Background Information
Posted: Wed May 29, 2013 3:24 pm
Background Information: ESTIMATING RHINO NUMBERS
Link: RAGE
Estimating the abundance of species in a particular area of interest presents significant challenges. There is ample room for error and this prevents conservationists from obtaining exact counts. For this reason several techniques are employed including strip-transects, block counts, distance sampling, dung counts, mark-recapture techniques, call-up surveys, registration studies and total counts. In many instances with regard to large mammals, aerial surveys are a key approach. These primarily make use of strip transects, distance sampling or total counts depending on the size of the area. The smaller the area, the more likely that total counts will be used. In the two rhino species' cases the following apply:
White rhino:
Annual survey of Kruger National Park as part of the herbivore survey with 100% coverage (total count) from 1977 to 1993 as well as 1997
Annual survey of Kruger National Park as part of herbivore survey, but only some southern parts from 1994 to 1996
Annual survey of Kruger National Park as part of the herbivore survey using statistical survey sampling techniques (i.e. distance sampling that covered between 15-22% and stratified Jolly-Seber fixed-width transect sampling with 7.5-14.9% coverage) from 1998 to 2008 as well as during 2010.
Black rhino:
Annual survey of Kruger National Park as part of herbivore survey with 100% coverage (total count) from 1977 to 1993 as well as 1997
Annual survey of Kruger National Park as part of the herbivore survey, but only some southern parts from 1994 to 1996
Annual survey of Kruger National Park as part of the herbivore survey using statistical survey sampling techniques (i.e. distance sampling that covered between 15-22% and stratified Jolly-Seber fixed-width transect sampling with 7.5-14.9% coverage) from 1998 to 2008 as well as during 2010.
The above are all aerial based, making use of primarily fixed-wing aircraft as observation platforms. SANParks have evaluated the applicability of approaches such as distance sampling using such observation platforms.
With regard to black rhino, rhino specialists suggested alternative approaches which have been tried as follows in Kruger National Park.
Trial block-based survey using 3kmx3km blocks intensely searched which is equivalent to a Jolly-Seber fixed-width transect sampling design in the areas South of the Sabie River during 2007.
Block-based survey south of the Olifants River with coverage of 21.5% of the area during 2008. The results of this survey and a summary of rhino trends have been accepted for publication. SANParks also noted white rhinos in this survey which allows for an additional estimate independent of the fixed-wing-based aerial survey.
Challenges when estimating animal abundance
Often biologists' survey a hundred percent of an area using a helicopter or sometimes use a fixed-wing platform for observation. Hundred percentage coverage of an area is usually referred to as a total count - this inherently assumes that it is a near exact estimate of the number of individuals of a specific species. The measure of how close an estimate is to the real number of individuals in a population is referred to as accuracy. Accuracy, however, has two components - bias and precision. Bias originates from several sources, but is captured in three broad types.
Availability bias – in this case animals are present on the landscape, but not available to be sampled. For example, koedoes under a tree, whales not at the surface of the sea and crocodiles in deep pools at the time when a helicopter or fixed-wing airplane flies over them. This type of bias is often also referred to as concealment bias, particularly in the crocodile literature for example.
Detectability bias – in this case animals are present and available, but there is considerable variation in detecting them. For instance, elephants that have just had a mud-bath walking across a black clay patch with no trees will be harder to detect at a distance than ones that had a mud-bath walking over green grasslands directly below the airplane.
Observer bias – even though availability bias and detectability bias may be accounted for, the third bias involves observers. Typically, observers have different capabilities. Some observers can handle flights better; have better eyesight and also the ability to concentrate for longer periods of times.
These three biases accumulate uncertainty in addition to the second component of accuracy – Precision is the likely spread of estimates given the uncertainties introduced by biases. An additional source of error comes from sampling, particularly when surveyors are not covering hundred percent of an area of interest. Individuals are not spread homogeneously across an area – some sample units may thus have lots of individuals and others few. This is captured in standard errors of an estimate, the normal statistical description of a mean and the data that supports that mean.
Rhino estimates
Given the above I summarize rhino estimates below for your interest.
The discrepancies in the fixed-wing strip transects and distance sampling estimates are of major concern to SANParks and originate from relatively low survey intensities which ranged from 12.8% to 14.9% coverage of Kruger National Park for the fixed-wing strip transects. For elephant survey intensities that define accurate estimates requires at least 5-20% coverage, but 50% coverage for precise estimates. The rhino block count surveys covered 21.7% of the area south of the Olifants River and will provide the most accurate and precise estimates for white rhinos. Given that SANParks encountered 89.8% of rhino observations during 2008 and 90.6% of observations during 2010 south of the Olifants River in the fixed-wing based aerial surveys, I can translate the 2009 block-based survey south of the Olifants River to a total park based estimate of 12027 (95% CI: 11318-12736). This estimate’s confidence interval overlaps with that estimated using Distance Sampling in 2010 (10544, 95% CI: 8743-12714), hence the indication that approximately 9000 to 12000 white rhinos live in Kruger.
SANParks will now conduct a short study to define the optimal survey requirements for white rhinos annually given the threats posed to white rhinos at present and conduct annual estimates in this way.
