Introduction

Meeting the increasing needs of human populations for food and other products from the world's crop and grasslands is a major global challenge (Pilling et al. 2020) that has led to agricultural intensification. Over the last half century, technological developments in agricultural production in Europe have fundamentally changed, including in Hungary where the number of tractors and harvesters per worker have steadily increased (Báldi & Faragó 2007). As a consequence, agricultural biodiversity has declined throughout Europe (Erisman et al. 2016), including bird, mammal, reptile and pollinator species. There has also been a severe decrease in the population abundance of small game species that occupy agricultural environments. This unfavourable trend appears to be associated with the transformation and elimination of former well-structured grassland habitats, large-scale cultivation and the introduction of intensive agricultural technologies (Edwards et al. 2000, Robinson & Sutherland 2002, Benton et al. 2003, Smith et al. 2005).

Traditionally, mowing on farms was accomplished through human and animal labour, and mowing larger areas (i.e. hundreds of hectares) could often take up to a month. Slow, step-by-step harvesting benefited wildlife because meadows developed a mosaic pattern, allowing individual animals to escape and different grass species to ripen (see e.g. Wan et al. 2016). However, the mechanization of agriculture has completely transformed and accelerated this traditional practice. Today, powerful and fast machines alter the structure and species composition of meadows, posing a serious threat to species occupying grasslands and hay fields.

In an 11-year study, Pépin & Angibault (2007) found that with the disappearance of vegetation patches (shrubs, alfalfa fields, fallow lands), hares were increasingly forced to occupy plough land and winter wheat fields. An agricultural landscape comprising less than 10% of such patches is suboptimal for brown hares, increasing not only the likelihood of predation but also mortality related to intensive agricultural activities (mowing, harvesting, other agrotechnical work). Similarly, mowing increases the vulnerability of European hares (Lepus europaeus), who otherwise hide from aerial and terrestrial predators in high vegetation (Fernex et al. 2011). Schai-Braun (2013) showed the disruptive effect of agricultural machinery in Switzerland, as the home range of hares during their active period increased after the harvest. In Sweden, roe deer (Capreolus capreolus) fawn mortality caused by mowing was estimated at 25-44% of the yearly recruitment (Jarnemo 2002).

The extent of mowing mortality is hard to quantify both from a nature conservation and game management perspective. In most cases this impact is not monitored, but even if investigated, the small size of many carcasses and animal remains are hidden under mown vegetation. A simple visual observation by humans along transects may, therefore, be minimally effective (see e.g. Kis et al. 1998). However, dogs are able to find carcasses more effectively than human searchers (Homan et al. 2001, del Valle et al. 2020).

There are several initiatives in Europe to employ dogs in conservation work. In Spain, Italy and Hungary dogs are successfully used to detect illegally poisoned raptors (Simón et al. 2007, Deák et al. 2020). Search dogs were also used to detect dead remains of birds and bats (Paula et al. 2011), threatened (Nussear et al. 2008) or aggressively invasive species (Vice & Engeman 2000), living (Cablk & Heaton 2006, Dahlrgen et al. 2010) or dead animals (Deák & Horváth 2018).

Our study examined whether detection dogs trained to find carcasses and poisons could also be effectively used to reveal mowing mortality. Our questions were: 1) How many species/individual carcasses could be detected by dogs after mowing in different agricultural crops? 2) Is there any difference in mowing mortality due to vegetation type or mown area? 3) Are there any difference in the number of carcasses revealed by morning (immediately after mowing) or afternoon searches using a dog?

Material and Methods

Study area

The effect of agricultural mowing machines was investigated in Jászság, Hungary (Fig. 1). The area is neither under legal conservation protection or part of the NATURA 2000 network. This landscape contains large cultivated fields of alfalfa or other leguminous crops, maize, winter wheat, rape and sunflower, grass meadows and some small forest blocks and shrubby patches. Tree and shrub alleys line agricultural parcels, but mainly it is grass strips that delineate the edges of fields.

In May and June 2018, we surveyed a total area of 94.2 ha, containing 49.6 ha of leguminous vegetation and 44.6 ha of grass meadows, and with parcel sizes ranging from 2.2 to 20 ha. We obtained six observations on mowing mortality for both vegetation types (see in Table S2).

Mowing was performed with a John Deere 6930 tractor with DISCO 3100 FRC front disc mower and DISCO 3100 TRC trailed disc mower. The mower consisted of seven discs and was 3 m wide. The cutting height was set at 8 cm in alfalfa and other leguminous fodder and at 5 cm in natural grasslands. The mowers were equipped with a stalk chopper that destroys the stem and leaf portions of the cut hay fodder, thereby opening surfaces for faster moisture release to hasten drying. All mowing was done during the day and only one machine at a time was working, except on one occasion.

