Malipati village (22°04’S, 31°25’E), located at the southern border of GNP on Sengwe communal lands (Gomo et al. 2012), was selected as the porous interface and the area has been described (Ndengu et al. 2017a). The village lies adjacent to GNP, and the veterinary fence, which was erected in 1975 to prohibit cattle/buffalo contacts (Dube et al. 2010), has been damaged extensively allowing free movement of livestock into the park and wildlife accessing human settlements. Cattle share grazing and watering sources with wildlife, particularly during the dry season when these resources are limited in the communal lands. It is therefore assumed that there is significant contact between humans, livestock including cattle and wildlife at this type of interface (Ndengu et al. 2017a).

Chizvirizvi village (20°59’S, 32°01’E), located on the periphery of the Malilangwe conservancy, was selected as the non-porous interface. The conservancy is located on the northern boundary of the GNP and is surrounded by a well-maintained game fence, which hinders direct contact between wildlife on one side of the fence and humans and livestock on the other. The conservancy is home to the full range of African wild ungulates occurring in the area, while on the other side of the fence, the Chizvirizvi village hosts livestock, mainly cattle, goats and sheep. The fence creates a physically defined linear interface, separating wildlife and cattle.

Chomupani communal lands (21°40’S, 31°19’E), located at least 15 km from the north western boundary of GNP, was the selected non-interface site. Wild ungulates are reportedly absent in the Chomupani area, and this site was considered to be a control site with no wildlife/livestock interactions as it was far away from the GNP.

A longitudinal study was designed targeting cattle populations in the three study sites of Chomupani, Chizvirizvi and Malipati, and populations of selected wild ungulate species in the neighbouring GNP. Communal cattle dip tanks were used as sampling frames in each of the selected study areas. In Zimbabwe, animal health regulations compel all cattle owners in rural areas to dip their cattle weekly during the rainy season and fortnightly during the dry season for control of ticks and tick-borne diseases (Chikerema, Matope & Pfukenyi 2013). For this reason, the government, through the Department of Livestock and Veterinary Services, has constructed communal dip tanks (plunge pools) in the rural animal health centres that are accessible by all farmers in the rural communities, with several villages sharing one dip tank. For each selected interface, one dip tank was chosen based on the history of compliance with previous research work and the appropriateness of its location to the interface type as defined in our study aims. Three dip tanks (Malipati, Chizvirizvi and Chomupani) were therefore selected representing the porous interface, non-porous interface and the non-interface, respectively. The cattle census was 1528 for Malipati, 1470 for Chizvirizvi and 2300 for Chomupani dip tanks. The minimum cattle sample size to be sampled was calculated using the formula: n=[z2×p(1−q)]/e2,[Eqn 1] where z is the value from standard normal distribution corresponding to the desired confidence level (z = 1.96 for 95% confidence interval [CI]), p is the estimated prevalence and e is the desired precision (Dohoo, Martin & Stryhn 2003). We estimated an individual prevalence of 12% for RVF based on previous studies in the same area (Caron et al. 2013) and a 5% error margin at 95% confidence level. A minimum of 104 cattle per dip tank was therefore targeted per sampling session. Samples were recovered at each site in the wet and dry seasons that mainly occur in Zimbabwe from November to March and April to October, respectively.

Three wildlife species, the African buffalo (Syncerus caffer), greater kudu (Tragelaphus strepsiceros) and impala (Aepyceros melampus), were targeted for sampling south of the GNP in Mabalauta close to the porous Malipati interface. The species were selected on the basis of their reported mixing with livestock during grazing and at water points as confirmed by communities in previous studies (Ndengu et al. 2017a) and GPS radio-tracking studies for African buffaloes (Miguel et al. 2013; Zengeya et al. 2015). Another herd of African buffaloes at Chipinda, north of GNP, was also targeted. This herd lived far from target communities in a non-porous interface and was therefore assumed to be in minimal or no direct contact with domestic animals. Because of the cost of the capture process for wildlife, animals were conveniently sampled and the numbers sampled were limited. The numbers and procedures for wildlife immobilisation and sampling followed the description given in the permit granted by Zimbabwe National Parks and Wildlife Management Authority (permit N° 59(4) (a) & (b) 24/2015) and was performed by experienced licensed practitioners.

