Welcome

Flaws in a method com­monly used in cen­suses of tigers and other rare wildlife put the accu­racy of such sur­veys in doubt, a new study suggests.

A team of sci­en­tists from the Uni­ver­sity of Oxford, Indian Sta­tis­ti­cal Insti­tute, and Wildlife Con­ser­va­tion Soci­ety exposes, for the first time, inher­ent short­com­ings in the ‘index-​calibration’ method that means it can pro­duce inac­cu­rate results. Amongst recent stud­ies thought to be based on this method is India’s national tiger sur­vey (Jan­u­ary 2015) which claimed a sur­pris­ing but wel­come 30 per­cent rise in tiger num­bers in just four years.

The team urges con­ser­va­tion prac­ti­tion­ers to guard against these sources of error, which could mis­lead even the best con­ser­va­tion efforts, and sug­gests a con­struc­tive way for­ward using alter­na­tive meth­ods of count­ing rare ani­mals that avoid the pit­falls of the index-​calibration approach.

A report of the research is first pub­lished online on 23 Feb­ru­ary in the jour­nal Meth­ods in Ecol­ogy and Evo­lu­tion.

This approach has been pop­u­lar among wildlife con­ser­va­tion agen­cies to gen­er­ate ani­mal num­bers at a regional and national level. These num­bers are then used to inform con­ser­va­tion efforts and direct resources worth mil­lions of pounds.

To inves­ti­gate index-​calibration the research team cre­ated a math­e­mat­i­cal model describ­ing the approach and then tested its effi­ciency when dif­fer­ent val­ues, rep­re­sent­ing vari­a­tions in data, were inputted. Under most con­di­tions the model was shown to lose its effi­ciency and power to pre­dict. The team then tested this math­e­mat­i­cal model on a real world exam­ple: attempt­ing to derive tiger num­bers from field­work data. The index-​calibration model was shown to be unre­li­able again, with any high degree of suc­cess shown to be down to chance, rather like being dealt a sin­gle incred­i­bly ‘high value’ poker hand, that could not be replicated.

Arjun Gopalaswamy, lead author of the report from the Wildlife Con­ser­va­tion Research Unit at Oxford University’s Depart­ment of Zool­ogy, said:

Arjun added: “Index-​calibration relies on the assump­tion that detec­tion rates of ani­mal evi­dence are high and unvary­ing. In real­ity this is nearly impos­si­ble to achieve. Instead, there are many flex­i­ble approaches, devel­oped over the past decade by sta­tis­ti­cal ecol­o­gists, which can cut through noisy ‘real world’ data to make accu­rate predictions.”

Dr Ullas Karanth, a co-​author from the Wildlife Con­ser­va­tion Soci­ety, and a mem­ber of India’s National Tiger Con­ser­va­tion Author­ity, said: “This study exposes fun­da­men­tal sta­tis­ti­cal weak­nesses in the sam­pling, cal­i­bra­tion and extrap­o­la­tions that are at the core of method­ol­ogy used by the Gov­ern­ment to esti­mate India’s num­bers, thus under­min­ing their reli­a­bil­ity. We are not at all dis­put­ing that tigers num­bers have increased in many loca­tions in India in last 8 years, but the method employed to mea­sure this increase is not suf­fi­ciently robust or accu­rate to mea­sure changes at regional and coun­try wide levels.”

Pro­fes­sor Mohan Delam­pady, a co-​author from the Indian Sta­tis­ti­cal Insti­tute, said: “The find­ings have wider con­se­quences for sev­eral applied sci­ences where sam­pling and direct extrap­o­la­tion is involved, espe­cially when sam­pling errors are influ­enced by unknown detec­tion probabilities.”

David Mac­don­ald added: “Index-​calibration can work well, if the cor­re­la­tions are tight and con­sis­tent, but often they aren’t, and many of us, myself included, for exam­ple in the con­text of esti­mat­ing num­bers of mink and water voles in the UK, have been using the tech­nique with­out appre­ci­at­ing its risks. Our inten­tion is to help con­ser­va­tion­ists by high­light­ing the con­di­tions when index cal­i­bra­tion can be mis­lead­ing. Every­body will ben­e­fit from greater accu­racy when it comes to count­ing rare animals.”

The team says that the aim of the study is to help ecol­o­gists and con­ser­va­tion­ists to address the global chal­lenge of count­ing rare and elu­sive ani­mals. The good news is that the math­e­mat­i­cal model cre­ated by the team pro­vides the cru­cial ‘link’ between some of the older meth­ods (which don’t esti­mate detec­tion rates) with some of the newer meth­ods (which do esti­mate detec­tion rates). The find­ings will help in the reanaly­sis of raw data from wildlife research. The study also rec­om­mends that esti­mates from future sur­veys will be most reli­able if designed, a pri­ori, keep­ing in mind the power of mod­ern, robust, mod­el­ling approaches.

(Source: Uni­ver­sity of Oxford news release, 23.02.2015)