Saurav Bhattarai Hydrologist
Research

From hydroclimatic extremes to floods you can model.

My research runs from how rainfall, heat, and their combinations are changing, through making open satellite data fit a valley, to an open physics-based model that turns those data into floods. It spans an M.S. on heat risk and urban cooling and a Ph.D. on open Earth data and flood modeling, plus collaborations on climate projections and multi-hazard risk. Below, the work is grouped by theme; each degree also has its own page.

Ph.D. dissertation · 2023–2027

Accessing, Refining, and Combining Open Earth Data for a Modular, Parsimonious, Physics-Based Hydrological Model

Five chapters, Access, Refine, Translate, Apply, and Extend, built around MRRpy. Defense February 10, 2027.

The dissertation, chapter by chapter M.S. thesis · May 2025

Heatwave Dynamics and Urban Cooling: From Risk Assessment to Nature-Based Solutions

Four studies, from heat risk in Mississippi counties and 50 U.S. cities to the cooling of parks, ponds, and forests in Kathmandu. Five papers.

The thesis, study by study
01

Hydroclimatic extremes

How rainfall and heat extremes are changing, and when they arrive: long records from ERA5 and gauges, and CMIP6 projections, from the whole globe down to single cities.

Ph.D.M.S.Collaborations
  • The timing of the wettest days changes independently of their intensity: the Amazon reaches its annual total in 15–18% fewer days, while eastern North America needs 12–25% more.
  • Heatwave months increased in all 44 IPCC land regions over 1950–2025, by 0.26 to 1.20 months per decade (Animation 1).
  • Across the 50 most populous U.S. cities, heatwave frequency rises by 200–300% by the end of the century under SSP5-8.5; coastal cities get more frequent but less intense heatwaves than inland ones.
  • Projected precipitation extremes across the Mississippi River Basin, and hydro-climatic extremes in the rain-fed rivers of western Nepal.
Animation 1Heatwave months per year in 10-year windows, 1950–2025, with the land-mean series (ERA5).
02

Compound and cascading hazards

Hazards rarely arrive alone: heat with drought, rain with storm surge, rain with humid heat. I measure how often they coincide, why that is changing, and what follows downstream.

Ph.D.M.S.Collaborations
  • Compound heatwave–drought months more than doubled, from 0.80 to 1.70 per year between 1950–1979 and 1996–2025, with the fastest rise in the driest regions (Animation 3).
  • At 79 U.S. tide gauges, joint extreme surge and rainfall occur 6.9 times more often than chance. Compound flood days rose from 1.03 to 1.79 per year, and with sea level held at its earlier level there is no increase.
  • The day after an extreme surge, river floods are twice as likely as chance for about a week: compound becomes cascade, as in Hurricane Florence (Animation 2).
  • Rain ends heatwaves but often leaves humid air behind, so heat stress can persist; heat and humidity during heatwaves intensified after 2000 across the eastern United States.
Animation 2Hurricane Florence, September 2018: surge at the tide gauges, rain, and the Cape Fear River rising days later at Wilmington, North Carolina.
Animation 3Compound heatwave–drought months per year in 10-year windows, 1950–2025, with the land-mean series.
03

Earth observation and data refinement

Making coarse satellite data fit a valley, and testing them against data they never saw: groundwater from GRACE, soil moisture from SMAP, rainfall from IMERG, and published ERDC algorithms in Earth Engine.

Ph.D.Collaborations
  • GRACE groundwater downscaled from 55 to 5 km by an explainable machine-learning ensemble that averages exactly back to GRACE; High Plains storage change matches USGS (r = 0.84) (Animation 4).
  • SMAP soil moisture at 1 km that preserves each 9 km pixel; agreement with radar-based labels (r ≈ 0.8) is not accuracy against independent data (r ≤ 0.35) (Animation 5).
  • IMERG V07 over the Kathmandu Valley shifted after 2016 because its gauge input disappeared; the volume correction keeps bias within ±4%.
  • SERVES root-zone soil moisture evaluated at 2,643 stations worldwide: useful where vegetation controls soil water, not in cold climates.
  • J3Satellite precipitation across the CONUS and Hawaii Remote Sensing, 2024
  • ReviewMultiscale evaluation of IMERG across the United States (with Howard University undergraduates)
  • Prep.Explainable GRACE groundwater downscaling; the 2017 IMERG shift over the Kathmandu Valley
Animation 4GRACE at 55 km beside the downscaled 5 km field, month by month; averaged back to 55 km, the fine field is exactly GRACE.
Animation 5Daily 1 km soil moisture (latest pass within 3 days), with a zoom on central Iowa where every 1 km cell is visible.
04

Physics-based flood modeling

MRRpy, the open, modular, parsimonious model I develop, turns open data into hydrographs. I test it on observed, design, long-term, and outburst floods, and against GSSHA.

Ph.D.Collaborations
  • Eight Bagmati floods at Khokana without event-wise fitting: median Nash–Sutcliffe efficiency 0.81, all peaks within 30%; for 30–31 July 2024 the efficiency at Khokana is 0.86 (Animation 6).
  • Runoff physics and routing switch with one setting; in the September 2024 storm the saturated area grew from 0.6% to 36% of the basin (Animation 7).
  • The 26 August 2026 outburst surge, routed as an upstream boundary, arrives within about 20 minutes of the reported times along the Trishuli; exploratory (Animation 8).
  • Flood risk beyond the model: local-scale flood hazard projections in vulnerable communities and flood risk in the Kathmandu Valley.

How MRRpy is built and verified

Animation 6MRRpy routing the 31 July 2024 flood through every river cell to Khokana (log-scale flow), with observed and simulated discharge.
Animation 7Saturated area (blue) spreading across the Bagmati basin in the September 2024 storm, with rainfall and the saturated share of the basin.
Animation 8The 26 August 2026 outburst surge entering as an upstream boundary and travelling down the Bhote Koshi and Trishuli, with modeled and reported arrival.
05

Heat risk and urban cooling

The core of my M.S.: who is most exposed to heat, how that will change, and how much parks, ponds, and forests cool a dense city.

M.S.Collaborations
  • A heatwave risk index that joins five heatwave indices with social vulnerability flags northwest Mississippi (Bolivar, Tallahatchie, Coahoma) as the highest-risk counties.
  • In the Kathmandu Valley, cooling grows exponentially with the size of a park, pond, or forest; forests cool up to 1.2 °C at 300 m, large parks about 1.15 °C within 120 m (Figure 1).
  • The same green space cools very differently in urbanized, transitional, and vegetation-dominated neighborhoods, so cooling strategies have to fit the urban fabric.
  • Urban form controls heat and moisture anomalies across U.S. cities.

The M.S. thesis, study by study

Figure 1Kathmandu Valley: blue-green spaces; population density; maximum temperature trend; land cover, 2000 and 2019.
06

Open tools and education

The tools that carry this work to other people: the GeoClimate Intelligence Platform, MRRpy with its documentation and course, and course platforms for students.

Ph.D.Teaching
  • 1,042 datasets and 23 climate indices without code in the GeoClimate platform; Best in Show at the Microsoft × AUC Data Science Initiative Data Challenge.
  • Used by 16 undergraduates in Advanced Water Resources Engineering and in a faculty workshop with 28 registered faculty; a 30-year daily rainfall analysis runs in under a minute (walkthrough).
  • MRRpy on PyPI with 11 worked examples and a free six-module course that runs in the browser (Animation 9).

All software

WalkthroughA real GeoClimate run: 30 years of CHIRPS rainfall around Jackson, Mississippi, from tool to download.
Animation 9The free interactive course: a D8 flow-direction explorer that runs in the browser.