SciELO - Scientific Electronic Library Online

Home Pagelista alfabética de revistas  

Servicios Personalizados




Links relacionados


Gayana (Concepción)

versión impresa ISSN 0717-652Xversión On-line ISSN 0717-6538

Gayana (Concepc.) v.68 n.2 supl.TIIProc Concepción  2004 

  Gayana 68(2): 578-585, 2004



E. Yáñez*, C. Silva, K. Nieto, M.A. Barbieri & G. Martínez

Escuela de Ciencias del Mar, Pontificia Universidad Católica de Valparaíso Casilla 1020, Valparaíso, Chile;


A work of gathering historical fishing (catch, fishing effort, CPUE, characteristics and operation of purseiner fleet) and satellite (NOAA, SeaStar and SeaWind) database of northern Chile, and fishing and oceanographic in situ sampling for validation purpose is made. With this information and the use of geographical information system (GIS) the distribution of pelagic resources and the associated environmental conditions (sea surface temperature, thermal gradients, chlorophyll concentration and wind fields) are analysed in order to develop a probable fishing ground (PFG) index prediction model. A PFG model of anchovy (Engraulins ringens) is operative for the northern Chile. The impact of the use of PFG charts was evaluated analyzing the operational efficiency of the involved purse seiner fleet. An availability index, independent from the present fishery activity, is estimated from the PFG index and images that depend on historical and present environmental conditions and historical catches. The management measures that assure the resources sustainability are present.

Key words: pelagic fisheries, northern Chile, remote sensing, GIS, PFG index and model.


Assessing of habitat selection is the result of diverse interactions in different spatial and temporal scales. It depends on biotic and abiotic factors. Anchovy is a pelagic fish located in a low trophic level of the marine food-web and is sensitive to environmental changes. Some factors responsible of fish habitat are easily predictable like circadian cycles, seasonal cycles and lunar periodicity, but other factors are difficulty predictable, like interannual variations and biotic factors.

resources support economically valuable purse seine fisheries in the Southeast Pacific, and are one of the major contributors to world fish production (FAO, 1997). The Chilean fishing activity is mainly sustained in the exploitation of small pelagic resources, such as anchovy, sardine, common sardine and jack mackerel, constituting more than 80% of total annual landing, mainly destined to the fish oil and fish meal.

The small pelagic are sensible to environmental fluctuations, affecting their distribution (Yáñez et al., 2001) and therefore the location of fishing grounds, reason for which the fishing industry must adapt its fishing strategies considering the environmental changes produced in different space and temporal scales. Relations between fishing parameter and environmental variables detected with satellites can be used to develop an expert system to assess probable fishing ground, which would decrease their search times and the operational costs in the fishing fleets (Nieto, 1999). In fact, environmental conditions and ocean features can be acquired with satellite-sensors, such as NOAA-AVHRR sea surface temperature (SST), SeaStar-SeaWiFS chlorophyll a (Chl_a) and QuikScat-SeaWinds ocean surface winds (OSW).

In this paper, PFG index and images are generated using remote sensing, GIS, mathematical modeling and a decision support system. The demonstration of the usefulness of the PFG images application is made during El Niño 1997-98 and La Niña 1999-2000.


Fishery database

Historical fishery database (1987-1998) were analyzed. The data come from operational records of purse seine industrial fleet of small pelagic resources that operated in the northern (18-24S; 70-73W) zone of Chile. The collected data provide information of georreferenced catches, fishing effort per day and vessel. The fishing information was used to calculate CPUE of anchovy, which was previously standardized using Lineal General Model (Yáñez et al, 1999). These daily CPUE data were mapped with GIS.

Satellite database

Historical database (1987-2000) period is composed of daily NOAA-AVHRR SST images, and thermal gradient (TGR) images derived from SST. These images were validated with in situ SST data obtained from fisheries oceanographic samples. The data is gathered in an HRPT satellite reception system operated in Remote Sensing and GIS (RS&GIS) Laboratory at ECM-UCV. Also satellite information from international cooperation project is acquired, like NASA, JPL, NAVO and GODDARD.

A 1999-2000 database of SeaStar-SeaWiFS Chl_a images are processed and analyzed. The data is acquired in the framework of SeaWiFS project, NASA Mission to Planet Earth Program. The Chl_a data was analyzed to explore association with c.p.u.e. (Silva et al., 2000). These images were validated with in situ Chl_a data obtained from fisheries oceanographic samples (Silva et al., 2000; Rocha, 2001).

Also, a 1999-2000 database of QuikSCAT-SeaWinds ocean surface wind (OSW) images are processed and analyzed.

In situ data

In situ data is gathered (1999-2000) through fishing and oceanographic samplings on board of purseiner vessels. To record data, a fishing and oceanographic spreadsheet is designed and used, which was filled by technicians and vessels captains in each trip. The recorded information indicate vessel name, date, geographical location and depth of fishing grounds, detection system, catch by species, SST (used to calibrate the satellite data), wind and ocean color. Also biological information about size structures and fish school description is registered. In situ measurements of Chl_a are made, mainly during bloom or high Chl_a concentration periods, with the purpose of SeaWiFS sensor validation. The collected data were analyzed in the Primary Productivity Laboratory ECM-UCV, using specialized equipment, like fluorimeter and spectrophotometer.

PFG prediction model

The PFG prediction model is based on relations between the environmental conditions and the resources distributions. Using the IDRISI GIS, daily anchovy CPUE distributions were superimposed to SST, TGR and Chl_a images to analyze associations between fishery and environmental variables. SST, TGR and Chl_a values for each fishing ground (grid of 10x10 nm) are extracted. The data were grouped monthly and given conditional probability distributions, determining the SST, TGR and Chl_a optimal ranges in the fishing grounds.

Conditional probability distributions are scaled between 0 and 1. This probability functions define the shape of the fuzzy set membership or evidence curve. Fuzzy Sets are sets (or classes) without sharp boundaries; that is, the transition between membership and non-membership of a location in the set is gradual (Zadeh, 1994). SST, TGR and Chl_a evidence images were generated applying the fuzzy logic to input images data of the PFG model according to the corresponding monthly evidence curve.

On the other hand, daily anchovy CPUE distributions were used to estimate the a priori knowledge of distribution and abundance of this resource, obtaining a representative a priori probability image for every month. The position of each pixel inside the area is evaluated and a value between 0 and 1 is assigned, depending on whether that pixel has registered high abundance levels in the past, and whether it has constantly been visited throughout the study period.

In order to obtain PFG images, the a priori and evidence images are integrated using the Bayesian theory approach. It is an extension of the Classical Probability Theory that allows combine new evidence about an hypothesis along with prior knowledge (or assumptions) in order to arrive to an estimate of the likelihood the hypothesis being true. The PFG images (a posteriori probability) for each evaluated environmental variable (SST, TGR and Chl_a) were calculated using the Bayesian theorem. This is described under a mathematical form by equation 1:

where h0 is the hypotheses is true, "it is a fishing ground"; h1 is the hypotheses is not true; p(h0/e) is the probability of fishing ground being true given the evidence, a posteriori probability; p(e/ h0) is the probability of finding the evidence given the fishing ground being true, the conditional probability; p(h0) is the probability of fishing ground being true regardless of the evidence, a priori probability.

The SST, TGR and Chl_a a posteriori probability images were integrated in one PFG image through a weighted linear combination.

The application of the PFG model, like an availability index, is demonstrated considering two environmental scenarios in an ENSO context: El Niño 1997-1998 and La Niña 1999-2000. The PFG model is applied during November 1997 and December 2000 in order to analyze simulation of the model during El Niño 1997-1998 and La Niña 1999-2000 conditions. A time series of weekly SST, Chl_a and TGR images is used as input of the PFG model. By the way, it is hoped to demonstrate the usefulness of the PFG model like an availability indicator of anchovy in both environmental conditions, considered like extreme situations within the interannual variability. A mean profile was extracted to the time series in the anchovy distribution (18-24°S ; 70-71.5°W), in order to extract the environmental and availability index.

The PFG model was validated with fishery operational information, taking into account the satellite images and georeferenced CPUE data collected during 2000. The validation was made applied a frequency analysis of coincidences between the grids with fishing and the high PFG.


Relationships between SST, TGR, Chl_a and CPUE

The period used to determine these relations and to evaluate the model (1987-2000) appears like positive phase for anchovy and negative for sardine. The Empirical Ortogonal Functions (EOF) time series analysis shows the alternate of predominance species (anchovy or sardine) on a decadal time scale (Fig. 1). The positive anomaly for anchovy after the mid 80's implied that the model is focused mainly in anchovy in the northern Chile (18-24S; 70-73W), target specie of the purseiner fleet.

Figure 1: EOF time series analysis of: a) catch, fishing effort, SOI and SST (in Arica coastal station, 18°28'S) for anchovy (1957-1999); and b) recruitment, biomass, SST and upwelling index (in Antofagasta coastal station, 2326'S) for sardine (1974-1995) (Yáñez et al., 2002).

Anchovy is distributed in SST ranging from 16 to 23C, the optimum range being between 19 and 20C, and in TGR from 0.3 to 3.5C/10nm with an optimum between 0.8 and 2.1C/10nm. In Chl_a the range were between 0.2 and 6 mg/m3, and the optimum range varied between 0.3 and 1.3 mg/m3 (Fig 2).

Figure 2: Frequency distribution of SST, TGR and Chl_a for anchovy

The optimum ranges vary monthly, giving warmer values during the summer months (January ­ March). Regarding the TGR, it is possible to observe that in the summer months the most frequent TGR values in fishing zones are higher than in the rest of the year. The wide range of SST is related to the greater differences in temperature produced between mid-spring and mid-autumn.

In winter (July ­ September), the fishing ranges with the lowest optimum SST values are recorded. The TGR begins to decrease towards the end of autumn, reaching its lowest values during the winter months. The lowest TGR values are likely to be associated with the fact that during the winter months there is a greater thermal homogeneity in the area, as the upwelling waters are more similar to the ocean water temperatures. In spring, (October ­ December), the optimum SST ranges in fishing zones begin to increase progressively once more, as does the area in which anchovy are found, due to the start of the upwelling season in northern Chile.

On the other hand, the anchovy shows a coastal distribution related to areas with steep gradients and high chlorophyll levels due to the permanent presence of coastal upwelling.

PFG images

The PFG model was applied on May 2000, using input satellite images received and processed in the RS&GIS lab ECM-UCV. The PFG image is the model output and indicates areas of high catch probability. The a priori knowledge of CPUE distribution of anchovy shows that this is located, preferentially, within 60 nm of the coast, and mainly within 30 nm. This behavior is more clearly observed in warmer months of summer and early autumn. In colder months, especially between late autumn to midwinter, a more oceanic distribution of the resources can be observed, concentrated in the Northeast of the study area. From spring onwards, the fishing zones return closer to the coast and are distributed throughout various areas of the zone of study.

An example of the PFG model application during 22 February 1998 (El Niño) and 22 February 2000 (La Niña), shows the spatio-temporal differences in the SST, Chl_a, TGR and PFG availability index in the two environmental conditions (Fig. 3). Variability in the distribution and spatial cover of the anchovy PFG was observed during ENSO conditions. Anchovy PFG simulations during El Niño 1997-1998 event, show a decrease in spatial cover of fishing grounds, concentrating in the first miles near the coast (Fig. 3a). Low catch probability dominate the area due to intrusion of warm and non-productive surface subtropical waters. On the other hand, the application of the model during La Niña 1999-2000 indicate that probable fishing grounds were distributed in all the area, but mainly in the first 60 nm, associated mainly to thermal fronts and chlorophyll blooms that form in specific coastal areas (Fig. 3b). During La Niña an increase of medium and high catch probabilities were simulated due a presence of thermal front area generated between the cold and productive upwelled waters and the oceanic waters.

Figure 3: Example of SST, TGR, Chl_a and PFG images during a) El Niño conditions of 22 February 1998 and b) La Niña conditions of 22 February 2000.

A profile was extracted from the time series (November 1997 - December 2000) of weekly SST, Chl_a, TGR and PFG images (Fig. 4).

