Fungus-growing termites and their symbiotic Termitomyces fungi are critically important carbon and nutrient recyclers in arid and semiarid environments of sub-Saharan Africa. A major proportion of plant litter produced in these ecosystems is decomposed within nest chambers of termite mounds, where temperature and humidity are kept optimal for the fungal symbionts. While fungus-growing termites are generally believed to exploit a wide range of different plant substrates, the actual diets of most species remain elusive. We studied dietary niches of two Macrotermes species across the semiarid savanna landscape in the Tsavo Ecosystem, southern Kenya, based on carbon (C) and nitrogen (N) stable isotopes in Termitomyces fungus combs. We applied Bayesian mixing models to determine the proportion of grass and woody plant matter in the combs, these being the two major food sources available for Macrotermes species in the region. Our results showed that both termite species, and colonies cultivating different Termitomyces fungi, occupied broad and largely overlapping isotopic niches, indicating no dietary specialization. Including laser scanning derived vegetation cover estimates to the dietary mixing model revealed that the proportion of woody plant matter in fungus combs increased with increasing woody plant cover in the nest surroundings. Nitrogen content of fungus combs was positively correlated with woody plant cover around the mounds and negatively correlated with the proportion of grass matter in the comb. Considering the high N demand of large Macrotermes colonies, woody plant matter seems to thus represent a more profitable food source than grass. As grass is also utilized by grazing mammals, and the availability of grass matter typically fluctuates over the year, mixed woodland-grasslands and bushlands seem to represent more favorable habitats for large Macrotermes colonies than open grasslands.

Data originates from 59 termite mounds representing two Macrotermes species occurring in the Tsavo Ecosystem, Kenya. Each termite mound was georeferenced using a handheld GPS and the inhabitant termite species was identified based on species-specific mound morphology. The results of stable isotope (δ¹³C, δ¹⁵N) and elemental (C%, N%) analysis originate from a single fungus comb specimen collected from each studied mound. The analysis were performed at the Laboratory of Chronology, Finnish Museum of Natural History (Helsinki, Finland) using an NC2500 elemental analyzer coupled to a Thermo Scientific Delta V Plus isotope ratio mass spectrometer and the raw isotope data was normalized with a multi-point calibration using certified isotopic reference materials (USGS-40, USGS-41, IAEA-N1, IAEA-N2, IAEA-CH3 and IAEA-CH7). The species identity of the cultivated Termitomyces fungi (unnamed species "A", "B" or "C") is based on DNA barcoding of the ribosomal ITS1–5.8S–ITS2 region extracted from the fungal nodules growing on the comb surface (Vesala et al. 2017, 2019). Canopy cover was assessed around each termite mound using LiDAR (Light Detection and Ranging) point clouds produced during a flight campaign in March 2014. Lidar point clouds were classified and converted to topographic and surface models representing either ground or vegetation canopy, respectively, using Lastools processing software (https://rapidlasso.com/lastools/). A Canopy Height Model at 1 m spatial resolution was derived from topographic and surface models and was used to calculate precentage canopy cover using QGIS 3.6. Canopy cover around each studied mound was calculated separately for three different tree height classes (> 1, 3 and 6 meters) within three different circular buffers (radiuses 35, 50 and 100 m) centered at mound centrum, leading to a total of nine different canopy cover estimates.

The principle data file Fungus_comb_stable_isotopes_and_canopy_cover.csv was used as the mixture dataset for the MixSIAR analysis. Additionally, the file includes the data needed to produce map illustration in Figure 2 and the regression models comparing fungus comb C:N stoichiometry and the canopy coverage or comb δ¹³C. Units of the stable isotope values (δ¹³C and δ¹⁵N) and fungus comb C and N contents (C_cont, N_cont) are per mille (‰) and percentages (%), respectively. Geographic coordinates (x_coord, y_coord) are in WGS84 coordinate system. Nine different canopy cover estimates are percentages, e.g., column “cov100_over1” means coverage (%) of over 1 meter height vegetation at 100 meters distance (circular buffer) from the mound centrum.

For the MixSIAR analysis, source and discrimination tables are given in two separate files named Source_table_C3C4.csv and Discrimination_table_C3C4.csv, respectively. For SIBER analysis, three different subsets (SIBER_site.csv, SIBER_termite.csv, SIBER_fungi.csv) have been extracted from the principle data file, where the compared groups (study site, termite species, Termitomyces species) are coded with the following numbers: Sanctuary = 1, Mwashoti = 2, Maktau = 3, Mbula=  4, M. subhyalinus = 1, M. michaelseni = 2, Termitomyces A = 1, Termitomyces C = 2. Observations from mounds where Termitomyces species could not be identified or groups had less than three individuals (colonies from study sites Latika and Mwashuma, or cultivating Termitomyces B) are omitted from the SIBER datasets. The fungus comb δ¹³C and δ¹⁵N are coded “iso1” and “iso2”, respectively.