Taiwan Reservoir Transparency

WRA publishes daily operational data for all officially announced major reservoirs in Taiwan via the consolidated dataset 45501. Each record includes reservoir name, observation timestamp, accumulated watershed rainfall, inflow, water level, full water level, effective storage capacity, storage percentage (蓄水率), outflow, and flood prevention operation status — 16 fields in total, confirmed available in JSON, CSV, and XML. The WRA Disaster Emergency Response System generates this data from self-recording instruments at each reservoir management unit. Selected individual datasets (Tsengwen 32733, Agongdian 32734) provide per-reservoir feeds with explicit hourly cadence and a richer schema (AccumulatedRainfall, EffectiveWaterStorageCapacity, PrecipitationHourly, ReservoirIdentifier, ReservoirName, WaterLevel, DateTime). Siltation data (32727) tracking DesignedCapacity, DesignedEffectiveCapacity, CurrentCapacity, CurrentEffectiveCapacity and TheLatestMeasuredTimeOfReservoirCapacity is published annually. The main gap is that Taipower-managed hydropower reservoirs (Deji/Dechi, Mingtan pumped storage, Minghu) are not integrated into the WRA central open data system; their operational data is not transparently published.

Taiwan provides multiple programmatic access routes of high quality. The primary WRA open data platform (opendata.wra.gov.tw) exposes a documented REST API conforming to the OpenAPI Specification (OAS), with a Swagger UI at opendata.wra.gov.tw/openapi/swagger/index.html and a machine-readable OAS document at opendata.wra.gov.tw/openapi/api/OpenData/openapi. Each dataset on data.gov.tw exposes stable, parameterised endpoints — e.g., the consolidated reservoir dataset is queryable at https://opendata.wra.gov.tw/api/v2/2be9044c-6e44-4856-aad5-dd108c2e6679 in JSON/CSV/XML, and the siltation dataset at https://opendata.wra.gov.tw/api/v2/572bda99-0593-4aee-9409-03c82423f8eb. The Flood Hazard portal (fhy.wra.gov.tw) offers a separate FHY General API v2 with documented reservoir endpoints. All data is released under the Open Government Data License v1.0 (free, no key required for most endpoints). The score is held below the highest tier because the FHY API documentation is almost entirely in Traditional Chinese, and the parallel data.wra.gov.tw Swagger surface is less prominent.

Conservative estimate — denominator includes Taipower-managed hydropower reservoirs (Deji/Dechi, Mingtan pumped storage, Minghu) not integrated into the WRA central open data system, plus small irrigation ponds managed by local irrigation associations and minor private water bodies. Methodology v1.3.0 (capacity-weighted). The WRA consolidated dataset 45501 (~1,900 hm³ across officially announced major reservoirs) captures the strategic drinking-water, irrigation and multipurpose reservoirs, but a realistic national denominator that includes Taipower hydropower storage and the long tail of small reservoirs reaches approximately 2,100 hm³. Score = round(100 × 1,900 / 2,100) = 90. The capacity-weighted result still scores well because Taiwan's strategic supply reservoirs are essentially fully tracked, but the systematic exclusion of Taipower hydropower data (a non-trivial slice of national impounded water) prevents a higher score under a conservative denominator.

Historical reservoir data exists but its accessibility through documented official channels is uneven. The fhy.wra.gov.tw disaster prevention portal includes an interactive reservoir comparison tool allowing custom date-range queries with daily 7 AM snapshots. The siltation dataset (32727) provides annual records of design vs. current capacity with the date of the most recent bathymetric survey, allowing multi-decade reconstruction of effective-capacity loss. Academic literature (MDPI, ScienceDirect) and the IHA case study confirm WRA archives volumetric records for Zengwen back to 1973 and detailed siltation surveys back decades. Third-party aggregators built on WRA data (reservoir.cqd.tw) chart series from roughly 2003 onwards but as of this evaluation the site shows '資料更新中' (data updating) status, indicating the public-facing historical aggregation is not entirely stable. The score is held at 78 because: (a) the official historical query API is not clearly documented in English, (b) bulk download of multi-decade daily records requires interactive navigation of the FHY portal rather than a single documented endpoint, and (c) the consolidated dataset 45501 is essentially a current-state feed without exposing a depth parameter.

Update cadence in Taiwan is tiered and asymmetric. Real-time stream/river water level stations (dataset 25768) record observations every 10 minutes — best-in-class. However, for reservoirs specifically, the WRA explicitly documents that the consolidated dataset 45501 has a nominal hourly update frequency but 'updates are generally made once daily, limited by the different data update frequencies provided by each reservoir management unit.' Only a subset of individual reservoir datasets (Tsengwen 32733, Agongdian 32734) deliver guaranteed hourly cadence end-to-end. During typhoon and flood events, refresh rates increase operationally within the Disaster Emergency Response System. This is solid daily-to-hourly coverage but does not match the consistent 10-minute reservoir cadence offered by Australia (BoM) or the sub-hourly federal cadence of the US (USGS NWIS / USBR HDB), so the score is reduced from a top-tier mark. Score reflects honest daily-to-hourly cadence for the bulk of reservoirs rather than the aspirational hourly figure on dataset metadata.

Taiwan is one of the few countries that publishes a dedicated, machine-readable siltation dataset (32727) reporting DesignedCapacity, DesignedEffectiveCapacity, CurrentCapacity, CurrentEffectiveCapacity, ReservoirSedimentationVolume and the date of the most recent bathymetric survey — making capacity degradation explicit and quantifiable. Data collection methods (automated self-recording instruments supplemented by manual observation by reservoir management units, with each unit responsible for inputting data to the WRA central system) are documented in dataset metadata on data.gov.tw. Full water level and dead water level are published alongside operational levels for each site. The Swagger API documents fields with English identifiers. The score is held below 85 because: (a) a unified technical document covering measurement standards, instrument calibration, and quality-control protocols is not consolidated into a single publicly accessible English document, (b) the operational dependence on each reservoir management unit's own QA introduces heterogeneity, and (c) Taipower hydropower facilities operate under different and less transparent methodological regimes.

Taiwan provides strong English-language access to its reservoir data ecosystem, though with notable rough edges. The national open data portal data.gov.tw has a complete English interface (data.gov.tw/en) with English dataset titles, descriptions, and field metadata for all reservoir datasets verified in this evaluation (45501, 32733, 32734, 32727). API field names use English identifiers (ReservoirName, EffectiveWaterStorageCapacity, WaterLevel, DateTime, StorageCapacityPercentage, AccumulatedRainfall, PrecipitationHourly), enabling international developers to consume the API without Traditional Chinese language skills. The Swagger UI at opendata.wra.gov.tw is bilingual-friendly. Deductions reflect: (1) the FHY disaster prevention portal (fhy.wra.gov.tw), its API documentation, the reservoir comparison tool, and the historical query interfaces are almost entirely in Traditional Chinese; (2) the English WRA portal (eng.wra.gov.tw) returned connection errors during this evaluation, indicating intermittent availability of the English-language gateway; and (3) the OpenData platform landing page title and several secondary menus remain in Chinese only.