trend is a web tool using D3 and Semantic UI for visualizing Gradient Nearest Neighbor (GNN) trends over time. GNN is a forest vegetation mapping method based on ordination (canonical correspondence analysis) and nearest-neighbor imputation.
We have run GNN models across Washington, Oregon and California from 1984 to 2012 (data avaialable online at http://lemma.forestry.oregonstate.edu/data). Using these models and ancillary stratifying variables (such as ownership, disturbance patterns, etc.), we create strata that include counts of pixels in a stratum and statistics from GNN continuous variables. The trend tool allows a user to:
- select a continuous variable to graph
- select a categorical variable to represent multiple series
- define the time range
- narrow the focus area based on other categorical (e.g. stratifying) variables
We anticipate that the tool will most likely be used by forest managers for assessing regional-scale trends, but the application should be adaptable to include any categorical/continuous variables.
trend relies on a JSON file to describe the data fields and point to a CSV file which holds the actual data. Right now it is hard-coded to read from a file called 'trajectory.json', which should be a user input in later iterations. For now, we have not posted the data files as they are big and change often, but they are currently available here:
- http://lemma.forestry.oregonstate.edu/sandbox/trajectory/trajectory.json
- http://lemma.forestry.oregonstate.edu/sandbox/trajectory/trajectory.csv
This is very much a research tool right now and apt to large changes.
We are currently hosting the example application at: