For large-bodied mammals living in biodiversity-rich tropics, mitigating human-wildlife conflicts (HWC) is a conservation priority. These conflicts have been widely documented in each tropical continent and involve problem animals such as elephants (Loxodonta africana) that raid crops in Kenya (Sitati et al., 2003; Chiyo et al., 2011) and Ghana (Monney et al., 2010), leopards (Panthera pardus) that kill livestock in Pakistan (Dar et al., 2009) and jaguars (Panthera onca) that attack humans in Brazil (Zimmermann et al., 2005). If solutions to alleviate the negative impacts of wildlife, notably elephants, are not found, persistent raiding of  crops  may compromise their conservation (Chiyo and Cochrane, 2005), like in many societies, where traditional farmers self-compensate for losses of human wildlife conflict by hunting and consuming the animals.

Mammals continue to be under pressure from high human population densities and as expanding human populations compete with these mammals for habitat, the future of mammal populations may soon depend entirely on protected areas (Barnes, 1999). On the other hand, as wildlife competes with humans for resources, mainly crops and livestock, mammal populations may also depend  on  the   extent   of   farmers’   co-operation   and vigilance, and the economic and social sacrifices involved in living next to a wildlife protected area.

Human communities are located adjacent to wildlife reserves, and interactions often have adverse effects on humans as well as wildlife, particularly when they compete for resources (Conover, 2002). Many wildlife species raid crops and livestock for food and cause considerable damage with corresponding impacts on community livelihoods. It is the same local communities whose livelihoods are directly or indirectly affected by the establishment and conservation of these wildlife reserves (Hens, 2006; Abbot et al., 2001). In addition, some forms of restrictions of access to resources from the reserves are put on members of the local communities (Bergseng and Vatn, 2009). Ironically, in most developing countries, the local communities who are affected by these restrictions are generally poor (Das Kanti, 2005) and so attempts by farmers to protect their crops from raids by wildlife from reserves may be relatively costly in terms of time and resources, and are often limited in effect. It may involve strategic barriers, buffers, and/or improving the efficacy of guarding, deterrents, and repellents (Conover, 2002); the costs of all these may also reduce local people’s tolerance to wildlife and support for conservation programmes.

Ensuring farmers’ livelihoods and food security through reduction of human-wildlife conflict is an internationally agreed goal (Parker et al., 2007) and conservation managers today are required to tackle this complex issue in collaboration with communities in order to achieve conservation objectives (Parker et al., 2007). To avert the current biodiversity extinction crisis, conservation practitioners require a clear understanding of the prevailing threats and the strategies to mitigate them. Whilst the threats are usually known and primarily relate to the competition between humans and biodiversity for limited space and resources (Sodhi et al., 2009), the recent calls for evidence-based conservation indicate that the appropriate solutions are not yet fully known (Sutherland et al., 2004; Ferraro and Pattanayak, 2006).

Conservation and development are linked and so can protected areas provide development opportunities for communities (Furze et al., 1996). In the developing world, increased concern over the burden that conservation often places on local communities has led to efforts to incorporate development goals into conservation practices (Hulme and Murphree, 2001). The coalescing of development and conservation gave rise to community-based conservation or community-based natural resource management (CBNRM), a participatory model which has provided the opportunity for conservation to produce tangible benefits for rural development (Wells et al., 1992; Munasinghe and McNeely, 1994; Western and Wright, 1994; Steiner and Rihoy, 1995; Brechin et al., 2002). In Ghana, the introduction of community-based conservation programs such as the Community Resource  Management Areas (CREMAs) and Community Based Tourism (CBT) ventures is an effort to gain local support for conservation through participatory management and benefit-sharing.

Given the recurring nature of conflict between conservation and local communities, it is critical that conservationists better understand local views with respect to wildlife and protected areas, particularly the cost of living next to a wildlife protected area, the sacrifices of accepting the establishment of a protected area and their readiness to tolerate raid emergencies. Unfortunately, no study has quantified losses of crops and livestock in terms of cash around the Mole National Park (MNP) and therefore the threshold to which communities can sacrifice and tolerate raids remains unknown. An important first step is to determine how much losses in cash are suffered every year through crop and livestock raids, and how this has affected collaboration with the local people, which is the basis of this study. From this, appropriate and, where necessary, alternative conservation intervention measures could be developed to serve as guidelines for outlining strategies for successful management schemes that may eventually lead to the improvement of livelihoods for the many households.