Figure 4: Profiles of (a) SST, (b) TGR, (c) Chl_a and (d) anchovy PFG indices extracted from time series images of November 1997 to December 2000. The anchovy landings (e) are also shown and rectangles indicate the ban times.

PFG index estimated for a strip of 30 nm in the longitudinal sense is show in Figure 5. Is observed a high spatio-temporal variability, from the end of November 1997 to June 1998 in the coastal area, showing medium and low PFG, catches not surpass the 10 thousand of ton (Fig. 4e). From August 1998 when La Niña period begins, appears in the study area the high PFG, this is not reflected in the landings levels because ban time are imposed by the authority (Fig. 4e).

Figure 5: Spatio-temporal distribution of anchovy PFG along the coast (30 nm offshore) and during November 1997 - Dicember 2000. Color palette: blue (low catch probability) to green (high probability).

PFG model validation

The analysis of values collected during May-December 2000 (situation with project) shows that 74% of the fishing grids coincided with high probability grids, and 26% with the medium probability. In view of this, the model's validation allows a conclusion of the reliability of the data entered into the model and the correct integration of environmental variables that influence anchovy distribution.


In the present study, the probable fishing grounds of anchovy were modeled in a simple way. The model designed was implemented in a GIS, due to the large amounts of geographic data analyzed, and the necessary analysis to develop the prediction model and generate a PFG chart as the end information product. The PFG model is supported by past evidence of the spatial and temporal distribution of anchovy, and by the optimal ranges of SST, TGR and Chl_a recorded in the fishing zones. The advantage of a modeling scheme is that the PFG model allows the integration of more variables for the estimation of fishing grounds. Therefore, the gathering and analysis of other variables can be integrated, thereby reducing uncertainty in the results produced by the model. Also, a permanent monitoring of the fishery and results obtained when applying the model, would allow the use of this new evidence for the feedback of the PFG model.

The anchovy is caught in a wide range of SST, from 16 to 23C with an optimum of between 19 and 20C. However, these ranges vary according to season. The wide range of SST is related to the greater differences in temperature produced between mid-spring and mid-autumn. The high gradient values in fishing zones are related to the season of upwelling events within the year, with the latter's intensity being greatest in summer and early autumn (November ­ April) (Barbieri et al., 1995).

The anchovy shows a coastal distribution related to areas with steep gradients and high chlorophyll levels due to the permanent presence of coastal upwelling. These coastal upwelling generate a frontal zone produced by the convergence of upwelling cold water and the warm ocean waters (Silva et al., 2000).

Nevertheless the predominance of anchovy since mid 80's when analyzing their abundance in an interdecadal time scale (Yáñez et al., 2001), are observed changes in the resource availability when analyzing the variability in an interannual scale. Anchovy is specie with lower tolerance to oceanographic change on an interannual time scale such as El Niño events that produces extremely adverse effects on population (Blanco et al., 2002; Segura et al., 2002; Castillo et al, 1996). In fact, when analyzing the PFG distribution and intensity during El Niño events, this availability indicator shows a decrease in values and spatial cover of the probable fishing grounds (Figs. 3 and 4). This implies that anchovy is less available for fishing during El Niño years, which is reflected in landings levels. This behavior could be explained by a change in the horizontal and vertical distribution of the resource, Bouchon et al., (2002) indicates that in Peru anchovy schools were detected 10 meters below the normal range (0-30 m) during El Niño events, coincided with the deepening of thermocline. In Northern Chile the anchovy is located between 40 and 80 m depth and up 120 m in some areas (the normal is 20 to 40 m deeper), far out of the range of the commercial fishing (Blanco et al., 2002; Yáñez et al., 2002). The latitudinal anchovy distributions moved south ward along the coast. During el Niño 1997-98 a deepening of thermocline was observed, exceeding the 60 meters in August 1997 and the 200 m in 1998 (Fuenzalida et al., 1999).

The reduction of probable fishing ground as well as its smaller probability of occurrence demonstrate that during El Niño event is still more necessary to have a previous knowledge of anchovy catch location to know the spatial strategy of anchovy.

