Hunting Ban Effects on Estuarine Crocodiles

EFFECTS OF A HUNTING BAN ON ESTUARINE CROCODILE OCCURRENCE IN NORTHERN TERRITORY RIVERS, AUSTRALIA

Abstract

The impact of the hunting ban on estuarine crocodile occurrence, colours, before–After Control–Impact (BACI) design and applied logistic regression, is examined. The analysis is based on the presence/absence data for the crocodile collected from the control and impact rivers from 1960 to 2020. Other contributions, from the environmental perspective of water and rainfall, include mean water temperature and annual rainfall. These results show that there is a significant BACI interaction effect (p = 0.0115) with significantly more crocodiles observed in protected impact rivers following hunting. Occurrence probability in impact rivers rose from 0.1500 before the legal protection measure to 0.7925 following the protection measures. Model selection (AICc) indicates that the BACI interaction model is the most likely to be the best statistical model – environmental variables add relatively little explanatory power. The residual test and likelihood ratio test further demonstrate the effectiveness of the model. The results illustrate the ecological significance of legal protection and long-term conservation management of estuarine crocodile recovery.

Table of Contents

Introduction. 4

Materials and methods. 5

Study Species and Context 5

Study Area and Survey Design. 5

Survey Methods. 5

Data Analysis. 6

Results. 7

Discussion. 21

References. 24

Introduction

River water is home to a large, important and highly conspicuous saltwater crocodile, Crocodylus porosus, which occurs in rivers in northern tropical regions. Crocodiles are also considered a keystone species; their presence and actions help maintain the balance of the ecosystem. In the middle of the 20th century, there are a lot of commercially organised hunts for saltwater crocodiles that resulted in significant loss of crocodiles in the Northern Territory at this historic time (Mureri et al. 2026). Large-scale commercial hunting was important in the Northern Territory in the mid-20th century, but it seriously affected the saltwater crocodile at this time. Due to the high demand for skins, many river systems are being over-collected, resulting in population decreases and patchy distribution of crocodiles.

As population numbers began to shrink and concerns became apparent about ecological effects, legal activities such as the hunting ban in a number of northern Australian areas in 1971 are adopted. After the croc hunts are made illegal, the recovery of croc after legal protection was considered. When determining ecological benefits and effectiveness of management interventions, appropriate statistical tools need to be employed.

The design is the Before–After Control–Impact (BACI) design, which compares temporal differences between impacted and control sites, and is commonly used in ecological research to assess disturbance or management (Lian & Islam 2026).  In this study, the occurrence of saltwater crocodiles in Northern Territory rivers was tested for the effect of the 1971 hunting ban on the basis of logistic regression. This work focused on the effect of the 1971 saltwater crocodile hunting ban based on a BACI logistic regression basis. To understand the effect of sampling period, river type of management and environmental factors, their presence or absence in particular are studied. It was expected that hunting in controlled rivers (where hunting was still allowed to continue) would not result in a significant change in the number of crocodiles, while in rivers where hunting was banned, the number would significantly increase.

Materials and methods

Study Species and Context

The estuarine crocodile (Crocodylus porosus) is Australia's largest living reptile and predaceous animal of estuarine and inland waterways in northernmost Australia (Cooke et al. 2025). In some rivers, however, there are special removal programmes that meet human safety and livestock protection needs (Ricketts et al. 2026). This provided a natural experiment where the hatchability of these rivers could be compared with the rivers that went on to be hunted after the hunting ban became fully operational (impact river) and the ones that did not become fully hunting free.

Study Area and Survey Design

Survey activity has been done at several rivers throughout the NT of Australia, both tidal and freshwater, including the northern coast. Survey sites are chosen to be representative of the two management categories - control rivers indicated areas where crocodile removal on estuaries continued up to 1971, and impact river areas where the ban has been applied with no crocodile removals.

