Land IQ, LLC 20170508 2017.05.08 vector digital data https://www.wildlife.ca.gov/Data/BIOS ftp://ftp.wildlife.ca.gov/BDB/GIS/BIOS/Public_Datasets/2600_2699/ds2677.zip https://gis.water.ca.gov/app/CADWRLandUseViewer/ This dataset presents the 2014 agricultural land use, managed wetlands, and urban boundaries for all 58 counties in California. This data is prepared by Land IQ, LLC and provided to the California Department of Water Resources (DWR) and other resource agencies involved in work and planning efforts across the state for current land use information. Delineated from 2014 NAIP Imagery. The data are derived from a combination of remote sensing and agronomic analysis and ground verification. This dataset represents a statewide, comprehensive, field-scale assessment of agricultural land use, as well as urban and managed wetland boundaries for the 2014 year. This data is prepared by Land IQ, LLC and provided to the California Department of Water Resources (DWR) and other resource agencies involved in work and planning efforts across the state for current land use information. This dataset is meant to provide information for resource planning and assessments across multiple agencies and serves as a consistent base layer for a broad array of potential users and multiple end uses. The associated data are considered DWR enterprise GIS data, which meet all appropriate requirements of the DWR Spatial Data Standards, specifically the DWR Spatial Data Standard version 2.1, dated May 11, 2016. This data set was not produced by DWR. Data were originally developed and supplied by LandIQ, LLC, under contract to California Department of Water Resources. DWR makes no warranties or guarantees — either expressed or implied — as to the completeness, accuracy, or correctness of the data. DWR neither accepts nor assumes liability arising from or for any incorrect, incomplete, or misleading subject data. The official DWR GIS Steward for this data set is John Lance, who may be contacted at 530-528-7442, or at john.lance@water.ca.gov. Comments, problems, improvements, updates, or suggestions should be forwarded to the official GIS Steward as available and appropriate. A full report is available as submitted from LandIQ. Contact John Lance for a file copy. 20140101 20141231 Unknown Complete Biannually -124.469095 -113.499687 42.069558 32.325102 None irrigated land fallow boundaries agriculture 2014 California agriculture crop land use ISO 19115 Topic Categories boundaries environment farming None Central Valley State of California None 2014 None None Land IQ, LLC Joel Kimmelshue Owner mailing and physical

2020 L St. Suite 110

Sacramento CA 95811 916-265-6358 jkimmelshue@landiq.com Land IQ, LLC Joel Kimmelshue Owner mailing and physical

