Modelling the influence of crop density and weather conditions on field drying characteristics of switchgrass and maize stover using random forest

Field drying trials were conducted using both field baskets as well as grab sampling techniques to study drying behaviour of switchgrass and maize (corn) stover (CS). Environmental conditions such as hourly solar radiation, vapour pressure deficit (VPD), average wind speed, rainfall amount, harvesting method, and field operations such as swath density were used as variables for model development. A powerful classification-based algorithm, which uses a collection of decision trees called random forest (RF) was utilised to predict moisture content (MC) of switchgrass and CS on wet basis. RF predicted the MC of switchgrass and CS with a coefficient of determination of 0.77 and 0.79, respectively. Rainfall, hours after harvest, average change in solar radiation in past 12 h, average solar radiation in past 12 h, and swath density were found to be the important variables affecting the MC of CS. Drying CS in low density (LD) and medium density (MD) swaths facilitated quick drying even in moderate drying conditions. Rainfall events ranging from 1.5 to 7.5 mm were experienced during the switchgrass drying period which delayed crop drying by one day to several days depending on the weather conditions after rainfall. Several rewetting events were also observed due to dew at night which increased the MC in LD switchgrass and CS by 5-15%. The models developed in the current study will help in decision-making of switchgrass and CS collection after harvest, based on forecast weather conditions in lower Midwestern states.