The periods of pre-hunting ban (1960–1971) and post-hunting ban (1972–2020) have created a ‘Before-After Control-Impact' (BACI) study design. This design provides an unbiased estimation of the impact of the hunting ban on crocodile occurrence, and can incorporate variables reflecting differences between river types preexisting in the site areas (Bateman & Gilson 2025).

Survey Methods

Standardised spotlight count techniques are used in estuarine crocodile surveys, following the method used by (Smith et al. 2025). The surveys are done at night and while the boat has been at a constant speed in the river transects. Routine spotlighting of crocs has been used to find the eyeshine characteristic of crocodiles. The presence/absence of estuarine crocodiles has been surveyed at all sites.

All data from each site are pooled, and each year's data are pooled with the data of the other sites to form a single binary outcome variable representing the presence or absence of estuarine crocodiles per survey record. The mean water temperature (°C) and mean precipitation (mm) are also measured at each of the sites surveyed, since previous studies indicate that temperature and precipitation have an impact on abundance and site use in the study area (Staines et al. 2025).

Data Analysis

All statistical analyses are carried out using the programming language R and RStudio. The response variable (crocodile presence/absence) has been categorical (binary), so a generalised linear model (GLM) with binomial error distribution and logit link function (logistic regression) has been used. A series of models has been fitted, first with no predictor variables (only an intercept) and then one at a time for the predictor variables included in this study: Before-After, Control-Impact, interaction term (BACI term), and two abiotic variables (mean temperature, mean rainfall), each with an intercept.

The model selection has been based on the corrected Akaike Information Criterion (Beasley et al. 2025), and the simplest model, in terms of AICc, has been selected, i.e. the model with the smallest AICc value and the highest Akaike weight. The model has been validated using the scaled Dunn-Smyth residual plot through the package mvabund to check whether any model assumptions are violated or systematic patterns in residuals exist.

Results

Figure 1: Dataset Inspection and Variable Overview

The first window which R shows after the data has been imported into R is the view of the data set “croc”. Figure 1 is the first look at the data that has been imported into “croc”. Five variables and 155 observations: crocodiles present or not, sampling period, river type, mean water temp, and annual rainfall.

Table 1. Summary Statistics of Study Variables

The first 6 records from the file can be seen after using head(), to verify that these are imported correctly and variables are named. This is one of the preliminary inspection stages to ensure the integrity of the data before statistically analysing. The figure also shows that there are both categorical and continuous predictors for the BACI logistic regression.

Figure 2: Summary Statistics of Crocodile Dataset

Figure 2 gives descriptive summary statistics of all the variables which have been included in the study. Crocodiles occurred in 35% of the observation transects across the transects. In the “After” period, there are 82 observations and in the “Before” period, there are 73 observations.

Table 2: Categorical Variable Counts

The range of the water temperature has been 26°C – 40°C, and the annual rainfall range has been 916 mm – 2102 mm. These statistics will be utilised to establish an overall picture of the environmental status of rivers in the control and impact period, as well as for the distribution and range of variables that would affect the occurrence of crocodiles.

Figure 3: Data Structure and Factor Releveling

In Figure 3 below, there are examples of how data are inspected when the data is categorical, and how categorical data is restructured with relevel. Structured inspection of data using str () and restructuring categorical data using relevel are shown in Figure 3 (Alarcón et al. 2025). Str (structural inspection) is used to inspect the contents of the dataset, and the process of restructuring categorical data using relevel is shown in Figure 3 below. Modelling with logistic regression using integer and factor variables. Two variables are recoded as factors (Before After = ‘B’ (Before), ‘C’ (Control); Control impact = ‘B’ (Before), ‘C’ (Control)). There is a fair amount of releveling.

Figure 4: Crocodile Presence Before and After Hunting Ban

In Figure 4, before and after the 1971 hunting ban, changes occur in the presence and absence of crocodiles that control and impact the rivers. There are always fewer crocodiles sampled in control rivers throughout the two sampling periods. The opposite trend has been observed for crocodiles; a dramatic increase in crocodile numbers occurred after the hunting ban was enforced in impacted rivers. In contrast, following the hunting ban, there has been a substantial increase in crocodile abundance in impacted rivers. The abundance of crocodiles, however, has been much higher during the post-hunting ban period for impact rivers. The BACI interaction effect is shown visually with increased occurrence of crocodiles in the areas where active removal ended, in the case of the hunting ban.