Study area

The MNP (Figure 1) lies between 9° 11’ and 10° 10’ N and between 1° 22’ and 2° 13’ W; it covers a land area of about 4,840 km² as Ghana’s largest wildlife refuge. The park is located in Northwest Ghana on grassland savanna and riparian ecosystems at an elevation of 150 m, with sharp escarpment forming the southern boundary of the park. The park's lands were initially set aside as a wildlife refuge in 1958, and later designated a National Park in 1971 after the small human population of the area was relocated.

 

 

Tree species of the park include Burkea africana (an important tropical hardwood), Isoberlinia doka, and Butyrospernum paradoxum (shea butter tree), which is important medicinally to treat infectious diseases. The savanna grasses are somewhat low in diversity, but known species include spike sedge, Kyllinga echinata, Ancilema setiferum, two endemic members of the Asclepiadaceae subfamily, the vine Gongronema obscurum, and the edible geophyte, Raphionacme vignei. The park is home to over 93 mammal species, and the large mammals of the park include the savanna elephant (Loxodonta africana africana), hippopotamus (Hippopotamus amphibius) and African buffalo (Syncerus caffer), which provide the park with a great potential for tourism and boost livelihood of the local people. Elephant (Blanc, 2008) and hippopotamus (Lewison and Oliver, 2008) are listed as vulnerable in IUCN’s conservation status red list, but the buffalo is listed as least concern (IUCN, 2008). Thirty-three species of reptiles and 344 bird species have been listed. To ensure effective management, the park is divided into four ranges, namely, Bawena, Dusie, Headquarters, and Jang (Figure 2). Patrol staff are deployed to patrol from each range camp. The park is bordered by 33 communities, with a total estimated population of about 35,000. Most  of  these  communities  are  poorly  accessible   and   depend almost entirely on the natural resource base for food and income. Virtually, all farming activities are concentrated in the rainy season between May and October and crops like cassava (Manihot esculenta), yam (Dioscorea species), maize (Zea mays), rice (Oryza sativa) groundnut (Arachis hypogea), cowpea (Vigna unguiculata), guinea corn (Sorghum bicolor) and millet (Urochloa ramosa) are commonly cultivated. Also, guinea fowls (Numida meleagris), domestic fowls (Gallus species), goats (Capra hircus), sheep (Ovis aries) and cattle (Bos primingenious) are the main livestock reared in the communities around MNP.

 

 

Data collection

Crops and livestock raids and estimated cost

Crops and livestock raid activities of mammals from MNP were monitored from 2005 to 2012 from communities that fringe the park. Data were obtained through reports from Community Resource Management Committees in collaboration with the park’s staff. Data for the 8-year period included species of animals that raided, their numbers and type of crop or livestock they raided, size of surface area raided, season and time of raids, the community and range in which raids occurred, estimated population of raided communities, nearest distance of communities where questionnaires were administered (including raided and non-raided communities) to the park’s boundary and shortest distance from raid occurrence to the park’s boundary.

Estimated population of the communities were obtained from Ghana Statistical Services. GPS coordinates were taken at sites where raid incidences occurred and at the centre of each raided community. The mean distances of raid occurrence and distance of communities from the park’s boundaries were calculated using the nearest-features extension in ArcGIS (v9.3), based on GPS coordinates of the raid occurrence. Geospatial data on the park’s boundaries were obtained from the Centre for Remote Sensing and Geographic Information Systems (CERSGIS), Accra. Costs of destroyed crops and livestock were estimated. In the case of seed crops, cost of crops per acre was calculated as the number of maximum-size bags of  seeds  obtainable  from  a  hectare  of  farm multiplied by the unit cost of bag. In the case of tubers, cost of yam (Dioscorea species)  was calculated as the average number of 100 tubers that can be obtained from a hectare of a farm multiplied by the unit price; unit price of yam calculated from the cost of 100 tubers of yam. Cassava (Manihot esculenta) is priced in terms of a full load of a maximum-size sac and the unit price multiplied by the number of sacs obtainable from a hectare of farm gives an estimate of loss per hectare. Unit prices of one bag of a seed of a crop, 100 tubers of yam, and a sac load of cassava were obtained from the District Agricultural Extension Office of the area of study.

