The strengths and limitations of vegetation maps for detecting conservation performance in napa county's natural vegetation and agricultural lands

We report the second edition of Napa County's vegetation map using California's Manual of Vegetation classification system and aerial imagery from 2016. It covers 2053.54 km2, increases polygon resolution by 23.8%, contains 35,244 polygons, and has 71 landcover types, 60 of which are dominated by natural vegetation. It has an overall error rate of 2.6%, includes canopy density measures for 24 vegetation types, and contains explicit measures of structures in the Wildland Urban Interface (WUI). We compared the new map to the one based on 1993 imagery to ask how effective Napa county's agricultural preservation and conservation regulations have been at preventing conversion of natural vegetation across the county and of agricultural lands within and beyond the county's Agricultural Preserve, established in 1968. We examined net change and change detected using a transition matrix, to seek concealed landcover conversion. The net change shows small increases in Agriculture and Urban extents. However, the transition matrix found substantial Urban expansion into Agriculture, 12.9 km2 overall and 226.9 ha (2.24%) within the Agricultural Preserve; and 30 km2 of agricultural expansion into natural vegetation, including 595 ha of high conservation-value oak woodlands. Due to increased map resolution in the 2016 map, we identified stands of natural vegetation (Oak Woodlands, Grassland, and Chaparral) in areas previously mapped as Agriculture and Urban, as well as Oak Woodlands in areas previously mapped as other vegetation types. Their inclusion in net measures of landcover change substantially obscures actual loss in natural vegetation types and agricultural lands. The net loss for Oak Woodlands is 191 ha, but the transition matrix identifies 782 ha converted to urban or agriculture, which suggests only moderate performance for the county's conservation measures. We also found rates of vegetation change in wildfire-burned areas three times higher than in unburned areas. Conversion of agricultural lands to urban was about 2.3% in the Agricultural Preserve, but 6.3% beyond, showing that county zoning slowed urban expansion. We found changes in spatial resolution and mapping details between vegetation map editions made net measures of landcover change inaccurate, and that transition matrix-analyses provided more accurate accounting and identification of mapping methods errors.