Elephants do not raid at random. In his studies in southern India, Sukumar (1990) found that adult males entered cultivation far more often than members of female-led family groups, that raiding followed the calendar of crops such as finger millet and paddy as they matured, and that cultivated plants offered more nutrition than wild forage at the same season. The pattern has been interpreted as a high-risk, high-gain feeding strategy of males (Sukumar, 2003). Work on African elephants reached a compatible conclusion: the level of conflict in an area was poorly predicted by elephant numbers and seemed to depend on the behaviour of particular males (Hoare, 1999). Because raiding is tied to particular animals, crops and months, mitigation can be targeted in place and season.

In fragmented landscapes elephants also depend on small features. In the plantation-dominated Anamalai hills they were found to move along rainforest fragments and riverine vegetation, using them as refuges by day within a matrix of tea and coffee (Kumar et al., 2010). Loss of such cover pushes animals into the open and closer to people.

A wide range of measures has been tried, and reviews agree that none works everywhere or for long without upkeep (Fernando et al., 2008; Shaffer et al., 2019). Table 1 summarises the main types. Electric fences and trenches are effective while maintained and fail when maintenance lapses. Evidence for deterrents is mixed: beehive fences reduced crossings by African elephants in Kenyan trials (King et al., 2009), whereas a controlled comparison in Sumatra found that chilli-based deterrents added nothing to conventional guarding (Hedges and Gunaryadi, 2010). A study in Assam compared the interventions that villagers used against crop-raiding elephants and found that effectiveness differed among them and depended on the combination applied (Davies et al., 2011).

How people respond to elephants is not a simple function of what they lose. Dickman (2010) showed across many conflicts that attitudes are shaped by trust in authorities, by beliefs and by the sense of having a say, as much as by the size of the damage. Costs that never appear in a damage register, among them nights spent guarding, children kept from school, fear and the effort of pursuing a claim, can outweigh the crop loss itself (Barua et al., 2013). Studies around southern Indian reserves have documented both the extent of losses and the shortcomings of compensation as experienced by households (Karanth et al., 2013; Gubbi, 2012).

The corridor is mapped from evidence of use, such as dung, tracks, camera-trap records and the knowledge of forest watchers, and not from the notified boundary alone. Around it a conflict zone is drawn to take in every village and estate whose land elephants reach from the corridor. Within the corridor the plan identifies the points where passage is narrowest and the features, such as stream beds and patches of cover, on which movement depends.

Before any new measure is installed, incidents are compiled for several past years from compensation claims, forest range records and village registers, each with its date, place, crop or property and, where known, whether a lone male or a herd was involved. Official records undercount, because many losses are never claimed, so the baseline is supplemented by a simple incident log kept in each village from the start of planning. The same definitions and forms are used throughout the life of the plan, since a change in recording is easily mistaken for a change in conflict.

Incidents are plotted against fields, crops, water, cover and elephant paths to show where and when risk is highest. The map is then taken to the villages. Meetings are held separately with groups whose exposure differs, for example landless labourers who walk at dawn, women who collect firewood, and farmers at the forest edge, and each is asked what it fears, what it already does and what it would be willing to maintain. This stage settles which problems the plan must address first, and safety of people takes precedence over crop loss.

Measures are assigned by zone. The governing rule is that barriers run parallel to the direction of elephant movement and never across it. Fences and trenches are placed to protect the outer edge of settlements and fields and to guide animals along the corridor, with the passage itself left open. A barrier that closes a corridor displaces conflict to the next village and defeats the purpose of protection, a trade-off between mitigation and connectivity that has been set out formally for heterogeneous landscapes (Goswami and Vasudev, 2017). Close to the passage, the plan favours early warning, lighting of paths, safe timing of travel and, where farmers agree, crops that elephants do not eat. Each measure is recorded with the party responsible for its upkeep.

A measure with no owner decays. The plan names who maintains each fence section, who operates the warning system, who responds at night and how they are paid or supplied. It sets a procedure for relief payments with a stated time limit and a local point of contact, so that claimants are not required to travel repeatedly to a distant office. Funds for maintenance are budgeted for the whole planning period and not only for construction.

Monitoring follows three things: incidents, using the baseline definitions; the working condition of each measure, checked on a fixed schedule; and continued use of the corridor by elephants, including females with calves. A review each year compares zones and seasons with the baseline and decides what to keep, repair, move or abandon. Because conflict varies from year to year with rainfall and cropping, conclusions are drawn from several seasons and not from one.