Spatial strategy of pelagic fish like anchovy, can be given by the relation between local density and population abundance (Swain and Sinclair, 1994; Freón and Misund, 1999). Mac Call (1976) and Ultang (1980) consider that catchability includes three components: technical (related to the fishing art efficiency), environmental (behavior with the fishing art) and spatial (spatially structuring the fish density). Petitgas (1994) conclude that the spatial component has the greater impact on catchability. The pelagic fish distribution is assessed preferably by environmental parameters.

The results indicate that after El Niño event, the environment begins to return to a normal condition, the PFG begin to extend his area and to stabilize the probability of fishing in high values. This agrees with the observed by Blanco (2002), which indicates that in this period the anchovy reappeared at their normal fishery locations and the biomass was relatively high. With the beginning of upwellings during La Niña, the stock begins his slow recovery.

Anchovy distribution is regulated by oceanographic parameters: biotic and abiotic, his regionalization varies in time and in distribution. Anchovy habitat depends on environmental conditions, the model considered one conservative variable like temperature and a biological productivity indicator like chlorophyll (indicate food availability), other variable considered are thermal gradients like an spatial variability indicator. High spatial and temporal variability observed on the PFG index allows the spatial occupancy strategy of resource in an area given the anchovy density is going to be variable, depending on habitat conditions.

To assess the PFG exist different type of index and models. First models were developed in mid 80's by Laurs (1989), Barbieri (1989), Stretta (1991) and Barrat et al., (2002) and were praxicologic models that consider the evolution of diverse environmental conditions in short time periods of days and weeks. The model proposed in this work is an expert system, based on the specialist knowledge, which applies the Bayesian approach for five years of information estimating a PFG index that is validated. Other forecasting models use the artificial intelligence methods focused on the simulation of animal behavior and the relations of the animal and its environment.



Barbieri, M.A., E. Yáñez, M. Farías & R. Aguilera (1989). Determination of probable fishing areas for the albacore (Thunnus alalunga) in Chile's central zone. In: Quantitative Remote Sensing: An Economic Tool for the Nineties. IGARSS '89, Vol 4, 2447-2450.         [ Links ] [1]

Barbieri, M.A, A. González & E. Yáñez (1995). Time-space variability of sea surface temperature and pelagic resources in northern Chile. IGARSS'95, Vol. I, 300-302.         [ Links ] [2]

Barrat, D., A. Bugg, S. Barry & B. Wise (2002). Modelling the distribution of sish resources using oceanographic variables, statistical methods and GIF technologies. Second International Symposium on GIS/Spatial Analyses in Fishery and Aquatic Sciences. University of Sussex, Brighton, UK, 105 pp.         [ Links ] [3]

Blanco, J.L., P. Barría, J. Castillo & L.P. Atkinson (2002). Response of anchovy (Engraulis ringens) off Northern Chile to the 1997-1998 El Niño-La Niña sequence. Invest. Mar., Valparaíso, 30(1): 107-108.         [ Links ] [4]

Bouchon, M., E. Díaz, S. Cahuín & M. Ochoa (2002). Effects of "El Niño" 1997-98 on the Schooling Behaviour of the Peruvian Anchovy (Engraulis ringens). Invest. Mar., Valparaíso, 30(1): 182-183.         [ Links ] [5]

Castillo J., M.A. Barbieri & A. González. (1996). Relationships between sea surface temperature, salinity, and pelagic fish distribution off northern Chile. ICES Journal of Marine Science, 53: 139-146.         [ Links ]

FAO (1997). Fishery statistics: capture production. FAO Yearbook, Vol. 84, 703 pp.         [ Links ] [6]

Fréon, P. & O. Misund (1999). Dynamics of pelagic fish distribution and behaviour: effects on fisheries and stock assessment. Fishing News Books, 348 pp.         [ Links ] [7]

Fuenzalida, R., J.L. Blanco, R. Núñez & S. Hormazabal (1999). Efectos oceanográficos y meteorológicos durante el evento El Niño 1997-98, en la zona de Iquique (20° S). Rev. Invest. Cient. Tecnol., 5: 56-59.         [ Links ] [8]

Laurs, R.M. (1989). Applications de la télédétection satellitaire du germon du Pacific Nord, Thunnus alalunga (Bonaterre). FAO Fish. Tech. Pap. 302, 87-97.         [ Links ] [9]

