Lake water quality: Data to 2024

Lake water quality variables trophic level index (TLI) – is a composite index developed for New Zealand lakes (Burns et al., 1999). It describes the state of nutrient enrichment, which can be understood as the life-supporting capacity of a lake (Burns et al., 2000). TLI takes into account levels of nitrogen, phosphorus, and chlorophyll-a (TLI3). An alternative version of TLI is also used in New Zealand, which includes clarity as a fourth variable (TLI4). We report TLI3 to maximise the number of lakes included in our analysis, as clarity data are not consistently available across monitoring sites. chlorophyll-a – is a measure of phytoplankton (algae and cyanobacteria) biomass in lake water. Chlorophyll-a concentrations are an indicator of trophic state (nutrient enrichment and productivity). Very high concentrations indicate ‘algal bloom’ conditions. When algae dominate, the water is persistently turbid (cloudy) and aquatic plants die out, which signals a degraded lake. Escherichia coli (E. coli) – are bacteria commonly found in the intestines of warm-blooded animals including humans. When found in fresh water, E. coli can indicate the presence of pathogens (disease-causing organisms) from animal or human faeces, which can cause illness when ingested. One of the most common such pathogens is Campylobacter, but it is difficult to measure. E. coli concentrations are used to infer Campylobacter infection risk in waterways, based on correlations between levels of E. coli and Campylobacter. total phosphorus (TP) – is the sum of all forms of phosphorus in water, including both dissolved reactive phosphorus (DRP), which is easily taken up by plants, and phosphorus bound to sediments or particles. dissolved reactive phosphorus (DRP) – is the phosphorus that is dissolved and available for aquatic plants and algae growth. High levels of DRP can cause rapid weed growth or algal blooms, which can choke aquatic life. total nitrogen (TN) – is the sum of all forms of nitrogen in water, including nitrate-nitrogen (NO3-N), nitrite-nitrogen (NO2-N), ammoniacal-nitrogen (NH4-N) and organic nitrogen, measured as nitrogen (N). At high concentrations, it can become toxic to fish and macroinvertebrates. It can also cause excessive growth of algae. ammoniacal nitrogen (NH4-N) – is a form of nitrogen in water that exists as either ammonia (NH3) or ammonium (NH4) and is usually derived from human or animal waste, and at high concentrations can be toxic. NH4-N is the concentration of ammoniacal nitrogen measured as nitrogen. nitrate-nitrite-nitrogen (NNN) – is the sum of two nitrogen forms: nitrate-nitrogen (NO3-N) and nitrite-nitrogen (NO2-N), measured as nitrogen. Nitrate is typically the dominant form and is highly soluble and mobile in water. At high concentrations, NNN can become toxic to fish and macroinvertebrates. It can also cause excessive algal growth. clarity – is a measure of underwater visibility, generally assessed by how deep a black and white disk can be seen from the surface of the water. Poor clarity can affect the habitat and food supply of aquatic life and the growth of aquatic plants. For clarity, increasing values indicate improving lake conditions; for all other variables, increasing values indicate worsening conditions. Accordingly, the terms ‘improving’ and ‘worsening’ are used to describe trends. However, the pop-up windows in the ArcGIS maps display trends as ‘increasing’ or ‘decreasing’. Data Monitored data We report on lake water quality monitoring sites operated by regional councils and unitary authorities (hereafter referred to as ‘councils’) in New Zealand. Monitoring data from these sites are compiled by Earth Sciences New Zealand (ESNZ, formerly NIWA) and provided to us for analysis and reporting. Some lakes are monitored at more than one location. In total, our analysis covers 159 monitoring sites across 119 lakes. To ensure our reports are nationally consistent, we only report data collected over consistent time periods. As a result, our evaluations may differ from those published separately by councils and Land Air Water Aotearoa (LAWA). If detailed regional-level information is required, we recommend consulting the relevant council’s environmental reports, or the LAWA website. Booker et al.(2025) provides more information about monitored data. Modelled data In addition to monitored data, ESNZ modelled water quality variables for approximately 4,500 lakes in New Zealand that are larger than 1 hectare in the Freshwater Ecosystems of New Zealand (FENZ) database based on measured data between 1 January 2020 and 31 December 2024, excluding Artificial Constructed or Mine (Wood et al., 2025). This allows us to estimate water quality state at lakes that do not have monitoring sites. Values of water quality variables were predicted for lakes using random forest models and predictors (explanatory variables) such as climate, geology, topography, hydrology, stock intensity, and landcover. We report modelled results for TLI only. Of the variables modelled, clarity, TLI, and total nitrogen demonstrated sufficiently good model performance (R² > 0.6 and NSE > 0.6). However, lake water clarity does not have an established national guideline, and TLI provides a more comprehensive measure of lake water quality by integrating nitrogen, phosphorus, and chlorophyll-a into a single index. Wood