Figure 5: Water Temperature and Crocodile Occurrence

Figure 5 (below) shows the boxplots for the mean annual water temp for sites with (N) and without (C) crocs. Similar temperatures are found in both groups, with the median temperatures being slightly cooler at the sites where crocodiles are present. There are some overlaps in the distributions of the two groups, but the distribution of crocodile occurrence suggests a wide range of thermal conditions in which the crocodile can be expected. The jittered and the values give more detail on the variability and the spread of the data.

Figure 6: Rainfall and Crocodile Occurrence

Boxplots and jittered observations between rainy and dry periods with and without crocodile locations (Figure 6) are used to compare the level of precipitation/annum to assess if there has been any difference. Rainfall tended to be slightly greater at sites that did not have crocodiles, as opposed to sites that did. Substantial overlap in both of the distributions, however, suggests that rainfall alone is not likely to have strong impacts on crocodile occurrence. Another striking aspect of the data values is their wide spread, which again indicates that there is high environmental variation between river systems.

Figure 7: Relationship Between Temperature and Rainfall

The scatter graph in Figure 7 illustrates the correlation between the mean water temperature at any point of the year and the annual rainfall with a fitted regression. The relationship between the two abiotic features is – as expected – very weak, and the slope of the regression is only slightly upward. Most of the observations are quite distant from the trend line, suggesting there is not a high amount of collinearity between the predictors. Another representation of what may have occurred during the estimation of the regression is the shaded confidence interval.

Figure 8: AICc Model Selection Results

Candidate logistic regression models are compared using the Akaike information criterion corrected for small sample size (AICc) and are displayed in Figure 8. None of the other models had a higher model weight than it nor lower AICc values (Sterckx 2025). The fitted models for the additional abiotic variables (temperature and rainfall) did not significantly affect the models when the abiotic variables are included in the model, with a slight increase in AICc. The BACI interaction had very little influence on the models; outside of it, they performed not so well.

Figure 9: Variance Inflation Factor (VIF) Analysis

The results of the Variance Inflation Factor (VIF) test for multicollinearity in predictor variables for the additive logistic regression model are shown in Figure 9. All VIF values are close to one versus each other, such as temperature, rainfall, sampling period, etc., and river type.

Table 3. Variance Inflation Factor (VIF) Results

There is apparently not much concern for collinearity among the explanatory variables. Thus, the predictors in the model are statistically independent and suitable for being used in the same regression analysis.

Figure 10: Analysis of Deviance for BACI Model

The BACI logistic regression model analysis of deviance table is shown in Figure 10. Significant differences (p value < 0.001) are found for sampling period (Before After) and river type (Control impact).

Table 4. Analysis of Deviance for the BACI Logistic Regression Model

The most important statistical analysis has been that of the interaction between sampling period and river type (p = 0.0115). This is a temporal interaction indicating that the temporal occurrence and abundance of crocodiles are different in both control and impact rivers.

Figure 11: Likelihood Ratio Test of Top Model

The likelihood ratio test compares the BACI interaction model with the null model, as presented in Figure 11. This yielded a highly significant result (p<0.001), which indicated that explanatory variables make a significantly greater contribution to the result than the null model.

Table 5. Likelihood Ratio Test Comparing the Top Model and Null Model

The decrease in deviance from 201.62 to 131.32 shows that a high proportion of the variance of the occurrence of crocodiles has been accounted for by the BACI predictors.

Figure 12: Logistic Regression Coefficient Summary

The summary of the logistic regression model BACI in the log-odds scale is given in Figure 12. There has been a significant difference between the interaction periods by impact rivers (p = 0.008), where there has been a significant increase following the hunting ban for impact rivers. The effect of the sampling period alone and that of the river type alone are not significant.

