Designing profitable and climate-smart farms using virtual reality

currently sustains between two and three stock units per hectare, which is below the 6.3 stock unit per hectare in average for New Zealand sheep and beef farms [20]. As such, we chose to adjust the stocking rate to match the industry average, so that the VR application we were developing was more broadly applicable to the national context and challenges of livestock farms. Data sources Geospatial data. Geospatial data have been used to map the spatial variation of the pas- ture production and to generate the realistic virtual landscape. Normalised Difference Vegeta- tion Index (NDVI) data from Sentinel-2 images have been used to classify the relative pasture production. To do so, two satellite tiles from the 2019 lambing season (October and Novem- ber) have been accessed through the EOS Data Analytics website [21]. To create the geospatial 3D terrain within the virtual environment, a one-meter spatial resolution Digital Elevation Model (DEM) has been downloaded from the Land Information New Zealand website [22] and comes from Lidar data captured between 2018 and 2019. A 3-band (RGB) orthophotogra- phy with 300 mm spatial granularity taken in October and November 2020 was obtained from Land Information New Zealand website [23] and was used to generate a base map that was dis- played in the virtual environment (described below). Carbon sequestration data. Carbon sequestration values for the different tree species rep- resent the values that are outlined in New Zealand’s Climate Change (Forestry Sector) Regula-

tion 2008 (SR 2008/355), available from the New Zealand Ministry for Primary Industries [24]. These sequestration values are available for three tree species including native trees (a broad category encompassing a range of New Zealand native species), Pine (Pinus radiata) and Euca- lytpus (Eucalyptus nitens). The amount of carbon that is sequestered by trees varies depending on age. However, in this study, we have considered the yearly average sequestration over a 30-year period. The establishment costs for the different types of trees were based on expert knowledge and, for each individual species, a uniform rate was applied (i.e., the cost of plant- ing on sloping terrain was assumed to be the same as planting on flat terrain). Carbon emission and profitability data. Carbon emission and finance data have been extracted from existing industry literature. Carbon emissions and financial profitability vary spatially across the farm according to stocking rate (e.g., number of animals per hectare). Therefore, carbon emissions and profitability values have been expressed per Stock Units (SU), respectively kgCO2eq/SU/year and NZD/SU/year. Gross carbon emissions from livestock farming primarily arise from methane generated by enteric fermentation in the rumen [25]. Between the years 2018 and 2020, the national SU average for the sheep and beef industry ranged from 41.9 million to 42.5 million with an aver- age of 42.2 [26]. Livestock emissions estimates for the same years have ranged from 13,551 kt

to 16,537 kt [26, 27]. These estimates suggest a gross carbon emission ranging between 327 and 395 kgCO2eq/SU/year for New Zealand sheep and beef farms. An emission of 370 kgCO2eq/SU/year has been assumed. In this study, profitability is defined as the Earnings Before Interest, Tax and Rent (EBITRm). Profitability varies from year to year and depends, amongst other things, on mar- ket and climatic conditions. Data from the literature suggests that the average national profit- ability for beef and sheep farms in New Zealand ranges between $42 and $62 per SU [20]. In the current study, NZD $56/SU has been assumed. Methods The main objective of the VR application was to enable farmers to explore different tree plant- ing strategies on their farms. Within the VR environment, the user can virtually ‘plant’ one of PLOS ONE Designing profitable and climate-smart farms using virtual reality PLOS ONE | https://doi.org/10.1371/journal.pone.0286723 June 2, 2023 4 / 16