MacCall, A.D. (1976). Density dependence of catchability coefficient in the California Pacific sardine purse seine fishery. Cal. Coop. Ocean. Fish. Invest. Reports, Vol. XVIII, 136-148.         [ Links ] [10]

Nieto, K. (1999). Determinación de zonas probables de pesca de pez espada (Xiphias gladius) en Chile central, a través de imágenes de temperatura superficial del mar de satélites NOAA. Tesis, Escuela de Ciencias del Mar, UCV, Valparaíso, 92 pp.         [ Links ] [11]

Petitgas, P. (1994). Spatial strategies of fish populations. ICES/CIEM, C.M./D14, 15 pp.         [ Links ] [12]

Rocha, E. (2001). Distribución de clorofila «a» en las aguas del norte de Chile (18°30'S - 24°S): Validación de imágenes satelitales de color del mar. Tesis, Escuela de Ciencias del Mar, UCV, Valparaíso, 56 pp.         [ Links ] [13]

Segura, M., A. Bertrand & L. Vasquéz (2002). Study of the pelagic fish environmental and behavioral adaptation to strong physicochemical induced by ENSO in Peruvian waters (1997-1998). Invest. Mar., Valparaíso, 30(1): 121-122.         [ Links ] [14]

Silva, C., E. Yáñez, M.A. Barbieri, K. Nieto, V. Mímica, F. Espindola & J. Acevedo (2000). Exploring the association between small pelagic fisheries and SeaWiFS chlorophyll and AVHRR sea surface temperature in the north of Chile. Proceeding 6th International Conference on Remote Sensing for Marine and Coastal Environments, USA. ERIM International, Vol. II: 81-88.         [ Links ] [15]

Stretta, J.-M. (1991). Contribution de la télédétection aérospatiales à l'élaboration des bases de l'halieutique opérationnelle: l'exemple des pêcheries thonières tropicales de surface (aspect prédictif). Doctoral Dissertation, Paris VI, Grenoble (France).         [ Links ] [16]

Swain, D.P. & A.F. Sinclair (1994). Fish distribution and catchability: what is the appropriate measure of distribution? Can. J. Fish: Aquat. Sci., 51: 1046-54.         [ Links ] [17]

Ultang, O. (1980). Factors effecting the reaction of pelagic fish stocks to explotation and requiring a new approach to assessment and management. Rapp. P.V. Réun. Cons. Int. Explor. Mer, 177: 489-509.         [ Links ] [18]

Yáñez, E., M.A. Barbieri & A. González (1995). Estructura térmica superficial asociada a la distribución espacio-temporal de sardina y anchoveta en la zona norte de Chile entre 1987 y 1992. Invest. Mar., Valparaíso, 23: 123-147.         [ Links ] [19]

Yáñez, E., F. Espíndola, P. Freón, M.A. Barbieri & I. Guerrero I. (1999). Estandarización de tasas de captura de pesquerías pelágicas de la zona norte de Chile durante 1987-92. Invest. Mar., Valparaíso, 27: 53-63.         [ Links ] [20]

Yáñez, E., M.A. Barbieri, C. Silva, K. Nieto & F. Espíndola. (2001). Climate variability and pelagic fisheries in northern Chile. Progress in Oceanography, 49: 581-596.         [ Links ] [21]

Yáñez E., M.A. Barbieri, C. Silva., J. Castillo, K. Nieto, G. Böhm, F. Gómez & J. Córdova (2002). Climate variability and pelagic fisheries in the south-eastern Pacific. Invest. Mar., Valparaíso, 30 (1): 175-177.         [ Links ] [22]

Yáñez, E., C. Silva, K. Nieto, M.A. Barbieri, G. Martinez & B. Ramírez (2002). Prediction of probable fishing grounds in northern Chile from pelagic fisheries distributions and environmental conditions. GLOBEC Report 16 (1-97): 24-26.         [ Links ] [23]

Zadeh, L.A. (1994). Fuzzy logic, neural networks, and soft computing. Communications of ACM, 37(3): 77-84.         [ Links ] [24]


Creative Commons License Todo el contenido de esta revista, excepto dónde está identificado, está bajo una Licencia Creative Commons