Table 6: Candidate Logistic Regression Models Compared Using AICc

Figure 13: Residual Validation of Top Model

This residual by fitted plot is shown for testing the suitability of the top logistic regression model in Figure 13 (Khimta 2025). The Dunn-Smyth residuals are relatively similar across the various classes, so that there does not appear to be any major concerns with heteroscedasticity or model bias. The distribution of the residual binomial modelling is satisfactory because of clustering, which is accounted for by the binary response structure.

Figure 14: Estimated Marginal Probabilities

The estimated MPOD of crocodiles for all the sampling periods and river types can be seen in Figure 14. The EE caused a spectacular increase in impact rivers (by ~79%). Confidence intervals can be used as a powerful tool to substantiate the increase.

Table 7. Estimated Marginal Probabilities of Crocodile Occurrence

The following predicted probabilities clearly show the positive ecological effect of the hunting ban and give an easily interpretable summary of the BACI interaction results from the logistic regression analysis.

Figure 15: Pairwise Comparisons of Estimated Means

Treatment combinations are contrasted pairwise graphically as odds ratios with p-values. Comparisons including impact rivers post hunting ban (A I) had very low p-values (<0.0001) and are highly significant.

Table 8. Pairwise Comparisons Among Treatment Groups

The results show that the occurrence of crocodiles has been significantly different between the protected condition and all other treatment conditions. Most of the comparisons, however, between impact and control conditions before and after the ban are insignificant.

Discussion

The discussion elucidates the ecology of the 1971 hunting ban on estuarine crocodile occurrence in Northern Territory rivers in Australia. The results of key statistics, environmental factors, model validation and conservation implications related to crocodile recovery in the surveyed rivers are discussed.

Impact of the Hunting Ban on Crocodile Recovery

The results strongly suggest that the hunting ban helped to substantially recover these impacted rivers of crocodiles. Legal protection resulted in a significant BACI interaction term (p = 0.0115), indicating that crocodile occurrence changed in different ways between control and impact rivers. In the impact of rivers, the estimated marginal probabilities went up from 0.1500 to 0.7925 before and after the ban (Gayo & Ngonyoka 2025). There is a slight rise in the control rivers, from 0.0909 to 0.1379. After Bonferroni correction (adjusted p-value < 0.001), or pairwise comparison, highly significant differences have been found for protected and impact rivers. The results of this study emphasise the ecological advantages of a reduction of commercial hunting activities.

Importance of the BACI Design in the Study

The pre–post–control impact design proved to be effective in assessing the impact of the hunting ban on the ecology. The analysis of deviation has been used in a significant result sampling period (deviance = 34.648, p < 0.001) and river type (deviance = 29.263, p < 0.001). The interaction effect is significant, revealing that the magnitude of the presence of crocodiles varied over the years in protected and control rivers (Campbell et al. 2025). The explanatory power is reflected in the reduction of residual deviance from 201.62 in the null model to 131.32 in the BACI model. The likelihood ratio test also gave a highly significant outcome (p< 0.001), indicating the statistical validity of the BACI.

 Influence of Environmental Variables on Crocodile Occurrence

The effect of environmental factors (temperature and rainfall) is compared to the effects of hunting management factors. The temperature of the water ranged from 26°C to 40°C, and the rainfall ranged from 916mm to 2102mm in the surveyed rivers (Taplin, 2023). The boxplots showed considerable overlap between crocodile presence and absence sites, indicating a low level of environmental separation. The BACI interaction model achieved the lowest AICc value of 139.6, whereas adding temperature and rainfall slightly increased AICc scores. VarINF values are also low (close to 1), which indicates that there is little multicollinearity among the variables, in particular: temperature (1.015) and rainfall (1.032). The results indicate that legal protection has a greater effect on the presence of crocodiles than abiotic environmental factors.