2020 L St. Suite 110

Sacramento CA 95811 916-265-6358 jkimmelshue@landiq.com N/A N/A N/A Land IQ, www.LandIQ.com Public domain Unclassified Available upon request Microsoft Windows 7 Version 6.1 (Build 7601) Service Pack 1; Esri ArcGIS 10.5.1.7333 N/A Data are considered logically consistent. Data are complete as of final delivery 5/11/2017. Positional accuracy for this mapping dataset was determined to be +/- 2.0 meters horizontal accuracy at a 95% confidence level when registered against NAIP reference imagery. The NAIP reference image has a reported positional accuracy of 6.0 meters. Therefore, the combined horizontal accuracy is 8.0 meters. Positional accuracy was evaluated by review of all polygon boundaries in the subject dataset against the NAIP reference dataset. Positional offset was measured in a randomly selected subset of approximately 10% of all fields. 8.0 meters John Lance 20170511 2014 None 20170511 publication date DWR contract, dataset steward Joel Kimmelshue, Land IQ, LLC, Owner 2014 onLine 20170511 publication date LandIQ producer Land IQ integrated crop production knowledge with detailed ground truth information and multiple satellite and aerial image resources to conduct remote sensing land use analysis at the field scale. The mapping approach employs advanced spatial statistical analysis approaches to determine prediction probabilities and inform QA/QC efforts. A rigorous QA/QC and analysis refinement process is employed to improve predictions on all lower probability fields. Individual fields (boundaries of homogeneous crop types representing true irrigated area, rather than legal parcel boundaries) are used so that each independent field could be analyzed independently and assigned to a crop class. The result represents the true irrigated area and not legal or other less detailed boundaries that may be available elsewhere. The classification legend was developed in coordination with DWR with consideration of the known crop variation, existing DWR legends used in current models, and Land IQ mapping classes. Two legend levels were selected in order to retain the detail in Land IQ’s base mapping while providing a rolled-up legend consistent with DWR’s classification that groups some crops into categories. The legends and crop classes can be related and cross-referenced and are summarized in Table 1. TABLE 1. 2014 STATEWIDE MAPPING LEGEND CROSS-REFERENCED Land IQ Class DWR Class Avocados Citrus/Subtropical Citrus Citrus/Subtropical Dates Citrus/Subtropical Kiwis Citrus/Subtropical Miscellaneous Subtropical Fruits Citrus/Subtropical Olives Citrus/Subtropical Almonds Deciduous Fruits and Nuts Apples Deciduous Fruits and Nuts Cherries Deciduous Fruits and Nuts Miscellaneous Deciduous Deciduous Fruits and Nuts Peaches/Nectarines Deciduous Fruits and Nuts Pears Deciduous Fruits and Nuts Pistachios Deciduous Fruits and Nuts Plums, Prunes and Apricots Deciduous Fruits and Nuts Pomegranates Deciduous Fruits and Nuts Walnuts Deciduous Fruits and Nuts Beans (Dry) Field Crops Corn, Sorghum and Sudan Field Crops Cotton Field Crops Miscellaneous Field Crops Field Crops Safflower Field Crops Sunflowers Field Crops Miscellaneous Grain and Hay Grain and Hay Wheat Grain and Hay Idle Idle Alfalfa and Alfalfa Mixtures Pasture Miscellaneous Grasses Pasture Mixed Pasture Pasture Rice Rice Wild Rice Rice Bush Berries Truck Crops Carrots Truck Crops Cole Crops Truck Crops Flowers, Nursery and Christmas Tree Farms Truck Crops Greenhouse Truck Crops Lettuce/Leafy Greens Truck Crops Melons, Squash and Cucumbers Truck Crops Miscellaneous Truck Crops Truck Crops Onions and Garlic Truck Crops Peppers Truck Crops Potatoes and Sweet Potatoes Truck Crops Strawberries Truck Crops Tomatoes Truck Crops Urban Urban Grapes Vineyards Young Perennials Young Perennial Managed Wetland Wetland 20170511 Land IQ, LLC Owner mailing

2020 L Street, Suite 110

Sacramento CA 95811 9162656330 jkimmelshue@landiq.com 2.1 DATA COLLECTION Both aerial and satellite data resources were used for the classification. Aerial imagery provided by the United States Department of Agriculture (USDA) National Agriculture Imagery Program (NAIP) was collected throughout the summer of 2014 by the USDA and used for field delineation, classification and QA/QC of the final product. Multiple Landsat 8 images were used for the initial crop classification. Imagery from the Landsat 8 satellite is free and available every 16 days and provided for temporal analysis throughout the growing season. Ground truth data were collected during the 2014 growing season prior to the initiation of this project. These data were used for training and validation of the mapping analysis (Table 2). Field data from over 15% of all irrigated land in the Central Valley was collected. This represented 32,698 data points and 56 crop classes. This dataset was split to provide for data training and to maintain a separate, independent validation dataset. 25% of the ground truth data were set aside for independent validation. Ground truth data was not collected outside of the Central Valley for this year. However, ground truth data collected in 2014 by the Bureau of Reclamation from Imperial County was used for the analysis of that specific county. Analysis in areas that lacked ground truth data was performed using imagery and classification approaches established in areas that were informed by training data. TABLE 2. TOP 25 CENTRAL VALLEY GROUND TRUTH DATA POINTS DISTRIBUTED AMONG CROP TYPE Crop Name Ground Truth Count Percent of Total Almonds 5121 15.7% Corn 3637 11.1% Grapes 2622 8.0% Alfalfa 2548 7.8% Walnuts 2238 6.8% Rice 2188 6.7% Fallow 2046 6.3% Citrus 1545 4.7% Irrigated Pasture 1380 4.2% Cotton 1229 3.8% Native 892 2.7% Tomatoes 775 2.4% Rangeland 771 2.4% Pistachios 701 2.1% Plums 677 2.1% Peaches/Nectarines 588 1.8% Wheat 497 1.5% Olives 471 1.4% Fallow Prep 351 1.1% Forage Grass 242 0.7% Beans 239 0.7% Cherries 219 0.7% Pomegranates 138 0.4% Melons 129 0.4% Others 1454 4.5% 20170511 Land IQ, LLC Owner mailing