 

Interviews

Structured questionnaires requiring “yes” or “no” answers were administered in the affected communities. The questions were designed to seek the views of the local people on their levels of acceptance for the National Park by allowing management regulations to prevail to protect wildlife (acceptance)  and  tolerance to the raids made by wildlife from the National Park irrespective of the attendant problems (tolerance). This was necessary to find out whether attitudes of the local people to the protected area were changing over a time span from 2005 when the study began to 2012 in the face of financial losses to farmers through crop raids by wildlife. 50% of the questions were intended to test the respondents on their acceptance and 50% on their tolerance. Each question was carefully designed in such a way that if a “yes” or “no” response was chosen it would always indicate a positive or negative tolerance, or a positive or negative acceptance, and each question was weighted in such a way that the percentage tolerance or acceptance could be calculated on a 100-scale for each respondent. The questionnaires were administered once every year by wildlife students from the University of Cape Coast. It was assumed that the results reflected the views of the local people throughout the year irrespective of which month they were interviewed. In each community, only heads of 20 households were interviewed. Households selected included those who had been affected by wildlife raids before; and if they were  not  up  to  20,  the  rest  were chosen randomly. The average percentages of responses indicating acceptance or tolerance were calculated for each year.

 

Analyses of data

Using MINITAB (v14), differences in seasonal raid occurrences (rainy or dry) and time of day (day or night) were tested by chi-square to find out whether they were due to chance or other factors. Correlation between the number of a particular species of animal that raided and the distance of raid occurrence from the park’s boundaries was analyzed with MINITAB (v14). Spearman’s rho was preferred for not requiring to be represented by a linear relationship (Myers and Well, 2003). The inferential Levene's test of homo-scedasticity (Zar, 2010) was used to ascertain if variances were equal for events around all the four ranges and Welch’s F-test which does not require equal variances (Welch, 1951; Field, 2009) was used to analyze the significance of the differences in variance.

 

Multiple regression

In order to ascertain how decisions of the local people were influenced, multiple regressions were carried out in SPSS v17.0 using the enter method. This was preferred to the stepwise methods (forwards, backwards and stepwise), because there is sound theoretical literature on which to base the model, precluding the need to rely on the computer to select variables based on mathematical criteria (Field, 2009; Groom and Harris, 2009). In addition, this method consistently produced the best model with regard to several different criteria (Groom and Harris, 2009). ‘Level of communities’ tolerance for wildlife’ was entered as the dependent variable (outcome) and six other variables, namely, ‘Years of study’, ‘period of time since the study began in 2005’, ‘estimated cost of raid’, ‘frequency of raids’, ‘number of farmers affected’ and ‘level of communities’ acceptance of the National Park’ specified hierarchically as the independent variables (predictors) in the second block; the first block specified by either ‘period of time since study began in 2005’, ‘estimated cost of raid’ or ‘level of communities’ acceptance of the National Park’ in turns for three separate outputs.

Assumptions of regression were checked to ensure that the model generalizes beyond the sample (Field, 2009). Variance inflation factor (VIF), tolerance (1/VIF) and eigenvalues were calculated to check the assumption of no perfect multicollinearity between predictors (Field, 2009). Assumptions of linearity, normality and homoscedasticity were ascertained by a histogram and a normal Q-Q probability plot. The case summaries of the final models (including Mahalanobis distances, Cook’s distance values, Leverage values  and  DFBeta  values)  were  examined  to  ensure there were no individual points which were having an especially strong influence on the model (Groom and Harris, 2009; Field, 2009).

The output of graph data by SPSS’s chart builder indicated a perfect collinearity between “Years of study” and “Period of time”. Collinearity was also observed between “Years of study” and “Number of farmers affected” and between “Cost of raids” and “Frequency of raids”. The regression correlation was significant in each of these pairings (Table 1). Collinearity issues made it necessary to remove “Years of study”, “Number of farmers affected” and “Frequency of raids”. Only one of “Years of study” and “Period of time” would be useful in this model and the “Period of time” was preferred for being enough as a continuous variable. Also, “Cost of raids” was preferred to “Number of farmers affected” because eigenvalues suggested “Number of farmers affected” appeared to be either co-linear with some other variables or non-linear enough for a linear regression model. The final model thus has “Period of time”, “Cost of raids” and “Level of communities’ acceptance” as the predictors. 