et al. (2025) provides more information about modelled data. State analysis We report on lake water quality state using summary statistics calculated from measured values for each variable for the five-year period between 1 January 2020 and 31 December 2024, alongside modelled TLI statistics for the same period. Only lakes with sufficient monitoring data that met the filtering rules were included in the analysis. To meet the rules, sites were required to have an adequate number of observations and an appropriate distribution in time. For example, Lake Taupō was excluded because its available data did not meet these requirements. The number of lake sites that met filtering rules for inclusion in the analysis using five-year medians was: 154 (TLI), 148 (chlorophyll-a), 147 (total nitrogen), 147 (total phosphorus), 145 (DRP), 144 (ammoniacal nitrogen - pH adjusted), 119 (E. coli), 112 (clarity), and 102 (nitrate-nitrogen). Booker et al. (2025) provides more information about data processing and filtering rules. Thresholds We compare water quality variable values to established thresholds to provide context on whether they are likely to impact ecosystem and human health: TLI results were reported against TLI categories, total nitrogen, total phosphorus, chlorophyll-a, ammoniacal nitrogen and E. coli were reported against bands of the National Objectives Framework (NOF) in the National policy statement for freshwater management 2020. Amended December 2025. TLI categories The TLI rating is used to place lakes into nutrient-enrichment categories known as trophic states: microtrophic (TLI rating 0-2; very good) lakes are very clean and often have snow or glacial sources oligotrophic (TLI rating >2–3; good) lakes are clear and blue, with low concentrations of nutrients and algae mesotrophic (TLI rating >3–4; average) lakes have moderate concentrations of nutrients and algae eutrophic (TLI rating >4–5; poor) lakes are murky, with high concentrations of nutrients and algae supertrophic or hypertrophic (TLI rating >5; very poor) lakes have extremely high concentrations of phosphorus and nitrogen, and are overly fertile; they are rarely suitable for recreation and lack habitats for desirable aquatic species. National Objectives Framework The NOF bands are designed to help communities make decisions on how to manage water quality. Each NOF band is defined by a numeric range and a description that corresponds to a scientifically determined range of effects. NOF Bands A through D/E represent different water quality states for lake sites, with A representing the highest rating band and D/E the lowest. For E. coli, the NOF bands describe the risk of Campylobacter infection (based on E. coli as an indicator). The predicted average infection risk is the overall average infection to swimmers based on a random exposure on a random day, ignoring any possibility of not swimming during high flows or when a surveillance advisory is in place (assuming that the E. coli concentration follows a lognormal distribution). Actual risk will generally be less if a person does not swim during high flows (MfE, 2025). For ammoniacal nitrogen, NOF bands indicate toxicity effects of ammonia but do not reflect the negative ecosystem health effects of nitrogen enrichment at lower concentrations such as promotion of excessive periphyton growth. As required for comparison with the NOF bands, ammoniacal nitrogen concentrations were adjusted to account for pH. For trend analysis, ammoniacal nitrogen concentrations were not pH-adjusted. Trend analysis We calculate and report on 20-year lake water quality trends, based on measured data between 2005 and 2024, using two methods: Mann-Kendall trend tests and Akritas-Theil-Sen slope estimates to assess overall trend direction, likelihood, and rate of change. We report trend likelihoods using categories describing the certainty of trends adapted from the Intergovernmental Panel on Climate Change (Mastrandrea et al., 2010). Generalised additive models (GAMs) to estimate smooth trends over years to show how trends change over time. Time was the only explanatory variable in the model. We use the Akritas-Theil-Sen (Helsel, 2012) slope estimator (an extension of the Theil-Sen slope estimate) to accommodate censored observations when calculating rate of change. The estimator may overestimate the rate of change in a small number of cases. We used the mgcv R package (Wood, 2011) to calculate the GAMs using the thin plate regression spline as the smoothing basis, the censored normal distribution family to account for censored observations, and restricted maximum likelihood (REML) as the method for estimating the smoothing parameters. Effective degrees of freedom are controlled by the degree of penalization selected during fitting, meaning the exact choice of basis dimension, k, is not generally critical. We found that for each site, setting k equal to the number of observations multiplied by 0.3 was sufficient. Values were log-transformed since the data are right-skewed. We applied inclusion criteria to manage censoring in addition to the inclusion criteria described in Booker et al. (2025). For each variable, trends were calculated for each monitoring site that had at least one observation in at least 90 percent of the sampling intervals (months, bi-months, quarters, or years) for the 20-year period between 2005 and 2024. Further, we