2020 L Street, Suite 110

Sacramento California 95811 9162656330 jkimmelshue@landiq.com 2.3 ACCURACY ASSESSMENT After completion of the final classification dataset, a comprehensive accuracy assessment is completed. Independent ground truthing samples set aside for this purpose (25% of the ground truth data) are used in this process. A stratified random sampling method is used for accuracy assessment sample selection. The datasets are stratified by land cover type and county boundary. In the 2014 analysis, more than 6300 samples were selected for accuracy assessment. These sites are not used to train the classification algorithm and therefore represent unbiased reference information. 20170511 Land IQ, LLC Owner mailing

2020 L Street, Suite 110

Sacramento California 95811 9162656330 jkimmelshue@landiq.com 2.2 ANALYSIS The Land IQ land use mapping unit is a field-scale layer focused on agricultural production areas greater than 2 acres across the state. More than 300,000 delineated fields are classified utilizing ground training examples and multiple image sources and dates. These images and ground truth data are used to develop classification algorithms for crop identification. Multiple selected image sources and timeframes serve as input data for the remote sensing classification process, along with comprehensive ground truth training samples. Ground truth data are reviewed and evaluated statistically to identify any samples considered unrepresentative (crops that are very stressed or sparse, for example). These data points are flagged and not used for training samples. The ground truthing data is then stratified based on Land IQ’s classification schema, with 75% of the data selected for model building and calibration, and the remaining 25% dedicated to independent accuracy assessment, These independent data are set aside and are not used during any stage of modeling process. A supervised classification algorithm was applied to classify delineated fields. The supervised classification used a random forest approach and is carried out county by county where training samples are available. Random Forest approaches are currently some of the highest performing for data classification and regression. They are advantageous because of their ability to classify large amounts of data with high accuracy. Random Forest approaches have other advantages over some more traditional classification methods like maximum likelihood algorithms and Classification and Regression Tree (CART). Random Forest algorithms are non-parametric and require no assumption of input data being normally distributed, and that they are flexible and can incorporate categorical and continuous input data and complex relationships within the dataset. Multiple geoprocessing tools were employed to assess the model dataset, including ArcGIS, ERDAS Imagine, and other open source statistical tools. These tools are used to generate spectral characteristics, textural characteristics, and temporal representations that are related to the specific attributes of each crop or land use. The input features are produced using satellite imagery from Landsat 8 OLI/TIRS sensors and NAIP collected during the growing season. Additional satellite imagery and ancillary inputs were used in some counties to supplement and improve the classification. These additional sources include the United States Geological Survey (USGS) National Elevation Dataset (NED) and the USDA Crop Data Layer (CDL 2012-2014). Selected ground truth data and feature data are fed into the Random Forest algorithm for model building and calibration. A portion of these data are used for model calibration and the remainder is used for training Random Forest models. Multiple Random Forest models are assessed and compared to determine which is the highest performing for classification. The preferred Land IQ model is applied to all delineated fields to predict land cover type, as well as prediction confidence, which is used to inform QA/QC efforts. Classified fields with a lower confidence level are carefully reviewed by reviewing image resources using photo interpretation methods. Results are also cross-validated with ancillary data sources such as the coinciding USDA Crop Data Layer and county agricultural surveys and county crop reports, to assess and evaluate significant differences. Differences do not always indicate incorrect classification but are used both to evaluate the classification result and explain deviation from other data sources if any exists. The geospatial database is attributed with field size in acres, relevant county, and the appropriate crop classification category per the Land IQ and DWR legends (Table 1). Table 3 summarizes the database attributes (columns) associated with the final mapping product and their definitions. TABLE 3. DEFINITION OF DATABASE FIELDS Field Data Type Description Object ID Object ID Auto generated by ArcMAP Crop2014 String Crop classification type based on 2014 DWR/Land IQ standard Land Use Legend dated August 2016. County String Indicated the county the centroid of each field resides in. Acres Double Area of the agricultural field. Comments String Any user-provided information. Source String Original source of the boundary and attribute information Last_Modified_Date Date Date record was last modified. Modified_By String Name of person who last modified the record. Date_Data_Refers_To String Date the data refers to. 20170511 Land IQ, LLC Owner mailing