 

 

Crops and livestock raids and estimated costs

The study recorded 239 raids over the eight-year period of study. These included 225 crop raids and 14 livestock raids (Table 2). All the crop raids reported were solely elephant activities that destroyed yam, cassava, plantain, maize, rice, guinea corn, groundnuts, millet and cowpea, and these occurred around all the four ranges, though raids occurred more frequently around the headquarters than the other ranges of the park (Figure 3). Larabanga community recorded the highest of 31.6% of all the crop raids followed by Mognori with 20.4%, Murugu 12.9%, Dusie 12.4%, Chasia 12%, and the other communities, namely, Jinfronu, Jang and Jentilpe altogether recording 10.6% with raids occurring only once at Jang and Jentilpe communities (Table 2). Livestock were raided fourteen times by leopards and hyenas in only two communities, Kananto (three times) and Mognori (11 times), both around the headquarters range, and only goats were casualties in all these raids.

 

 

 

The most frequently raided crops were yam and cassava which were raided 105 and 76 times, respectively around the park. Maize followed with 19 times, guinea corn 11 times, and each of the other five crops was raided less than ten times (Table 3). Even though the trend of raid showed an initial discrepancy and was unpredictable from 2005 to 2008, raids have been predictable and increasing from 2008, and very steeply in the last two years of the study (Figure 4); 2012 recorded the highest of at least three raids per month on average, showing an average increase of 1 raid per month over the 2011 raids. It is suggested that the trend of raids during the study corresponds with the trend of population growth of the elephants over the period; and that increases in raids by the elephants indicate increasing population of the elephant. The difference in crop raids between the rainy and dry seasons was significant (c²=26.372, df=6, p= 0.0002). Only 12.1% of crop raids occurred during the dry season; the remaining 87.9% suggests that elephants preferred to raid in the rainy season during which most crops are available. Also, the difference between diurnal raids and nocturnal raids was significant (c²=22.393, df=6, p=0.0001). Only 4.7% of the crop raids occurred during the day; the remaining 95.3% suggests that elephants from the park raided at night. The results of Levene’s test, p=0.0024, rejected the hypothesis of equal variance for events around the four ranges of the park and Welch F test was not significant (F_{3, 27}=7.89, df=11.05, p=0.0043); thus, chance alone may account for the differences in raid distribution around the various ranges of the park. There was a very weak negative correlation between distance of raid occurrence from the park’s boundary and the number of elephants that   raided   (Y=  -0.0504x  + 5,   R=   -0.038);   and    as   Spearman’s rank correlation (Rho = 0.082, p=0.4060) indicated a no significant relationship, the distance of raid occurrence from the park’s boundary could not be said to depend on number of elephants that raided. Also, the results of t-test indicate a significant difference between the communities’ populations and communities’ nearest distances from the park’s boundaries (t = 6.5057; p < 0.001). But communities’ estimated populations and nearest distance of communities from the park’s boundaries were poorly correlated and not significant; and so is the regression equation (Y = 0.003x + 17.601; r = -0.091105, R² = 0.008; p > 0.05). With such a small contribution by raid events (0.8%) accounting for the relationship between the two variables, it could be said that some factors other than raid events may account for the differences between them. Also, the results for only those communities affected by raids indicate a significant difference (Y = 0.003x + 0.797; t = 2.77; r = 0.5; R² = 0.25; p < 0.05), implying that there is a positive correlation between the estimated population of the communities affected by raids over the study period and the communities’ nearest distances from the park’s boundaries. 25% of raid events account for the relationship between the two variables but other factors may be more important in explaining the relationship between them. The relationship between estimated populations of communities raided over the study period and the frequencies of raids is also significant (Y = 0.022x – 0.672; t = 8.159; r = 0.86; R² = 0.743; p < 0.001). This suggests a strong correlation between the two variables. About 74% of raid events account for the differences. It could be said that the frequency of raids depend on the population size of the community raided. This may be due to increase in the number of farms, the expanse of farms and the diversity of crops in the farms that may accompany large number of farmers in large communities. It implies also that small communities are at a lower risk of crop raids around the park. However, there is no significant difference between the nearest distance of communities from the park and the frequencies of raid occurrence (Y = 0.785 + 7.304; t = 0.931; r = 0.19; R² = 0.036; p < 0.05).

