of corn and sorghum were combined due to similar spectral <br />signatures and relatively low proportion of sorghum within the SPDSS study area (less than one percent). <br />4.2.1 Irrigated Lands Masking <br />In the first step, non-irrigated areas were eliminated from the multitemporal NDVI image composite by <br />applying the irrigated lands mask generated as described in Section 4.1.1 of this memo. This procedure <br />eliminated areas associated with other land-cover types (e.g., dry-land crops, riparian vegetation, urban <br />areas, etc.). <br />4.2.1.1 Spectral Signature Analysis and MLC Classification <br />The second step began with an unsupervised classification of the masked NDVI composite into 50 <br />classes. Due to the irrigated lands masking in step one, each class of this 50-class unsupervised <br />classification corresponded to different varieties of the selected crop groups. <br />Page 12 of 59 ,Rereradde ~eshn~P~gy, ~os~ <br />