Some variables in X 1, X2, X3 and X4 are related. For example, the affected area may be related to the number of reservoirs (an increase in the number of reservoirs may reduce the affected area), or there is an unobserved factor affecting Y in the remaining part, and this factor is related to one or more of X 1, X2, X3 and X4.

The existence of this correlation will lead to the failure to meet the exogenous hypothesis in OLS hypothesis, thus making the estimation result inaccurate.

Therefore, under this correlation condition, X 1, X2, X3 and X4 may not be significant, or the estimation results may not satisfy the theory. For example, in the first regression equation, x 1 (sown area) should be significantly positively correlated with y (crop yield) in theory, but the estimated result is not significant; In addition, in the second regression equation, x4 (number of reservoirs) should be positively correlated with Y (crop yield) in theory, but the estimated results are significantly negatively correlated.

So there is something wrong with both your regression equations.

Processing methods include using panel data, GMM regression, or adding control variables.