 

 

 

Elephants raided a total surface area of 85.7 ha over the seven-year study period and destroyed crops worth Cedi (¢) 249,504.00 ($77,365.58) at a rate of ¢1,917.79 per hectare and ¢31,188.00 ($9,670.70) per year (Table 3). Yam alone was destroyed over a total surface area of 39 ha, which amounted to an estimated cash value of ¢140,400 ($43,582.99), and recording a loss of ¢17,550.00 ($5,441.86) per year that accounted for 56.3% of the annual loss of crops due to elephant raids. Cassava followed with ¢10,582.50 ($3,281.24) annual loss representing 33.9%; maize, 4.2%; and all of the rest accounting for less than 5.6% (Table 3). Each of the 14 goats raided cost ¢1,200.00 ($372); on average that amounted to ¢1,680.00 ($520.9).

 

Interviews

The average percentage level of communities’ support for the park and average level of communities’ tolerance to wildlife raids were 74.2% (range = 65.6 - 81.3%) and 18.9% (range = 10.1 - 31.1%), respectively. Out of the 180 households interviewed, 94 (52.2%) have not experienced crop or livestock raids by wildlife before. This suggests that some raids were not reported. For example, at Jang, reported cases involved one household but results of the questionnaire indicated that 6 households had been raided before within the time frame. These erratic circumstances in the data, however, do not invalidate the analyses and results.

 

Multiple regression

The R square values from the model summary statistics of three separate outputs indicate that “Period of time” is the most significant predictor among the variables used alternatively as model 1 predictors in the hierarchical method. It accounts for 89.3% of the variation in tolerance to wildlife raids and this is significant (p<0.001). The other two variables (when included in model 2) could only account for an additional 3%. “Cost of raids” also accounts for a significant 19.5% of the variation (p<0.001) and the other two variables when included in model 2 account for an additional 72.8% that explained quite a large amount of the variation in communities’ tolerance to wildlife raids. “Level of communities’ acceptance” accounts for 0.3% but this is not significant (p>0.05) and the F change statistic is also not significant (F∆ (94) = 0.240, p=0.625). The F change statistic is significant (F∆ (94) = 22.769, p<0.001) when “Cost of raids” is the initial predictor, and also significant (F∆ (94) = 786.965, p<0.001) for the model in which “Period of time” is the initial predictor. With such a large F change statistic which is very highly significant, the importance of the variable “Period of time” in predicting the outcome (that is, communities’ tolerance for wildlife raids) cannot be overemphasized. The adjusted R Square value is very close to the R Square value; the difference for the final model is small (0.002), suggesting that, as a characteristic of a good model, if it were derived from the population rather than a sample it would account for approximately 0.2% less variance in the outcome. 

From the output of coefficients of the model parameters (Table 4), as the period of time increases by one standard deviation (27.857 months) (Table 5) tolerance decreases by 0.977 standard deviation (Table 4). The standard deviation for tolerance is 6.2379 (Table 4) and this constitutes a change of 6.09% (that is, 0.977 × 6.2379). Therefore, it could be said that for every 27.857 months, tolerance reduces significantly by 6.09% if the effects  of   cost   of   raids   and   level   of   communities’ acceptance of the park are maintained constant. The same way, every ¢3313.56 ($1027.46) loss of crops increases tolerance by 1.25% (Table 5) but this is not significant. Also, every 5.2374% (Table 5) increase in communities’ acceptance of the park decreases communities’ tolerance to wildlife raids by 1.1% if the effect of the other variables remains constant. But this can be ignored as there is no correlation between communities’ tolerance to wildlife and the communities’ tolerance to wildlife raids (Table 1).