Fig. 1.

Map of the study area in Jászság, Hungary showing the positions of the studied parcels of land (M1-6: mowed parcels of grass meadows, L1-6: mowed parcels of leguminous vegetation).

Results

Falco searched the fields with an average speed of 4.38 ± 0.71 km/h and the km index of the found carcasses was 2.74 ± 1.68 remains/km. The dog found 11.51 ± 6.78 carcasses/h on average.

We detected 149 killed individuals (Table S1) across 12 surveyed parcel of land representing 158.2 carcasses/100 ha (Table S2). Reptiles, which are all protected in Hungary, were the most affected vertebrate group (57%). Reptiles were mainly represented by sand lizards (91%), though some grass snakes were also observed. Mammals and birds comprised 30% and 6% of carcasses, respectively.

The dog also found brown hares, pheasants (Fig. S2) and roe deer (11% of all carcasses), indicating damage for local game managers. In the case of legumes and grassland, four dead European hares were found to have died as a result of mowing in each crop type. Several rodents (the second most frequent group, 21% of total remnants and 70% of mammals), some small passerines (2.7% of total remnants) and quails (1.3% of total remnants) were also detected by the dog. A detailed list of the species that were killed to mowing, and their proportions, can be found in Table S1.

Fig. 2.

A) Difference between leguminous vegetation and grass meadow in the density of animal remains found. B) Difference between leguminous vegetation and grass meadow in the density of reptile remains found.

No significant difference was observed in the density of animals (Table S2) killed by mowing, either in legume crops (median mortality = 73.55 remains/100 ha) or grassland (median mortality = 222.22 remains/100 ha; Mann-Whitney U-test: U = 24, n_1,2 = 6, P = 0.394; Fig. 2A). Both, the smaller- (skylarks, rodents, lizards, snakes) and larger-bodied (quails, pheasants, hares, roe deer) categories showed no significant difference between the two habitat types (Mann-Whitney U-test: U = 12, n_1,2 = 6, P = 0.394, U = 20, n_1,2 = 6, P = 0.818, respectively). However, if we separated the carcasses among reptiles, birds and mammals, more reptile remains were found on grassy meadows than on leguminous fields (Mann-Whitney U-test: U = 5, n_1,2 = 6, P = 0.041; Fig. 2B). The other animal groups showed no significant differences (Mann-Whitney U-test: U = 22, n_1,2 = 6, P = 0.589 for birds and U = 19.5, n_1,2 = 6, P = 0.818 for mammals).

Considering search times, there was no significant difference between searches immediately after mowing in the morning (mean ± SD = 134.45 ± 106.17) and later in the afternoon (mean ± SD = 309.21 ± 249.85; independent samples t-test with equal variances: t = –1.577, df = 10, P = 0.146).

The size of the mown area did not appear to affect the number of animals killed (negative binomial GLM: the effect of the area size on the carcass density was –0.087, Chi² = 3.195, df = 1, P = 0.074).

Discussion

We found that using a dog trained to find carcasses as part of anti-poisoning surveys can also be valuable in finding the remains of animals that have died as a result of mowing. Matthews et al. (2013) and del Valle et al. (2020) also demonstrated that dogs that received prior training were much more effective than human observers in finding dead bats and birds.

Our results show that mowing leguminous vegetation or grassland poses a threat to a variety of vertebrate species. The finds made during the dog-handler surveys confirm, and previous studies have shown, that grassy meadows and legume cultivation offer foraging, hiding and nesting sites for several animal species during spring and summer, resulting in disturbances by agricultural works and leading to significant mowing mortality when these crops are harvested (e.g. Tyler et al. 1998, Viszló 2012, Vadász & Lóránt 2014, Faria et al. 2016).

Visual searches for small animals, like lizards, whose remains can be hidden under the cut vegetation, are extremely challenging. Moreover, some organisms can be obliterated during mowing, which occurs when smaller animals cling to the vegetation and are drawn into the stalk-breaking equipment, which entirely crushes them. This may explain why the remains of arthropods and amphibians were not found during searches. For example, Humbert et al. (2010a) observed that meadow harvesting caused the direct mortality of 65-85% orthopterans (21-57% mortality caused by mowing), which suggests there may be a high impact on invertebrates. In the case of amphibians another explanation for the lack of carcasses, particularly toads, was that surveys were conducted during a dry period. Toads may have been dormant, hidden in cracks of the soil, and thus protected from the disc-mower.

