<p>Assessment Metrics.</p>

Among the flora that are harvested are palms that are distributed across swidden agriculture landscapes. Here we draw on imagery derived from unmanned aerial vehicles (UAVs) and convolutional neural network (CNN) object detection techniques to identify individual palms within a localized region of swidden agriculture in Guyana.

A total of 10,194 palms were identified from six different species over 255 hectares. Land disturbed by swidden agriculture saw a 136% increase in palm density compared to undisturbed forest (53.3 versus 22.6). Local Moran’s <i>I</i> cluster analysis showed that <i>Attalea maripa</i> was the dominant species in both undisturbed forest, representing 77.1% of palms, and swidden-disturbed land, representing 89.0% of palms. <i>Astrocaryum vulgare</i> declined in relative representation from 17.3% in undisturbed forest to 10.0% in disturbed forest, accompanied by reduced representation of other less abundant palms.