 

 

 

The final model in which the “Period of time” is the model 1 predictor is successful in predicting the level of communities’ tolerance to wildlife; this model is a good fit and is generalized. The confidence interval is small and does not cross zero (-.227 to -.195). Furthermore, the maximum VIF value is 1.315, the average is 1.157 and the average tolerance (1/VIF) is 0.864. It is a rule of thumb that for no multicollinearity the  VIF  values  should not be greater than 10 (Bowerman and O’Connell, 1990; Myers, 1990), the average VIF should not be substantially greater than 1 (Bowerman and O’Connell, 1990) and the 1/VIF should not be less than 0.2 (Menard, 1995). Therefore, the model does not violate the assumption of no multicollinearity. It is reasonable to expect about 5 cases (5%) out of the 96 samples to have z-scores outside the limit of ±2 and 1% outside the limit of ±2.5 (Field, 2009). This model listed 4 cases (4%) outside the limit of ±2, but none outside ±2.5 (Table 6). Therefore, the results of these diagnostics give no real cause to worry about extreme residual cases and so the sample appears to conform to what we would expect for a fairly accurate model.

 

 

In all cases, Cook’s distance is <1 suggesting that none of the cases has undue influence on the model (Cook and Weisberg, 1982). Hoaglin and Welsch (1978) recommends, as a measure of  influence,  that  twice  the  average leverage value {2(k+1)/n; k=number of predictors and n=number of cases} and Stevens (2002) recommends three times the average as cut-off points for identifying cases having undue influence. Three cases (91, 92 and 93) were identified as greater than the limit of Hoaglin and Welsch (1978) and two (92 and 93) were greater than the limit of Stevens (2002). Mahalanobis distance criteria (Barnett and Lewis, 1978) also identified 92 and 93 cases as being too far from the means of predictor variables. By the criteria of Belsey et al. (1980), a case with a covariance ratio lying outside a minimum of 0.875 and a maximum of 1.125 may have an influence on the variances of the model parameters. Again, only cases 92 and 93 are outside the upper limit. All DFBetas lie between ±1 and this confirms that cases do not have undue influence on the regression parameters. However, the evidences suggest that of all the 96, only cases 92 and 93 probably need to be re-examined in order to confirm that there are no influential cases within the data. 

In designing integrated conservation projects, an understanding of the relationship between local people and protected areas is critical. In particular, it is important to understand the conservation attitudes of local people. Protected area managers have traditionally relied upon law enforcement techniques to resolve conflicts with local people. However, given the nature of these conflicts in Africa, such techniques will be insufficient and in many cases inappropriate. Alternative approaches to reduce conflict will need to be developed that provide tangible benefits to local communities and empower local people to manage natural resources (Govan et al., 1998). Such programmes are now being developed in Africa (Martin, 1984; Lewis et al., 1990), but before they can be designed and implemented successfully, the attitudes and problems of the local people, their relationships with protected areas, the essence of the programme and possible trade-offs must be clearly understood. For instance, a serious consideration of how much annual loss of local people’s crops and livestock to wildlife in relation to how much the local people benefit from a protected area need to balance in an agreement to a sustained collaborative conservation.

This study has confirmed that living next to a protected area can be very expensive, especially when farming is the main occupation, and can be very frustrating with the progressive trend of the number of farmers affected, the frequency of raids and estimated costs. Moreover, most raids take place at night that make preventive measures difficult or expensive to implement as elephants raided mostly during the rainy season. For example, the popular chili-pepper method (Osborn, 2002; LeBel et al., 2010; Hedges and Gunaryadi,  2010;  Monney  et  al.,  2010),  if  implemented as a deterrent to prevent elephant raids, is limited in the rainy season when the rain washes away the pepper. Therefore, with such heavy annual crop and livestock losses, it would be worth thinking of compensatory gains in other forms. For the fact that most of the communities are poorly accessible, their remoteness further severely limits access to services of development agencies and depends almost entirely on natural resource base for food and income; such gains in the form of social amenities, scholarships, alternative livelihood ventures, etc., are worth considering and these should be on the government agenda instead of agencies. Unfortunately, this study failed to quantify these gains that would have facilitated sound comparisons and assessments. However, loss of tolerance over the time span of the study provided a clue of what could be missing in this assessment. Contradictory results of percentage level of communities’ tolerance to wildlife raids in relation to percentage level of communities’ acceptance of the park probably suspect some deception to have characterized the results of acceptance. While it was expected that the relationship would be positively linear, both either reducing or increasing over time, this study rather revealed a negative relation in which tolerance reduces and acceptance increases. Though this was not significant and also as acceptance contributed ignorable and little account for tolerance, it could be suggested that the people revealed their true sentiments in respect of tolerance rating but not for acceptance. Also, with such a heavy financial loss through crops and livestock raiding every year, the expectation for estimated cost to account for a high percentage contribution of tolerance should be normal. However, instead, a small and non-significant contribution was revealed. Importantly, less than 50% of the households interviewed had been raided before and this may suggest that the majority of households did not bother much about the cost of raids. This also may imply that if raids become extensive to involve many farmers, tolerance will reduce markedly. Censuses carried out before the study (Wilson, 1993; Bouché, 2002, 2006; Mackie, 2004) published by IUCN (2010) indicated reducing elephant population from 1993 to 2004 and increasing population from 2004 to 2006. Assuming that increased elephant population would increase crop raids by elephants and decreased elephant population would decrease crop raids, then, if management would step up conservation efforts for the elephants to increase their population, there is the likelihood that there would be more raids and reduced tolerance of farmers. Indeed, elephant raids increased over the period of study, and though censuses have not been carried out currently, this observation lends credence to the prediction that raids will further increase with increasing population after 2006. However, because “period of time” and “estimated cost” did not account for all the tolerance, it could be suggested  that the model probably missed some variables outside the domain of this study. Such variables as mistrust for the protected area employees could possibly account for loss of tolerance as observed elsewhere in Tanzania (Songorwa, 1999) where the local people oppose the suggested abolishment of adjacent protected areas, view poachers as law-breakers, but generally do not hold positive attitudes towards protected area employees mainly due to mistrust (Songorwa, 1999). Comparably, around MNP, it could be suggested that the local people approved the park’s establishments, but harbor bitter sentiments; and the odds are always against the protected area employees who may be innocent.

