The temperature live chart may contain some patterns that I did not spell out or seem not to correct/expected. This is the place that I offer my best observation.
First, let look at the temperature from www.timeanddate.com of Hanoi, Vietnam.
The temperature curve in one day time is symmetrical reflecting the solar energy close the Earth's ground level. With an all-day sunshine, the maximum temperature is observed around 12 PM (or the noon). The solar energy influences strongly to the observed temperature. For example, in Oct 10, a rainy morning and a late sunshine around noon skewed the pattern and pushed the peak temperature in the after noon, which is different than one in Oct 13. The peaks were a moderate increment to neighboring values. These patterns would interpret the temperature recorded by shaded and outdoor station.
I hope to this point, by comparing a more trusted source than my source, you can pick something out. My temperature station is placed outdoor in a balcony that is west facing. Around 3-4 PM, the Sun shines directly to the area of the station, and the temperature sensors are now picking the temperature with direct sunlight. The needle-shape in Oct 12 in the middle the afternoon is a result of my location. A cloudy day like Oct 13, the temperature pattern is more symmetrical, and the effect of the direct sunlight is smaller and the peak was moderate and no needle-shape increment was observed.
The self-heating effect on the recorded temperature is challenging to eliminate. The some sensors uses resistance to measure temperature change and most others uses a voltage change to extrapolate the temperature. This all means that the operation of sensors generates some heat. In my case, a temperature reading is averaged from 4 independent sensors, two DS18B20s, one thermistor from DHT22, and another one from SHT21. The DS18B20 sensors are used direct-to-digital temperature patented by Dallas Semiconductor Corp. The DHT22 is used of a thermistor with a fix-base resistor and SHT21 with a bandgap temperature sensor. Those sensors are not known for a high self-heating probes. The ESP8266 MCU is operated with the deepsleep mode and the heat generated by its operation is minimal as well. I noted that charging the Lithium battery generates extra heat and was picked up by the sensors as seen in the graph in Oct 12. The gray-shaded area marked for an out-of-power period followed up by a charging for 1.5 hours.
It is easy to say the sunlight was the cause without a concrete evidence. The figure below shows the data captured by the same weather station in a last couple of days.
The light sensors used are a low-cost light resistant diode (LDR), and the delay between the peak of light intensity and the peak of temperature is logical. It would take certain amount of time to heat up the box, and in turn the delay of lowering the temperature after the light was low. The magnitude of light intensity is more different with luminosity sensor (the live data is here). And now we have a rather conclusive evidence of causation between the direct sunlight and the recorded temperature.