excluded sites where 80 percent or more of the observations were left-censored. This aligns with our approach for Groundwater quality: Data to 2024. Sites were downsampled to standardise the time series when sampling frequency changed. Trend duration A 20-year period was chosen to include more observations for trend calculation while limiting the influence of natural climate cycles. The assessment period can have a noticeable effect on the estimated trend. Natural climate cycles may mask the effects of anthropogenic drivers of water quality trends at shorter time scales (Snelder et al., 2021). Trend certainty We classify trends as ‘likely’ when the probability of an increasing or decreasing trend is above 66 percent, and as ‘very likely’ when the probability is above 90 percent. We use the term ‘indeterminate’ when there is either no trend direction determined or not enough statistical certainty to determine trend direction (less than 66 percent certainty). The categories used to describe the certainty of trends are adapted from Mastrandrea et al. (2010). Limitations Results from monitored data provide insight into lake water quality only at the monitoring sites and only for the periods covered by the data. They do not represent conditions at unmonitored lakes or outside the monitoring period. The small number of lake monitoring sites and limited geographic coverage limit the extent to which results represent New Zealand lakes in general. Some regions like the far north of the North Island have more monitoring sites than others, therefore may be over-represented. Lake water quality measurements can be made using a range of laboratory and field methods. Differences in methods used to measure the same variable at different sites may introduce systematic differences between sites or regions, affecting comparability of estimated state. Changes in methods over time can also create ‘step changes’ in the data, which may complicate interpretation of trends. Trend analyses assume that observations are comparable through time and that changes in measured values reflect real environmental change. While efforts are made to ensure consistency, these assumptions may not always be met in practice, and apparent trends may arise from methodological changes rather than changes in water quality. (Davies-Colley and McBride, 2016; Wood, 2024; Booker et al., 2025). The modelled data are intended to give predictions of water quality states for all New Zealand lakes, but the performance of the model used was not sufficiently good for some water quality variables, which limits confidence in those predictions. References Booker, D., Smith, R., Wood, D., Fraser, C., & Snelder, T. (2025) Water quality state and trends in New Zealand lakes. Analysis of national lakes data ending in 2024. Earth Sciences New Zealand client report prepared for Ministry for the Environment. ESNZ, Christchurch. https://environment.govt.nz/publications/water-quality-state-and-trends-in-new-zealand-lakes-analysis-of-national-lakes-data-ending-in-2024/? Burns, N.M., Rutherford, J.C., Clayton, J.S. (1999) A monitoring and classification system for New Zealand lakes and reservoirs. Lake and reservoir management, 15, 255-271. Burns, N., Bryers, G., & Bowman, E. (2000) Protocol for monitoring trophic levels of New Zealand lakes and reservoirs. https://environment.govt.nz/publications/protocol-for-monitoring-trophic-levels-of-new-zealand-lakes-and-reservoirs/ Davies-Colley, R., McBride, G. (2016) Accounting for changes in method in long-term nutrient data: recommendations based on analysis of paired SoE data from Wellington rivers. NIWA Client Report, HAM2016-070: 34. Helsel, D. R. (2011).Statistics for censored environmental data using Minitab and R(Vol. 77). John Wiley & Sons. https://onlinelibrary.wiley.com/doi/book/10.1002/9781118162729 Ministry for the Environment (MfE). (2025) National policy statement for freshwater management 2020. Amended December 2025. https://environment.govt.nz/publications/national-policy-statement-for-freshwater-management/ Mastrandrea, M. D., Field, C. B., Stocker, T. F., Edenhofer, O., Ebi, K. L., Frame, D. J., Held, H., Kriegler, E., Mach, K. J., Matschoss, P. R., Plattner, G.-K., Yohe, G. W., & Zwiers, F. W. (2010) Guidance Note for Lead Authors of the IPCC Fifth Assessment Report on Consistent Treatment of Uncertainties. Intergovernmental Panel on Climate Change (IPCC). https://www.ipcc.ch/site/assets/uploads/2018/05/uncertainty-guidance-note.pdf Snelder, T.H., Fraser, C., Larned, S.T., Whitehead, A.L. (2021) Guidance for the analysis of temporal trends in environmental data. NIWA Client Report 2021017WN prepared for Envirolink (MBIE). NIWA, Christchurch. Wood, S. N. (2011) Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society (B), 73(1), 3–36. Wood, D. (2024) Assessing and accounting for the influence of changes in laboratory measurement methods on the interpretation of long-term time-series data. NIWA Client Report, 2024055CH: 54. Wood, D., Booker, D., & Smith, R. (2025) Spatial modelling of lake water quality state. Incorporating monitoring data for the period 2020 to 2024. Earth Sciences New Zealand client report prepared for Ministry for the Environment. ESNZ, Christchurch. https://environment.govt.nz/publications/spatial-modelling-of-lake-water-quality-state-incorporating-monitoring-data-for-the-period-2020-to-2024/? en-NZ