After uploading data for a particular site or sites, authorized users can clean, visualize, and download the data. The user who first uploads data for a new site is automatically granted this authorization. If she or he wishes, other users can be given access too (Currently this is done by emailing the data portal administrator.).
To clean data, first choose a river site from the dropdown menu on data.streampulse.org/clean. Time series of all available variables will appear, with the most recent four weeks of data visible.
By clicking and dragging over these time series plots, you can select intervals in time, as well as any data points they contain. At the release of the mouse, a popup window will appear. Here you can choose one of three labels with which to flag the selected points : "Interesting", "Questionable", or "Bad Data". These labels can later be used to filter data prior to model fitting. You can include additional information in the comments box and then click "Flag all selected" to apply your flag. You can also zoom in to better distinguish which points mark the boundaries of the anomaly you wish to flag.
Any red points that appear in the plots represent data automatically highlighted by our "potential outlier detector". The detector is tuned to identify certain kinds of anomalies for the purpose of speeding up manual flagging. If you agree with the detector that a point is an outlier, and should therefore be flagged, you can check the "Flag potential outliers (red)" button in the popup menu, and only the red points under your selection will be flagged with your designated label and comment.
Sometimes an anomaly can be shared across all time series for a site (e.g. if all sensors are attached to the same rig, which becomes buried in sediment). In this case, you may want to check "Apply to all variables", rather than only flagging points within one plot window.
See Modeling Metabolism with streamMetabolizer to find out how to deal with flagged data.
To clean data, first choose a river site from the dropdown menu on data.streampulse.org/clean. Time series of all available variables will appear, with the most recent four weeks of data visible.
By clicking and dragging over these time series plots, you can select intervals in time, as well as any data points they contain. At the release of the mouse, a popup window will appear. Here you can choose one of three labels with which to flag the selected points : "Interesting", "Questionable", or "Bad Data". These labels can later be used to filter data prior to model fitting. You can include additional information in the comments box and then click "Flag all selected" to apply your flag. You can also zoom in to better distinguish which points mark the boundaries of the anomaly you wish to flag.
Any red points that appear in the plots represent data automatically highlighted by our "potential outlier detector". The detector is tuned to identify certain kinds of anomalies for the purpose of speeding up manual flagging. If you agree with the detector that a point is an outlier, and should therefore be flagged, you can check the "Flag potential outliers (red)" button in the popup menu, and only the red points under your selection will be flagged with your designated label and comment.
Sometimes an anomaly can be shared across all time series for a site (e.g. if all sensors are attached to the same rig, which becomes buried in sediment). In this case, you may want to check "Apply to all variables", rather than only flagging points within one plot window.
See Modeling Metabolism with streamMetabolizer to find out how to deal with flagged data.
