To record the frequency and patterns of crop-raiding in the study area, I visited each farmer every two weeks from December 2009 to November 2010. This is similar to the frequency used in other studies of crop-raiding (e.g., weekly records: Naughton-Treves 1997 and Webber 2006;
monthly records: Hill 2000 and Perez and Pacheco 2006), and allowed me to monitor one regroupement per week while allowing time for anthropological investigations. Initially, I provided farmers with calendars to record crop-raiding events, as used in Linkie et al (2007).
However, the high rate of illiteracy meant that this method was not successful. I surveyed all new fields at each visit, but visited old fields only when informants told us damage had occurred.
I surveyed farms in Ntchongorové and Idjembo 17 and 16 times, respectively, and corrected for this difference when comparing the two sites. I made additional field visits occasionally after elephant raiding events, increasing the frequency of survey for this pest species to 18 for both Ntchongorové and Idjembo. In la Haute, I surveyed new fields and old fields opportunistically because I had so few opportunities to survey. I pooled data from all regroupements in la Haute due to their similarity (e.g., no direct access through land, fields surrounded by mature forest, maximum of five households and all families relying heavily on fishing activities), and because samples in Ntchonimbani and Iloupi were too small to be analysed independently.
4.2.2.1 Frequency of damage
I considered a crop-raiding event as the occurrence of damage caused by one animal species between two visits. Because I was primarily interested in mammalian crop-raiding, I recorded diseases and damage caused by insects as presence/absence. For damage caused by mammals I recorded the following data for each pest and crop species (manioc, banana and taro) at each visit to old and new fields:
Location/GPS point of the central point of damage per pest/crop
Distance from centre of damage to the forest edge
Pest and crop species
Species and stage of crop damaged
Part of the plant damaged and type of damage
Frequency of damage
Estimated area damaged and intensity of damage (see sections 4.2.2.2 and 4.2.2.3).
My observations and discussions with farmers during the pilot phase suggested that some species damaged crops more at the field/forest boundary. Furthermore, Naughton-Treves (1998) showed that sight distance from the forest edge was an important factor in primate crop-raiding. Based on this result, I recorded the distance of the central point of the damage to the forest edge as close (≤ 20m) or far (> 20m), which represents the mean sight distance at the study site. In order to determine sight distance, my assistant and I assessed the maximum distance at which we could see each other, at hip height for all new fields at the beginning and the end of the study, and I took the mean. While different species are likely to have different visual limits at different heights, this provides information on possible fine-scale movement by raiders in fields.
I identified pest species through examination of tracks, dung, tooth-marks, and eating strategy. I identified pests to species-level when possible, and lumped identifications into groups otherwise. It was often difficult to differentiate damage caused by different rodent species, so I combined them under the label “rodents” for analysis. I also combined the different species of duikers and antelopes under the label “antelopes” for the purpose of analysis. It was impossible to tell whether guenons (Cercopithecus sp.) crop-raided, but interviews and my observations suggested that only red-capped mangabeys raided. I therefore lumped all observations of monkeys along with gorillas and chimpanzees under “primates”. I analysed elephants individually and lumped all other pest species under the label “other species”.
I noted the maturity of a damaged manioc plant as immature when it was a seedling up to about 1.5 m high, and mature afterwards, based on farmer’s descriptions. This reflects the presence of palatable manioc tuber rather than the maturity of the plant as harvested by farmers. I recorded part of plant damaged as root/tuber, stems, leaves and fruit.
I estimated the frequency of raiding events by looking at signs of decay on plants, dung or track, such as change in colour, dryness, and whether the signs had been washed by rain. I scored frequency as 1 (1 raid between visits), 2 (at least 2 between visits), 3 (at least 3 between visits) or 4 (at least 4 different raids between visits). I attributed scores 3 and 4 only when farmers said they had suffered more than two raids and tracks in fields confirmed their statements. To prevent overestimation of damage frequency, I gave a frequency score of > 1 only when there was clear evidence of multiple raids, such as repeated damage on the same plant. In cases of doubt, I gave a score of 1 (the minimum). I also used a more conservative approach, which was to record frequency score using only non-raided/raided. While both these
4.2.2.2 Area damaged
Estimating actual crop damage is difficult, time consuming, and controversial. The methods used represent a trade-off between accuracy and labour intensity (Hoare 1999; Hill, Osborn and Plumptre 2002). No systematic survey method currently exists to allow precise estimation of damage, and various methods have been used, making comparisons across studies difficult (Hill, Osborn and Plumptre 2002). Following Zadoks (1985), I defined the area damaged as the area that suffered from the action of any harmful agent leading to a reduction in the quality or quantity of yield. To obtain estimates of damage that were both statistically robust and allowed comparison with previous studies, I used a transect-like method. My assistant and I walked through the fields in a systematic way, following parallel lines (Fig. 4.3).
Figure 4.3: Crop-raiding survey methods.
Visibility in fields decreased over time due to plant growth. To ensure a constant level of surveillance across visits, I determined the number of transects per field, and the distance between the two investigators, at each visit based on visibility. Each transect started at one corner of the field and followed the edge(s) of the field (Fig. 4.3). I set the distance between the two investigators so that all stems between the two investigators were visible to at least one investigator. When one of us spotted damage, we assessed it together to avoid double counting.
During transects, I recorded the number of damaged stems for manioc, banana and taro. This method was labour intensive, and the gain in accuracy compared to a simple visual estimation of damage is debatable. However, systematic transects were the only way to provide a reliable account of damage by rodents and medium size herbivores, which is difficult to spot from a distance (Fig. 4.4).
Start/end
Investigator 1
Investigator 2
Figure 4.4: Example of rodent damage to manioc roots, while the stems are untouched. This type of damage is typically difficult to spot from a distance.
For manioc, I estimated the area damaged in two different ways. Where possible, I estimated the damaged area by counting the number of damaged stems, which I then converted into area using the mean planting density (Hill 2000). When the damage was over an extended area, or when we could not identify individual stems, I estimated the area damaged directly as a proportion of the field damaged to the nearest 5 %, and converted this into m2. To do this, I took the mean of independent estimates made by my assistant and myself. To ensure estimate accuracy, I tested our ability to estimate areas by first estimating damage and then counting the exact number of stems damaged and comparing the resulting values during the first month of the study. For villages in la Haute, an écogarde was also present, so I estimated the area damaged as the mean of the three observers’ estimates. During transects, I was unable to estimate the area devoted to crops other than manioc accurately, because banana and taro were often very patchy in distribution and stems were interspersed with manioc. Hence, I
4.2.2.3 Intensity of damage
We estimated the intensity of damage for each pest/crop pair using the same index as Parker and Osborn (2001):
Low: some severely damaged plants and damage not critical to the plant Medium: some (25-75 %) critically damaged plants
Intense: most (> 76 %) of the plants are critically damaged
Intensity is a valuable tool to assess how damage events lead to actual loss of plants and post-production. It also allows comparison between pest species, crop species and sites. When plants were damaged regularly, preventing them from maturing, I set the intensity as medium to reflect the possible loss of production to the farmer.