Lying at an altitude of 920 metres in the rain-shadow area of the Western Ghats, Bengaluru’s geography and climate do not support a perennial natural water source. Instead, a north-south ridge divides the city into three primary valleys: Hebbal, Koramangala-Challagatta, and Vrishabhavathi. Rainwater draining down these slopes feeds a network of small streams — forming a cascading lake system that the city historically relied on for its water supply.

During the Open City biodiversity datajam held in December 2025, our group examined the ecological trends around Bengaluru’s lakes. Data on insects, spiders and flora on platforms like iNaturalist were insufficient for meaningful analysis, hence we focused on waterbirds in the city’s lakes and its cascades using the popular citizen science platform eBird.

Our analysis suggests a majority of waterbirds are declining across Bengaluru lakes. What could be driving these declines? Are some lakes within a cascade more impacted than others? Are all waterbirds getting affected equally, or are there some that are benefiting? How are lake management decisions affecting the different guilds of birds? 

We address these questions across a series of four articles. In this first one, we compare how migrant and resident waterbirds are faring across 14 of Bengaluru’s lakes, and explore if these patterns are driven by cascade connectivity or individual lake management. 

The cascade system and the changing role of lakes

The 16th century saw a significant expansion of the city’s water infrastructure. Streams were dammed with earthen embankments to create lakes; sluices and drains distributed the stored water. Surplus water from the upstream lakes flowed downstream through canals called Rajakaluves, creating a cascade of interconnected tanks. The lakes were rain-fed and seasonal, yet they contributed up to 50% of the water supply.

With the arrival of piped supply in the early 20th century (from Arakavathi and later Kaveri), lakes stopped being a fresh water source for the city. Instead, they became valued for the land and real estate they occupied. Rapid urbanisation further accelerated the encroachment of lakes and Rajakaluves. 

Moreover, the number of water bodies has dwindled from over 2,300 in 1897 to around 200 today. More than 90% of the surviving lakes have been encroached upon besides becoming dumping grounds for industrial effluents and sewage. Urbanisation has also transformed the catchment area of the lakes. The wetlands, orchards, farming landscapes which once surrounded a lake have given way to concrete built-up areas. Our land use change analysis around the major cascades captures this transformation: 

---

Read more: The wild in the city: What citizen scientists tell us about Bengaluru’s biodiversity

---

Lakes and the effect of urbanisation 

Explosive urbanisation has increased the sewage influx into lakes. This is reflected in our analysis of the Karnataka State Pollution Control Board’s 2025 water-quality results. Every monitoring station exceeds the ideal biochemical oxygen demand (BOD) threshold of 3 mg/L , and 85% of the stations exceed 6 mg/L. Most stations fluctuate between water quality labels D (fish/wildlife propagation) and E (irrigation/industrial cooling), the two lowest tiers in the official classification scheme.

Rural communities relying on lakes for livelihoods are being displaced by gated communities and residential apartments. Wetlands and reedbeds, which once sustained fishing and grazing, are seen as mosquito breeding dumps and replaced by urban gardens with ornamental, non-native vegetation. The lack of rural community involvement also affects desilting and other maintenance activities.

Bengaluru’s groundwater crisis has brought lakes back into the centre of the city’s water planning. Treated sewage flowing into the lakes now recharges local aquifers, and has become an important part of its groundwater security blueprint. Major lakes today are being restructured to hold water year-round. 

A recent study has found that 95% of Hebbal valley lakes today are sustained by treated sewage. The sewage treatment plants mitigate pollution and replenish the aquifers. But the year round-supply of water keeps the lake perennially full, with consequences for their hydrology and biodiversity.

Consequently, most lakes in the city today are sewage-fed, concretised soup bowls, fortified by walking paths, dotted with non-native ornamental trees and shrubs and managed through non-ecological lens. How these changes have affected biodiversity forms the crux of our analysis.

Birds as indicators of a lake’s health

Bengaluru’s lakes are important biodiversity hotspots, harboring diverse bird, insect, arthropod and plant life, and putting the city on the global winter bird-migration map. But assessing whether a lake is “healthy” is not straightforward. Water quality and land-use data tells you only half the story. Vegetation, fish and aquatic invertebrates, each add a different layer. The challenge is finding indicators that reflect how these components function together.

Birds are widely used as indicators of ecosystem health because they depend on many parts of a wetland simultaneously. Their presence is shaped by factors like food availability, vegetation structure, nesting sites and the degree of human and predator disturbance. Different species have evolved to exploit different ecological niches in a lake: open water, shallows, vegetation, and exposed edges. This is pictorially represented below:

Using eBird data

eBird is a citizen science platform for recording the birds they observe in a simple ‘checklist’. Each checklist captures the species, their counts, the timing and the location of the observation. These checklists are curated and stored in eBird database, forming a rich source of biodiversity data. Our study is based on analysing the changing detection trends of waterbird species in eBird checklists around lakes in the Bengaluru Urban district from 2013 onward.

For waterbodies, we used OpenCity’s dataset of 181 lakes in the BBMP limits. We defined a buffer around the waterbodies, as biodiversity observations are often concentrated in the areas surrounding a waterbody. The buffer area chosen was dependent on the lake area, the table here has the details.

We then geomapped the eBird checklists to their respective waterbodies. The final dataset comprised 837,356 total bird observations across 181 lakes, recorded in 34,818 complete checklists. Out of these, there were 348,113 waterbird observations across 25,868 checklists, with 113 unique waterbird species.  