Each cattle dip tank was visited four times for sampling, as highlighted previously. Farmers enrolled their herd in the sampling protocol on a voluntary basis. Cattle were systematically randomly selected as they passed through the handling facilities of the dip tank. As cattle passed through the handling pen, each fourth mature (>1 year) and intact (if male) individual was targeted for sampling. The targeted individual would then be physically restrained and 15 mL – 20 mL of blood collected using coccygeal venipuncture and a 20 mL disposable plastic syringe and an 18-gauge needle. Upon collection, the blood was immediately transferred to 4 mL plain tubes. To allow clot separation, all blood samples were left to stand for approximately 15 min in the shade at ambient temperature (25 °C – 30 °C). Clotted blood samples were then centrifuged at 3000 g for15 min and 2 mL of serum were collected into cryo-tubes and stored in liquid nitrogen at −196 °C en route to the laboratory where they were stored at −20 °C until the time of analysis. As each individual animal was being sampled, epidemiological data pertaining to that individual was simultaneously collected. Among the data collected were the date and season of sample collection, interface type, owner of the animal and the village of origin, the sex and the parity in case of females and any history of previous abortion(s). All sampled animals were ear-tagged to avoid resampling them on subsequent visits.

Within GNP, two buffalo groups were selected by aerial spotting from a helicopter in an area as close as possible to the porous park border (Mabalauta area), and in the northern part of the park (Chipinda area) closer to the non-porous border, respectively. The buffaloes were immobilised using a standard protocol similar to Burroughs et al. (2006): one to four individuals were anesthetised via a dart gun from a helicopter using a combination of etorphine hydrochloride and xylazine. The kudus were immobilised using similar standard procedures (Burroughs et al. 2006), although they were darted from the ground after being driven into a boma structure, using pole syringes. The impalas were captured less than 2 km from Malipati dip tank, right at the edge of the porous interface, using nets followed by physical restraint without anaesthesia (Kock & Morkel 2006). Following immobilisation, blood was collected using jugular venipuncture and processed for laboratory analysis as described for cattle above. After sample collection from buffaloes and kudus, anaesthesia was reversed by injection of diprenorphine hydrochloride. The animals were released at the site of capture and monitored from the air or the ground until they recovered fully. Epidemiological data collected for each wild animal sampled included the date of capture, the age as estimated by dentition and the location using GPS. All sampled animals were classified as adults, subadults or juveniles.

Specific anti-RVFV antibodies were tested in serum samples using the commercial ID Screen® RVFV competition multi-species enzyme-linked immunosorbent assay (ELISA) (IDvet, Grabels, France [www.id-vet.com]) with a level of sensitivity and specificity of 91% – 100% and 100% (95% CI: 99.58% – 100%), respectively, based on livestock samples (bovine, ovine, caprine) (Comtet et al. 2010; Kortekaas et al. 2013). Based on competition, this test is able to detect specific RVFV antibodies regardless of the species tested (Gür et al. 2017; Moiane et al. 2017). The detection of RVFV-specific anti-N antibodies by ELISA indicates exposure to RVFV either by natural infection or by vaccination. Briefly, wells on the ELISA plate were coated with a recombinant RVFV nucleoprotein. Samples to be tested and the controls were added to the microwells. Anti-nucleoprotein antibodies, if present, would form an antibody–antigen complex, which masks the nucleoprotein epitopes. An anti-nucleoprotein-peroxidase conjugate was added to the microwells to fix the remaining free nucleoprotein epitopes forming an antigen-conjugate-HRP complex. After washes aiming to eliminate excess antigen–conjugate complex, the substrate solution was added and the resultant colouration depended on the quantity of specific antibodies present in the test sample. The optical density (OD) was measured using an ELISA microplate reader at 450 nm. For each sample, the competition percentage (S/N%) was calculated using the formula: S/N% = OD sample/OD negative control × 100. An SN% of less than or equal to 40% was considered positive for RVFV antibodies.

Data recording and editing were performed in Microsoft Excel^®, and the descriptive statistics were performed using EpiCalc 2000 version 2. The overall number of seropositive animals was calculated from the total number of samples tested over the study period and expressed as a percentage. For cattle, seropositivity was examined in relation to interface type, sex, parity, abortion history and season, while seropositive wild animals were examined in relation to species, age and site. Interface type, sex, parity, abortion history and season categories were generated as follows: three for interface type (porous, non-porous and non-interface), two for sex (females and males), four for parity (heifers, 1–2, 3–4 and ≥ 5), two for abortion history (history of abortion and no history for abortion) and two for season (wet and dry). For wildlife, species, age and site categories were as follows: three for species (buffalo, kudu and impala), three for age (juveniles, subadults and adults) and two for site (Mabalauta and Chipinda). Descriptive statistics on abortion history, parity and season were restricted to female cattle and because of a small sample size, seasonal and interface type descriptive statistics were not assessed for male cattle. The chi-square test was used to measure differences between categories, and values of p < 0.05 were considered as significant.