In principle, reptiles would be expected to be mobile at ground level rather than on the surface of vegetation. As a result, the killed animals were not totally obliterated by the mower. Durbian (2006) reported high mortality of the massasauga (Sistrurus catenatus), a snake species in Missouri, caused by mowing, as a result of crushing by tractor tyres, contact with the mower blades, or indirectly through depredation in the freshly mown area.

Comparison of vegetation types showed that 84% of reptile remains were found on natural grassland. This habitat offers more varied micro-topography and diverse vegetation patches and tends to promote taxonomic diversity. The greater diversity of the grassy meadows might, therefore, result in more carcasses in that habitat type. However, the difference was not significant, though with a larger sample of each habitat type this difference may be supported. Reptiles occurred in grassland meadows in greater numbers than arable crops. Notably, the cutting height of the mower was set at 5 cm on grassland, lower than when mowing leguminous crops, in order to produce larger amounts of fodder. For this reason farmers do not want to raise the blades, even if it would result in lower mortality. Additionally, farmers do not utilize a chain curtain (Fig. S3), which alerts wildlife to the mower so that they can escape, because they do not perceive it is effective in decreasing wildlife mortality. A lower cutting height and failure to use a chain curtain probably contribute to a greater number of mortalities of small-sized species.

One of the main challenges of the study in relation to using dogs was the difficulty of planning the exact time and place of searches due to unpredictable weather and changes in the work schedule and priorities of farmers. These uncertainties meant that it was impossible to search fields for carcasses prior to mowing. It was important that searches are made immediately after mowing, because scavenging animals will constantly visit the area for the expected food during and after agricultural work. Scavenging species can rapidly remove a significant number of small animal remains without any trace (Balcomb 1986). For this reason we predicted that we would detect more carcasses in the morning than in the afternoon. In fact we found the reverse, perhaps due to the increased concentration of volatiles associated with longer decomposition. Although the work of the search dog was sometimes hampered by high temperatures (if the dog is panting, it is harder to sniff), we did not notice any difference in the temporal distribution of detections during searches.

Conclusions and recommendations

Our study provides a case study for dogs to be trained and used effectively to assess mowing mortality in wildlife. Assessing the exact damage caused by mowing is a complex task. Ideally, surveys of living fauna before and after mowing should be conducted, using dogs or other complimentary techniques (e.g. strip census or netting), so that a baseline is obtained for comparison with mortalities associated specifically with mowing thereby enabling estimation of the influence of mowing mortality on population dynamics.

Because the work was, of necessity, conducted in warm weather, it could also be beneficial to work with two dogs, in shifts. Poison/carcass-seeking dogs are taught to use active signalling (i.e. barking), because the search is often carried out in dense vegetation and bushes. However, for inspecting mown arable fields, the dog works in an open area within continuous view of the handler. Therefore, passively alerting (i.e. either a sit or down position) to signal carcass detection will be less exerting for the dog. Although for poison detection work Falco was trained solely with decomposed carcasses, the dog was able to find fresh remains. For a nature conservation application, we recommend training dogs on a combination of fresh and decomposed carcasses. The training of the dog and the handler contributes to the success of the search. The handler must guide and support the dog, and the dog must perform a focused search on an area to find each carcass. For this task an inexperienced person using a pet dog will have limited success. Using a dog across projects, to achieve multiple conservation purposes and benefits, makes even more effective use of the resources invested in their training. Comparative studies on the effectiveness of searching between different trained dogs and between human surveyors and dog-handler groups would be valuable to refine the method.

In order to introduce appropriate conservation measures to reduce mowing mortality, it is necessary to examine the impact of mowing in different periods of the day or year (e.g. Broyer et al. 2020) and using different harvesting equipment (e.g. Humbert et al. 2010b). Leguminous crops and, in high-rainfall years, natural grassland may be mown several times in a given year, which can also affect the results. These activities are of special importance on the self-managed grasslands of the National Parks, which support particularly abundant wildlife.

In addition to technological factors, the attitudes of farmers also plays an important role in limiting damage to wildlife. It was notable that farmers do not consider the chain curtain on mowers to be effective in protecting wildlife and, since there is no unequivocal legal obligation for its use in these areas, they rarely utilise it. It would also be valuable if farmers notified conservation and hunting associations in advance of mowing to enable them to survey wildlife (e.g. using aerial thermography, Cukor et al. 2019), or potentially to enable some species to be driven away, removed or protected locally (e.g. by enclosing or marking a nest). A well-functioning wildlife alert system, with monitoring and rescue interventions acceptable by all partners, will be essential to decrease this kind of avoidable, non-targeted impact on agricultural biodiversity.