December 2011
Dr Sam Ferreira, Large Mammal Ecologist, SANParks
Link: RAGE
Estimating the abundance of species in a particular area of interest presents significant challenges. There is ample room for error and this prevents conservationists from obtaining exact counts. For this reason several techniques are employed including strip-transects, block counts, distance sampling, dung counts, mark-recapture techniques, call-up surveys, registration studies and total counts. In many instances with regard to large mammals, aerial surveys are a key approach. These primarily make use of strip transects, distance sampling or total counts depending on the size of the area. The smaller the area, the more likely that total counts will be used. In the two rhino species' cases the following apply:
White rhino:
Annual survey of Kruger National Park as part of the herbivore survey with 100% coverage (total count) from 1977 to 1993 as well as 1997
Annual survey of Kruger National Park as part of herbivore survey, but only some southern parts from 1994 to 1996
Annual survey of Kruger National Park as part of the herbivore survey using statistical survey sampling techniques (i.e. distance sampling that covered between 15-22% and stratified Jolly-Seber fixed-width transect sampling with 7.5-14.9% coverage) from 1998 to 2008 as well as during 2010.
Black rhino:
Annual survey of Kruger National Park as part of herbivore survey with 100% coverage (total count) from 1977 to 1993 as well as 1997
Annual survey of Kruger National Park as part of the herbivore survey, but only some southern parts from 1994 to 1996
Annual survey of Kruger National Park as part of the herbivore survey using statistical survey sampling techniques (i.e. distance sampling that covered between 15-22% and stratified Jolly-Seber fixed-width transect sampling with 7.5-14.9% coverage) from 1998 to 2008 as well as during 2010.
The above are all aerial based, making use of primarily fixed-wing aircraft as observation platforms. SANParks have evaluated the applicability of approaches such as distance sampling using such observation platforms.
With regard to black rhino, rhino specialists suggested alternative approaches which have been tried as follows in Kruger National Park.
Trial block-based survey using 3kmx3km blocks intensely searched which is equivalent to a Jolly-Seber fixed-width transect sampling design in the areas South of the Sabie River during 2007.
Block-based survey south of the Olifants River with coverage of 21.5% of the area during 2008. The results of this survey and a summary of rhino trends have been accepted for publication. SANParks also noted white rhinos in this survey which allows for an additional estimate independent of the fixed-wing-based aerial survey.
Challenges when estimating animal abundance
Often biologists' survey a hundred percent of an area using a helicopter or sometimes use a fixed-wing platform for observation. Hundred percentage coverage of an area is usually referred to as a total count - this inherently assumes that it is a near exact estimate of the number of individuals of a specific species. The measure of how close an estimate is to the real number of individuals in a population is referred to as accuracy. Accuracy, however, has two components - bias and precision. Bias originates from several sources, but is captured in three broad types.
Availability bias – in this case animals are present on the landscape, but not available to be sampled. For example, koedoes under a tree, whales not at the surface of the sea and crocodiles in deep pools at the time when a helicopter or fixed-wing airplane flies over them. This type of bias is often also referred to as concealment bias, particularly in the crocodile literature for example.
Detectability bias – in this case animals are present and available, but there is considerable variation in detecting them. For instance, elephants that have just had a mud-bath walking across a black clay patch with no trees will be harder to detect at a distance than ones that had a mud-bath walking over green grasslands directly below the airplane.
Observer bias – even though availability bias and detectability bias may be accounted for, the third bias involves observers. Typically, observers have different capabilities. Some observers can handle flights better; have better eyesight and also the ability to concentrate for longer periods of times.
These three biases accumulate uncertainty in addition to the second component of accuracy – Precision is the likely spread of estimates given the uncertainties introduced by biases. An additional source of error comes from sampling, particularly when surveyors are not covering hundred percent of an area of interest. Individuals are not spread homogeneously across an area – some sample units may thus have lots of individuals and others few. This is captured in standard errors of an estimate, the normal statistical description of a mean and the data that supports that mean.
Rhino estimates
Given the above I summarize rhino estimates below for your interest.
The discrepancies in the fixed-wing strip transects and distance sampling estimates are of major concern to SANParks and originate from relatively low survey intensities which ranged from 12.8% to 14.9% coverage of Kruger National Park for the fixed-wing strip transects. For elephant survey intensities that define accurate estimates requires at least 5-20% coverage, but 50% coverage for precise estimates. The rhino block count surveys covered 21.7% of the area south of the Olifants River and will provide the most accurate and precise estimates for white rhinos. Given that SANParks encountered 89.8% of rhino observations during 2008 and 90.6% of observations during 2010 south of the Olifants River in the fixed-wing based aerial surveys, I can translate the 2009 block-based survey south of the Olifants River to a total park based estimate of 12027 (95% CI: 11318-12736). This estimate’s confidence interval overlaps with that estimated using Distance Sampling in 2010 (10544, 95% CI: 8743-12714), hence the indication that approximately 9000 to 12000 white rhinos live in Kruger.
SANParks will now conduct a short study to define the optimal survey requirements for white rhinos annually given the threats posed to white rhinos at present and conduct annual estimates in this way.
December 2011
Dr Sam Ferreira, Large Mammal Ecologist, SANParks