2020 L Street, Suite 110

Sacramento California 95811 9162656330 jkimmelshue@landiq.com Vector GT-polygon composed of chains 361023 NAD 1983 California Teale Albers 34.0 40.5 -120.0 0.0 0.0 -4000000.0 coordinate pair 0.0001 0.0001 meter D North American 1983 GRS 1980 6378137.0 298.257222101 ds2677 Source Original source of the boundary and attribute information DWR LandIQ, LLC Name of source LandIQ, LLC 20140101 20140101 None planned County Indicates the county that the centroid of each crop field resides in. Due to the size of many managed wetland and urban areas we did not attribute the county/counties for these features because some extended beyond a single county. Land IQ California Counties 20140101 20140101 As needed Shape Feature geometry. Esri Coordinates defining the features. 20140101 20140101 N/A GlobalID ESRI-defined ESRI ESRI-assigned 20140101 20140101 None planned DWR_Standard_Legend Used for symbolizing the data with DWR's standard legend. DWR Contains DWR's land use code from previous survey years and the full crop type name 20140101 20140101 Unknown Date_Data_Refers_To Date the data refers to DWR July, 2014 date of analysis LandIQ 20140701 20140701 None planned Comments Any user-provided comments DWR comments 20140101 20140101 None planned Acres Area of the agricultural field, urban area, or managed wetland Land IQ acreages 20140101 20140101 None planned OBJECTID Internal feature number. Esri Sequential unique whole numbers that are automatically generated. 20140101 20140101 As needed Last_Modified_Date Date record was last modified DWR 5/7/2017 date of last pre-delivery modification LandIQ 20170507 20170507 None planned Crop2014 Crop classification type for the year 2014 Land IQ explicit crop types each of the crop types are explicitly named LandIQ 20140101 20140101 None planned Modified_By Name of person who last modified the record DWR Zhongwu Wang Name of last modifier LandIQ 20140101 20140101 None planned Shape_Length Length of feature in internal units. Esri Positive real numbers that are automatically generated. Shape_Area Area of feature in internal units squared. Esri Positive real numbers that are automatically generated. N/A DWR California Department of Water Resources John Lance Data Steward mailing

2440 Main Street

Red Bluff CA 96080 530-528-7442 john.lance@water.ca.gov None 20171016 California Department of Water Resources John Lance Data Steward mailing and physical

2440 Main Street

Red Bluff CA 96080 530-528-7442 john.lance@water.ca.gov FGDC Content Standard for Digital Geospatial Metadata FGDC-STD-001-1998 local time