Using the earliest and latest sighting of a species at a location, we inferred the migratory status of a species. Uncommon visitors (detected in fewer than 20% of valid years), termed vagrants, were excluded from the analysis. The remaining species were classified based on the following criterion:

• Resident: seasonality ratio < 4.0 (detection is spread across seasons)
• Migrant: seasonality ratio > 4.0 (strongly concentrated within a single season)

For migrants, we defined two primary migration windows:

• Winter Migration: October–May (Extended to capture late-departing migrants)
• Monsoon Migration: June–September (primarily local movement)

Each waterbird observation was assigned a lake and a resident/migrant label. 

Individual years that didn’t meet the requirements were dropped from the analysis. The stringent requirements are necessary for a meaningful analysis. Consequently, only 14 of 181 lakes dataset, shown below, met this criterion.

Sampling effort

The number of eBird checklists varies widely across lakes. For example, Jakkur and Kasavanahalli, have more than 10 years of data with a reasonable spread. In contrast, lakes like Sowl and Doddanekundi have seen a dramatic surge in the number of checklists in the past two to three years. Within a lake, the checklist count can vary across seasons. Winters typically see a spike in checklists compared to monsoons and summers. To account for this uneven effort, we assigned each lake a data quality label based on these criteria. These labels provide the necessary nuance when interpreting the results.

Interpreting trends

The annual detection frequency of a species– the fraction of checklists with the bird recorded– was recorded at every qualifying lake over the range of valid years. Tracking how this frequency changes over time gives us the long term trend of the species.

But a simple year-on-year trend can be misleading. A year with only 20 checklists gives a much shakier estimate than a year with 200 checklists. When there is less data, a few chance sightings can swing the trend disproportionately. A trend affected by these unreliable years is a reflection of how much birding has happened at a lake rather than how the bird is actually doing.

This is why we use bootstrapping. Instead of trusting a single trend line, we resample the data thousands of times, and redraw the trend line each time. If it tilts consistently in the same direction across all these redraws, we have high statistical confidence that the trend is real. If it flips around, it means the original line was distorted by a few low-checklist years.

By looking at the trend direction and our statistical confidence, we sort each species into one of six distinct trend labels. Details of the implementation are explained here.

To compare bird trends across lakes, we assigned scores to each label. Definitive trends ( Increasing or declining) are scored higher compared to speculative (like Likely increasing) or unreliable ones (Indeterminate). 

Trend Label	Trend Score
Increasing	0.9
Declining	-0.9
Likely Increasing	0.7
Likely Declining	-0.7
Stable	0.5
Indeterminate	0

We defined two additional labels based on the absence or presence of a species at a lake. Because the sudden absence or appearance of a species at a lake conveys a significant ecological signal, the highest scores are assigned to these labels.

Trend Label	Criteria	Trend Score
Gained	Absent in first half of study period, present in both 2024 and 2025	1
Lost	Absent in both 2024 and 2025	-1

Residents vs migrants

The results of the trend label analysis for lakes in the Hebbal cascade are shown in the figure below:

Individual lakes within the same cascade can have very different ecological trajectories. For example, the decline of migrants in Yelahanka is striking, with almost 74% of them either declining or already lost.  But, in Kalkere and Rachenahalli, migrants are faring better than residents, while Jakkur shows no discernible difference. Species richness across the lakes showed a similar pattern, indicating that local management of lakes influence ecological health more than cascade connectivity.

This is consistent with the evolution of the city’s Rajakluves over the years. While they once served as vegetated buffers providing wetland connectivity between lakes, satellite imagery shows that areas around them have been completely built out. Further, the Rajakaluves themselves have been concretized, transforming them into channels delivering pollutants.

Despite these variations, the cascade framework remains useful for investigating ecological health across different kinds of lakes in the city. 

Migrants are showing a sharp decline in the Sowl cascade. In Sowl, a migratory hotspot, 50% of the migrant species are declining while 22% have been lost. Migrants also fare worse than residents in Kasavanahalli and Kaikondranahalli. 

The results are the starkest in South Bengaluru lakes where migrant populations are crashing across all the three lakes in our study.

To establish if the faster decline of migrants is statistically significant, we applied the Mann-Whitney test. Details of the implementation are provided here. The results are shown below:

Pooled across all 14 lakes, migrants average a trend score of -0.29 (slight decline), compared to -0.045 for residents (essentially flat). This gap is statistically significant establishing that migrants are declining significantly faster than resident species across these lakes. While this analysis does not cover all of Bengaluru’s 181 lakes, the consistency of the migrant decline signal across most of the 14 geographically distributed lakes with diverse management histories suggests a genuine city-wide pattern rather than a statistical artefact.

Migration is a complex phenomenon affected by multiple external factors.The 2023 State of India’s birds survey also points to the decline of several migratory waterbirds across India. However, the fact that lakes within Bengaluru show different ecological trajectories points at local factors at play. 

Are the declines concentrated in specific guilds? Do some lakes support certain microhabitats better? We tackle these questions in our next article.

[The team would lie to acknowledge:

• Rashmi Kulranjan, Hydrologist at WELL Labs, for sharing datasets related to cascade connectivity and pointing out the pertinent ecological issues associated with Bengaluru lakes today.
• Ashwin Viswanathan, Scientist at Nature Conservation Foundation, for advising us on waterbird guild classification and interpreting eBird data analysis results.]

Also read:

• Why all Bengaluru lakes need a biodiversity survey: Lessons from Dorekere
• Photos: A decline in birds in Navi Mumbai
• Flooded lands, disappearing birds: Pallikaranai wetland tour highlights impact of shrinking lakes