Enhancing data-driven decision making for flood governance in Bihar and Uttar Pradesh, India

Around 130 million people in the Indian states of Uttar Pradesh and Bihar, situated along the Ganges and its tributaries, are regularly exposed to floods – equivalent to about seven percent of the global flood-prone population.

With disaster risks set to intensify due to climate change, the governments of both states are introducing measures to mainstream data-driven decision-making and climate resilience into their planning and public finance systems. At the interdisciplinary research lab  CivicDataLab, we are working with Open Contracting Partnership to support this shift through our Intelligent Data Solution for Disaster Risk Reduction (IDS-DRR), a data-driven platform that makes it easier for authorities to prioritize spending based on the needs of the communities most vulnerable to flood risks. Uttar Pradesh and Bihar are the fourth and fifth states to adopt IDS-DRR in India. The platform is already being used in the state of Assam to inform the allocation of US$11.7 million in disaster response funding for the 2025 monsoon season, while Himachal Pradesh and Odisha are preparing to deploy full-scale versions of the solution after piloting it successfully in 2024.

Scaling smart disaster management

The five states have similar data models since our long-term ambition is to make the IDS-DRR easy for others to adopt and adapt. However, each location has unique characteristics that affect vulnerability or resilience and we worked hard to tailor the solution to local needs, starting by conducting comprehensive landscape studies. In Uttar Pradesh and Bihar – two of India’s most populous states – sheer size is a key factor. We also worked closely with officials and communities to validate the data models, improve them, and promote long-term uptake of the solution among public institutions in their decision-making processes. This included consultations with the Uttar Pradesh State Disaster Management Authority (UPSDMA) and the Bihar State Disaster Management Authority (BSDMA). 

Developing the data models

We completed the initial scoping of primary baseline datasets, which have been made publicly available through the open UP Flood Data and Bihar Flood Data repositories. 

Data models are being developed for the 75 districts in Uttar Pradesh (UP Risk Score Model) and 38 districts in Bihar (Bihar Risk Score Model) using the internationally recognized Sendai Framework. Through our experience developing models for other states, we defined a model for replicability, focusing on datasets that are easily accessible across geographies and for the time period of interest.

These datasets include:

• Mission Antyodaya for socio-economic indicators,
• WorldPop for population estimates,
• Bhuvan for flood inundation, and
• Indian Meteorological Department (IMD) for rainfall. 

To this, we added government response datasets in the form of State Disaster Response Funds and tender funds that were accessible through public sources. 

We categorized indicators into factors that were used to model various aspects of flood risk. The table below lists the factors and their associated indicators.

An alpha version for both states is ready for testing internally with the baseline datasets feeding into the respective factor scores for Exposure, Vulnerability, Hazard and Government Response. 

Data quality issues undermine flood risk analysis 

We identified several issues with the publicly available datasets that constrained the analysis of the effectiveness of the states’ flood response measures. 

There was very little overlap in time periods for the different datasets we needed (e.g. losses and damages versus flood response), so we invested significant time to standardize this. 

The two states’ procurement platforms are also different from other states. We tailored our data scraping methodologies, which limited opportunities for streamlined data integration. Data was also contained in PDF documents, often in Hindi, requiring advanced extraction methods based on Optical Character Recognition (OCR) and Large Language Models (LLMs).

Data suggests flood relief is not reaching all affected districts

We explored the relationship between flood impact and flood-related spending by comparing available losses and damages data (e.g. houses damaged, lives lost and crop damages) with available data for tenders and State Disaster Response Funds (SDRF). The goal was to use geospatial analysis to assess the correlation and draw insights from data available through the limited public and official sources. 

The data sources and time period of available data are listed in the table below:

We analyzed the correlation in such a way that makes it possible to compare the flood impact and funding received between different locations.

We visualized our analysis through maps to identify high-risk districts that face high flood impact but have low funds to address it (marked in red). Districts that have low damages but high rates of flood-related funding were classified as very low risk, indicating mismatched funding (marked in blue).

Using the analysis of flood risk and flood funding indicators, in Bihar, we identified that districts such as Muzaffarpur, Vaishali, Sitamarthi, Madehepura, Supaul, Samastipur and Sitamarhi were not recipients of SDRF funds commensurate with the level of losses and damages they faced. On the other hand, districts such as Kaimur, Jehanabad, Nawada and Jamui were identified as having a low overall flood risk, but still receiving a disproportionate amount of funds. 

In Uttar Pradesh, the inundation dataset indicated that Siddharthnagar, Gorakhpur, Sambhal, Maharajganj, and Ghazipur were among the most flood-impacted districts. However, between 2020 and 2024, districts like Aligarh, Ayodhya, Agra, and Prayagraj with less severe historical flood records were the top recipients of flood-related tender funds. The allocation gap analysis also identified Siddharthnagar, Maharganj, Sambhal and Agra as potentially requiring attention towards equitable flood resource allocation.

This suggests that current fund distribution mechanisms are not sufficiently aligned with ground-level flood risk, potentially undermining the effectiveness and equity of flood response and resilience measures. The potential mismatches in resource allocation identified by the analysis require more evidence and supporting datasets to generate actionable insights. Future analysis will be augmented using more locally relevant datasets transformed for interoperability and a more robust data model.

Way forward 

The introduction of IDS-DRR has the potential to support flood resilience initiatives in Uttar Pradesh and Bihar. We have cleaned and opened $59 million and $255,000 worth of tenders in Bihar and Uttar Pradesh, respectively. We have developed the risk score models across 113 districts of both states and have received commitments from the respective governments to continue collaboration towards data-driven disaster governance.

Scope

In the coming months, we will work to update the data models for Uttar Pradesh and Bihar by adding more relevant datasets. These will include Digital Elevation Models (DEM), complete tender datasets, SDRF, and losses and damages. In case they are not available through public sources, we will work with relevant departments to obtain them at the sub-district level. Based on expanded datasets, we will run additional analyses to draw conclusive insights by interlinking completed procurement datasets with disaster datasets.

Process

Based on our experience of collaborating with the stakeholders in the states, we will work with the respective disaster management authorities to identify concrete actions that will help further improve the states’ risk data capacity towards data-driven decision-making. We will engage with relevant departments associated with disaster response (both finance and procurement) to identify ways in which their data processes could be improved to assist with flood risk analysis. This could involve identifying datasets, transforming them for interoperability, recommending pipelines and data formats towards addressing the challenges highlighted by UPSDMA and BSDMA of collating standardized and interoperable data.

Capacity building

We are working to develop the capacity of the Bihar and Uttar Pradesh governments in terms of embedding IDS-DRR insights into their disaster preparedness and planning systems.

These practical steps will help institutionalize IDS-DRR as a versatile tool for flood risk reduction and climate resilience in Uttar Pradesh and Bihar, enabling state and district administrations to make informed decisions and demonstrating the potential of data-driven systems to strengthen disaster preparedness and climate action.

Our heartfelt thanks to the Patrick J McGovern Foundation whose leadership and continued generosity helped us scale this work to Bihar and Uttar Pradesh, two of the largest Indian states – with a combined population of over 370 million people. Our appreciation also goes to The Rockefeller Foundation who previously supported this work in Assam, Himachal Pradesh and Odisha.