Chiayi County is located in the largest alluvial plain of Taiwan with extensive aquaculture and rice farming sustained by water extracted from groundwater wells. Chiayi is a typical aquaculture area affected by land subsidence, yet such lands worldwide combine to provide nearly 90% of global aquaculture products, greatly reducing oceanic overfishing problems. This study uses precision leveling, multi-layer compaction monitoring well (MLCW) and spaceborne SAR interferometry (InSAR) to examine the cause and effect of land subsidence in Chiayi associated with groundwater extractions and changes. Heights at benchmarks in a leveling network are measured annually and soil compactions at 24–26 layers up to 300-m depths at 7 MLCWs are collected at one-month intervals. Over 2007–2011, 15 ALOS/PALSAR images are processed by the method of TCPInSAR to produce subsidence rates. All sensors show that land subsidence occur in most parts of Chiayi, with rates reaching 4.5 cm/year around its coast, a result of groundwater pumping from shallow to deep aquifers. MLCWs detect mm-accuracy seasonal soil compactions coinciding with groundwater level fluctuations and causing dynamic compactions. Compactions near Taiwan High Speed Rail may reduce the strength of the rail’s supporting columns to degrade its safety. The SAR images yield subsidence rates consistent with those from leveling and compaction wells after corrections for systematic errors by the leveling result. Subsidence in Chiayi’s coastal area leads to relative sea level rises at rates up to 15 times larger than the global eustatic sea level rising rate, a risk typical for world’s aquaculture-rich regions. At the fish pond-covered Budai Township, InSAR identifies subsidence spots not detected by leveling, providing crucial geo-information for a sustainable land management for aquaculture industry.

Specific yield (S_y) and specific storage (S_s) are two hydrogeological parameters often estimated in a costly pumping test (hydraulic method). Pumping or a natural aquifer recession can cause gravity change and surface deformation, useful for S_y and S_s determination by the more economic geodetic method. From 2013 to 2017, over a few days we measured gravity changes at four sites and subsidence at one site in Taiwan by a FG5 gravimeter and a precision level to estimate S_y and S_s. Using short‐time gravity and level records avoids complicated logistic supports and temporal effects. The measured gravity changes are associated with the MODFLOW‐modeled groundwater depletion. We succeeded in estimating S_y at two sites and failed at the other two. One successful case uses rapid postrain declines of gravity and groundwater level. We did leveling only at one site, but here the 500‐min, continuous, submillimeter subsidence records result in a S_s consistent with the hydraulic result. Lessons on the failed cases of S_y are suggested to avoid disrupting factors in gravity surveys. A semiautomatic leveling procedure is proposed to maximize the chance for S_s determination. Our simulations show that gravity changes increase with S_y and decrease with initial hydraulic head and horizontal gravimeter‐pumping well distance, helping to predict if the gravity‐based method works at a site and to best plan a gravity survey. Drilling data show that the S_y values from the hydraulic and geodetic methods represent aquifer storage coefficients over different aquifer layers.

In this study, we attempt to establish an alternative method for estimating the groundwater levels and the specific yields of an unconfined aquifer for the evaluation of potential groundwater reservoirs. We first converted the inverted resistivity into the normalized water content. Then, we inverted the parameters of the Brooks-Corey model from the vertical profiles of the water content by assuming that the suction head was in proportion to the elevation regarding a predefined base level. Lastly, we estimated the groundwater level, the theoretical specific yield, and the specific yield capacity from the Brooks-Corey parameters at every survey site in the study area. The contour maps of the time-lapse groundwater levels show that the groundwater flows downstream, with a higher hydraulic gradient near the river channel than in the area away from the main channel. We conclude that the estimated maximum specific yield capacities are consistent with that derived from the pumping tests in the nearby observation well. Additionally, the specific yield capacities are only three quarters to two thirds of the theoretical specific yields derived from the difference between the residual and saturated water contents in the Brooks-Corey model. We conclude that the distribution pattern of the specific yields had been subjected to the distribution of natural river sediments in the Minzu Basin, since the modern channel was artificially modified. Although we had to make some simple assumptions for the estimations, the results show that the surface resistivity surveys provide reasonable estimations of the hydraulic parameters for a preliminary assessment in an area with few available wells.

In the paper, we propose an acceleration-based and angular-velocity-based vehicle attitude recognition method by using a popular deep learning algorithm, i.e., Convolution Neural Network (CNN). We use an Inertial Measurement Unit (IMU) to collect six-axial signal of a vehicle. In particular, we construct a CNN model to learn the characteristics of six-axial IMU signal and the model can be used to predict vehicle attitudes. We constructed training data consists of 800 package from six attitudes. In addition, we preprocess the 800 package that each package will be broken down. Finally, our training data is 59200 sample-train. The experiment results show that the CNN works well, which can reach an average accuracy of 98% by the time of 1/5 of the overall action without any feature extraction methods. Because we use CNN model that it's the number of convolution kernel is less, we can reach real-time. Each estimated time is less than 0.5 sec based on the raspberry pi3.

Accurate and densely covered specific yields (S_y) are essential for estimating the storage capacity of a groundwater reservoir. A cross‐well pumping test can determine S_y, but its high cost often makes it unsuitable for sampling high‐resolution S_y. The gravity‐based method (GBM) based on gravity changes near existing groundwater wells may outperform cross‐well pumping tests in estimating S_y. We established 10 gravity sites close to groundwater wells in the aquifer‐rich Choushui River Alluvial Fan and Mingchu Basin in central Taiwan and measured gravity changes with two FG5 gravimeters from 2012 to 2017. Thirty‐nine S_y values with formal errors are determined by natural rises and falls in gravity and groundwater level. The representative S_y values (0.04 to 0.29) from GBM are in general consistent with those from cross‐well pumping tests (0.03 to 0.24). Repeated groundwater level changes over similar depth range at different times serve as revisit tests, showing that GBM can reproduce a reliable S_y value at a given site and depth. Soil moisture and compaction data show that the effects of gravity changes originating from unsaturated zones and deep aquifer layers are minor. Using the cylinder model for aquifers with limited lateral extents, we assess the validity of the Bouguer model by quantifying gravity differences and relative S_y differences originating from the model assumption. Improvements in environmental resilience and transportability achieved by recent atomic gravimeters may increase the potential of GBM to replace or supplement cross‐well pumping tests in densifying S_y point densities for an improved groundwater resource management.