Reliability and Performance of the Logistic Regression Model

The logistic regression provided sound statistical backing for the study objectives. The Dunn–Smyth residual validation revealed that there is no systematic bias or heteroscedasticity in the residuals, even after they are standardised. The BACI interaction model outperformed simpler models, with the highest Akaike weight (0.447) and showing higher performance than the simpler models (Reynolds & Morgan, 2023). There is a very poor explanatory strength for environmental-only models (AICc>200). Logistic regression with a binary presence-absence data structure is also efficient. The overall survey statistical results of this study confirm the suitability of BACI logistic regression for long-term ecological assessment and conservation research of estuarine crocodile populations.

Ecological and Conservation Implications

The findings of this study offer insights into the significance of conservation policies to bring back ‘apex' predator populations and confirm the ecological benefit. Estuarine crocodiles serve as keystone species in the estuarine ecosystems, helping to promote their balance and stability in the tropical river systems (Ayadi & Ayedun 2025). This increase in the occurrence of crocodiles in the protected rivers illustrates the benign impact and benefits of having long-term crocodile protection legislation undertaken. These results indicate that science-led legal action is significant to species recovery and to clean up ecological conditions in historically over-exploited and commercially hunted estuarine river systems.

References

Alarcón, D., Smith, S., Hsieh, I.S., Colon, J., McElhinney, J., Mahone, J., Mellgren, S., Weaver, K. and Pérez, J.P.M., 2025. Citizen scientist come out of their shells. In Proceedings of the Forty-Second Annual Symposium On Sea Turtle Biology And Conservation (pp. 61-61). International Sea Turtle Society. Retrieved from: https://research.usc.edu.au/esploro/fulltext/conferencePaper/Citizen-scientist-come-out-of-their/991128804002621?repId=12288624540002621&mId=13289003450002621&institution=61USC_INST

Bateman, P.W. and Gilson, L.N., 2025. Bad dog? The environmental effects of owning dogs. Pacific Conservation Biology, 31(3), p.PC24071. Retrieved from: https://connectsci.au/pc/article-abstract/31/3/PC24071/200023

Beasley, I., Amepou, Y., Mavera, J., Mavea, W., Anamiato, J., Baird, K., Lawrence, A. and Masere, C., 2025. Demand for fish swim-bladders driving inshore dolphin populations in Papua New Guinea towards local extinction. Pacific Conservation Biology, 31(3), p.PC23060. Retrieved from: https://connectsci.au/pc/article-abstract/31/3/PC23060/200022

Campbell, M. A., Udyawer, V., White, C., Baker, C. J., Kopf, R. K., Fukuda, Y., ... & Campbell, H. A. (2025). Quantifying the ecological role of crocodiles: a 50-year review of metabolic requirements and nutrient contributions in northern Australia. Proceedings of the Royal Society B: Biological Sciences, 292(2042). Retrieved from:https://royalsocietypublishing.org/rspb/article/292/2042/20242260/104874

Cooke, S.J., Bett, N.N., Hinch, S.G., Adolph, C.B., Hasler, C.T., Howell, B.E., Schoen, A.N., Mullen, E.J., Fangue, N.A., Todgham, A.E. and Cheung, M.J., 2025. Co-production and conservation physiology: outcomes, challenges and opportunities arising from reflections on diverse co-produced projects. Conservation Physiology, 13(1), p.coaf049. Retrieved from: https://academic.oup.com/conphys/article-pdf/doi/10.1093/conphys/coaf049/63792992/coaf049.pdf

Khimta, A.C., 2025. Fiscal policy, gender equity, and development: a study of Gender Budgeting in Himachal Pradesh. Humanities, 13(2), pp.63-76. Retrieved from: https://www.hpuj.in/content/hpuj%20dec%202025%20-%204.pdf

Reynolds, S., & Morgan, D. (2023). Review of toxicity of agricultural chemicals and implications for aquatic fauna of the Keep River. Retrieved from: https://library.dpird.wa.gov.au/lr_consultrpts/13/

