Читать книгу Earth Observation Using Python - Rebekah B. Esmaili - Страница 20

The missions mentioned in the previous section provide open and free data to all users. However, data delivery, the process of downloading sensor data from the satellite and converting it into a usable form, is not trivial. Raw sensor data are first acquired on the satellite, then the data must be relayed to the Earth’s ground system, often at speeds around 30 Mbits/second. For example, GOES satellite data are acquired by NASA’s Wallops Flight Facility in Virginia; data from the Suomi NPP satellite is downloaded to the ground receiving station in Svalbard, Norway (Figure 1.4). Once downloaded, the observations are calibrated and several corrections are applied, such as an atmospheric correction to reduce haze in the image or topographical corrections to adjust changes in pixel brightness on complex terrain. The corrected data are then incorporated into physical products using satellite retrieval algorithms. Altogether, the speed of data download and processing can impact the data latency, or the difference between the time the physical observation is made and the time it becomes available to the data user.

Data can be accessed in several ways. The timeliest data can be downloaded using a direct broadcast (DB) antenna, which can immediately receive data when the satellite is in range. This equipment is expensive to purchase and maintain, so usually only weather and hazard forecasting offices install them. Most users will access data via the internet. FTP websites post data in near real time, providing the data within a few hours of the observation. Not all data must be timely – research‐grade data can take months to calibrate to ensure accuracy. In this case, ordering through an online data portal will grant users access to long records of data.

While data can be easily accessed online, they are rarely analysis ready. Software and web‐based tools allow for quick visualization, but to create custom analyses and visualizations, coding is necessary. To combine multiple datasets, each must be gridded to the same resolution for an apples‐to‐apples comparison. Further, data providers use quality flags to indicate the likelihood of a suitable retrieval. However, the meaning and appropriateness of these flags are not always well communicated to data users. Moreover, understanding how such datasets are organized can be cumbersome to new users. This text thus aims to identify specific Python routines that enable custom preparation, analysis, and visualization of satellite datasets.

Figure 1.4 NOAA‐20 satellite downlink.