Collaborative management is the recognition that people will manage wildlife and other resources when they are given sufficient economic incentives to do so. Through collaborative management, the wildlife division tries to provide the right conditions and incentives for people to manage their resources sustainably, according to Songorwa (1999), the decision to accept and join collaboration for wildlife conservation could be largely influenced by promises of socioeconomic benefits. But if these promises remain unfulfilled, expectations not met, the problems wildlife had long caused the communities remain unsolved and some existing problems such as crop damage increasing and new ones emerging (Songorwa, 1999), then communities can hold strong mistrust for the management. Various protected area participatory platforms have emerged around MNP to enhance wildlife conservation through management-community collaboration. These include Protected Area Management Advisory Board (PAMAB), Protected Area Management Units (PAMAU) and Community Resource Management Committees (CRMC). Common objectives of all these are to identify and integrate local people’s concerns into park management; to win local support for park management and wildlife; to collaborate with local people to try to ensure better park management and to reduce conflicts relating to the park and its natural resources. In relation to these, CREMAs have been identified to be a sustainable land use option to secure community resources and judicious use in few communities; CBT ventures offering canoe safari, village walk and cultural drumming and dancing have been based in one of the 33 fringed communities to provide permanent and temporary employment to over 30 community members. Perhaps these projects have been operational for too short a period to be adequately evaluated as observed elsewhere (Newmark et al., 1994; Kiss, 1990; Wells et al., 1992); the communities have to wait for as long as the CREMAs and CBTs are not viable while they still suffer heavy crop losses. It is plausible that the communities did not understand these programs from the beginning and have run out of patience soon. It seems, however, that the government should shoulder the  responsibility  of  providing  long-term   and   tangible  benefits to such communities, while management opens up for donor support.

The potential for such mistrust to influence tolerance for wildlife raids around the Mole National Park may be latent, but precarious and in consequence the local people could re-evaluate their decision and start opposing the park and employees. This study missed any chance of evaluating temporal variation in prices of raided crops and livestock over the study period. This would be necessary to establish the effect loss of tolerance could have on the market values of the products which are raided, how this could impact on the economy of the affected people and how it feeds back unto the tolerance of the people in the communities around the park; though it is indubitable that only negative consequences could emanate from this evaluation. The study also missed the opportunity to find out what the local people could do in reaction to the raids under very critical low tolerance level but suffice it to say that these must be avoided. Therefore, it is suggested that the plan for management-community collaboration should be reviewed taking cognizance of the local people‘s views and concerns and, especially, the area of mistrust, if any should be addressed; some long-term and tangible benefits should be proposed for government imple-mentation and donor support in fringe communities. Evidence from this study suggests that review measures need expedited action because tolerance reduces yearly as period of time progresses.

The author(s) have not declared any conflict of interests.