Ricketts, J., Brianne, D., Letter, A., Ford, R., Steen, D. and Deem, V., 2026. An apparent case of Pb toxicosis in an American crocodile Crocodylus acutus in South Florida, USA. Endangered Species Research, 59, pp.1-6. Retrieved from: https://www.int-res.com/journals/esr/articles/esr01474

Smith, C., van de Merwe, J., Finlayson, K., Barraza, A.D., Young, E., Gilby, B. and Townsend, K., 2025. Disentangling the impacts of contaminants on green sea turtle physiology. In Proceedings of the Forty-Second Annual Symposium on Sea Turtle Biology And Conservation (pp. 30-30). International Sea Turtle Society. Retrieved from: https://research.usc.edu.au/esploro/fulltext/conferencePaper/Disentangling-the-impacts-of-contaminants-on/991128804102621?repId=12288624600002621&mId=13289003400002621&institution=61USC_INST

Staines, M.N., Versace, H., Laloë, J.O., Hof, C.M., Smith, C., Haskin, E., Lawrence, A., Tibbetts, I., Booth, D., Pilcher, N. and Hays, G., 2025. Prevalence of male-producing nesting sites for endangered sea Turtles in the Asia-Pacific region and globally. In Proceedings of the Forty-Second Annual Symposium on Sea Turtle Biology and Conservation (pp. 246-247). International Sea Turtle Society. Retrieved from: https://research.usc.edu.au/esploro/fulltext/conferencePaper/Prevalence-of-male-producing-nesting-sites-for/991128803702621?repId=12288624260002621&mId=13289003600002621&institution=61USC_INST

Sterckx, R., 2025. Fish Farming in Pre-modern China: A Study and Translation of Two Texts. T'oung Pao, 111(3-4), pp.353-396. Retrieved from: https://brill.com/view/journals/tpao/111/3-4/article-p353_4.xml

Taplin, L. (2023). Modelling population dynamics of estuarine crocodiles on Queensland’s northern populated east coast. Retrieved from:  https://www.qld.gov.au/__data/assets/pdf_file/0023/607163/modelling-population-est-crocs-qld-north-east-coast-report.pdf

Lian, Y. and Islam, M.Z., 2026. The Neocolonial Tightening of CITES: How Northern Narratives Marginalize Southern Conservation. Conservation Letters, 19(2), p.e70029. Retrieved from: https://conbio.onlinelibrary.wiley.com/doi/abs/10.1111/con4.70029

Gayo, L. and Ngonyoka, A., 2025. Do wildlife management areas help to mitigate negative human-wildlife interactions? A case of Eastern Bufferzone of Selous Game Reserve, Tanzania. Tropical Conservation Science, 18, p.19400829251340581. Retrieved from: https://academic.oup.com/conphys/article-pdf/doi/10.1093/conphys/coaf049/63792992/coaf049.pdf

Ayadi, P.O. and Ayedun, H., 2025. Health and environmental footprints of spent lubricating oil (SLO). COAST JOURNAL OF THE SCHOOL OF SCIENCE OAUSTECH OKITIPUPA, 7(2), pp.1402-1417. Retrieved from: https://coast.oaustech.edu.ng/index.php/coast/article/view/144

Mureri, A., Musavengane, R. and Manyakaidze, P., 2026. Indigenous knowledge systems in the collaborative governance of wetlands conservation in rural areas of Zimbabwe. South African Geographical Journal, 108(1), p.2630956. Retrieved from: https://journals.co.za/doi/abs/10.1080/03736245.2026.2630956

Gebremedhin, G.A., 2025. The Environmental Impact of Italy’s Chemical Weapons Use in the 1935–1936 Italo-Ethiopian War: A Case Study of Tigray’s Scarce Water Bodies. ITYOPIS: Northeast African Journal of Social Sciences and Humanities, 7(1), pp.37-53. Retrieved from: https://journal.mu.edu.et/index.php/ityopis/article/view/1117